Representational Deficits in SLA: Studies in honor of Roger Hawkins (Language Acquisition and Language Disorders)

Representational Deficits in SLA

Language Acquisition and Language Disorders (LALD) Volumes in this series provide a forum for research contributing to theories of language acquisition (first and second, child and adult), language learnability, language attrition and language disorders.

Series Editors Harald Clahsen

University of Essex

Lydia White

McGill University

Editorial Board Melissa F. Bowerman

Max Planck Institut für Psycholinguistik, Nijmegen

Katherine Demuth Brown University

Wolfgang U. Dressler Universität Wien

Nina Hyams

University of California at Los Angeles

Jürgen M. Meisel

Universität Hamburg

William O’Grady

University of Hawaii

Luigi Rizzi

University of Siena

Bonnie D. Schwartz

University of Hawaii at Manoa

Antonella Sorace

University of Edinburgh

Karin Stromswold Rutgers University

Jürgen Weissenborn Universität Potsdam

Frank Wijnen

Utrecht University

Mabel Rice

University of Kansas

Volume 47 Representational Deficits in SLA. Studies in honor of Roger Hawkins Edited by Neal Snape, Yan-kit Ingrid Leung and Michael Sharwood Smith

Representational Deficits in SLA Studies in honor of Roger Hawkins

Edited by

Neal Snape Gunma Prefectural Women's University

Yan-kit Ingrid Leung University of Hong Kong

Michael Sharwood Smith Heriot-Watt 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 Representational deficits in SLA : studies in honor of Roger Hawkins / edited by Neal Snape, Yan-kit Ingrid Leung, Michael Sharwood Smith.        p. cm. (Language Acquisition and Language Disorders, issn 0925-0123 ; v. 47) Includes bibliographical references and index. 1.  Second language acquisition. 2.  Grammar, Comparative and general--Syntax. 3.  Generative grammar.  I. Hawkins, Roger (Roger D.) II. Snape, Neal. III. Leung, Yan-kit Ingrid. IV. Sharwood Smith, Michael, 1942- V. Series: Language acquisition & language disorders ; v. 47. P118.2.R45   2009 418.0071--dc22 isbn 978 90 272 5308 8 (Hb; alk. paper) isbn 978 90 272 8999 5 (eb)

2008045265

© 2009 – 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

Table of contents Dedication Preface Acknowledgements Introduction Neal Snape, Yan-kit Ingrid Leung and Mike Sharwood Smith Prosodic transfer and the representation of determiners in Turkish-English interlanguage Heather Goad and Lydia White Exploring Mandarin Chinese speakers’ L2 article use Neal Snape Successful features: Verb raising and adverbs in L2 acquisition under an Organic Grammar approach Anne Vainikka and Martha Young-Scholten Non-permanent representational deficit and apparent target-likeness in second language: Evidence from wh-words used as universal quantifiers in English and Japanese speakers’ L2 Chinese Boping Yuan

vii ix xiii xv

1

27

53

69

Acquisition of the local binding characteristics of English reflexives and the obligatory status of English objects by Chinese-speaking learners Lin Jiang

105

Selective deficits at the syntax-discourse interface: Evidence from the CEDEL2 corpus Cristóbal Lozano

127



Representational Deficits in SLA

Clitic doubling and clitic left dislocation in Spanish and Greek L2 grammars 167 Teresa Parodi Aspect and the interpretation of motion verbs in L2 Greek Ianthi Maria Tsimpli and Despina Papadopoulou

187

Associating meaning to form in advanced L2 speakers: An investigation into 229 the acquisition of the English present simple and present progressive Sarah Ann Liszka Name index

247

Subject index

249

Dedication This volume is dedicated to someone who has played, and is playing a key role in both European and international second language acquisition studies. Roger Hawkins, along with a small handful of researchers in the UK has not only helped to nurture a local tradition of scholarship in this widely scattered but still numerically modest community of researchers, he has made his mark internationally, both in his close association with one major hypothesis about the nature of second language acquisition but also in his extremely important contributions to the journal, Second Language Research, both as reviews editor and latterly, and especially, as editor. Effectively, to have what is, to all intents and purposes, a ‘festschrift’ dedicated to you, it is not sufficient to have excelled in your own field, both in terms of your personal research record, your contributions as a guest-editor, and the kind of valuable gate-keeping service you deliver as the editor of a leading international journal – I add this last bit with the briefest of blushes: you also have to have a long line of ‘academic children’, postgraduate students, that is, who can bear witness to the dedication and skill with which you have brought new generations of SLA researchers into the world, who subsequently go on themselves to become established international figures in the field. As Florence Myles makes abundantly clear in her preface to this volume, Roger scores excellently on all counts and does this with characteristic meticulousness and modesty. In a sense much of the discussion in the following chapters is about failure, failure that is, in a very technical, theoretical sense. Let there be no misunderstanding, however: if there is one feature on the SLA landscape that never fails, it has to be Roger Hawkins and, hopefully, the contributors to this volume will be seen to be following in his footsteps. Mike Sharwood Smith Edinburgh, March 2008

Preface I first met Roger Hawkins in 1982, when I arrived as a fresh faced undergraduate entering the second year of an almost exclusively literary French BA at the University of Sheffield (being a French national, I was exempt from the first year of the degree). I had never heard of linguistics before, having missed the first year introductory courses, but the two modules Roger offered, one in syntax and one in psycholinguistics, aroused my curiosity and I signed up for both of them. I have never looked back! Anyone who has ever been taught by Roger will agree that he is an outstanding teacher; in his self-effacing and unassuming manner, he enthused and guided us in new ways of thinking about language. No question was ever too obtuse to deserve an answer (and we DID ask some brainless questions!!), and his dedication to his students was second to none. I was hooked! During my final year, Roger convinced me that I should consider applying for funding to do a new MA in French Language and Linguistics which he had just set up. The thought was rather far from my mind, as my priority at that stage, given my status as a mature student and my somewhat precarious financial circumstances, was to train as a teacher and get a job as quickly as I could in order to finally earn a living. But thanks to Roger, the study of Linguistics had truly captivated me and I agreed. Pretty clueless about how to fill in a British Academy application, Roger did most of the work, and I received funding! One thing led to another, and I became his first PhD student the following year, again with his help in securing British Academy funding. So I blame him squarely for where I am now!! If these introductory paragraphs are rather convoluted and self-indulgent, and somewhat removed from my brief of introducing Roger as an academic, it is not only because of Roger’s profound and lasting effect on my own career and thinking, but also because I believe those early years shaped Roger’s future career trajectory in some significant ways. Roger had arrived in the French Department at Sheffield in 1979, after a couple of years as a temporary lecturer in Portsmouth and as research fellow in Leeds, still fresh from his BA in French and Linguistics at the University of Edinburgh and his PhD in theoretical linguistics at the University of Cambridge, on The semantic structure of adverbs in English. He found himself as the lone linguist in an overwhelmingly literary department, a situation so characteristic of French Departments in UK universities at the time. Such appointments were commonplace, not because of a sudden awakening about the intellectual



Representational Deficits in SLA

excitement linked to the study of linguistics, but because it was assumed that academics who are knowledgeable about language and its intricacies should be able to help teach languages better. Invariably, such scholars were given a prominent role in the development and overseeing of the language teaching curriculum. Finding himself in a situation where he had to marry his interest in theoretical linguistics with his responsibilities as a French language teacher undoubtedly shaped Roger’s subsequent research interests. If literary specialists in Modern Languages departments often think linguistics and language teaching are very closely related (even today!), we as linguists know this to be far from the truth. But trying to establish a relationship between the two led Roger to the relatively new field of SLA, which was rather exciting at the time. Investigating the language produced by learners as a language in its own right was still a very new idea, and Contrastive Analysis proponents were still by far in the majority. But the work of Corder, Selinker, Krashen etc. was starting to have an impact, and I remember our excitement at reading Language Two when it came out in 1982! During my MA year, he started work on a mini-project to which I participated, investigating the relative merits of aural versus written input for the development of listening and speaking skills, leading to what I believe was his first SLA publication (Hawkins and Myles 1986). This topic might seem a far cry from Roger’s subsequent research agenda, but the (highly tentative) conclusions of that research were already trying to understand the ‘mechanisms of learners’ language faculties’, and their role in acquisition and processing. And it is this deep-seated interest in trying to understand the mental grammars underlying L2s and how they develop over time, which has continued to drive Roger’s research forward. On his web page, he describes his research interests as follows: My research interests are in the development of theories of human linguistic ability. My work focuses on how insights from generative linguistic theories help us understand the development of grammatical knowledge by second language learners. A specific hypothesis that I have explored in empirical work is that abstract features that give rise to surface differences between languages, but appear to have no semantic function (so-called ‘uninterpretable features’), are subject to a critical period.

Unsurprisingly, we can see here a rather more sophisticated research agenda than in the 80s. In fact, I cannot think of many scholars who have been as influential in moving forward the generative research agenda in SLA research. This is due not only to his undeniable enthusiasm for this subject and for finding answers to the questions currently being asked, as well as his genuine openness of mind (i.e., not minding being proven wrong if this advances our understanding), but also to his ability to enthuse the next generation of young researchers. Twenty eight students have completed PhDs under his supervision, and a further fifteen are currently in



Preface 

progress. He has been able to achieve this incredible feat by having a very clear research programme – outlined very broadly above – and engaging some bright young researchers in contributing to this agenda. Many well-known publications have arisen from this very fruitful collaboration with his colleagues and PhD students, e.g., Hawkins, Towell and Bazergui 1993, Hawkins et al. 2002, Hawkins and Liszka 2003, Hawkins and Francheschina 2004, Hawkins et al. 2004, Hawkins and Hattori 2006, Hawkins et al. 2006, Hawkins and Lozano 2006, Hawkins et al. 2008, Hawkins and Casillas 2008, and I am sure that my own experience of unfailing guidance and support as his first PhD student has been repeated many times since. Another important partnership in Roger’s life as an academic also came from his early days in a Modern Languages department, where linguistics specialists were very rare, never mind SLA specialists. But Richard Towell was working just across the Pennines, and they soon started collaborating, first on a three year ESRC-funded project in the early 90s on The development of linguistic knowledge and language processing capacity in second language learning, which gave rise to their co-authored book Approaches to Second Language Acquisition (1994, Multilingual Matters) as well as many articles. This collaboration also produced a number of editions of a highly acclaimed French Grammar and workbook, and the co-editing in 2004 of a Special Issue of Transactions of the Philological Society on Empirical evidence and theories of representation in current research into second language acquisition. He moved from his position as a lone linguist in a Modern Languages department to a senior lectureship in a large internationally reputed Language and Linguistics department at the University of Essex in 1989. This enabled Roger to develop further the theoretical dimensions of his research agenda, and he has thrived in this environment. The number of PhD students he has trained is testimony to that, as is his impressive record of publications and scholarly activities. He heads the Second Language Acquisition Research Group at the University of Essex which currently includes some thirteen academics and doctoral students working on second language acquisition from a generative perspective. He has published in all the top journals in the field, and is co-editor of Second Language Research. His 2001 book, Second Language Syntax: a generative introduction is a masterpiece of lucidity and clarity, and a must for any student interested in second language syntax. He has been in great demand as a plenary speaker at international conferences throughout the world, and has taught on graduate summer/winter schools in Spain (University of the Basque Country, Vitoria) and the Netherlands (LOT). He is a contributor to the forthcoming new edition of the Handbook of Second Language Acquisition edited by Ritchie and Bhatia, a highly regarded state of the art review of the field. I could get carried away, but let me conclude by saying what a pleasure it has been to write the preface to this volume, regrouping some of the best scholars in

 Representational Deficits in SLA

generative SLA, including some of Roger’s past students. The publication of this volume in his honor is possibly the clearest indication yet of Roger’s standing as a truly inspirational teacher and researcher. Florence Myles Newcastle, February 2008

Acknowledgements We are grateful to all the contributors. Without their generosity and effort the idea of publishing the volume would not have been realised. We would also like to thank the LALD series editors, Lydia White and Harald Clahsen, as well as Kees Vaes at John Benjamins for their support and advice. Finally we would like to thank the following colleagues (in alphabetical order) for reviewing the contributions: María del Pilar García Mayo, Ayşe Gürel, Henriëtte Hendriks, Tania Ionin, Fufen Jin, Silvina Montrul, Philippe Prévost, Roumyana Slabakova, Elena Valenzuela, Shigenori Wakabayashi and Virginia Yip. Neal Snape, Yan-kit Ingrid Leung and Mike Sharwood Smith

Introduction Neal Snape, Yan-kit Ingrid Leung and Mike Sharwood Smith Gunma Prefectural Women’s University / University of Hong Kong / Heriot-Watt University

This book sets out to explore the question of “representational deficits” in second language acquisition (SLA), currently a much-debated topic. The volume is intended as a tribute to Roger Hawkins, a leading researcher in generative second language acquisition, whose research has been devoted to explaining lack of native-like success in terms of representational deficits. The volume includes papers written by alumni of the University of Essex, as well as leading L2 researchers. Roger Hawkins’s research has contributed to fundamental questions in SLA research. Over the years, his research agenda has focused on how learners’ interlanguage grammars develop and whether learners can ever reach native-like attainment. He has addressed the role of the L1, the nature of linguistic representations in interlanguage grammars and the issue of Critical Periods. His contributions to the field have included the investigation of L2 speakers’ comprehension and production of English tense morphophonology, article interpretation, the (non) acquisition of gender, the (non) acquisition of features associated with wh-elements, amongst other topics. Such issues have been addressed in his own research and in collaboration with his students. Roger Hawkins has made numerous contributions to the field, in terms of books and articles. The book co-authored with Richard Towell Approaches to Second Language Acquisition (1994) and Roger Hawkins’s book Second Language Syntax: a generative introduction (2001) are both used extensively in general SLA courses and more specialized second language syntax courses, and are recognized as important sources for researchers in SLA. In addition, Roger Hawkins has published in peer-reviewed international journals such as Second Language Research and Journal of Linguistics. He has contributed chapters to edited volumes, many of which are co-authored with his graduate students. He has also edited special issues of international journals, such as the 2008 Lingua 118, which looks at current emergentist and nativist perspectives on second language acquisition.

 Neal Snape, Yan-kit Ingrid Leung and Mike Sharwood Smith

Roger has supervised a large number of MA and PhD graduate students since starting at the University of Essex in 1989. Roger’s students all comment on his patience, understanding and support; all of them feel that they have greatly benefited from his guidance. Roger, at any given time, has no less than fifteen PhD graduate students and six MA graduate students, making his attention to each one all the more remarkable. His ability to manage and have time for each individual student is a testament to Roger’s patience and attention to detail. In addition, Roger also leads a discussion group once a week where all his PhD students come together to discuss a current paper on SLA or to give presentations in order to gain feedback from the group. 1. Representational deficits in second language acquisition An ongoing theme in Roger Hawkins’s research is the apparent failure of most adult L2 learners to achieve native-like success. His approach to explaining this problem has been to focus on the issue of parameter resetting, the claim being that this is impossible in adult SLA. Various positions have been advanced by researchers in the generative SLA field as to whether second language (L2) learners transfer their first language (L1) grammar at the initial state and whether they have full access to Universal Grammar (UG), including new parameter settings, functional categories and feature values. White (2003a) categorizes these positions as Full Transfer/No Access (Bley-Vroman 1989), No Transfer/Full Access (Epstein, Flynn and Martohardjono 1996, Flynn 1996, Epstein, Flynn and Martohardjono 1998), Partial Transfer/Full Access (Vainikka and Young-Scholten 1996, 1998), Full Transfer/Full Access (Schwartz and Sprouse 1994, 1996) and Full Transfer/Partial Access (Tsimpli and Roussou 1991, Hawkins and Chan 1997). A crucial assumption shared by these researchers is that L2 interlanguage grammars are UG-constrained, but not necessarily in exactly the same way as L1 grammars are constrained. Framed within this so-called UG access debate, the view that L2 grammars are in some way impaired or defective as compared to native speakers’ grammars, originated with a claim of no parameter resetting in L2. Based on Principles and Parameters Theory (Chomsky (1981), Tsimpli and Roussou (1991) argued that learners are unable to reset parameters; as a consequence the L2 will be unattainable if the L1 and L2 parameter values are different. This became known as the ‘No Parameter Resetting Hypothesis’. This position was further developed by Smith and Tsimpli (1995), who focused on the parameterization of functional categories, claiming that, beyond childhood, L2 learners cannot acquire new functional categories, features or feature strength after the end of a critical period. Roger Hawkins and Cecilia Y.-h. Chan (1997) extended this claim, developing the notion that L2



Introduction  

learners have different mental representations from those of native speakers. They proposed the Failed (Functional) Features Hypothesis (FFFH) which claims that certain subcomponents of UG are inaccessible to L2 learners, while others are fully available. Their study focused on the extent to which Chinese-speaking learners could acquire English operator movement in restrictive relative clauses (claimed to be lacking in Chinese). The FFFH predicts significant differences between the Chinese speakers and the French speakers at comparable ages and proficiency levels as Chinese lacks a [wh] feature, hence learners will be unable to acquire the [wh] feature in the L2. In contrast, learners whose L1 has the [wh] feature, such as French, should have no problem as French and English share this property. Hawkins and Chan found that Chinese speakers are significantly different from the French speakers, even at advanced levels. In other words, in the case of learners whose L1 functional feature specifications are different from the L2, fossilization will occur, such that grammatical development stabilizes short of the target grammar (cf. Lardiere 1998a, 1998b, Franceschina 2001, White 2003b). More specifically, Hawkins (2000) refers to non-convergence as ‘persistent selective fossilization’. In other words, it is not possible under the FFFH for an L2 learner to reach native-like attainment in the L2 syntax due to a deficit relating to L1 transfer. The idea of a syntactic deficit in the interlanguage grammars of speakers whose L1s lack the corresponding functional categories and features was further developed, following a shift in generative linguistics from the Principles and Parameters Theory to Chomsky’s (2005, 2008) work on a Minimalist approach to narrow syntax. In this context, Tsimpli (2003), Tsimpli and Dimitrakopoulou (2007) and Hawkins (2005) have theorized that it is uninterpretable syntactic features which remain problematic for L2 learners, rather than interpretable features, contra Sorace and colleagues (Sorace 2004, 2005, 2006, Sorace and Filiaci 2006, Tsimpli and Sorace 2006, Tsimpli et al. 2004). L2 learners are subject to a critical period affecting uninterpretable features; beyond this critical period, the uninterpretable features, with the exception of those already available via the L1 grammar, become inaccessible for modification. This updated version of the original proposal by Tsimpli and Roussou (1991) has been referred to as the ‘Interpretability Hypothesis’ (Tsimpli 2003, Tsimpli and Dimitrakopoulou 2007) or the ‘Representational Deficit Hypothesis’ (RDH). In a number of papers, Roger Hawkins and his former students (Hawkins et al. 2002, Hawkins and Liszka 2003, Hawkins and Franceschina 2004, Hawkins and Hattori 2006, Hawkins et al. 2008) have argued that L2 learners’ interlanguage grammars lack uninterpretable features; hence, L2 learners have partial access to UG. The nine papers in this volume (contributed by leading L2 researchers, as well as by four of Roger’s former students) address the debate over representational deficits in L2 grammars, with authors taking different positions on the nature of

  Neal Snape, Yan-kit Ingrid Leung and Mike Sharwood Smith

deficits, if any. The papers are arranged in terms of linguistic domains. These include the syntax-phonology interface (Goad and White, Snape), syntax-morphology interface (Vainikka and Young-Scholten) the interfaces between syntax and semantics (Yuan, Jiang), syntax and pragmatics (Lozano) and a combination of the syntax and semantics/pragmatics interfaces (Parodi, Tsimpli and Papadopoulou, Liszka). A number of different languages are investigated. For example, L1 English L2 Spanish (Lozano), L1 French L2 English (Liszka, Vainikka and Young-Scholten), L1 Chinese L2 English (Snape, Jiang), L1s English and Japanese L2 Chinese (Yuan), L2 Greek (Tsimpli and Papadopoulou), L1 English L2 Spanish/L2 Greek (Parodi) and L1 Turkish L2 English (Goad and White). 2. A summary of the papers in this volume Goad and White’s chapter addresses the issue of representational deficits in L2 acquisition from a phonological perspective, namely the Prosodic Transfer Hypothesis. The authors argue against representational deficits on the syntactic level (contra Hawkins 2005, Trenkic 2007), and contend that it is the absence of certain prosodic representations in the L1 that results in L2 learners’ difficulty with functional morphology, including articles (the and a). Their study involves adult Turkish speakers with beginner and intermediate proficiency in L2 English. The target property under investigation is DPs, with and without adjectives. Using an oral picture description task the authors find that there are a number of omission errors (deletion of articles) and non-target forms. The authors argue that the errors are the result of the L2 learners being unable to acquire the target phonological structures in English. Instead, they have to adapt L1 prosodic structures in order to represent articles in the L2. The authors claim that the Prosodic Transfer Hypothesis can account for production errors involving articles, such as non-target forms involving stress, fillers and deletion. They propose that their results are consistent with a weaker interpretation of the Prosodic Transfer Hypothesis and the Full Transfer/Full Access model (Schwartz and Sprouse 1996). Snape addresses the issue of the acquisition of English articles by L2 learners. He focuses on the acquisition of semantic and syntactic aspects of English articles by speakers of an article-less language, Mandarin Chinese. Snape sets out to test a number of existing hypotheses in the L2 literature; the Fluctuation Hypothesis, proposed by Ionin, Ko and Wexler (2004), the Representational Deficit Hypothesis (Hawkins 2005), and the strong interpretation of the Prosodic Transfer Hypothesis (Goad and White 2004) by providing data collected from thirty eight adult Mandarin Chinese L2 English speakers using a written forced choice elicitation task (cf. Ionin et al. 2004) and an oral picture description task (Goad and White 2006).



Introduction 

If the L2 learners are found to fluctuate in article selection this would be consistent with the Fluctuation Hypothesis. However, if deletion of articles is found in article + noun contexts e.g., the/a man, perhaps the L2 learners are unable to build new prosodic structures for L2 functional morphology. This finding would be consistent with the strong interpretation of the Prosodic Transfer Hypothesis. If learners are unable to acquire uninterpretable features of DP then omission of articles is expected to occur in obligatory singular definite and indefinite contexts under the Representational Deficit Hypothesis. The results are consistent with the Fluctuation Hypothesis and the weaker version of the Prosodic Transfer Hypothesis. However, due to a high suppliance of definite and indefinite articles in obligatory contexts, the results are inconsistent with a syntactic deficit account. The paper by Vainikka and Young-Scholten argues for Organic Grammar, the re-branded version of the authors’ earlier theory variously known as “Minimal Trees”, “Structure Building” and “Weak Continuity”. According to the authors, Organic Grammar is meant to clarify some confusion associated with their approach to understanding (adult) L2 development, and to incorporate the spirit of minimalism in explaining interlanguage grammars. Organic Grammar posits that only lexical categories transfer from L1 to L2 initial state; functional categories and properties dependent on them, such as verb raising, do not transfer. Re-interpreting a set of problematic L1 French-L2 English beginning interlanguage verb raising data reported in White (1991a, b), Vainikka and Young-Scholten contend that L1 influence which has previously been used to explain White’s data is only apparent. Based on the ten general assumptions of Organic Grammar as well as the specific assumptions that adverbs (e.g., frequency and temporal adverbs) occupy various specifier positions in the functional architecture (e.g., [spec AgrP]) and that learners might thus prematurely project a functional category (e.g., Agr) void of syntactic features, White’s observation that verb raising happened only in the case of adverbs but not in negation and questions (which head different functional projections) is fully captured. Vainikka and Young-Scholten conclude that their approach neatly accounts for non-L1-based verb raising observed in White’s data, even assuming Hawkins and colleagues’ Failed Functional Features (or Representational Deficit) Hypothesis, in that although non-L1 uninterpretable features might indeed be (permanently) inaccessible in (post-critical period) L2 grammars, the Full House Principle (Vainikka and Young-Scholten 1994) will nevertheless force non-L1 verb raising to fill an empty functional category in L2 grammars in any case. Yuan investigates the L2 acquisition of the use of Chinese wh-words as universal quantifiers. Following Adger’s (2003) account for Tense, Yuan assumes that dou heads a functional category Distribution Phrase (DisP) with an interpretable movement-forcing [dis*] feature while the wh-word in [spec, vP] has an uninter-

 Neal Snape, Yan-kit Ingrid Leung and Mike Sharwood Smith

pretable [udis] feature; the wh-word as a universal quantifier is forced to move to [spec DisP] to be licensed and valued by dou. Yuan aims to test (1) whether L2 learners of Chinese demonstrate knowledge of wh-word movement to the left of dou and (2) whether these same learners interpret the wh-words as universal quantifiers, thus implicating the projection of DisP and a successful feature checking/ valuing mechanism in place in interlanguage grammars. Native speakers of English and Japanese who started learning Chinese as an L2 after the age of 17 participated in the study. They were divided into various proficiency groups. The experiment involves two tasks, namely an acceptability judgement task testing (1) and a sentence interpretation task testing (2). Results indicate that learners of both L1 groups at more advanced stages of L2 development have native-like performance in both tasks. Yuan interprets this as successful projection of DisP and corresponding feature checking/valuing mechanism in end state interlanguage grammars. However, less advanced L1 Japanese speakers show an asymmetry between subject and object wh-words in judgement and interpretation. Yuan conjectures that this is due to a non-movement-forcing [dis] feature in their grammars, resulting in a non-targetlike representation at earlier stages of L2 development. Based on this, Yuan argues for a non-permanent representational deficit in (adult) L2 acquisition. Jiang addresses the issue of whether Chinese speakers can acquire the local binding characteristics of L2 English reflexives and the obligatory status of English objects. Chinese differs to English in that it allows both the long-distance and the local binding of reflexives, whereas English only allows the local binding of reflexives. With regard to objects, Chinese allows both overt and null objects, whereas English only allows overt objects. Jiang argues that though the two properties involve different types of antecedent-anaphora and operator-variable binding relations, both properties potentially involve a ‘poverty of stimulus’ for Chinese learners of English. The results lead Jiang to claim that the obligatory status of English objects is not acquirable because of the lack of relevant triggering data. However, the learners’ interpretation of binding of reflexives is more or less native-like because of the availability of suitable triggering data. Jiang argues that learners at advanced stages are unlikely to ever become native-like because of the lack of the relevant triggering data; thus, the findings are not consistent with fossilization in the learners’ interlanguage grammars i.e., a representational deficit. Lozano investigates the distribution of null and overt subject pronouns in terms of formal and pragmatic constraints in very advanced L1 English L2 Spanish learners. In Spanish, the null pronoun pro and overt pronoun él are in apparent free variation e.g. él/pro tiene mucho dinero (he/Ø has a lot of money). The null and overt pronominal subject can alternate, depending on the context. In English only an overt pronoun is possible. Theoretically, Lozano seeks to answer the question of whether syntax-discourse related features related to pronominal sub-



Introduction 

ject use, i.e., +/-topic-shift, are problematic for L2 Spanish learners and if so, whether the ‘deficit’ is associated with a subset of members of the paradigm or with the whole set. The data are drawn from a written corpus of English learners of Spanish, annotated and analysed using the corpus tools referred to and described in the text. Lozano argues that the learners have selective deficits at the syntax-discourse interface, as just a subset of features seems to be affected. Ultimately, Lozano suggests that the path of development is guided by Feature Geometry, which is constrained by Universal Grammar. Parodi’s study looks at L1 English L2 Spanish/L2 Greek grammars with a focus on clitic doubling and clitic left-dislocation. Parodi adopts the theoretical assumption that clitics are instances of agreement, rather than being pronouns, and are bearers of uninterpretable phi-features. Parodi aims to investigate the issue of real vs. apparent optionality with evidence from various stages of L2 development. An experiment employing an acceptability judgement task testing contrasts involving case (dative in Spanish and genitive in Greek as against accusative in both languages) and definiteness of full NPs in clitic doubling and clitic left-dislocation structures was implemented with both native speakers and L2 learners of different proficiency levels of Spanish and Greek. Findings from native speaker data are consistent with the theoretical assumption that clitics in both Spanish and Greek languages are agreement markers, as native speakers of the respective languages show differentiated acceptability judgements of corresponding clitic structures based on case and/or definiteness. Findings from the L2 learner data, however, suggest that interlanguage grammars may not be the same as native grammars. L2 learners do not seem to be sensitive to effects of case, demonstrating, as Parodi contends, real optionality. Definiteness, on the other hand, seems to pose less of a learning problem. In sum, Parodi concludes that her results are compatible with a Representational Deficits view in that the lack of object agreement and associated uninterpretable features in L1 English might have led the L2 learners of Spanish and Greek to mis-analyze clitics as pronouns rather than agreement markers, posing problems in L2 representations. Tsimpli and Papadopoulou’s paper is concerned with the interaction of aspect and argument structure of manner-of-motion verbs in native and L2 Greek. The authors assume that aspect in Greek is a grammaticalized, interpretable feature and propose a syntactic analysis of manner-of-motion structures for Greek in line with Zubizarreta and Oh (2007). An experiment which comprised a sentencepicture matching task and an oral production task testing the comprehension and production of potentially ambiguous manner-of-motion verbs was conducted on ten native speakers of Greek and ten L2 learners with L1 Slavic and L1 Germanic backgrounds. It aimed to investigate the syntax-semantics interface (i.e., the use the perfective/imperfective distinction to differentiate between complement and

 Neal Snape, Yan-kit Ingrid Leung and Mike Sharwood Smith

adjunct PPs) and the syntax-discourse interface (i.e., the interpretation of telic vs. atelic readings of motion events) in both native and L2 grammars. The L2 learners also completed a morphology pre-test successfully, demonstrating good knowledge of correct perfective/imperfective verb forms in Greek. Findings of the comprehension and production tasks show that both native speakers and L2 learners make effective use of aspect for argument structure and telicity interpretation, suggesting that aspect as an interpretable feature poses no problem for the L2 Greek learners, thereby lending support to the Interpretability Hypothesis. However, L2 learners diverge from native speakers in that they seem to prefer a one-toone mapping between morphological form (perfective vs. imperfective) and meaning (telic vs. atelic) and thus are less able to resolve the ambiguities involved in the syntax-discourse interface. Liszka’s chapter investigates the acquisition of the English present simple e.g., She works at home, and present progressive e.g., She is working at home, by L1 French-L2 English speakers. In French, the present licenses both interpretations e.g., Elle travaille chez elle (she works/is working at home). Liszka discusses whether overuse of the simple form in progressive contexts is directly attributable to a permanent L1 syntactic influence. Liszka employs three tasks, one written and two oral, to test whether advanced French L2 learners are successful in assigning target-like meaning to present simple and present progressive. Lizska discovers that the L1 feature setting cannot be modified from strong to weak for thematic verbs, thus allowing a deictic interpretation to be mapped onto the simple form. Although a strong uninterpretable feature is established in the L1 grammar of the French L2 learners, Lizska posits that a weak uninterpretable inflectional feature has not been established in the L2 learners’ grammars due to a representational deficit. In conclusion, Roger Hawkins’s research has focused on the central issue of whether L2 learners can ever attain native-like competence, his hypothesis (as implemented in the RDH) being that they cannot. His contributions to the field of SLA, along with his supervisory role, have been remarkable. Roger truly deserves a volume dedicated to the research that he has inspired many academics and future academics to undertake. References Adger, D. 2003. Core syntax: A minimalist approach. Oxford: OUP. Bley-Vroman, R. 1989. What is the logical problem of foreign language learning? In Linguistic perspectives on second language acquisition, S. M. Gass & J. Schachter (eds.), 41–68. Cambridge: CUP.



Introduction   Chomsky, N. 1981. Principles and parameters in syntactic theory. In Explanation in Linguistics: The logical problem of language acquisition, N. Hornstein & D. Lightfoot (eds.), 123–146. London: Longman. Chomsky, N. 2005. Three factors in language design. Linguistic Inquiry 36: 1–22. Chomsky, N. 2008. On phases. In Foundational issues in linguistic theory: essays in honor of JeanRoger Vergnaud, R. Freidin, C. P. Otero & M. L. Zubizarreta (eds.), 133–166. Cambridge MA: The MIT Press. Epstein, S.D., Flynn, S. & Martohardjono, G. 1996. Second language acquisition: Theoretical and experimental issues in contemporary research. Behavioural and Brain Sciences 19: 677–714. Epstein, S.D., Flynn, S. & Martohardjono, G. 1998. The strong continuity hypothesis: Some evidence concerning functional categories in adult L2 acquisition. In The generative study of second language acquisition, S. Flynn, G. Martohardjono & W. O’Neil (eds.), 61–77. Hillsdale NJ: Lawrence Erlbaum. Flynn, S. 1996. A parameter-setting approach to second language acquisition. In Handbook of second language acquisition, W. Ritchie & T. Bhatia (eds.), 121–158. San Diego CA: Academic Press. Franceschina, F. 2001. Morphological or syntactic deficits in near-native speakers? An assessment of some current proposals. Second Language Research 17(3): 213–247. Goad, H. & White, L. 2004. Ultimate attainment of L2 inflections: Effects of L1 prosodic structure. In EUROSLA Yearbook 4, S. Foster-Cohen, M. Ota, M.A. Sharwood Smith and A. Sorace (eds.), 119–145. Amsterdam: John Benjamins. Goad, H. & White, L. 2006. Prosodic transfer: L1 effects on the production of L2 determiners. In Proceedings of the 30th Annual Boston University Conference on Language Development, D. Bamman., T. Magnitskaia & C. Zaller (eds.), 213–224. Somerville MA: Cascadilla Press. Hawkins, R. 2000. Persistent selective fossilisation in second language acquisition and the optimal design of the language faculty. Essex Research Reports in Linguistics 34: 75–90. Hawkins, R. 2001. Second language syntax: A generative introduction. Oxford: Blackwell. Hawkins, R. 2005. Explaining full and partial success in the acquisition of second language grammatical properties. Second Language 4: 7–25. Hawkins, R. & Chan, C-Y. 1997. The partial availability of UG in SLA: The failed functional features hypothesis. Second Language Research 13(3): 187–226. Hawkins, R., Al-Hamad, M., Al-Malki, E., Casillas, G., Franceschina, F., Hawthorne, J., Karadzovska, D., Kato, K., Liska, S. Lozano, C., Ojima, S., Okuwaki, N. & Thomas, E. 2002. Interpretation of English tense morphophonology by advanced L2 speakers. In EUROSLA Yearbook 2, S. Foster-Cohen, T. Ruthenberg & M.-L. Poschen (eds.): 49–70. Hawkins, R. & Liszka, S. 2003. Locating the source of defective past tense marking in advanced L2 English speakers. In In The lexicon-syntax interface in second language acquisition, R. van Hout, A. Hulk, F. Kuiken & R. Towell (eds.), 21–44. Amsterdam: John Benjamins. Hawkins, R. & Franceschina, F. 2004. Explaining the acquisition and non-acquisition of determiner-noun gender concord in French and Spanish. In the acquisition of French in different contexts, J. Paradis & P. Prévost (eds.), 175–205. Amsterdam: John Benjamins. Hawkins, R. & Hattori, H. 2006. Interpretation of English multiple wh-questions by Japanese speakers: A missing uninterpretable feature account. In Fossilization and ultimate attainment, D. Lardiere (ed.). Special issue of Second Language Research 22(3): 269–301.

  Neal Snape, Yan-kit Ingrid Leung and Mike Sharwood Smith Hawkins, R., Casillas, G., Hattori, H., Hawthorne, J., Husted, R., Lozano, C., Okamoto, A., Thomas, E. & Yamada, K. 2008. The semantic effects of verb raising and its consequences in second language grammars. In The role of formal features in second language acquisition, J. M. Liceras, H. Zobl & H. Goodluck (eds), 328–351. New York NY: LEA. Ionin, T., Ko, H. & Wexler, K. 2004. Article semantics in L2 acquisition: The role of specificity. Language Acquisition 12(1): 3–69. Lardiere, D. 1998a. Case and tense in the ‘fossilized’ steady state. Second Language Research: 14: 1–26. Lardiere, D. 1998b. Dissociating syntax from morphology in a divergent end-state grammar. Second Language Research 14: 359–375. Schwartz, B. D. & Sprouse, R. A. 1994. Word order and nominative case in nonnative language acquisition: A longitudinal study of (L1 Turkish) German interlanguage. In Language Acquisition Studies in Generative Grammar, T. Hoesktra & B. D. Schwartz (eds.), 317–68. Amsterdam: John Benjamins. Schwartz, B.D. & Sprouse, R.A. 1996. L2 cognitive states and the Full Transfer/Full Access model. Second Language Research 12(1): 40–72. Smith, N. V. & Tsimpli, I.-M. 1995. The mind of a savant: Language learning and modularity. Oxford: Blackwell. Sorace, A. 2004. Native language attrition and developmental instability at the syntax-discourse interface: Data, interpretations and methods. Bilingualism: Language and Cognition 7: 143–145. Sorace, A. 2005. Selective optionality in language development. In Syntax and variation: Reconciling the biological and the social, L. Cornips & K. P. Corrigan (eds.), 55–80. Amsterdam: John Benjamins. Sorace, A. 2006. Possible manifestations of shallow processing in advanced second language speakers. Applied Psycholinguistics 27: 88–91. Sorace, A. & Filiaci, F. 2006. Anaphora resolution in near-native speakers of Italian. Second Language Research 22: 339–368. Towell, R. & Hawkins, R. 1994. Approaches to second language acquisition. Clevedon: Multilingual Matters. Trenkic, D. 2007. Variability in L2 article production – beyond the representational deficit vs. processing constraints debate. Second Language Research 23(3): 289–327. Tsimpli, I.-M. 2003. Clitics and determiners in L2 Greek. In Proceedings of the 6th Conference on Generative Approaches to Second Language Acqusition, J. Liceras, H. Zobl & H. Goodluck (eds.), 331–339. Somerville MA: Cascadilla Proceedings Project. Tsimpli, I.-M. & Roussou, A. 1991. Parameter-resetting in L2? UCL Working Papers in Linguistics 3: 149–70. Tsimpli, I.-M. & Sorace, A. 2006. Differentiating interfaces: L2 performance in syntax-semantics and syntax-discourse phenomena. BUCLD Proceedings 30: 653–664. Tsimpli, I.-M. & Dimitrakopoulou, M. 2007. The Interpretability Hypothesis: Evidence from wh-interrogatives in second language acquisition. Second Language Research 23(2): 215–242. Tsimpli, I.-M., Sorace, A., Heycock, C. & Filiaci, F. 2004. First language attrition and syntactic subjects: A study of Greek and Italian near-native speakers of English. International Journal of Bilingualism 8: 257–277.



Introduction   Vainikka, A. & Young-Scholten, M. 1994. Direct access to X’-theory: Evidence from Korean and Turkish adults learning German. In Language acquisition studies in generative grammar, T. Hoekstra & B. D. Schwartz (eds.), 265–316. Amsterdam: John Benjamins. Vainikka, A. & Young-Scholten, M. 1996. Partial transfer, not partial access. Behavioural and Brain Sciences 19: 744–5. Vainikka, A. & Young-Scholten, M. 1998. The initial state in L2 acquisition of phrase structure. In The generative study of second language acquisition, S. Flynn, G. Martohardjono & W. O’Neil (eds.), (pp. 17–34). Mahwah NJ: Lawrence Erlbaum. White, L. 1991a. Adverb placement in second language acquisition: some effects of positive and negative evidence in the classroom. Second Language Research 7: 133–160. White, L. 1991b. The verb-movement parameter in second language acquisition. Language Acquisition 1: 337–360. White, L. 2003a. Second language acquisition and universal grammar. Cambridge: CUP. White, L. 2003b. Fossilization in steady state L2 grammars: Implications of persistent problems with inflectional morphology. Bilingualism: Language and Cognition 6(2): 129–141. Zubizarreta, M. L & Oh. E. 2007. On the syntactic composition of manner and motion. Cambridge: The MIT Press.

Prosodic transfer and the representation of determiners in Turkish-English interlanguage Heather Goad and Lydia White McGill University

Problems with the second language (L2) acquisition of articles in TurkishEnglish interlanguage are examined, specifically omission, substitution of other determiners in place of articles, overuse of stressed articles and differential treatment of articles in DPs with and without adjectives. Trenkic’s (2007) syntactic misanalysis account is argued against, that difficulties with articles for L2 English speakers from L1s like Turkish and Serbian stem from the absence of DP in the L1 grammar. Instead, in accordance with the Prosodic Transfer Hypothesis, it is proposed that L1 prosodic representations underlie these problems; L2 learners are argued to have difficulty in producing functional morphology in target-like fashion when the L2 requires a prosodic representation that is unavailable in the L1 grammar.

1. Introduction Problems with the second language (L2) acquisition of articles are well-documented, especially in the case of learners whose mother tongue (L1) lacks articles (Huebner 1985; Thomas 1989; Robertson 2000; Leung 2002; Snape 2006; amongst others). Two kinds of problems have commonly been identified, namely omission of articles in obligatory contexts, as in (1a), and overuse of one article in place of the other (usually definite in place of indefinite), as in (1b). (For both examples, the L1 is Turkish.) (1) a. So brain is already shaped. (White 2003a) b. She was holding the fork. (Goad and White 2006a) There have been various different approaches towards explaining and predicting these kinds of problems. Some focus on semantic factors, including the issue of overuse (e.g., Ionin 2003; Ionin, Ko and Wexler 2004), or on discourse/pragmatic functions (e.g., Robertson 2000). In this paper, we offer an account in terms of



Heather Goad and Lydia White

prosodic representation, an account which explains article omission and a particular kind of substitution, namely the use of other determiners in place of articles, as well as other non-native aspects of article production related to stress.1 The issue of article underuse and overuse is potentially relevant to the debate over whether L2 speakers suffer from representational deficits and, if so, of what kind. In recent years, proposals have focused on various sub-components of the grammar; properties of functional categories and their features have become a particular area of concern, with claims centring round presence or absence of impairment in this domain (e.g., Eubank 1996; Schwartz and Sprouse 1996; Hawkins and Chan 1997; Beck 1998; Lardiere 1998; Duffield et al. 2002; Tsimpli 2003). An influential proposal in this vein is the Representational Deficit Hypothesis (RDH) (originally the Failed Functional Features Hypothesis) (Hawkins and Chan 1997; Hawkins 2000, 2003; see also Tsimpli and Roussou 1991; Tsimpli and Mastropavriou 2007). According to the RDH, adult L2 speakers fail to acquire uninterpretable formal features which are not realized in the L1 grammar, resulting in permanent misrepresentation of L2 features. This syntactic deficit has a number of consequences as far as overt inflection is concerned: Chinese-speaking learners of English are argued to be unable to acquire a tense feature (due to presumed absence of such a feature in Chinese) and, hence, have problems realizing tense morphology on verbs (Hawkins and Liszka 2003); similarly, English-speaking learners of Spanish are argued to be unable to acquire gender (due to lack of a gender feature in English) and, hence, fail to fully master gender agreement within the DP (Franceschina 2001). The inability to acquire L2 uninterpretable features leads to omission of morphology in the case of tense, or inappropriate substitutions of one form for another in the case of gender. In the present paper, we examine article production in L2 English speakers whose mother tongue, Turkish, has no definite article. English articles appear in DP and are associated with a definiteness feature (see Section 2). If the L1 lacks this functional category and/or its associated features, then impairment might be anticipated, as Leung (2002) suggested for Cantonese-speaking learners of English and French. However, the RDH does not in fact expect features to fail in this case, the crucial feature – definiteness – being interpretable, rather than uninterpretable (Hawkins et al. 2006). A different kind of syntactic representational deficit involving English articles is proposed by Trenkic (2004, 2007, 2008). Trenkic assumes that learners whose L1s lack DP fail to analyse English articles as D elements, instead treating them as adjectives. She assumes (contra Hawkins et al. 2006) that definiteness is an uninterpretable feature in SpecDP. If so, according to the RDH, definiteness should not 1.

Our account does not speak to misuse of one kind of article for the other.



Prosodic transfer and determiners

be acquirable. However, Trenkic does not attribute acquisition difficulties to uninterpretablility of features. Rather, according to her, the problem lies with the category D itself, as will be discussed below. In the present paper, we also offer a representational account of L2 article production. We argue against Trenkic’s syntactic misanalysis proposal. Instead, in accordance with the Prosodic Transfer Hypothesis (PTH) (Goad, White and Steele 2003; Goad and White 2004, 2006b), we propose that L1 prosodic representations are a major source of problems in L2 article production; furthermore, we will show that some aspects of non-native article suppliance can only be accounted for in prosodic terms. 2. Articles and other determiners in English and Turkish Articles are functional elements which head the category D (Abney 1987);2 associated with them are a number of features. Choice of articles in English depends on the feature [±definite], the being definite and a indefinite. Turkish has only an (optional) indefinite article, unstressed bir, as shown in (2a) (Kornfilt 1997). In fact, there is disagreement amongst linguists working on Turkish on the status of unstressed bir. Kornfilt (1997) considers it to be an article, while Underhill (1976) treats it is a numeral. Lyons (1999) argues that it is neither; based on its meaning, he refers to it as a ‘quasi indefinite article’, but argues that definiteness is not grammaticalized in Turkish, a position adopted by Öztürk (2005) as well. We return to this issue below. When bir is stressed, as in (2b), it is interpreted as the numeral one. In other words, depending on stress, the interpretation of bir varies between article and numeral (Erguvanli 1984; Kornfilt 1997; Öztürk 2005; amongst others). Like English, Turkish has other determiners which are stressed, such as quantifiers, demonstratives, numerals and possessives, as in (2c). Finally, in contexts where English would require an overt article, Turkish also allows bare nouns, as in (2d). The interpretation of a bare noun as definite or indefinite depends on a number of factors, including word order and case-marking. Bare nouns typically receive a definite interpretation in subject position and are potentially ambiguous elsewhere. (2) a. bir kitáp ‘a book’

2. There is disagreement as to whether both articles head D. On some accounts, definite and indefinite articles head different functional projections (e.g., Lyons 1999).





Heather Goad and Lydia White

b. bír kitap ‘one book’ c. bú kitap ‘this book’ d. kitáp ‘a/the book’ 2.1

Prosodic representation of articles and other determiners in English and Turkish

In this paper, we are particularly concerned with cross-linguistic differences in the prosodic organization of articles and the consequences this has for L2 acquisition. In (3a), the prosodic representation for English definite and indefinite articles is shown; these link directly to the phonological phrase (PPh) as ‘free clitics’ (Selkirk 1996). Phonological considerations have led us to conclude that Turkish does not permit free clitics anywhere in the grammar (see Section 3.1); that is, Turkish lacks the representation in (3a) for functional material at either edge of its host. Instead, unstressed bir is adjoined to the prosodic word (PWd) of its host as an ‘affixal clitic’, as shown in (3b), the typical representation for prefixes across languages (e.g., Peperkamp 1997).

(3) Articles: a. English definite and indefinite: free clitic:

b. Turkish indefinite: affixal clitic/prefix:

PPh

PWd

PWd

the/a

bóok

PWd

bir

kitáp

‘a book’

Other determiners in both English and Turkish, including the numeral bir, are stressed and thus form independent PWds, as shown in (4a-b).3

3. The Turkish examples in (4)-(6) mark phrasal stress only, because Turkish speakers disagree on the status of secondary stress. We have followed Kabak and Vogel (2001) who state that main stress falls on the leftmost word in the phrase, as this corresponds to the judgments received from the native speaker informants that we consulted about stress. There is, however, disagreement in the literature on this; see Inkelas and Orgun (2003) for a different view.



Prosodic transfer and determiners



(4) Other determiners: independent PWds: a. English:

b. Turkish: PPh

PWd

PWd

óne

bóok



PPh



PWd

PWd

bír

kitap

‘one book’

As we have previously argued (Goad and White 2004), word order in DPs containing adjectives provides evidence for the differing representations of English and Turkish articles. Turkish bir appears in a different position in relation to the adjective, depending on whether it is a numeral or an article (e.g., Kornfilt 1997; Öztürk 2005). When bir is a numeral (i.e., stressed), like English one, it precedes the adjective, as shown in (5).

(5) PPh PPh

PWd

PWd

PWd

bír one

iyi good

adam man

‘one good man’

However, when bir is an indefinite article (unstressed), it follows the adjective. Thus, (6a) is a possible representation but (6b) is not because, as a prefix, indefinite bir must cliticize onto the noun that it modifies.4

4. The representation in (6a), where indefinite bir adjoins to the PWd, can be motivated for determiners in other languages as well. Poser (1990) argues that in Japanese, Aoyagi “pre-nominal modifiers”, some of which have the semantics of determiners, are organized as prefixes; one piece of evidence for this is that adjectives cannot intervene between these “modifiers” and nouns, in contrast to the behaviour of other determiners in the language (see also Snape 2006).





Heather Goad and Lydia White



(6) a.

b. *PPh

PPh

PWd

PWd

PWd

PWd

PWd iyí bir good a

adam man

PWd

‘a good man’

bir iyi a good

adam man

In English, no such change in word order is observed: the presence of an adjective has no effect on the position of articles, indicating that articles link higher in the structure, at the level of the PPh, as shown in (7). (7)

PPh

a

PWd

PWd

góod

mán

In short, the non-canonical word order observed in Turkish constructions with adjectives and unstressed bir would be unexpected if unstressed bir had the same prosodic representation as English articles. Öztürk (2005) states that other numerals and determiners in Turkish also alternate, appearing before or after adjectives, as shown in (8). (8) a. bu kırmızı kitap this red book b. kırmızı bu kitap red this book c. iki kırmızı kitap two red books d. kırmızı iki kitap red two books However, we have observed that this alternation differs from the alternation involving bir. Firstly, there is a preference for the word order where the determiner precedes the adjective, as in (8a) and (8c). Secondly, while there are subtle meaning



Prosodic transfer and determiners

differences associated with this alternation, they do not involve the determiner itself, nor do they depend on whether or not the determiner is stressed. We conclude, therefore, that only bir shows an alternation which is prosodically conditioned. 2.2

Syntactic representation of articles in Turkish

There is considerable debate amongst linguists working on Turkish as to the representation of noun phrases, in particular, as to whether or not Turkish has a DP projection. Öztürk (2005) proposes that Turkish lacks DPs altogether. She treats unstressed bir as an adjective-like modifier within NP.5 A related analysis is assumed by Trenkic (2004, 2007, 2008) for Serbian. She argues that there is no DP in languages where definiteness is not grammaticalized in an article system. Like Öztürk, she assumes that Turkish is such a language. Trenkic follows Lyons (1999) in assuming that the D position and a definiteness feature are associated only with definite articles, which Turkish lacks; indefinites are associated with a different functional projection, namely cardinality. This is not the place to go into details on the debate over NP versus DP in Turkish. Suffice it to say that Trenkic advances the NP analysis for interlanguage grammars as well, in situations where the L1 lacks a DP projection. She claims that L2 learners of English whose L1 is Serbian or Turkish fail to project a DP, misanalysing English articles as adjective-like modifiers. This misanalysis, together with pragmatic factors (see Section 4), accounts for problems with L2 articles. If languages like Serbian and Turkish indeed lack DP and if article-like elements are in fact adjectives, this analysis has important prosodic consequences. The crux of the matter is that adjectives, as lexical elements, must form independent PWds. As detailed in Zec (2005), lexical elements (including adjectives) in Standard Serbian always form their own PWds, making them bearers of both pitch and stress (Serbian is a pitch accent language). Bisyllabic free functional elements, including a subset of determiners, also form PWds; they receive pitch accent but cannot be stressed (unless under focus). Monosyllabic free functional elements, including another subset of determiners, do not form PWds; they are free clitics, and are thus represented like articles in English, namely with the representation in (3a). As expected, they do not receive pitch accent and cannot be stressed (focus aside). If determiners in general are analysed as adjectives in Serbian, these contrasts would be lost; all determiners should form PWds and receive pitch accent. In other words, under this view, Serbian would not permit any determiners to be organized as free clitics, contrary to fact. 5.

Öztürk (2005) considers stressed bir to be an adverbial modifier of a complex predicate.





Heather Goad and Lydia White

Turning to Turkish, if bir is analysed as an adjective, it should also form its own PWd. Such an analysis, however, cannot account for the fact that bir, when interpreted as the indefinite article, is unstressed. Rather, bir should always receive main stress when in phrase-initial position, regardless of its interpretation. In other words, the difference in prominence pattern between bír kitap ‘one book’ in (2b) and bir kitáp ‘a book’ in (2a) would be lost. To summarize, English and Turkish differ morpho-syntactically and prosodically. English realizes a [±definite] distinction overtly in the article system (the versus a). Articles are prosodically realized as free clitics, attached directly to the PPh. In Turkish, on the other hand, there is no morphological realization of the definite article (although other determiners such as possessives and demonstratives are definite, as they are in English). Unstressed bir is [-definite]; prosodically, it is organized as an affixal clitic, adjoined to the PWd. As far as the syntactic representation of Turkish unstressed bir is concerned, we suggest that the evidence from interpretation, prosodic characteristics and positioning relative to nouns with and without adjectives favours the article analysis over an analysis as some other type of modifier, and that this is true of interlanguage articles as well. 3. Prosodic Transfer Hypothesis (PTH) In earlier work, we proposed the Prosodic Transfer Hypothesis (Goad, White and Steele 2003; Goad and White 2004, 2006b), according to which L1 prosodic representations are (at least in part) the determinants of success or lack of success in spoken production of L2 inflection and function words. L2 learners are predicted to have difficulty in accurately producing functional morphology when the L2 requires a prosodic representation which is unavailable in the L1 (or unavailable for the target construction). In previous work, we have shown that prosodic transfer constrains the production of English articles, as well as tense and agreement morphology, for Mandarin speakers (Goad, White and Steele 2003; Goad and White 2006b, 2008). We have also shown in a case-study of an endstate L2 Turkish-English speaker (known as SD) that article suppliance was comparatively low and that there was a significant contingency between presence of an adjective and omission of articles (White 2003a; Goad and White 2004). In other words, there was some support for the claim that the representation in (3a) is problematic for Turkish-speaking learners of English, thereby supporting the PTH. However, the data came from only one subject and we did not examine other properties of her article suppliance, such as whether or not articles were stressed. In this paper, we examine article suppliance in the L2 English of a number of Turkish speakers, at various levels of proficiency,



Prosodic transfer and determiners

using a task designed to more systematically elicit some of the patterns we observed in the original study, particularly the interaction of articles and adjectives. 3.1

General predictions of PTH

Given the differences in the prosodic organization of articles in English and Turkish discussed above, the PTH predicts that Turkish speakers should have difficulty constructing the target representation for articles in English. This should affect definite and indefinite articles alike, since they are prosodified in the same way in the L2; that is, difficulties should hold for both articles, independent of the fact that only an indefinite article is realized in the L1 grammar. When the target prosodic representation is not present in the L1 grammar, as in the case of Turkish, high rates of article deletion are expected, especially at low levels of proficiency. As proficiency improves, we expect learners to become aware of the need for overt articles in the L2; hence suppliance should increase. However, we anticipate various non-target strategies in outputs, all of which have as their source the prosodic constraints of the L1 (see Section 3.2). Concerning the potential for native-like suppliance of English articles, we have argued in earlier work that target prosodic representations can be built in the interlanguage grammar under two conditions (see (9)), both of which involve minimally adapting L1 representations that occur in other constructions (Goad and White 2004). Specifically, new structures are possible:

(9) a. when they can be built through combining pre-existing (L1) licensing relations; or b. when they involve L1 structures being licensed in new positions. Support for this proposal came from a comparison of SD’s rates of suppliance of tense, agreement and plural morphology versus articles in English. In the former case, it was possible to minimally adapt structures from Turkish; target-like prosodic representations could thereby be built and rates of suppliance in English were correspondingly high. In the case of articles, however, no adaptation of L1 structures is possible to allow the representation in (3a) (see discussion below) and rates of suppliance were correspondingly lower. As mentioned in Section 2, Turkish does not permit free clitics at either edge for other constructions in the language, thus precluding the adoption of the structure for English articles through condition (9b). Left-edge free clitics seem to be absent altogether from Turkish; that is, every PPh must start with a PWd (either a lexical element, a stressed functional element (e.g., stressed bir), or a prefix (e.g., unstressed bir)). While Kornfilt (1997) analyses a number of





Heather Goad and Lydia White

bound elements at the right-edge as syntactic clitics, there is phonological evidence that they are not directly linked to the PPh. A case in point is provided by inflected forms of the copula -y. Kornfilt analyses these as clitics because they undergo vowel harmony but display exceptional ‘prestressing’ behaviour (e.g., Hasan dün hastá-y-dı Hasan yesterday sick-coppast ‘Hasan was sick yesterday’ (p. 436)). Kornfilt does not address how these elements are organized into prosodic structure. However, to capture their prestressing behaviour, Kabak and Vogel (2001) and Newell (2005) argue that they are adjoined to the PWd ([[hastá]PWd -y]PWd);6 indeed, Kabak and Vogel explicitly argue against their being attached to the PPh (*[[hastá]PWd -y]PPh). Since the domain of vowel harmony is the lower PWd in the vast majority of languages (cf. van der Hulst and van de Weijer 1995) and evidence from compounds supports this view for Turkish (cf. Kabak and Vogel 2001), we consider that elements such as copular -y must be internal to the domain of their host in order to fall within the scope of harmony ([hastá-y]PWd). In short, although there is some disagreement about the organization of these elements, as PWd-internal or as adjoined, it is agreed that they are not linked to the PPh as free clitics. Turning to minimal adaptation under condition (9a), in Goad and White (2006b), we detail how L1 licensing options cannot be combined to build a structure where articles link directly to the PPh because languages will only allow a syllable to associate directly to this constituent if the syllable, underlined in (10a), is functional. (10) a. Free clitic: PPh

σ

b. Affixal clitic: PWd

PWd

PWd

…

…

σ

Consider the category PWd for comparison. This category is unique in its ability to organize both lexical and functional material. Accordingly, prosodic representations for lexical material in the L1 can be adapted to represent functional morphology in the interlanguage. Concerning English inflection, for example, the necessary adjunction structure for the inflectional syllable, underlined in (10b), can be built in the Turkish-English interlanguage grammar by combining existing 6. This is somewhat of a simplification. Kabak and Vogel argue that clitics as well as suffixes that display exceptional stress are organized into the Clitic Group. For present purposes, this can be equated with the recursive PWd structure provided in the text for hastá-y.



Prosodic transfer and determiners

representations from the L1 grammar: PWd recursively dominating PWd (PWd– PWd), a structure required for lexical compounds, and PWd directly dominating σ (PWd–σ), a structure required at the right edge to capture non-final stress in a subset of the Turkish vocabulary sometimes referred to as ‘Sezer stems’ (mostly place names and foreign proper names) (cf. Sezer 1981). As most languages are like Turkish in having both lexical compounding and permitting syllables in lexical words to link directly to the PWd under some condition or other, learners from most L1s lacking adjunction will be in a position to build the structure required for inflection in English-type languages under condition (9a). Returning to the structure in (10a) required for English articles, this structure is impossible to build from existing L1 licensing relations because – unlike the case of the PWd – a syllable can only link directly to the PPh if the syllable is functional. Hence, if an L1 grammar lacks PPh­–σ, as Turkish appears to, it will be impossible to build this licensing relation under condition (9a). To summarize, the absence of free clitic constructions in Turkish should prevent learners from building this structure for English articles under either of the conditions in (9). In other words, the strongest interpretation of the PTH is that the target prosodic structure for English articles will not be acquirable by Turkish speakers. A weaker interpretation is that this structure should be particularly difficult to acquire, only emerging in the grammars of particularly proficient learners. Under either view, for all but perhaps the most proficient speakers, we expect a number of non-target productions in article contexts, which have as their source the prosodic constraints of the L1. 3.2

Predicted non-target productions

In this Section, we detail the various strategies that speakers are predicted to adopt for the interlanguage production of articles. If Turkish-speaking learners have not acquired (or possibly cannot acquire) the free clitic structure for representing English articles, these strategies should be robustly attested in outputs, depending in part on level of proficiency. Perhaps the ‘easiest’ way to avoid the English free clitic representation in (3a) is to omit articles altogether, instead producing bare nouns. The result is the prosodic representation in (11) which is, of course, well-formed for bare nominals in Turkish and English, although inappropriate in contexts requiring articles. As mentioned earlier, we expect deletion to be particularly common at low levels of proficiency.





Heather Goad and Lydia White

(11)

PPh PWd bóok

Alternatively, the representation in (3a) can be avoided by stressing the article or by substituting some other stressed determiner. In both cases, the presence of stress requires that the determiner be organized as an independent PWd, as shown in (12); this representation is permitted in Turkish as well as in English (see (4)). (12)

PPh PWd

PWd

thé óne

bóok bóok

It is also possible that learners will be able to produce seemingly target-like unstressed articles, by using the Turkish affixal clitic representation in (3b) required for unstressed bir. In most contexts where unstressed articles are produced, it will be impossible to determine which structure is adopted, that is, whether L2 speakers have acquired the target representation in (3a) or whether they continue to use the transferred representation in (3b). However, given the word order alternation shown in (5) and (6) above, English DPs containing adjectives can be used to determine the interlanguage representation of unstressed articles. If subjects have acquired the appropriate representation for prosodifying articles as free clitics, namely (3a), no problem should arise in the context of adjectives. If, on the other hand, they represent articles as affixal clitics, as in (3b), the PTH predicts problems in DPs containing adjectives. We do not expect speakers to adopt the Turkish word order (adjective article noun, as in (6a)), since this would violate the syntactic structure of English.7 Instead, we predict an asymmetry in suppliance of unstressed (target-sounding) articles in contexts with and without adjectives. In particular, articles that are unstressed should be supplied less frequently in DPs with adjectives than in DPs without adjectives. This is because affixal clitics must prefix onto 7. On Trenkic’s account, articles are treated as adjectives in Serbian-English and TurkishEnglish interlanguage because of L1 transfer. The adjectival status of determiners in the L1 is, in part, motivated by the fact that word orders like (8b) and (8d) are possible in Serbian (Trenkic 2004: 1412), as well as in Turkish. Given such a syntactic misanalysis, then, one might expect adjective article noun word order to show up in the interlanguage. Trenkic does not discuss this possibility and it is not clear how she could rule it out.



Prosodic transfer and determiners

the head noun, something which can only be achieved if the article directly precedes the noun, and not if it directly precedes an adjective, as shown in (13), repeated from (6b). (13)

*PPh

PWd

PWd PWd

bir a

iyi good

adam man

In consequence, deletion and stressing of articles, as well as substitution of other determiners in place of articles, are all expected to be higher in the presence of adjectives. 4. An alternative hypothesis on the representation of L2 articles 4.1

Syntactic misanalysis

Trenkic (2007) argues that treatment of articles in L2 English cannot be accounted for by the PTH because learners whose L1 is Serbian omit articles more in the context of an adjective, even though the representation required for English articles is available in the L1 for a subset of determiners (Zec 2005). Since prosodic transfer should help rather than hinder in this case, an asymmetry between contexts with and without adjectives is unexpected. Further, the asymmetry is found in written as well as oral production, suggesting that prosodic structure cannot be implicated. Trenkic accounts for the observed behaviour in her own and previous studies in terms of a representational deficit together with processing pressures largely relating to pragmatic factors. In particular, learners whose L1s lack DP fail to analyse English articles as D elements and instead misanalyse them as adjectives, with meanings associated with identifiability (definite, in the case of the, and indefinite, in the case of a). Deletion of articles occurs because pragmatically redundant information is liable to be omitted under communication pressure. Overt expression of definiteness is largely redundant because there will be other cues to identifiability in the context. Hence, articles in general are subject to omission. Furthermore, when another adjective is present, this increases the complexity of the utterance, making it even more likely that articles will be omitted, since they are generally less informative than other adjectives. In addition, Trenkic states that indefinite





Heather Goad and Lydia White

articles are more likely to be omitted than definite because they require greater computational resources than definite articles, since they encode not only identifiability but also number and countability. Finally, Trenkic argues that there are certain contexts which favour redundancy; hence, article omission is particularly likely to occur. This includes ‘given’ or old information; that is, articles are more likely to be omitted from NPs that are topics or that have previously been introduced into the discourse than from NPs introducing new information.8 Note as well that if articles are in fact adjectives, they should invariably be produced as prosodically prominent (bearers of pitch or stress), as discussed in Section 2.2. Concerning Serbian-English interlanguage, we assume that this is not the case, since Trenkic does not mention it. As we will see below, it is also not the case in Turkish-English interlanguage. Stress on articles is common but it is not inevitable. 4.2

Predictions of two accounts of article representation

Table 1 summarizes the predictions of the two approaches to article representation under discussion. Article omission is expected, although for different reasons, by both the PTH (in order to avoid a particular prosodic representation) and by the syntactic misanalysis account (in response to communication pressure). Stressing of articles and substitutions of other determiners in place of articles is predicted by the PTH; the reason for these options is again prosodic, in order to avoid (3a). Stressing of all articles is in fact predicted by the syntactic misanalysis account, although Trenkic does not consider this prediction. Problems with articles in adjectival contexts are predicted by both the syntactic misanalysis account (which expects article deletion) and by the PTH (which expects a variety of options to be displayed in learners’ productions, including – but not limited to – deletion). Greater omission of indefinite articles in general is predicted by the syntactic misanalysis approach, particularly differential treatment of articles in the context of old versus new information, such as topics or previous mention in the discourse. Table 1.  Predictions of PTH and syntactic misanalysis

PTH Art as adj

omission

stressing

√ √

some all

substitution adj problems indef > def old info √

√ √

√

√

8. In fact, as far as we can tell, these claims about effects of redundancy do not depend on the misanalysis of articles as adjectives and could just as easily be made if articles are analysed as articles (see, for example, Robertson 2000).



Prosodic transfer and determiners

In the following Sections, we report on an experiment testing the predictions of these two approaches to article representation. We examine the treatment of articles in the L2 English of Turkish speakers, considering the extent to which their behaviour supports a prosodic account over syntactic misanalysis. 5. Experiment 5.1

Methodology

The experiment involved 18 Turkish-speaking learners of English, whose proficiency level (determined by means of a cloze test and self-report) was low (n=9), intermediate (n=7) and advanced (n=2).9 Mean age of the subjects at time of testing was 22.5 years, range 16–33. Six subjects were living in Canada and were tested there; mean length of residence was 2.5 years, range 6 months to 6 years. Twelve subjects (none of whom had ever resided in an English-speaking country) were tested in Turkey. With one exception, all subjects were currently university students; all of them had learned some English in secondary school (in Turkey) and the majority were currently taking an English course at university (in Turkey or in Canada).10 The experiment used an elicited production task, involving a sequence of pictures telling a story. The story was designed so that a variety of nouns and adjectives would (hopefully) be elicited, including the use of pictures which contrasted two items (e.g., a small girl and a taller girl; a red balloon and a blue one). Subjects had to describe what was happening in their own words. They were taped using a SONY PCM-M1 DAT recorder. The data were subsequently transcribed and coded in two different ways: (i) syntactic coding, which coded presence or absence of articles, substitutions, what article would be expected in the context, etc.; (ii) narrow phonetic transcription, which focused on correlates of stress.11 Orthographic 9. Unfortunately, numbers in each group are unbalanced as we were unable to find sufficient subjects at an advanced level of proficiency. 10. One of the intermediate-level subjects, T13, attended a German-medium secondary school in Turkey and reported himself as a near-native speaker of German. German has articles which, in at least one construction (in sentences with topicalized NPs), can be organized in the same fashion as in English (Kupisch et al. in press). This may have contributed to his relatively strong performance. 11. The vowel in an article was coded as stressed when it was perceived as having higher than target-like pitch, greater than target-like duration and/or greater than target-like loudness (based on narrow transcription). Our transcribers were trained to pay particular attention to all of these factors. Note that F0 contour is the principal acoustic cue for stress in both Turkish and English; English relies secondarily on duration while Turkish relies on intensity (e.g., Fry 1958; Lehiste 1976 versus Konrot 1981; Levi 2005).





Heather Goad and Lydia White

transcription and syntactic coding were carried out by native speakers or near native speakers of English and checked by three native speakers; coders showed a high degree of consistency.12 Phonetic transcription and coding was carried out by a native speaker of English with extensive training in phonetic transcription. A second native speaker of English with significant training in phonetic transcription re-transcribed data from every subject (20% to 100% of the data, depending on sample size). The inter-transcriber reliability was 88.4% (range 80%-100%). 5.2

Results

We restrict our analysis to DPs with singular count nouns, since these require an article (or other determiner) in English; in other words, they cannot be bare, unlike indefinite plurals and mass nouns. The task yielded an average of 111 such contexts per subject (range: 61–204). Results from DPs without adjectives are presented in Figure 1, which shows the patterns of suppliance for definite and indefinite contexts taken together (since they have the same prosodic representation). Subjects are organized by proficiency level; within each proficiency group they are organized in terms of proportion of suppliance of target-like forms. Non-target productions include omission of articles, as well as stressing of articles and substitution of other determiners in place of articles.13 Target-like productions are unstressed.14 As can be seen, omission of articles is robustly observed for all of the low proficiency subjects and for three of the intermediate group (T16, T8, T15), ranging from 20% to 80% of their article contexts. The omission rates of the remaining subjects were low (10% or less). Almost all subjects (with the exception of T15 and T2) produce a sizeable proportion of stressed articles, ranging from 15% to 40% of their productions. As for target-like articles, both of the advanced subjects (T6, T2) and five of the intermediate group (T15, T3, T9, T13, T1) produced a majority of DPs with unstressed articles, ranging from 61% to 87% of their DPs. 12. The study included no native speaker controls. However, the coding by native speakers effectively provided controls, since coders indicated whether or not utterances were grammatical, which articles were appropriate and which not, etc. 13. In addition, articles were sometimes produced followed by pauses or fillers which has the effect of breaking up the representation and thereby averting the target prosodic structure we are interested in. However, there are many different reasons why L2 speakers (and indeed native speakers) might insert pauses or fillers, for example, for reasons of lexical retrieval. For this reason, these items have been excluded from the analyses. 14. Target-like productions are unstressed, but they may not, in fact, involve the target representation; recall that article + noun constructions will sound target-like if either of the representations in (3a) or (3b) is used. DPs with intervening adjectives will help to tease apart these options.



Prosodic transfer and determiners 

% omitted 100.00 90.00 80.00 70.00 60.00 50.00 40.00 30.00 20.00 10.00 0.00

% stressed

% substitution

% targetlike

T18 T10 T11 T17 T5 T12 T4 T7 T14 T16 T8 T15 T3 T9 T13 T1 T6 T2 Low

Intermediate

High

Figure 1.  Patterns of article suppliance for DPs without adjectives

Amongst non-target productions, omission of articles (as in (14)) and stressing of articles (underscored, as in (15)) were the most frequently-attested patterns. Inappropriate substitutions, such as one in place of a (see (16)) and this or that in place of the, were less common.15 (14) Susan... reading book (T18) (15) a man and a girl (T10) (16) and one girl is serving something (T17) For five of the seven subjects who produced more target-like articles than nontarget (T3, T9, T13, T1, T6), stressing predominated over deletion in non-target productions (20%-29% stress versus 4%-9% deletion); one subject (T15) produced more deletion (28%) than stress (8%); the seventh subject (T2) produced very few deletions or stressed forms. The behaviour of most of these subjects suggests that they are aware that English singular count nouns require overt articles. There is a significant positive correlation (r=.813; p<.001) between rate of target-like suppliance in DPs without adjectives and proficiency level (as determined by the cloze test alone), with low proficiency subjects showing the lowest proportion of target-like forms and higher proficiency subjects producing correspondingly more. This raises the question of whether the higher proficiency subjects have indeed acquired the target representation. For this reason, we turn now to a consideration of DPs with singular count nouns containing adjectives, in order to see whether target-like productions drop in the context of an adjective; these 15. We restrict the term substitution here, reserving it for substitutions of other determiners in place of articles. We return to overuse of definite for indefinite, and vice versa, below.



Heather Goad and Lydia White

% stressed

% omitted 100 90 80 70 60 50 40 30 20 10 0

T10

T11

T7

T14

T16

T8

Low

T15

T3

T9

T13

Intermediate

% targetlike

T1

T6

T2

High

Figure 2.  Patterns of article suppliance for DPs containing adjectives

results are presented in Figure 2. (Note that only one subject produced substitutions of other determiners in DPs with adjectives, so substitutions have been omitted from this analysis.) Several subjects are excluded because they did not produce sufficient adjectives for a comparison to be possible, despite our attempts to elicit them, leaving 13 subjects for this analysis. Excluded subjects were all from the lowest proficiency group, namely T18, T17, T5, T12, T4.16 Recall that we predict greater problems with these DPs if L2 speakers are adopting the L1-based affixal clitic representation in (3b). While this representation is viable in the absence of an adjective and will in fact sound target-like, it is illicit in contexts where adjectives intervene between the article and noun. A comparison of figures 1 and 2 reveals that a number of subjects show behaviour in the predicted direction. In other words, in the context of an adjective, their suppliance of unstressed articles falls and they produce correspondingly more non-target forms involving deletion or stressing. The examples in (17) illustrate deletion and stressing in the presence of adjectives. (17) a. She has pink … skirt. (T15) b. Susan uh bought the … uh red one and the yellow hat. (T16) Chi-square tests show that for 7 of the 13 subjects for whom the adjective analysis was possible, the contingency between presence of an adjective and non-target16. Exclusion criteria were as follows: fewer than 12 adjectives produced and/or production of DPs with adjectives was less than 12% of total DPs.



Prosodic transfer and determiners 

like article production was significant or neared significance (T10: χ2=3.795, p<.05; T11: χ2=5.674, p<.02; T7: χ2=3.337, p<.0677; T14: χ2=4.326, p<.05; T8: χ2=11.792, p<.001; T3: χ2=16.536, p<.0001; T6: χ2=4.798, p<.05).17 However, in the case of T14, the contingency is in the wrong direction, for which we have no explanation. The observed effect cannot be attributed to proficiency, amount of L2 input or to type of exposure to English, since a contingency was found at all levels of proficiency and for two of the six subjects who were living in Canada (T3, T6). Let us consider the contingency data in more detail, in terms of how targetlike overall performance was on DPs without adjectives. These results are summarized in Table 2. Six of the subjects for whom this comparison is possible (T10, T11, T7, T14, T16, T8) produce unstressed articles 49% of the time or less in DPs without adjectives. A comparison of their target-like productions in Figures 1 and 2 shows that four of these subjects (T10, T11, T7, T8) have the predicted asymmetry in the right direction, in other words, omitting or stressing articles significantly more in the presence of an adjective. As previously mentioned, T14 shows the asymmetry in the wrong direction, while T16 shows no significant contingency. Four subjects (T15, T3, T9, T13) showed target-like productions ranging from 61–68% when no adjective was present; of these subjects, two (T3, T6) show the expected asymmetry. Finally, three subjects (T13, T1, T2) produced a substantial majority of target-like forms (70% or higher) when no adjective was present. If their target-like productions are indeed truly on target (that is, using the representation in (3a) rather than (3b)), no asymmetry is expected between performance on DPs with and without adjectives. This expectation holds for all three of them. (Note that T13 is the subject who reported being near-native in German, so possibly we are seeing beneficial effects from L2 to L3 here.) To summarize, results show that non-target production is common, especially in subjects of lower proficiency. Omission and stressing are found, as well as some cases of substitution. Furthermore, when apparently target-like unstressed articles are produced, a comparison of DPs involving adjectives with DPs that do not contain adjectives suggests that unstressed articles are represented with the Turkish structure (3b) rather than the English one (3a) for approximately half of the 13 subjects for whom this analysis was possible. The number of subjects who show no contingency between presence of an adjective and target-like production of articles increases as target-like performance on articles in DPs without adjectives increases.

17. We recognize that the chi-square results must be interpreted with caution, as the observations are not fully independent.



Heather Goad and Lydia White

Table 2.  Contingency or no contingency on ±adjective Target-like performance on DPs without adjectives

Contigency on ± adj

No contigency

Not enough DPs with adjectives T18, T17, T5, T12, T4

0 – 49%

T10, T11, T7, T8, (T14)

T16

50 – 69%

T3, T6

T15, T9

70 – 100%

5.3

T13, T1, T2

Syntactic misanalysis revisited

We turn now to omission of articles in more detail, looking at two claims of Trenkic’s (2007) syntactic misanalysis account, namely: (i) indefinite articles should be omitted more than definite; (ii) articles are more likely to be omitted from NPs which are topics or which have been previously mentioned in the discourse. Table 3 compares omission of definite and indefinite articles. It can be seen that there are some subjects who omit indefinites more than definites, some who omit definites more than indefinites and some who show no difference. The subjects for whom the contingency between definiteness and omission is significant are T18 (χ2=3.869, p<.05), T10 (χ2=13.178, p<.001), T11 (χ2=13.973, p<.001), T15 (χ2=4.73, p<.05) and T13 (χ2=7.979, p<.01), who, contrary to Trenkic’s prediction, omit a higher proportion of definite articles, and T4 (χ2=12.885, p<.001), T16 (χ2=3.986, p<.05) and T2 (χ2=5.337, p<.02), who omit a higher proportion of indefinites. Perhaps crucially for Trenkic’s claim that articles are more often deleted in the context of adjectives for reasons of computational pressure, of the six subjects who showed the relevant contingency between presence of adjectives and omission/stressing of articles (T10, T11, T7, T8, T3, T6; see Table 2), only two (T10, T11) show a significant difference in omission depending on definiteness, both in the wrong direction. This suggests that omission of articles before adjectives is independent of supposed difficulties in computation for the Turkishspeaking learners of English under discussion. Table 3.  Omission of definite and indefinite articles in obligatory contexts Low

Intermediate

High

T18 T10 T11 T17 T5 T12 T4

T7 T14

T16 T8 T15 T3

T9 T13 T1

T6

Omission of a

22/ 11/ 37 45

3/ 25

22/ 28/ 61 91

1/ 46

Omission of the

20/ 27/ 19/ 20/ 11/ 10/ 24 43 26 60 40 33

8/ 12/ 10/ 33 49 33

8/ 20/ 48 31

8/ 55

7/ 12/ 18/ 34 51 78

4/ 34

9/ 83

16/ 23/ 13/ 12/ 77 112 39 90

T2

0/ 32

4/ 65

4/ 10/ 46 73

8/ 25/ 75 120

7/ 97

11/ 4/ 71 100

% zero indef

59

24

24

24

30

17

65

12

15

36

31

12

11

2

0

6

9

14

% zero def

83

63

73

33

28

30

21

24

23

21

21

33

13

11

21

7

15

4



Prosodic transfer and determiners 

Table 4.  Suppliance of articles and topic status (in %) Low

Intermediate

T18 T10 T11 T17 T5 T12 T4

High

T7 T14

T16 T8 T15 T3

T9 T13 T1

T6

T2

% overt article, subject

0

27

0

71

76

73

86

74

94

73

82

71

94

89 100

94

88 100

% overt article, non subject

6

59

52

71

77

79

61

83

77

69

70

75

87

93

93

91

79

90

In addition, article omission does not correlate with whether or not the NP is a topic (Table 4) or whether or not first mention is involved (Table 5). According to Trenkic’s account, article deletion is expected to be higher in the case of NPs expressing topics and with NPs involving subsequent mention, because any lexical meaning conveyed by the article is less critical in these cases, since they convey old information. Table 4 compares the percent article suppliance in the case of preverbal subjects (standardly taken to be topics) with NPs that are not subjects. It can be seen that the proportion of overt articles with subjects and non-subjects is, by and large, similar. Only three participants (T10: χ2=4.013, p<.05; T13: χ2= 9.114, p<.01; T2: χ2=4.747, p<.05) show a significant contingency between suppliance of an article and the topic/subject status of the NP, and for two of these (T13, T2), the contigency is in the opposite direction to that predicted by Trenkic. Table 5 looks at whether an NP provides first or subsequent mention of the referent in question. Suppliance of an overt article is independent of first mention for almost all participants. Again, only three participants (T10: χ2=12.201, p<.001; T11: χ2=8.542, p<.01; T15: χ2=5.775, p<.02) show a significant contingency between suppliance of an article and whether first or subsequent mention is concerned, in the direction predicted by Trenkic. Table 5.  Suppliance of articles and first vs. subsequent mention (in %) Low

Intermediate

T18 T10 T11 T17 T5 T12 T4

High

T7 T14

T16 T8 T15 T3

T9 T13 T1

T6

T2 88

% overt article, 38 first mention

69

64

75

71

80

52

79

81

73

70

83

87

95

79

95

89

% overt article, 14 subsequent

28

20

54

68

67

65

85

81

68

81

56

92

89

98

88

89 100



Heather Goad and Lydia White

6. Discussion To summarize, Turkish-speaking learners of English at various proficiency levels have revealed linguistic behaviour most consistent with an account of interlanguage articles in terms of prosodic transfer. An analysis simply in terms of overt suppliance might have given the misleading impression that many of our subjects’ representations of articles are relatively unproblematic, since the majority of them supply articles more often than not. By dividing overt articles into unstressed and stressed, we have been able to show that deletion is not the only strategy which learners adopt to circumvent the target English representation. While omission of L2 articles can in principle be explained on other accounts, the relatively high level of stressed articles is not susceptible to explanations that do not invoke prosody. There is no morphological, syntactic or semantic reason to stress articles;18 hence, this finding supports a phonological approach to the interlanguage representation of functional structure, in particular the PTH. With respect to the production of target-like articles, several subjects show a higher incidence of article deletion or stressing in DPs containing adjectives, suggesting that unstressed articles, at least for these subjects, do not involve the target L2 representation (in (3a)) and, instead, are represented by means of the L1 representation (in (3b)). Of course, articles should always be produced as stressed if they are indeed adjectives (as Trenkic maintains) because they must form independent PWds (Section 2.2), something which is not observed in our data. In other words, the observation that stressed and unstressed articles co-occur in our data goes unexplained on Trenkic’s account. The syntactic misanalysis proposal of Trenkic expects article omission to be particularly robust in adjectival contexts because of communication pressure. Thus, finding higher instances of article omission in the context of adjectives does not serve to differentiate between Trenkic’s proposal and the PTH, although higher incidence of other non-target strategies, such as stress, does. However, we have shown that several other predictions of the syntactic misanalysis account are not supported in our data, namely that articles should be omitted more when new information is involved and that indefinite articles should be omitted more than definite.19 The absence of such findings suggests that syntactic misanalysis is not implicated here. In short, the only explanation which accounts for the cluster of

18. Of course, articles under focus are stressed in English, but the contexts in which the learners produced stressed articles in our experiment were not ones where stressing would normally be appropriate. 19. Although we do not find greater omission of indefinites in the present study, we did find this pattern with SD (see White 2003; Goad and White 2004).



Prosodic transfer and determiners 

phenomena we have found (omission, stressing, substitution, differential treatment of articles before adjectives) is provided by the PTH. It should be noted that the PTH does not exclude the possibility of multiple influences on L2 article production (see White 2007). For example, there might be semantic or discourse factors involved as well as phonological ones, which could explain why a few of the subjects did omit indefinite articles more than definite, for example. In contrast, according to Trenkic, the syntactic misanalysis account explains all aspects of L2 article production, the claim being that the misanalysis of adjectives as articles can explain the phenomena argued by Goad and White (2004) to be prosodic. In other words, Trenkic aims to replace the PTH as opposed to supplementing it. Our results provide evidence against this position. Finally, we consider the implications of our findings for the question of whether or not L1-based prosodic representations are permanent. Although as originally proposed (Goad, White and Steele 2003), the PTH might be seen as a phonological version of the RDH, that is, that L2 speakers cannot appropriately modify L1 prosodic representations in response to L2 input, in subsequent work (Goad and White 2004), we found evidence for restructuring under the two conditions in (9). In Goad and White (2006b), we predicted that ultimate attainment of appropriate L2 prosodic representations will differ depending on how the particular functional material must be organized, as an affixal clitic (usually acquirable under the conditions in (9)) or as a free clitic (often not acquirable under the conditions in (9)). For the case under discussion, the necessary structure for English articles is not acquirable under either of the conditions in (9) (see Section 3.1). However, the results from T13, T1 and T2 indicate that new prosodic structures can in fact be built under conditions other than those in (9), thereby providing support for a weaker interpretation of the PTH. Three subjects of intermediate proficiency (T16, T15, T9) also appear to be in the process of acquiring the target representation for English articles, while at the same time showing considerable L1 effects. In conclusion, since we do have some evidence that target prosodic representations are acquirable in L2 under conditions other than those in (9), this suggests that the PTH is the prosodic equivalent of Full Transfer Full Access (FTFA) (Schwartz and Sprouse 1996) rather than the RDH or other Full Transfer Partial Access models (see White 2003b). We note that development in the phonological domain lags behind the syntactic domain as far as article acquisition is concerned. In the sub-domain of prosodic phonology, the explanation for this is presumably that learners must acquire the syntax of new functional material before they can begin to figure out how to organize it prosodically. Given that the same morphosyntactic element can be prosodified in different ways across grammars (e.g., prepositions in different NeoŠtokavian dialects of Serbo-Croatian (Zec 1993)), within a single grammar depending on prosodic criteria (e.g., determiners in



Heather Goad and Lydia White

Standard Serbian (Zec 2005)), or within a single grammar depending on syntactic considerations (e.g., presence or absence of focus; final versus non-final position in the phrase (e.g., Selkirk 1996)), the presence of a lag in the acquisition of prosodic phonology is not surprising. Acknowledgements We would like to thank the following research assistants: Wid Al Bayaty, Erin King, Moti Lieberman, Cornelia Loos, Jen Mah, Mizuki Mazzotta, Corrine McCarthy, Luisa Meroni, Monika Molnar, Jen Morehouse, Adèle-Elise Prévost, Chen Qu, Mari Umeda. Thanks as well to Shabana Ali for her artistic skills and to Natalie Batmanian, Ayşe Gürel, Nihan Ketrez, Öner Özçelik and Özge Öztürk for comments and judgments on Turkish. Special thanks to Ayşe Gürel for organizing the testing in Turkey. This research was funded by grants from SSHRCC and FQRSC. References Abney, S. 1987. The English noun phrase in its sentential aspect. PhD dissertation, MIT. Beck, M.-L. 1998. L2 acquisition and obligatory head movement: English-speaking learners of German and the local impairment hypothesis. Studies in Second Language Acquisition 20: 311–348. Duffield, N., White, L., Bruhn de Garavito, J., Montrul, S. & Prévost, P. 2002. Clitic placement in L2 French: Evidence from sentence matching. Journal of Linguistics 38: 487–525. Erguvanli, E.E. 1984. The function of word order in Turkish grammar. Berkeley CA: University of California Press. Eubank, L. 1996. Negation in early German-English interlanguage: More valueless features in the L2 initial state. Second Language Research 12: 73–106. Franceschina, F. 2001. Morphological or syntactic deficits in near-native speakers? An assessment of some current proposals. Second Language Research 17: 213–247. Fry, D. 1958. Experiments in the perception of stress. Language and Speech 1: 126–152. Goad, H. & White, L. 2004. Ultimate attainment of L2 inflection: Effects of L1 prosodic structure. In EUROSLA Yearbook 4, S. Foster-Cohen, M. Sharwood Smith, A. Sorace & M. Ota (eds), 119–145. Amsterdam: John Benjamins. Goad, H. & White, L. 2006a. Prosodic transfer: L1 effects on the production of L2 determiners. In Proceedings of the 30th Annual Boston University Conference on Language Development, D. Bamman, T. Magnitskaia & C. Zaller (eds), 213–224. Somerville MA: Cascadilla Press. Goad, H. & White, L. 2006b. Ultimate attainment in interlanguage grammars: A prosodic approach. Second Language Research 22: 243–268. Goad, H. & White, L. 2008. Prosodic structure and the representation of L2 functional morphology: A nativist approach. Lingua 118: 577–594.



Prosodic transfer and determiners  Goad, H., White, L. & Steele, J. 2003. Missing inflection in L2 acquisition: Defective syntax or L1-constrained prosodic representations? Canadian Journal of Linguistics 48: 243–263. Hawkins, R. 2000. Persistent selective fossilisation in second language acquisition and the optimal design of the language faculty. Essex Research Reports in Linguistics 34: 75–90. Hawkins, R. 2003. ‘Representational deficit’ theories of (adult) SLA: Evidence, counterevidence and implications. Plenary paper presented at EUROSLA, Edinburgh, Sept. Hawkins, R., Al-Eid, S., Almahboob, I., Athanasopoulos, P., Chaengchenkit, R., Hu, J., Rezai, M., Jaensch, C., Jeon, Y., Jiang, A., Leung, Y.-k.I., Matsunaga, K., Ortega, M., Sarko, G., Snape, N. & Velasco-Zárate, K. 2006. Accounting for English article interpretation by L2 speakers. In Eurosla Yearbook 6, S. Foster-Cohen, M.M. Krajnovic & J.M. Djigunovic (eds), 7–25. Amsterdam: John Benjamins. Hawkins, R. & Chan, C. Y-H. 1997. The partial availability of Universal Grammar in second language acquisition: The ‘Failed Functional Features Hypothesis’. Second Language Research 13: 187–226. Hawkins, R. & Liszka, S. 2003. Locating the source of defective past tense marking in advanced L2 English speakers. In The interface between syntax and lexicon in second language acquisition, R. van Hout, A. Hulk, F. Kuiken & R. Towell (eds), 21–44. Amsterdam: John Benjamins. Huebner, T. 1985. System and variability in interlanguage syntax. Language Learning 35: 141–163. Hulst, H. van der & Weijer, R. van de. 2004. Vowel harmony. In The Handbook of Phonological Theory, J. Goldsmith (eds), 495–534. Oxford: Blackwell. Inkelas, S. & Orgun, C.O. 2003. Turkish stress: A review. Phonology 20: 139–161. Ionin, T. 2003. Article semantics in second language acquisition. PhD dissertation, MIT. Ionin, T., Ko, H. & Wexler, K. 2004. Article semantics in L2 acquisition: The role of specificity. Language Acquisition 12: 3–69. Kabak, B. & Vogel, I. 2001. The phonological word and stress assignment in Turkish. Phonology 18: 315–360. Konrot, A.K. 1981. Physical correlates of linguistic stress in Turkish. University of Essex Language Centre Occasional Papers 24: 26–53. Kornfilt, J. 1997. Turkish. London: Routledge. Kupisch, T., Anderssen., M., Bohnacker, U. & Snape, N. In press. Article acquisition in English, German, Norwegian and Swedish. In Georgetown University Round Table on Languages and Linguistics (GURT) 2007 Proceedings, R.W. Leow, H. Campos & D. Lardiere (eds.). Washington DC: Georgetown University Press. Lardiere, D. 1998. Dissociating syntax from morphology in a divergent end-state grammar. Second Language Research 14: 359–375. Lehiste, I. 1976. Influence of fundamental frequency pattern on the perception of duration. Journal of Phonetics 4: 113–117. Leung, Y.-k.I. 2002. Functional categories in second and third language acquisition: A crosslinguistic study of the acquisition of English and French by Chinese and Vietnamese speakers. PhD dissertation, McGill University. Levi, S.V. 2005. Acoustic correlates of lexical accent in Turkish. Journal of the International Phonetic Association 35: 73–97. Lyons, C. 1999. Definiteness. Cambridge: CUP. Newell, H. 2005. The phonological phase. In McGill Working Papers in Linguistics 19, R. Mercado & Y. Furukawa (eds), 21–63.



Heather Goad and Lydia White Öztürk, B. 2005. Case, referentiality and phrase structure. Amsterdam: John Benjamins. Peperkamp, S. 1997. Prosodic words. The Hague: HAG. Poser, W.J. 1990. Word-internal phrase boundary in Japanese. In The phonology-syntax connection, S. Inkelas & D. Zec (eds), 279–287. Chicago IL: University of Chicago Press. Robertson, D. 2000. Variability in the use of the English article system by Chinese learners of English. Second Language Research 16: 135–172. Schwartz, B. D. & R. Sprouse. 1996. L2 cognitive states and the full transfer/full access model. Second Language Research 12: 40–72. Selkirk, E.O. 1996. The prosodic structure of function words. In Signal to Syntax, J.L. Morgan & K. Demuth (eds), 187–213. Mahwah NJ: Lawrence Erlbaum. Sezer, E. 1981. On non-final stress in Turkish. Journal of Turkish Studies 5: 61–69. Snape, N. 2006. The acquisition of the English determiner phrase by Japanese and Spanish learners of English. PhD dissertation, University of Essex. Thomas, M. 1989. The acquisition of English articles by first- and second-language learners. Applied Psycholinguistics 10: 335–355. Trenkic, D. 2004. Definiteness in Serbian/Croatian/Bosnian and some implications for the general structure of the nominal phrase. Lingua 114: 1401–27. Trenkic, D. 2007. Variability in L2 article production – beyond the representational deficit vs. processing constraints debate. Second Language Research 23: 289–327. Trenkic, D. 2008. The representation of English articles in second language grammars: Determiners or adjectives. Bilingualism: Language and Cognition 11: 1–18. Tsimpli, I.-M. 2003. Clitics and determiners in L2 Greek. In Proceedings of the 6th Generative Approaches to Second Language Acquisition Conference (GASLA 2002): L2 links, J.M. Liceras, H. Zobl & H. Goodluck (eds), 331–339. Somerville MA: Cascadilla Press. Tsimpli, I.-M. & Mastropaviou, M. 2007. Feature interpretability in L2 acquisition and SLI: Greek clitics and determiners. In The role of formal features in second language acquisition, J. Liceras, H. Zobl & H. Goodluck (eds), 143–183. New York: LEA. Tsimpli, I.-M. & Roussou, A. 1991. Parameter resetting in L2? UCL Working Papers in Linguistics 3: 149–169. Underhill, R. 1976. Turkish grammar. Cambridge MA: The MIT Press. White, L. 2003a. Fossilization in steady state L2 grammars: Persistent problems with inflectional morphology. Bilingualism: Language and Cognition 6: 129–141. White, L. 2003b. Second language acquisition and Universal Grammar. Cambridge: CUP. White, L. 2007. Some puzzling features of L2 features. In The role of formal features in second language acquisition, J. Liceras, H. Zobl & H. Goodluck (eds), 301–326. New York: LEA. Zec, D. 1993. Rule domains and phonological change. In Studies in lexical phonology, S. Hargus & E. Kaisse (eds), 365–405. San Diego CA: Academic Press. Zec, D. 2005. Prosodic differences among function words. Phonology 22: 77–112.

Exploring Mandarin Chinese speakers’ L2 article use Neal Snape

Gunma Prefectural Women’s University This paper attempts to explore the difficulties Chinese speakers have with the acquisition of English articles by situating itself within three hypotheses by using two tasks; a written forced choice elicitation task and an oral elicited picture description task. Our findings from the first task are consistent with the Fluctuation Hypothesis. Our findings from the second task are not fully consistent with the Representational Deficit Hypothesis or the strong interpretation of the Prosodic Transfer Hypothesis as omission of articles is low. This indicates that L2 learners are perhaps able to access uninterpretable features and target-like prosodic representations even at intermediate levels, which is consistent with the view of full access to syntax and phonological structures via Universal Grammar.

1. Introduction During the past few years there have been numerous studies investigating the acquisition of articles by second language (L2) learners (White 2003, Lardiere 2004, Hawkins et al. 2006, Snape, Leung and Ting 2006, Trenkic 2008). The authors of the various studies have attempted to offer explanations as to why L2 learners continue to have difficulties with L2 acquisition with reference to articles in English. These difficulties can be attributed to three factors; (1) fluctuation between semantic features [+definite] and [+specific] according to Ionin, Ko and Wexler’s (2004, 2008a) Fluctuation Hypothesis (FH); (2) the acquisition of uninterpretable formal features, which are not present in the L1 grammar (Tsimpli and Roussou 1991) referred to as the Representational Deficit Hypothesis (RDH) (Hawkins 2000, 2005); (3) the Prosodic Transfer Hypothesis (PTH) (Goad and White 2004, 2006, 2008 and in this volume) which predicts that L2 learners who lack corresponding L1 prosodic structures will fail to produce target-like phonology. However, it is not all L2 learners who have trouble acquiring articles as the role of transfer from the



Neal Snape

first language (L1) plays a large part in successful L2 article acquisition (Thomas 1989, Murphy 1997, Wakabayashi 1997, Liu and Gleason 2002, Trademan 2002, Snape 2006). The study we report on in the current paper presents data from a written forced choice elicitation task and an oral elicited picture description task. The participants are Mandarin Chinese speakers whose L1 does not have articles. We will discuss how the findings in the present paper are consistent with the FH that L2 learners from L1s without articles fluctuate between the interpretable features of articles [+definite] and [+specific] in a forced choice elicitation task. The findings are also consistent with the RDH as interpretable features [+definite] and [+specific] are acquirable even though [+specific] is the incorrect feature for English articles. However, omission of articles is unexpectedly low compared to previous findings (Robertson 2000), which is inconsistent with the RDH that the uninterpretable count noun number feature is not acquirable due to a representational deficit (Hawkins et al. 2005), but consistent with the weaker interpretation of the PTH because learners are not deleting articles in Art+N contexts (Goad and White in this volume). The paper is organised as follows. Section 2 provides a definition of definiteness and specificity. Section 3 outlines the three hypotheses which are going to be tested in our studies. In Section 4 the nominal domain of Mandarin Chinese is presented. Section 5 describes the methodology used and the results. Section 6 is a discussion of the results. Section 7 presents the conclusion. 2. Articles: definiteness and specificity It is well documented in second language acquisition observational studies that learners of English have difficulty acquiring articles. Previous studies of L2 learners use of articles in English, such as Huebner (1985), Parrish (1987), Thomas (1989) and Murphy (1997) have attempted to explain this difficulty by employing Bickerton’s (1981) binary semantic system [±Specific Referent, ±Hearer Knowledge] for noun phrase reference. Recent studies of L2 learners by Ionin et al. (2004, 2008a) have tried to tease apart the definiteness and specificity distinction by proposing that learners from article-less languages need to select the correct feature via Universal Grammar (UG). In order to discuss what it means to select the correct feature we first must define what we mean by definiteness and specificity. For Ionin et al. (2004) definiteness is based on the Fregean analysis of definites and specificity is based on Fodor and Sag’s (1982) proposal of speaker intent to refer. Ionin et al’s informal definitions of definiteness and specificity are provided in (1):



L2 article use 

(1) Definiteness and Specificity If a Determiner Phrase (DP) of the form [D NP] is… a. [+definite], then the speaker and hearer presuppose the existence of a unique individual in the set denoted by the NP. b. [+specific], then the speaker intends to refer to a unique individual in the set denoted by the NP, and considers this individual to possess some noteworthy property. (taken from Ionin et al. 2004: 5) 2.1

Definiteness in English

Standard English has the articles the and a. Ionin et al. (2004) follow Heim`s (1991) definition of definiteness where [+definite] is a semantic feature expressed morphologically as the. Some uses of the definite article are illustrated in (2) based on J. Hawkins’s (1978) taxonomy of definites e.g., anaphoric, associative, larger situation. (2) Anaphoric use a. I saw a cat. I gave the cat some milk Associative use b. A man drove past our house in a car. The exhaust fumes were terrible. Larger (cultural) situation c. (When invited to a wedding) Have you seen the bridesmaids? Other such examples come from Lyons’s (1999) discussion of definiteness as in (3).

(3) a. The winner of this competition will get a week in the Bahamas for two. b. The man who comes with me will not regret it. (Lyons 1999: 9) The uniqueness presupposition is satisfied in examples (2) and (3) where the speaker and hearer are referring to a referent which is hypothetical, potential or takes place in the future. Other examples such as those in (4) demonstrate that the definite article can be interpreted as specific or non-specific. The indefinite article can also be interpreted as specific or non-specific as illustrated in (5).



Neal Snape



(4)

[+definite]

We can’t start the seminar because the student who’s [+specific] giving the presentation is absent – typical of Bill, he’s so unreliable. b. We can’t start the seminar because the student who’s [-specific] giving the presentation is absent – I’d go and find whoever it is, but no-one can remember, and half the class is absent. (5) [-definite] a. A dog was in here last night – it’s called Lulu and Fred [+specific] always lets it sit by the fire on wet nights. b. A dog was in here last night – there is no other expla[-specific] nation for all these hairs and scratch marks. (Lyons 1999: 171 and 172) a.

Because articles in English do not encode specificity, they are potentially ambiguous with respect to this property. According to Ionin et al. (2004) specificity is the result of the speaker identifying a noteworthy property of the referent and that if an L2 learner has a representation where the can signal both definiteness and specificity, then the cases where this will become evident are where a definite DP is non-specific or an indefinite DP is specific. In (4a) the speaker has a particular individual in mind with the noteworthy property of being unreliable. The conditions on specificity are therefore satisfied in (4a) but this is not the case in (4b). The speaker does not have a particular individual in mind with a noteworthy property because no-one can remember who is supposed to be presenting. The speaker only knows that someone is presenting from the class. In (5a) the conditions on specificity are satisfied but not in (5b). The difference between (4a) and (5a) is that the uniqueness presupposition is not satisfied in (5a) as it is indefinite. For Ionin et al. (2004) the specificity distinction is separate from the definiteness distinction. In English, definiteness cuts across the specificity distinction, but in languages like Samoan and Sango specificity cuts across the definiteness distinction. This is discussed in the next section. 2.2

Specificity in Polynesian languages (Samoan and Sango)

In Samoan, according to Mosel and Hovdhaugen (1992), articles are used to mark specificity in the language; the article le is used to mark DPs as [+specific] while se is used to mark DPs as [-specific]. An example of le is given in (6).



L2 article use



(6) [+definite, +specific] Sa i ai le ulugāli’i ‘o Papa le tane a ‘o PAST exist ART couple PRES Papa ART husband but PRES Eleele le fafine. Eleele ART woman ‘There was a couple, Papa, the husband, and Eleele, the wife.’ (Lyons 1999: 58) Sango is another language that seems to have a marker for specificity – ní. It is used to identify and single out a particular object in the discourse and it has an anaphoric function to refer to an object already mentioned in the discourse (Samarin 1967). It can also co-occur with demonstratives and be used pronominally as a third person pronoun for inanimate reference (Lyons 1999). An example from Sango is provided in example (7). (7)

[-definite, +specific] Auto afáa mbéní mέrεngέ ní car had-killed certain child ART ‘A car had killed a child’

(taken from Samarin (1967: 63)

Lyons (1999) provides further examples of Polynesian languages which encode specificity rather than definiteness. However, he argues that some languages have articles which encode definiteness and determiners that encode specificity. One such language is English, where referential this/these is specific. Examples of this are in (8) and (9).1

(8)

[-definite]

Peter intends to marry a/this merchant banker – even [+specific] though he doesn’t get on at all with her. b. Peter intends to marry a/??this merchant banker – [-specific] though he hasn’t met one yet. (taken from Lyons 1999: 176) a.

(9) a. John has a/this weird purple telephone. [+specific] b. John has a/??this telephone, so you can reach me there. [-specific] (taken from Maclaran (1982: 88)

1.

For further discussion of referential this see Prince (1981) and Ionin (2006).





Neal Snape

In (8a) the speaker intends to refer to a particular individual from a set of individuals (merchant bankers) and the individual has the noteworthy property of Peter not getting on with her. Equally in (9a) there is a particular telephone, which has the noteworthy property of being weird and purple. In (8b) and (9b) there is no particular individual or no particular telephone in the mind of the speaker and no noteworthy property of the referent to refer to. The conditions for this as a [+specific] marker are satisfied in (8a) and (9a), but this cannot be used to mark [-specific] contexts. The indefinite article a can be used in either [-definite, +specific] contexts or [-definite, -specific] contexts, thus does not encode specificity. 3. Three hypotheses relating to L2 article use 3.1

The Fluctuation Hypothesis

Ionin et al. (2004) claim that L2 learners from languages which lack articles have full access to UG, thus have access to the features [+definite] and [+specific]. However, given the absence of articles there is no L1 transfer from article semantics. L2 learners from article-less languages are expected to fluctuate between the two settings of an article choice parameter. They propose that L2 learners from articleless languages may fluctuate between parameter settings that are not instantiated in the L1 nor the L2 but that may be instantiated in some third language. In the absence of direct transfer from articles in the L1, L2 learners are predicted to fluctuate as they have no initial preference. However, Hawkins et al. (2006) argue that there is no need to propose a construct specific article choice parameter as fluctuation can be accounted for by L2 learners fluctuating between features made available by UG. Ionin et al. (2008a) make no reference to an article choice parameter. Instead, they claim that L2 learners of English fluctuate between the two features [+definite] and [+specific] overusing the in [-definite, +specific] contexts and overusing a in [+definite, -specific] contexts. They are not expected to overuse the in [-definite, -specific] contexts and a in [+definite, +specific] contexts. The predictions are given in Table 1: Table 1.  Predictions for article choice in L2 English: [+definite, +specific] [+definite] (target: the) [+specific] [-specific]

correct use of the overuse of a

[-definite] (target: a) overuse of the correct use of a (from Ionin et al. 2004: 19)



L2 article use 

Ionin et al. (2004, 2008a) tested participants from languages without articles – 30 L1 Russian speakers and 40 L1 Korean speakers. They used a forced choice elicitation task containing 76 written dialogues in English. The L2 learners were asked to choose between the most appropriate article a, the and – (null article) to fill a gap in the dialogue, basing their choice on the preceding context. There were equal numbers of [±definite, ±specific] combinations in the task. They found that the Russian and Korean L2 learners overused the in [-definite, +specific] contexts and a in [+definite, –specific] contexts, thus supporting the FH. The finding that there is variation between the uses of the features [+definite] and [+specific] means, according to Ionin et al. (2004), that specificity plays a role in the interlanguage grammars. They found that learners who overused the in [–definite, +specific] contexts were also likely to overuse a in [+definite, –specific] contexts. They also looked at individual performances within the groups to see if there was a similar pattern in use of the and a. They found that the group results conceal different individual patterns of use. The definiteness and fluctuation patterns are expected patterns of use of articles in English under the FH, but the specificity, partial and miscellaneous patterns are unexpected patterns of use (see Table 4 in Section 5.4.5). Recent follow-up studies have tested the Fluctuation Hypothesis with different groups of L2 learners from article-less languages. Reid, Battaglia, Schuldt, Narita, Mochizuki and Snape (2006) show Japanese speakers perform differently to Spanish and English native speakers, which was expected given that Japanese is a language that lacks article semantics. García Mayo (2007) and Ionin, Zubizarreta and Maldonado (2008b) have since replicated Ionin et al’s (2004) study with Spanish speakers and found that intermediate learners are highly accurate with suppliance of articles in all contexts i.e., no fluctuation. Ting (2005), reported in Snape et al. (2006), compared Chinese and Spanish native speakers. The results were unexpected as it demonstrated that the four intermediate Chinese speakers could perform as well as the Spanish and English native speakers in a forced choice elicitation task. However, given that the study reports on findings from four participants it is premature to argue that L1 Chinese do not fluctuate between definiteness and specificity in article choice. 3.2

The Representational Deficit Hypothesis

In the previous section it is predicted under the FH that L2 learners will fluctuate between interpretable features (Ionin et al. 2008a); here, definiteness (see Abney 1987) and specificity are interpretable. Fluctuation in article choice is consistent with the RDH as interpretable features are acquirable. But, the RDH claims that L2 learners’ syntax is selectively impaired, lacking parameterized uninterpretable



Neal Snape

features, not present in the L1 which are no longer accessible following a critical period for acquisition (Hawkins and Chan 1997, Hawkins 2005). For Hawkins et al. (2005) definiteness is interpretable in the syntax but all count nouns have an uninterpretable number feature. The licensing of count nouns in English is discussed in the following section. 3.2.1 The syntactic licensing of count nouns in English Chierchia (1998a) proposes that there are three types of languages in the world; 1. e.g., English that has articles, count nouns and bare mass nouns 2. e.g., Italian that has articles and count nouns 3. e.g., Chinese that only has mass-like nouns or ‘kinds’ (see Section 4). Languages like English, Italian and Chinese have different parameter values according to Chierchia’s (1998a, 1998b) Nominal Mapping Parameter (NMP). The NMP setting for English is [+argument, +predicate]. Languages like English allow NPs to be argumental [+arg], i.e., to merge directly in argument positions in syntactic derivations without further modification. For Chierchia (1998a), mass nouns and count plural nouns are potentially [+arg]. But, count singular nouns require a determiner of some kind. Count singular nouns in English and Italian are predicative [+pred], and need to be modified (that is, licensed) by determiner elements like articles, but languages like Chinese are [+arg] and do not require licensing as all nouns are ‘kinds’.2 Clearly, Chierchia’s (1998a) proposal has clear implications for the syntax of each language type. Hawkins et al. (2005) and Snape (2008a) adopt a syntactic analysis of count nouns that is based on the notion of `Agree’ defined by Radford (2000) and Radford et al. (2007) as the checking, valuing and deleting of uninterpretable features. Radford (2000) provides some empirical evidence for arguing that number is an uninterpretable feature of nouns in English. It comes from noun gapping structures such as ‘she tried on both dresses, but the blue dress was too big’, where the noun dress in the second clause is gapped: “clearly gapping cannot require PF-identity, since the gapped noun is the singular form dress whereas its antecedent is the plural form dresses; if gapping requires identity of LF-interpretable features, it follows that the number feature carried by dress/dresses in adult English must be uninterpretable” (Radford 2000: 13). The licensing of count nouns takes place in the following way, as shown in (10).

2. An anonymous reviewer commented that number is marked on nouns therefore it is interpretable, but we argue that before the noun enters into the syntactic derivation it could either be count as in ‘a dog is in the road’ or mass ‘dog is all over the road’. Thus, number is assigned via the syntax forcing one interpretation over another (Barner and Bale 2002).



L2 article use 

(10) the/a [D, Num] [D, Num]

boy [noun, animate, masc, 3p, u-Num: ] → [noun, animate, masc, 3p, u-Num: +singular]

Under a Minimalist account (Chomsky 1998) the feature [+definite] is considered to be interpretable at LF and carry an interpretable number feature. Following Lyons (1999: 298–301) we assume that D is the locus of interpretable [definiteness], and is expressed in English through the and a. The article is selected from the lexical array and enters the derivation carrying the interpretable feature [+definite].3 The noun boy and the article subsequently merge via Concord and enter into an Agreement relation where the interpretable number feature of the article checks and values the uninterpretable feature of the noun boy. Once the noun boy has been valued, the uninterpretable feature is deleted as it is uninterpretable at LF. The prediction is that L2 learners of English from languages without articles will be unable to acquire the syntactic [u-Num] feature as it has not been activated in the primary grammar. We adopt the position outlined by Chierchia (1998a) and Hawkins et al. (2005) for English. 3.3

The Prosodic Transfer Hypothesis

A prosodic approach to variability in L2 production has recently been suggested for L2 learners’ omission of inflectional morphology. Goad and White (2004, 2006, 2008) have made predictions about free forms, such as the prosodic structure of determiners. If the L1 is an article-less language it is possible to make predictions about the suppliance of articles in the L2 if the L1 lacks or has a different corresponding prosodic structure. In English, articles (the/a) are represented in the phonology as free clitics (Selkirk 1996). Representations of free clitics for Art+N constructions are illustrated in Figure 1. a. English rep: Art+N PPh PWd the/a

hat

Figure 1. 3. The indefinite article a is inherently singular but the definite article the may appear with plural nouns e.g., the boys, and with mass nouns e.g., the furniture. However, we will limit our discussion to singular count nouns.



Neal Snape

The strong interpretation of the PTH is that L2 learners from a language that does not have articles will delete articles in L2 English because they lack the corresponding L1 prosodic structures. In the case of Mandarin Chinese it is expected that Chinese learners will be able to transfer structures from the L1 in order to represent articles in the L2 if nage (that) has been grammaticalized in the L1 (Goad and White 2008). Therefore, the target representation for articles is available in the interlanguage grammars of learners, but because Mandarin is quite restricted in using determiners to head an NP (see Section 4) it is possible that alternative prosodic representations may be adapted to represent L2 articles. 4. Mandarin Chinese nominal phrases The difference between English type and Chinese type languages within Chierchia’s (1998a, 1998b) theoretical framework is that English is a language with articles and a count – mass distinction but Chinese lacks articles and count nouns, hence all nouns are ‘kinds’ i.e., treated mass-like. This is because semantically all nouns in Chinese have a category-type mapping turning all NPs into arguments, whereas in English count nouns are predicates, and D must be projected to convert them into arguments. Examples of Chinese nominal phrases are in (11). (11) a. yí lì mĭ one cl rice ‘one (grain of ) rice’ b. liăng lì mĭ two cl rice ‘two (grains of ) rice’ c. yí zhāng zhuōzi one cl table ‘one (piece of ) table’ d. liăng zhāng zhuōzi two cl table ‘two (pieces of ) tables’ e. ranhou nage lingban yizhi gaosu then that boss always tell ‘then that boss always says that’ (taken from Chierchia 1998a: 354 and Huang 1999: 85)



L2 article use 

Huang (1999) claims that nage (that) has started to be used as a definite article by Taiwanese speakers. Chen (2003, 2004) argues that the numeral yi (one) and the demonstrative nei (that) are emerging as grammatical categories in Mandarin Chinese. The fact that Chinese has some determiners which are beginning to take on the function of definite/indefinite articles does not contradict the claim by Chierchia (1998a) that Chinese is a [+arg] type language because nage, nei and yi + classifier are not used frequently by all speakers and demonstratives and numerals in Chinese do not yet have the same range of functions as articles in a language like English (see Partee 2006). We adopt Chierchia’s (1998a, 1998b) analysis of Chinese for our study. Given that Chinese is a type of language that lacks articles the FH predicts that Mandarin Chinese L2 learners will fluctuate between the features [+definite] and [+specific]. The RDH predicts that as Chinese lacks the [uNum] feature, Chinese L2 learners will be unable to acquire it as it is an uninterpretable feature. As discussed in Section 3.3, Goad and White (2008) argue that Mandarin Chinese has prosodic structures available which can be used to represent L2 articles. One such structure is illustrated in Figure 2. The free clitic representation is available for Art+N constructions in Mandarin, so in short, Mandarin speakers of English should be able to represent articles in a target-like fashion, though Goad and White (2006) found that there were many non-target-like productions and deletion amongst individual L2 learners. PPh

PPh

PWd nei the

ge CL

máo4zi hat

‘the hat’ (taken from Goad and White 2008: 585)

Figure 2.



Neal Snape

5. Empirical study 5.1

Predictions

Firstly, we want to test the generalizability of Ionin et al’s (2004) FH claim on a new group of L2 learners from an article-less language. The prediction is that L1 Mandarin Chinese speakers will perform as the L2 learners in Ionin et al’s (2004, 2008a) studies on a written forced choice elicitation task. They will fluctuate between the features [+definite] and [+specific] as Mandarin Chinese lacks articles; hence no L1 transfer is possible from article semantics. Secondly, we want to test the generalizability of the PTH by using an oral elicited picture description task. Goad and White (2006, 2008) tested Mandarin Chinese speakers using an oral elicited picture description task. Here, we replicate the study using the same task as Goad and White (2006, 2008) with another group of Mandarin speakers from mainland China.4 If the claim that grammaticalization of the demonstrative nage (that) in Mandarin Chinese has taken place in these speakers’ dialects it is likely that there are corresponding L1 prosodic structures, so we predict a high suppliance of target-like articles in Art+N contexts. However, according to the RDH, omission is expected due to an inability to acquire the uninterpretable number feature. Chinese speakers may not have acquired the syntactic licensing of count singular nouns in English; i.e., there is no direct semantic equivalent of articles in Chinese, and these speakers may not (yet) have established such a syntactic position in their interlanguage grammars. 5.2

Participants

All of the 38 adult participants in this study are Mandarin Chinese speakers currently residing in Canada. Most of the participants are enrolled in an English for Academic Purposes course at the University of Calgary. Their ages range from 18 to 40 (mean age is 21.7). The mean length of residency in Canada is 14 months. Very few of the participants have knowledge of a third language. The ones that do have knowledge of another language are currently studying Japanese or Korean. Ten of the participants are currently EFL students at the University of Toronto. Most of the participants started learning English at around 12 years of age at a high school in China. Five native speaker controls were also recruited from the 4. The data collected is part of a larger study comparing L1 Mandarin Chinese L2 English learners in an English speaking environment with L1 Mandarin Chinese L2 English learners currently living and studying in mainland China. Goad and White (2007) claim that differences between learners in reaching target-like phonology could be attributed to the learning environment.



L2 article use 

University of Calgary. The reason for testing L1 Chinese L2 English learners is because Mandarin Chinese lacks articles. 5.3

Tasks

The Chinese L2 learners were asked to complete the Oxford Quick Placement Test (2001) and two tasks. The first task was a written forced choice elicitation task based on Ionin et al’s (2004) task. The second task was Goad and White’s (2008) oral elicited picture description task. The results of the placement test allowed all the learners to be placed into an intermediate proficiency level. 5.3.1 The forced choice elicitation task The forced choice elicitation task contained a total of 86 short dialogues in English. In each dialogue there was a target sentence with an article missing. The participant was instructed to read through the dialogue and assume the role of the first speaker. The participant was asked to read each dialogue once and then choose between the articles the/a/an and the null article Ø. 36 target dialogues in the task were the same singular dialogues used in Ionin et al’s (2004) task. Examples of the target dialogues are in (12) and (13). For the L2 learners to be fluctuating between definiteness and specificity it was expected that some of the time the article a would be selected in the [+definite, -specific] context and the would be selected in the [-definite, +specific] context. For native speakers of English the target article for [+definite, -specific] contexts is the and for [-definite, +specific] contexts the target article is a. (12) [+definite, -specific] Conversation between a police officer and a reporter Reporter: Several days ago, Mr. James Peterson, a famous politician, was murdered! Are you investigating his murder? Police officer: Yes. We are trying to find ____ murderer of Mr. Peterson – but we still don’t know who he is. a an the Ø (13)

[-definite, +specific] In an airport, in a crowd of people who are meeting arriving passengers Man: Excuse me, do you work here? Security guard: Yes. Man: In that case, perhaps you could help me. I am trying to find ____ red-haired girl; I think that she flew in on Flight 239. the a Ø an



Neal Snape

There were 24 distracter items featured in the task. Distracters were generic singular with the target article being either the or a, plus plural and mass contexts, where the target article was the null article Ø and other uses of definites based on J. Hawkins’s (1978) definite types e.g., encyclopaedic use. The final 26 short dialogues were based on Trenkic (2008). Trenkic (2008) argues that the operationalisation of specificity in Ionin et al’s (2004) study conflates two unrelated factors; (1) explicit speaker knowledge (ESK) or familiarity of the person or object being referred to and (2) the intention to refer to a specific referent. In example (13) above, she classifies the [–definite, +specific] context as [–definite, +specific; +ESK] as the speaker has a specific referent in mind and explicitly states that she knows the person referred to in the conversation i.e., a red-haired girl. But, if the speaker denies any knowledge of the identity of the person the context is considered to be [–definite, +specific; –ESK] (see Trenkic 2008 for details). She argues that her findings can be explained by the stated / denied familiarity with ‘identifying attributes’ of individuals being talked about, thus considered being extra-linguistic rather than fluctuation between the semantic features [+definite] and [+specific]. Due to considerations of space the results of the 26 dialogues combining definiteness and specificity with explicitly stated knowledge will be discussed in a future paper. In our study all of the dialogues were randomized throughout the task. 5.3.2 Procedure The Oxford Quick Placement Test (2001) was administered to the participants in a classroom setting. A copy of the forced choice elicitation task was given to each participant and they were given forty five minutes to complete the task. Most participants managed to complete the task within forty minutes. 5.3.3 Coding procedure The results of the forced choice elicitation task were coded in SPSS data analysis software. The use of a and an were collapsed into indefinite use. 5.3.4 Group results of the forced choice elicitation task The five native controls chose the target article around 95% and above in the forced choice elicitation task. The results of the 38 L2 learners are presented in Table 2 below.



L2 article use 

Table 2.  Results of definiteness vs. specificity [+definite]

Intermediate Chinese L2 learners (n=38) [+specific] [-specific]

the

*a

90% 72%

9% 27%

[-definite] *Ø

*the

a

*Ø

1% 1%

23% 5%

77% 90%

0% 5%

* indicates non-target like use

The results in Table 2 show that the Chinese L2 learners perform differently in article choice to the native speakers. The findings are consistent with the FH, as the learners are fluctuating between definiteness and specificity (see the shaded areas in Table 2). Native speakers do not select the for [–definite, +specific] contexts and a for [+definite, –specific] contexts. But, the L2 learners mistakenly select a in contexts which are [+definite, –specific] and the in contexts which are [–definite, +specific]. As a group, the L2 learners performed similarly to the Russian and Korean L2 learners in Ionin et al’s (2004) study and the Japanese L2 learners in Reid et al’s (2006) study. Repeated measures ANOVAs were performed to determine whether there was a significant difference between definiteness and specificity on the use of the and a by category. The results are presented in Table 3. The results show that definiteness and specificity had significant effects on performance for the Chinese L2 learners. There is a significant effect of definiteness since most learners are able to distinguish definites from indefinites (i.e., the appears in [+definite, +specific] contexts and a appears in [–definite, –specific] contexts). However, specificity had a highly significant effect on article choice where the use of a in [±specific] contexts and the use of the in [±specific] contexts was measured. Paired-samples t-tests revealed that the L2 learners were fluctuating between definiteness and specificity in the use of a (–definite, +specific and -definite, -specific contexts, t (37) = 3.173, p <.01 and +definite, -specific and +definite, +specific contexts, t (37) = 4.259, p <.01) and in the use of the (+definite, +specific Table 3.  Effects of definiteness and specificity Use of the

Use of a

Intermediate Chinese L2 learners (n=38) Definiteness

F(1, 37) = 420.998***

F(1, 37) = 414.089***

Specificity

F(1, 37) = 27.583***

F(1, 37) = 18.996***

Definiteness x Specificity

F(1, 37) =.038

F(1, 37) = 1.152

***p <.001



Neal Snape

and +definite, –specific contexts, t (37) = –4.452, p <.001 and –definite, +specific and –definite, –specific contexts, t (37) = 4.674, p <.01). In other words, the results in Table 2 show that there is fluctuation in the predicted contexts and the pairedsamples t-tests show that the L2 learners are incorrectly selecting the and a as markers of specificity. The results are consistent with the fluctuation pattern of article choice. But, individual results show that it is not all learners who fit the fluctuation pattern. Individual results are discussed in the next section. 5.3.5 Individual results of the forced choice elicitation task The individual results, like Ionin et al’s (2004) results, show that the L2 learners do not all fit into the definiteness pattern (the correct pattern for English) or the fluctuation pattern (the predicted pattern under the Fluctuation Hypothesis). Individual results of article use are divided into five patterns presented in Table 4. None of the L2 learners fall into (c) the specificity pattern. This is expected given that English is [+definite] and that input should lead learners towards definiteness rather than specificity. However, as Ionin et al. (2004) found, (d) the partial fluctuation pattern and (e) the miscellaneous patterns of article use are unexpected. A learner fits the partial fluctuation pattern when the is correctly chosen for [+definite, +specific] and [+definite, –specific] contexts but incorrectly chosen for [–definite, +specific] contexts. Conversely, the learner would also fit the partial fluctuation pattern when a was correctly chosen for [–definite, +specific] and [–definite, –specific] contexts but incorrectly chosen for [+definite, –specific] contexts. The miscellaneous patterns are when learners incorrectly used the articles the and a in all four [±definite, ±specific] combinations. Table 4.  Chinese L2 learners: individual patterns of use Response type (a) (b)

Definiteness pattern Fluctuation pattern

(c) (d) (e)

Specificity pattern Partial fluctuation pattern Miscellaneous patterns

No. of individuals

%

8/38 19/38

21 50

0/38 6/38 5/38

0 16 13



L2 article use 

5.3.6 The elicited picture description task The elicited picture description task consisted of a series of pictures, which are based around one main character. The goal was to elicit singular count NPs headed by either a definite or indefinite article. The learner was instructed to create a story about the main character’s day using each of the pictures. The task was originally designed to elicit vocabulary items of different prosodic shapes (see Goad and White 2008 for further discussion). Each participant produced between 40 to 100 DPs requiring articles in singular contexts. Some of the DPs included article + adjective + noun e.g., a/the red jacket and noun compounds e.g., a/the store assistant. Here, we focus only on the article + noun DPs. 5.3.7 Participants All 38 participants took part in the elicited picture description task but only 15 audio files have been transcribed and coded to date. The results from the 15 transcribed files are discussed in Section 5.4.9. 5.3.8 Coding procedure The audio files were transcribed using Express Scribe software. All article + noun DPs were coded into definite/indefinite obligatory contexts. The coding procedure followed Snape (2006) whereby all contexts were judged definite or indefinite by native speaker controls. Substitution errors (definite for indefinite or vice versa) may have arisen as a result of the task used. As both the researcher (hearer) and the participant (speaker) could see the pictures they have a shared knowledge of what exists so if the participant said in the first mention use “??the boy is playing guitar’ (?? meaning pragmatically odd within the context) instead of the indefinite “a boy is playing guitar” it was still counted. If the definite article is used in first mention contexts it does not necessarily mean that it is incorrect as both the speaker and hearer can see the referent in the picture. There is a unique individual or object, which the speaker refers to so the definite article would satisfy the uniqueness presupposition condition. One could argue that if the Chinese speakers were producing substitution errors then it may be the result of fluctuation, as defined by Ionin et al. (2004). However, as the task is free production it is not possible to tell if the learner is fluctuating in article choice. 5.3.9 Group results of the elicited picture description task The five native controls supplied articles in Art+N contexts 100% of the time. The results of the 15 L2 learners are presented in Table 5.



Neal Snape

Table 5.  Suppliance of articles in Art+N contexts Chinese speakers (n=15)

Art+N

+def

-def

Supplied Bare Ns

406 4

202 13

%Supplied

98%

89%

The results in Table 5 show that the Mandarin Chinese L2 learners were able to supply articles in obligatory Art+N contexts. Overall, there are very few omission errors. Examples of suppliance of the/a are shown in the following examples. (14)

a. b. c. d. e. f.

put it in an envelope look at the receipe she went to a musuem she’s driving a car the doctor is coming a man was singing

Most occurrences of the definite and indefinite article are in object position (14 a–d) due to the task design. Subsequently, there are very few occurrences in subject position (14 e–f). Examples of omission of the/a are in (15). (15)

a. b. c. d. e. f.

in front of Ø computer send Ø letter is better he walked by Ø restaurant what’s that Ø gorilla outside of Ø store I think it’s Ø monkey

The Mandarin Chinese speakers are significantly more accurate with the suppliance of definites than indefinites (χ2=13.7, p <.001). The significant difference in the suppliance of definites and indefinites is not predicted to happen under the RDH or PTH as omission or deletion is expected in definite and indefinite contexts. We do not offer an explanation here for why indefinites are omitted more than definites.



L2 article use 

There were few substitutions of the definite article by the demonstrative this or of the indefinite article by the numeral one, as found in Robertson’s (2000) and Goad and White’s (2008) studies. Other contexts where an article could appear were coded to see whether the L2 learners produced definiteness effect (DE) violations. Very few existential there constructions were found in the data. In total, there were 11 uses of the existential there construction in Art+N contexts and the article was always supplied. (16)

a. b. c. d.

there is a boat in her book there is a problem there is a man there’s an accident

There were no DE violations e.g., *there is the problem. Other studies by White (2003, 2008) have shown that L2 learners do not demonstrate DE violations. Lardiere (2004, 2006) found that Patty, a L1 Chinese-L2 English speaker, did not produce any such DE violations in her written or spoken production. 6. Discussion 6.1

Possible explanations for fluctuation

The results of article use in the current study by 38 Chinese L2 learners are consistent with the Fluctuation Hypothesis. As predicted, the intermediate L2 learners fluctuated between the two features [+definite] and [+specific], sometimes incorrectly selecting the as a marker for [+specific] and a as a marker of [-specific]. The RDH predicts that L2 learners who lack the interpretable features in their L1 grammars are able to acquire them in any subsequent language. Only uninterpretable features in the L2 are unacquirable due to age effects (Hawkins 2000, 2005). L1 Chinese L2 English learners in the current study are able to acquire interpretable features but have problems in selecting the correct feature for English i.e., [+definite] for articles, not [+specific] because there are no L1 transfer effects i.e., no articles in Chinese so L2 learners have to work out whether the L2 encodes definiteness or specificity for articles. Because Mandarin Chinese is still a language which lacks articles i.e., equivalent semantic functions to article languages, there is no direct L1 transfer as there is from Spanish to English.5

5. Spanish is a language with the articles un/una/unos/unas (a) and el/la/los/las (the) which have similar semantic functions to the articles in English.



Neal Snape

Ionin et al. (2004) argue that L2 learners from article-less languages continue to fluctuate even at advanced levels of English because they “need to evaluate the discourse situation and decide whether the is marking the presupposition of uniqueness (from the hearer’s perspective) or the existence of a noteworthy property (from the speaker’s perspective)” (2004: 51). Maybe it is possible for learners to evaluate the discourse situation and improve in accuracy in article use, but clearly some learners are more accurate than others (see Table 4 above in Section 5.4.5) and continue not only to fluctuate but misuse articles in all contexts. Hawkins et al. (2006) provide a feature-based account as to why individual learners continue to have problems with articles in different contexts. Though a feature-based account may account for some of the errors it does not explain why certain L2 learners end up with very different interlanguage grammars given that they have received as much input as the learners in the current study. Hawkins et al. (2006) argue that further exposure to the target language may lead learners to determine that English is a language that selects [+definite] for articles. As the learners have not yet reached an end-state grammar, restructuring of the interlanguage grammar may take place, or for some learners progress may be limited due to fossilization in their interlanguage grammars. Therefore, it is possible that the 11 learners in the current study do not fit the predicted patterns of article use as they are at a less advanced stage in their interlanguage grammars where article acquisition is concerned. Previous findings from Ting (2005), reported in Snape et al. (2006), who compared L1 Mandarin Chinese and L1 Spanish L2 English learners are comparable with the findings from the current study as both sets of Chinese participants are within the same age range and their length of stay in an English speaking country is around the same. One reason why the four L1 Chinese participants in Ting’s (2005) study performed much better on the forced choice elicitation task may be due to further development in their interlanguage grammars, but it is not clear at what stage the target input triggers the correct feature selection as even advanced L2 learners continue to fluctuate (see Ionin et al’s 2004 findings). Further research is currently being carried out by Ionin, Zubizarreta, Maldonado and Philippov (2007) and Snape (2008b) to find out whether exposure type (English as a Foreign Language and English as a Second Language) has an effect on reaching target-like achievement in article use. 6.2

The interpretable and uninterpretable features of DP

Hawkins et al. (2005) argue in Minimalist terms that all count nouns in English carry an uninterpretable number feature which need to be checked, valued and deleted before they can go off to Spell-Out. The RDH proposes that L2 learners who lack the uninterpretable features in their L1 grammars are unable to acquire



L2 article use 

them in any subsequent language due to age effects. The findings in our study are consistent with the RDH as interpretable features are acquirable.6 The difficulty is selecting the correct interpretable feature; definiteness for English articles, which is predicted by the FH. However, the Chinese L2 learners’ omissions of articles were relatively low in both tasks. This is unexpected given that Chinese has the Nominal Mapping Parameter setting [+arg], according to Chierchia (1998a), where all nouns in Chinese are ‘kind’ denoting. Mandarin Chinese lacks count nouns, so there is no uninterpretable number feature in the L1 grammar. The implication, based on the findings from our study, is that the Chinese L2 learners’ interlanguage grammars do not suffer from a syntactic deficit whereby the uninterpretable number feature cannot be acquired. On the contrary, it could be argued that the learners have full access to interpretable and uninterpretable features and it is expected that they will become more target-like as their proficiency progresses.7 6.3

Prosodic representation of articles

Goad and White (2006, 2008) argue that Mandarin Chinese speakers may have L1 prosodic structures available to them in order to represent articles in L2 English. It depends on whether nage (that) has been grammaticalized in the speaker’s dialect. We argued that because our participants, like Goad and White’s (2008) participants, originate from mainland China it would be likely that learners would have L1 representations available in their grammars. Our findings are consistent with Goad and White’s (2008) findings as there was very little deletion of articles in Art+N contexts. However, the data needs to be analysed further to see if the articles supplied are non target-like e.g., stressed. A detailed analysis using Praat (Boersma and Weenink 2006) software will be carried out and reported in a future paper.

6. An anonymous reviewer points out that learners are able to supply articles in obligatory contexts as a result of overt instruction rather than the presence of features at a deeper level in the grammar. But, Robertson (2000), Goad and White (2006) and Lardiere (2004, 2006) found that Chinese L2 learners omit articles in oral production. Other studies of L2 learners (Trenkic 2000, Hawkins et al. 2005) have found that omission of articles is high. Therefore, omission rates are not necessarily due to a lack of overt instruction. In fact, the learners in the aforementioned studies had all received overt instruction over the course of many years, so overt instruction alone does not automatically lead to target-like suppliance of articles. 7.

For discussion of parameter resetting in L2 acquisition see Snape (2008a).



Neal Snape

7. Conclusion In conclusion, the current study contributes to the existing literature on L2 article acquisition having tested three recent hypotheses. The findings from 38 L1 Chinese L2 English learners are consistent with the Fluctuation Hypothesis as proposed by Ionin et al. (2004). As a group, the Chinese L2 learners fluctuated between the two features [+definite] and [+specific]. The Chinese L2 learners overused the in [–definite, +specific] contexts and overused a in [+definite, –specific] contexts. The findings are not fully consistent with the Representational Deficit Hypothesis, but are consistent with the weaker interpretation of the Prosodic Transfer Hypothesis as target-like articles were supplied in obligatory contexts. This implies that L2 learners have full access to the L2 syntax and phonology which is consistent with the Full Transfer/Full Access model proposed by Schwartz and Sprouse (1996). Acknowledgements I wish to thank a number of people who helped make the current study possible; the Language Research Centre at the University of Calgary for acquiring funds from the Province of Alberta’s ACCESS grant in second language learning, the Ethics Committees at the University of Calgary and the University of Toronto, Dr. Bruce Clark, Dr. Anuradha Sengupta and everyone involved in the EAP program at the University of Calgary, Prof. David Barner and Katherine Chow at the University of Toronto and all the participants. I would also like to thank two reviewers and Lydia White for their constructive comments and suggestions. Any shortcomings are mine alone. References Abney, S. P. 1987. The English noun phrase in its sentential aspect. PhD dissertation, MIT. Barner, D. & Bale, A. 2002. No nouns, no verbs: Psycholinguistic arguments in favor of lexical underspecification. Lingua 112: 771–791. Bickerton, D. 1981. Roots of language. Ann Arbor MI: Karoma Press. Boersma, P. & Weenink, D. 2006. Praat: doing phonetics by computer (Version 4.5.05) [Computer program]. Retrieved November 5, 2006. (From http://www.praat.org/). Chen, P. 2003. Indefinite determiner introducing definite referent: A special use of ‘yi ‘one’ + classifier’ in Chinese. Lingua 113: 1169–1184. Chen, P. 2004. Identifiability and definiteness in Chinese. Linguistics 42(6): 1129–84.



L2 article use  Chierchia, G. 1998a. Reference to kinds across languages. Natural Language Semantics 6: 339–405. Chierchia, G. 1998b. Plurality of mass nouns and the notion of ‘semantic parameter’. In Events and Grammar 70, S. Rothstein (ed.), 53–103. Dordrecht: Kluwer. Chomsky, N. 1998. Minimalist inquiries: The framework. MIT Occasional Papers in Linguistics, 15, (also published in R. Martin, D. Michaels & J Uriagereka (eds.), Step by step: Essays on minimalism in honor of Howard Lasnik, The MIT Press, Cambridge MA, 89–155). Fodor, J. & Sag, I. 1982. Referential and quantificational indefinites. Linguistics and Philosophy 5: 355–398. García Mayo, M. del P. 2007. Article choice in L2 English by Spanish speakers: Evidence for Full Transfer. Paper presented at the 17th European Second Language Acquisition EUROSLA Conference, Newcastle, UK. Goad, H. & White, L. 2004. Ultimate attainment of L2 inflections: Effects of L1 prosodic structure. In EUROSLA Yearbook 4, S. Foster-Cohen, Ota, M., Sharwood Smith, M. A. & Sorace, A. (eds.), 119–45. Amsterdam: John Benjamins. Goad, H. & White, L. 2006. Prosodic transfer: L1 effects on the production of L2 determiners. In In Proceedings of the 30th Annual Boston University Conference on Language Development, D. Bamman., T. Magnitskaia & C. Zaller (eds.), 213–224. Somerville MA: Cascadilla Press. Goad, H. & White, L. 2007. Prosodic transfer and determiners in Turkish-English interlanguage. In In Proceedings of the 31st Annual Boston University Conference on Language Development, H. Caunt-Nulton, S. Kulatilake & I.-h. Woo (eds.), 230–238. Somerville MA: Cascadilla Press. Goad, H. & White, L. 2008. Prosodic structure and the representation of L2 functional morphology: A nativist approach. Lingua 118: 577–594. Hawkins, J.A. 1978. Definiteness and indefiniteness: A study in reference and grammaticality prediction. London: Croom Helm. Hawkins, R. 2000. Persistent selective fossilisation in second language acquisition and the optimal design of the language faculty. Essex Research Reports in Linguistics 34: 75–90. Hawkins, R. 2005. Explaining full and partial success in the acquisition of second language grammatical properties. Second Language 4: 7–25. Hawkins, R. & Chan, Y. 1997. The partial availability of UG in SLA: The failed functional features hypothesis. Second Language Research 13(3): 187–226. Hawkins, R., Al-Eid, S., Almahboob, I., Athanasopoulos, P., Chaengchenkit, R., Hu, J., Rezai, M., Jaensch, C., Jeon, Y., Jiang, A., Leung, I., Matsunaga, K., Ortega, M., Sarko, G., Snape, N. & Velasco-Zarate, K. 2005. Non-target-like article use in English: Implications for current UG-based theories of SLA. Ms, University of Essex. Hawkins, R., Al-Eid, S., Almahboob, I., Athanasopoulos, P., Chaengchenkit, R., Hu, J., Rezai, M., Jaensch, C., Jeon, Y., Jiang, A., Leung, I., Matsunaga, K., Ortega, M., Sarko, G., Snape, N. & Velasco-Zarate, K. 2006. Accounting for English article interpretation by L2 speakers. In EUROSLA Yearbook 6, S. H. Foster-Cohen, M. M. Krajnovic & J. M. Djigunović (eds.), 7–25. Amsterdam: John Benjamins. Heim, I. 1991. Articles and definiteness. In Handbook of Semantics. an international handbook of contemporary research, A. Stechow & D. Wunderlich (eds.), 487–535. Berlin: Walter de Gruyter. Huang, S.1999. The emergence of a grammatical category definite article in spoken Chinese. Journal of Pragmatics 31: 77–94.



Neal Snape Huebner, T. 1985. System and variability in interlanguage syntax. Language Learning 35: 141–163. Ionin, T. 2006. This is definitely specific: Specificity and definiteness in article systems. Natural Language Semantics 14: 175–234. Ionin, T., Ko, H. & Wexler, K. 2004. Article semantics in L2-acquisition: The role of specificity. Language Acquisition 12(1): 3–69. Ionin, T., Zubizarreta, M. L., Maldonado, S. B. & Philippov, V. 2007. Exposure type doesn’t matter: Similar patterns of English article use among ESL and EFL learners. A paper presented at SLRF 2007. UIUC. Ionin, T., Ko, H. & Wexler, K. 2008a. The role of semantic features in the acquisition of English articles by Russian and Korean speakers. In The role of formal features in second language acquisition, J. M. Liceras, M, H. Zobl & H. Goodluck, (eds.), Mahwaw NJ: Lawrence Erlbaum Associates. Ionin, T., Zubizarreta, M. L. & Maldonado, S. B. 2008b. Sources of linguistic knowledge in the second language acquisition of English articles. Lingua 118: 554–76. Lardiere, D. 2004. Knowledge of definiteness despite variable article omission in second language acquisition. In Proceedings of the 28th Boston University Conference on Language Development, A. Brugos, L. Micciulla & C. E. Smith (eds.), 328–339. Somerville MA. Cascadilla Press. Lardiere, D. 2006. Ultimate attainment in second language acquisition: A case study. Mahwah NJ: Lawrence Erlbaum Associates. Liu, D. & Gleason, J. L. 2002. Acquisition of the article the by nonnative speakers of English: An analysis of four nongeneric uses. Studies in Second Language Acquisition 24: 1–26. Lyons, C. 1999. Definiteness. Cambridge: CUP. Maclaran, R. 1982. The semantics and pragmatics of the English demonstratives. PhD dissertation, Cornell University, Ithaca NY. Mosel, U. & Hovdhaugen, E. 1992. Samoan reference grammar. Oslo: Scandinavian University Press. Murphy, S. 1997. Knowledge and production of English articles by advanced second language learners. PhD dissertation, University of Texas, Austin. Parrish, B. 1987. A new look at methodologies in the study of article acquisition for learners of ESL. Language Learning 37: 361–383. Partee, B. H. 2006. A note on Mandarin possessives, demonstratives, and definiteness. In Drawing the boundaries of meaning: Neo-Gricean studies in pragmatics and semantics in honor of Laurence R. Horn, G. Ward & B. Birner (eds.). Amsterdam: John Benjamins. Prince, E. 1981. On the inferencing of indefinite-this NPs. In Elements Of Discourse Understanding, A. K. Joshi, B. L. Webber and I. A. Sag (eds.), 231–50. Cambridge: CUP. Radford, A. 2000. Children in search of perfection: Towards a minimalist model of language acquisition. Essex Research Reports in Linguistics 34: 57–74. Radford, A., Kupisch, T., Köppe, R. & Azzaro, G. 2007. Concord, convergence and accommodation in bilingual children. Bilingualism: Language and Cognition 10(3): 239–256. Reid, J., Battaglia, P., Schuldt, M., Narita, E., Mochizuki M. & Snape N. 2006. The article choice of learners of English as a second language. Ms, University of Essex. Robertson, D. 2000. Variability in the use of the English article system by Chinese learners of English. Second Language Research 16(2): 135–172. Samarin, W. 1967. A grammar of Sango. The Hague: Mouton.



L2 article use Schwartz, B. D. & Sprouse, R. A. 1996. L2 cognitive states and the Full Transfer/Full Access model. Second Language Research 12(1): 40–72. Selkirk, E. 1996. The prosodic structure of function words. In Signal to Syntax: Bootstrapping from speech to grammar in early acquisition, J.L. Morgan & K. Demuth (eds.), 187–213. Mahwah NJ: Lawrence Erlbaum Associates. Snape, N. 2006. L2 acquisition of definiteness and specificity in English by advanced Japanese and Spanish learners. In Language Acquisition and Development. Proceedings of the Generative Approaches to Language Acquisition Conference 2005, 500–4. Cambridge: Cambridge Scholars Press. Snape, N. 2008a. Resetting the nominal mapping parameter: Definite article use and the count – mass distinction in L2 English. Bilingualism: Language and Cognition 11(1). 63–79. Snape, N. 2008b. Fluctuation in L2 article acquisition: Interpreting the findings from L1 Chinese L2 English participants. A paper to be presented at PacSLRF 2008. FLTRP International Convention Centre. Snape, N., Leung, Y.-k. I. & Ting, H. C. 2006. Comparing Chinese, Japanese and Spanish speakers in L2 English article acquisition: Evidence against the fluctuation hypothesis? In Proceedings of the 8th Generative Approaches to Second Language Acquisition Conference (GASLA 2006), M. Grantham O’Brien, C. Shea & J. Archibald (eds.), 132–9. Somerville MA: Cascadilla Proceedings Project. Syndicate, U. C. L. E. 2001. Quick placement test. Oxford: OUP. Thomas, M. 1989. The acquisition of English articles by first- and second-language learners. Applied Psycholinguistics 10: 335–355. Ting, H. C. 2005. The acquisition of articles in L2 English by L1 Chinese and L1 Spanish speakers. MA thesis, University of Essex. Trademan, J. E. 2002. The acquisition of the English article system by native speakers of Spanish and Japanese: A cross-linguistic comparison. Albuquerque NM: University of New Mexico. Trenkic, D. 2000. The acquisition of English articles by Serbian speakers. PhD dissertation, University of Cambridge. Trenkic, D. 2008. The representation of English articles in second language grammars: Determiners or adjectives? Bilingualism: Language and Cognition 11(1): 1–18. Tsimpli, I.-M. & Roussou, A. 1991. Parameter-resetting in L2? UCL Working Papers in Linguistics 3: 149–170. Wakabayashi, S. 1997. The acquisition of functional categories by learners of English. PhD dissertation, University of Cambridge. White, L. 2003. Fossilization in steady state L2 grammars: Implications of persistent problems with inflectional morphology. Bilingualism: Language and Cognition 6(2): 129–141. White, L. 2008. Different? Yes. Fundamentally? No. Definiteness effects in the L2 English of Mandarin speakers. In Proceedings of the 9th Generative Approaches to Second Language Acquisition Conference (GASLA 2007), R. Slabakova, J. Rothman, P. Kempchinsky & E. Gavruseva, 251–261. Somerville MA: Cascadilla Proceedings Project.



Successful features Verb raising and adverbs in L2 acquisition under an Organic Grammar approach Anne Vainikka and Martha Young-Scholten Johns Hopkins University and Newcastle University

Under Organic Grammar (Vainikka and Young-Scholten 2005; 2007), the building of L2 syntactic structure is constrained by UG in interaction with primary linguistic data. On the basis of verb raising, we argue against the partial UG availability pursued in Hawkins and Chan (1997) in their Failed Functional Features Hypothesis, whose refinement by Hawkins and Hattori (2006) states that only those uninterpretable syntactic features present in the learner’s L1 remain accessible for L2 acquisition. After summarizing UG-driven L2 German verb raising, we consider French/English adverb placement data pointing to L1-driven verb raising (White, 1991a, b; Schwartz, 1996). However, absence of Frenchlike raising in questions and negation (White 1992; Eubank 1994) requires an alternative account. Based on Vainikka’s (2007) framework which dispenses with Cinque’s (1999) prolific projections, we provide an alternative analysis.

1. Introduction If Universal Grammar (UG) is fully available throughout the lifespan, why do we routinely observe the systematic failure of post-puberty learners to attain nativelike morpho-syntactic competence in a second language? Unlike the Fundamental Difference Hypothesis (Bley-Vroman 1990; Clahsen and Muysken 1986), hypotheses such as Minimal Trees/Structure Building (Vainikka and Young-Scholten 1994) and Full Transfer/Full Access (Schwartz and Sprouse 1996) assume complete and direct access to UG by younger and older learners alike. However, Hawkins and Chan (1997) have contested the view that UG is available in toto. Unlike some proponents of the Fundamental Difference Hypothesis, for example Clahsen and Muysken (1986), Hawkins and Chan take the position that second language grammars are UG-constrained, while nonetheless arguing that UG is only partially available to adults. For the latter, partial availability revolves around



Anne Vainikka and Martha Young-Scholten

the idea that after the end of the critical period, certain subparts of UG start to become less accessible or are completely inaccessible. The hypothesis Hawkins and Chan put forward, Failed Functional Features, refers to the idea of a Representational Deficit (Hawkins 2005), which has since been further refined as a hypothesis that rules out the learner’s access to uninterpretable, syntactic features in the L2 which are not present in the L1 (e.g., Hawkins and Hattori 2006; Tsimpli and Dimitrakopoulou 2007). Unlike Full Transfer/Full Access proponents such as Schwartz and Sprouse (1996), Hawkins (2001) pursues the idea that the L2 learner gradually builds up syntactic structure from an initial point at which there are only lexical projections. That is, the learner’s initial state in L2 acquisition is not equivalent to the learner’s entire knowledge of the L1. For Hawkins, during development, syntactic projections can then be modulated by the learner’s L1. Under the Failed Functional Features Hypothesis modulation entails the L2 learner only being able only to posit those functional projections that draw on features existing in the learner’s L1. We return to this point below. 2. Verb raising and the minimal trees and structure building of Organic Grammar We argue that not all functional features are equal; consider verb raising. In a new paper that draws on comprehension data from Arabic, Chinese, French, German, Japanese and Spanish speakers of English Hawkins et al. (2008) conclude that the systematic differences found between the main verb raising in L1 group (Arabic, French, German and Spanish speakers) and the non-raising in L1 group (Chinese and Japanese) point to adult L2 learners’ failure to access uninterpretable features not represented in their L1s. Hawkins et al. assume verb raising involves narrow syntax and they adopt Adger’s (2003) Minimalist account under which the agree operation involves the interaction between the interpretable features present, past and progressive and the uninterpretable feature INFL. Like Hawkins et al. we assume verb raising is syntactic. But we part ways in terms of evidence for application of non-L1 uninterpretable features. There is ample evidence of the successful acquisition of verb raising, where data from studies of learners’ production have shown that, regardless of the learner’s age upon L2 exposure, non-L1-based verb raising is acquired (e.g., Vainikka and Young-Scholten 1994; 2002). Data from adult English, Korean, and Turkish learners of German suggest that the Failed Functional Features/Representational Deficit hypotheses are too strong. In the three L1s in question, thematic verb raising either does not exist (English) or all lexical and functional projections are head final (Korean and



Organic Grammar and verb raising 

Turkish). These data, as well as the data from speakers of Italian and Spanish in whose L1s thematic verb raising does occur, point to a non-L1-based, common pattern of acquisition for all L2 German learners. In one particularly crucial respect, the pattern is similar to that of children learning German as their first language: the first functional projection is of an underspecified head-initial functional projection, eventually followed by the adult/target head-final German AgrP. Under what is now referred to as Organic Grammar (Vainikka & YoungScholten 2005; 2007),1 all language learners at the earliest stages of acquisition project only a Minimal Tree, i.e., only lexical categories such as VP. Learners at the initial state rely exclusively on such a projection. For beginning L2 learners, these lexical projections are L1-based as shown in (1) where a Korean learner produces a head-final Korean VP, and in (2) where an English learner produces a head-initial English VP (data from Vainikka & Young-Scholten 1994; 2007). (1)

Eine Katze Fisch alle essen. a cat fish all eat-INF [Eine Katze hat den ganzen Fisch gefressen.] ‘A cat ate the whole fish.’

(2)

Peter lernen die Buch. Peter learn-INF the book. [Peter liest das Buch.] ‘Peter reads the book.’

(Changsu/Korean L1)

(Paul/English L1, month 2)

If and when the L2 learner receives sufficient target language input, the next step in his/her syntax is a target-language-based Minimal Tree, as shown in (3), where Salvatore has switched his previously head-initial VP to a head-final one (see Vainikka and Young-Scholten 1996b). (3)

Vielleicht Schule essen. maybe school eat-INF [Vielleicht isst sie in der Schule.] ‘Maybe she eats at school.’

(Salvatore/Italian L1, Session 6)

At the early stages, as exemplified by the data in (1) – (3), functional projections are absent, but the interaction of primary linguistic data with UG enables the learner to progress further to build functional structure that begins to resemble 1. Introduction of the term Organic Grammar stems from the need to arrest the terminological confusion about Structure Building, and to encompass both the Structure Building and the Economy (under Minimalism) aspects of this approach. The new term is also a response to variable and often incorrect use of terms such as “Minimal Trees/Structure Building” and “Weak Continuity Hypothesis” applied to our approach.



Anne Vainikka and Martha Young-Scholten

that of the target language. For German this involves raising the thematic verb to second position in declaratives, and to final position (due to a head-final AgrP) in embedded clauses. Learners in an input-rich environment from the start of initial exposure reach this final stage, as in the case of the American secondary school exchange student, Joan, who like the many of the immigrants studied, had had no prior exposure to German upon arrival in that country and received no regular instruction in German once there (see Vainikka and Young-Scholten 2002). Apart from a case error with im in her twelfth month in Germany Joan’s morpho-syntax is completely German. (4)

Und ich habe mit Markus geredet, so weil Kosta schon ins Bett war. and I have with M. spoken, so because K. already into bed was [Und ich habe mit Markus geredet, weil Kosta schon im Bett war.] ‘And I spoke with Markus because Kosta was already in bed.’

Unlike under Hawkins’ (2001) Modulated Structure Building, within an Organic Grammar framework, there is no recourse by the learner to the L1 during the acquisition of functional syntax. With structure building constrained only by UG, no differences for a given L2 are expected for learners from languages which differ in terms of functional features. In the L1 English/L2 German data one might, for example, expect to see evidence that the value of auxiliary ‘be’ or ‘have’ transfers from English to German, since equivalent auxiliaries (sein and haben) with similar functions exist. Thematic verb raising, since it does not occur in English, should proceed more slowly than for, say, Spanish and Italian learners of German. There is no evidence of either sort of influence in the data, and indeed no other differences have been found across native languages in the adult L2 German data. At the stage of development immediately following the bare VP, learners of German – the L1 children as well as the L2 adults we have studied – project an underspecified head-initial functional projection to which they optionally raise verbs. This is shown in (5), where the Turkish native speaker Sevinc adds to his interlanguage German new syntactic structure which is not present in strictly head-final Turkish (see Vainikka and Young-Scholten 1994). (5)

Jetzt brau Wohnungsamt fragen. now need-0/1SG housing.authority ask-INF [Jetzt brauche ich das Wohnungsamt fragen.] ‘Now (I) need to ask (the) housing authority’

(Sevinc/Turkish L1)

At the earliest stages of the acquisition of German, all learners produce non-finite verb forms when the verb has not raised, pointing to a common Optional Infinitive Stage (Wexler 1994) in German at which Root Infinitives dominate, i.e., syntactic structure is truncated (Rizzi (1993/4). Joan’s production of thematic verbs in



Organic Grammar and verb raising 

default (-n) form is typical of the prevalence of such forms in the early L2 German data. In the first two samples (her second and third month in Germany) when she has only projected her head-initial English–based bare VP, of the 47 thematic verbs she produces, 25 end in –n and a further 11 in –e or no suffix and are incorrect with respect to the intended person and number. (6)

Ich trinken Tee immer morgen. I drink tea always morning [Ich trinke morgens immer Tee.] ‘I always drink tea in the morning.’

(Joan/L1 English)

Adult L2 learners differ, however, from child L1 learners, and appear to differ as well from child L2 learners, in terms of their production of default forms in raised positions of the sort shown in (7) and (8). While learners such as Joan may hardly evidence default forms in raised positions due to the rapidity of their development, for those second language learners in acquisition-poor environments, use of default forms persists in the context of stalled syntactic development. (7) Und dann nachher kommen die Sonne nochmal wieder. (Maria/L1 Spanish) and then afterwards come-INF the sun yet again [Und dann nachher kommt die Sonne nochmal wieder.] ‘And then afterwards the sun comes out again’ (8)

Ich sehen Schleier. I see-INF veil [Ich sehe den Schleier.] ‘I see the veil’

(Kemal/L1 Turkish)

The default form data point to a deficit not captured by Hawkins and Chan’s (1997) FFFH. The deficit responsible might relate to surface morphology, as discussed in Prévost and White (2000a, b, c). Alternatively, the deficit might derive from the prosodic constraints of the learner’s L1 phonology, as pursued in Goad, White and Steele (2003). And the deficit could well be a combined one (see e.g., Bayley 1996). For now, however, we leave further consideration of the source of such non-target morphological patterns aside (but see also Vainikka and Young-Scholten (1998; 2007) on the interface of phonology, morphology and syntax under Organic Grammar).



Anne Vainikka and Martha Young-Scholten

3. Verb raising across languages: the case of French learners of English On the basis of the German data, we conclude that L1 verb raising and what verb raising entails does not transfer in L2 acquisition. However, there is a set of problematic data that points to L1 influence. Dating back to White (1991a/b), it has been argued that the position of finite thematic verbs in data from French learners of English reveals the systematic allowance of non-target utterances such as in (9), which indicate erroneous, L1-based raising of thematic verbs, vs. target-languageconforming auxiliary placement, as in (10). In other words, French verb raising transfers (see Hawkins, Towell and Bazergui’s (1993) analysis of this as a problem with Agr in the L2, based on systematic differences between English learners of French and French learners of English.)

(9) *John reads frequently linguistics books.

(10) John is frequently reading linguistics books when the library closes. In French, but not in English, the thematic verb is raised to a functional head (see e.g., Emonds 1978; Pollock 1989). Thus the errors found in learners’ L2 English could be explained by assuming that the French speakers have transferred from their L1 the information that verbs raise, contrary to the English input. If only a VP were transferred from French to English, as claimed under Organic Grammar, how could information about raising to a functional projection be transferred? The French data sit well under an approach such as Schwartz and Sprouse’s (1996) Full Transfer/Full Access Hypothesis. They also support the Failed Functional Features Hypothesis; an account for these data involves precisely the sort of uninterpretable features that guide L2 acquisition: those that are L1-based. The data under consideration here seem to constitute a problem for Organic Grammar, which ascribes neither failure nor success to L1 influence. Yet if we look a bit more closely at previous analyses, we find that the basis on which transfer of verb raising is argued to occur is shaky. While White’s (1991a, b) and Schwartz’s (1996) analyses of the data are based purely on the adverb context shown in (9), the full L2 English data reveal that French-like verb raising does not occur in questions or in negation, as pointed out in White (1992).2 Other, longitudinal data indicate there is no stage for French learners of English where French-type verb raising occurs, as detailed in Eubank (1994) in reference to studies by Gerbault 2. Interestingly, the reverse situation does not seem to hold, if L1 verb raising facts are transferred. The English speakers acquiring French studied by Grondin and White (1993) raised main verbs to INFL, prompting the authors to conclude that they used “the functional projections in their grammars in ways that are appropriate to the L2, rather than in ways that are appropriate to the L1.” (1993:143).



Organic Grammar and verb raising 

(1978) and Tiphine (1983). And finally, Ayoun’s (2005) study indicates that adult French-speaking learners of English cease misplacing adverbs at more advanced stages of acquisition. 4. Adverbs and syntactic projections How, then, do we explain the placement of adverbs in French-English interlanguage? These interlanguage data can be demystified by adopting the three assumptions shown in (11): (11) a. Adverbs occur in a number of specifier positions (rather than adjoined) in various locations in the syntactic tree. b. Adverbs are a lexical category. L1-based patterns in the interlanguage data are evidence for lexical transfer, in line with Organic Grammar. c. Functional projections need to be filled by syntactic material. Absence of material in the head position may result in verb raising. These assumptions point to the appearance of verb raising in declaratives and explain its non-occurrence with negation and in questions. Let us expand on these three points by looking at previous accounts of adverb position. In his drastic departure from the traditional analysis of adverbs, Cinque (1999) argues that rather than being adjoined to the VP as under traditional approaches, adverbs (in Italian) occur in a number of specifier positions in various locations of the syntactic tree. Under Cinque’s approach, adverbs are no longer strictly speaking adjuncts but occupy non-adjoined positions. The impetus for this analysis comes from Kayne’s (1994) proposal to restrict adjunction in general. Alexiadou (1997) and Laenzlinger (1998) also apply Kayne’s idea to adverbs (in Greek and French), suggesting that adverbs occupy specifier positions. In order to explain certain word order and scope patterns, Cinque proposes that each adverb is basegenerated in a separate functional projection. However, there are major problems with his approach. First, there is often no other evidence for the projection other than the adverb. Second, such an analysis of adverbs requires an unconstrained approach to functional projections, and by the end of his 1999 book, Cinque has posited over 30 functional projections universally for straightforward declarative clauses alone. He then comments that even more projections are likely to emerge in the theory as more data are considered. While theorizing is possible within the confines of formal linguistics, at some point the proposed structure must be accountable to acquisition.

 Anne Vainikka and Martha Young-Scholten

The child would start off with a great Ur-baobab tree which covers all possibilities in all languages, and this is how the tree would remain (under the Strong Continuity view this entails; see e.g., Boser, Lust, Santelmann and Whitman, 1992; Hyams, 1992; Poeppel and Wexler, 1993; Weissenborn, 1990). We argue that such a story is untenable. If every child is born with myriad projections, evidence must exist in the primary linguistic data for the realization only of subsets of these adverbial projections in a given language, i.e., for the establishment of a specific grammar. It can be argued that, given what we know about the general uniformity coupled with parsimony with which children undertake the task of language learning, children do not operate with such congenitally dense trees. 4.1

A reanalysis of adverbs and a reanalysis of the L2 French data

In the first application of Organic Grammar to syntax, Vainikka (2007) presents a new analysis of adverbs, using English as a starting point. This approach is related to Cinque’s, to Alexiadou’s and to Laenzlinger’s in that adjunction is curtailed, with adverbs instead occupying specifier positions. Vainikka’s analysis goes the required step further in its adoption of an Organic-Grammar approach to functional projections based on acquisition, resulting in roughly 10 functional projections in the most complex sentences. Sentences and languages vary in exactly which projections are posited, and under this approach, adverbs typically occur either in the VP, or are raised to specifier positions of (existing) functional projections. Importantly, the strong association between the semantics of a particular adverb and its projection that Cinque posits is no longer necessary and not maintained in Vainikka’s approach. Schwartz (1996) claims the French-English adverb data provide compelling evidence against the structure building component of Organic Grammar (Schwartz 1996), but there are important ramifications of an Organic Syntax-based adverb analysis for these classical L2 acquisition data. Schwartz presents the following putative counterargument to structure building. The Structure Building Hypothesis predicts that verb raising (or any other type of raising involving a functional projection) would not be transferred under L2 acquisition, since the hypothesis entails that functional projections are not transferred, but projected solely through the interaction of UG with the primary linguistic data. Under Schwartz and Sprouse’s (1996) Full Transfer/Full Access Hypothesis, functional syntax is transferred along with verb raising during L2 acquisition; evidence of this is the position of adverbs in the French-English data. The flipside of the Failed Functional Features coin similarly predicts transfer; those features present in the learner’s L1 are not only accessible but also accessed during L2 acquisition. The former (accessible) implies success, but of course when the L1 and L2 differ, as in the case of



Organic Grammar and verb raising 

French and English the latter (accessed) results not in success, but also in failure, i.e., in non-target like structure. Arguing on the basis of the data in White’s (1991a, b; 1992) study, Schwartz points out that French learners of English tend to allow sentences in which the thematic verb appears to have been raised, as in the L1 French (but not in English, see e.g., Emonds 1978, Pollock 1989). White’s data show that verb raising sometimes occurs in the learners’ L2 English when the sentence contains an adverb such as frequently; that is, the finite verb precedes the adverb as in the L1 French (but not in the target language, as in example (9) repeated here as (12).

(12) *John reads frequently linguistics books.

In allowing such constructions, it appears that the French-speaking L2 English learners studied are raising the verb from the VP to a position higher in the tree, preceding the adverb. But as pointed out above, there are problems with this analysis. The obvious solution to explain these data – and, as also noted above, the one invariably entertained – is that verb raising transfers from the L1 to the L2. Yet verb raising appears to occur only in the data in the adverb context, contrary to native French. Indeed White (1992) shows that French-based verb raising does not occur in questions or negation in French-English interlanguage, as would be expected if French-type verb raising were operative.3 Here White (1992) admits that apparent ‘verb raising’ in the adverb data could be due to a different process, if Iatridou (1991) is correct in claiming that adverb placement differs across languages. Bolstering the position that the accepted analysis is actually incorrect, we refer to Eubank (1994), who points out that there is no stage with French-type verb raising in the longitudinal L2 English (L1 French) data from Gerbault (1978) and Tiphine (1983).4 Evidence of transfer of verb raising thus rests solely on adverb position, and it is important to consider whether the new developments in adverb analysis described above undermine the original argument. 4.2

Reassessing the verb raising argument: Applying Organic Syntax

While Cinque wishes to argue that adverb position is universally uniform, if we take an Organic Syntax approach, the immediate result is considerably more 3. Given the lack of verb raising in question or negation contexts, our argument against Schwartz (1996) in Vainikka and Young-Scholten (1996a) does not hold. There we suggested that based on the distribution of auxiliaries in the L2 English input, the learner might posit verb raising. However, if that were the case, we would expect to find verb raising also in question and negation contexts. 4. Note that Håkansson (2001) and Sayehli (2001) claim that verb raising does not transfer from L1 Swedish to L2 German, although both languages have verb raising.



Anne Vainikka and Martha Young-Scholten

flexibility. Under Organic Grammar, there is a language-specific Master Tree whereby the set of specific functional projections varies across languages. In Vainikka and Young-Scholten (2007), ten assumptions underlying this notion are introduced. The first seven assumptions define the backbone of syntactic structure, the Master Tree, while the last three deal with related issues. Central to these assumptions is the idea that there is no fixed universal CP tree that contains all possible functional projections. Cross-categorial generalizations can be stated in the grammar based on the Master Tree, as under Kayne (1994), but unlike Kayne, there is no assumption of common (initial) headedness across languages. The second assumption captures the essence of structure building. Here we draw on Baker’s (1985) Mirror Principle, where inflectional morphology mirrors syntax, in particular, functional projections. Cross-linguistic differences in projections and in their headedness (contra Kayne 1994) are thus allowed. We also follow those who limit functional projections to those which are warranted, i.e., Fukui and Sakai (2003), Giorgio and Pianesi (1997), Grimshaw (1997), and particularly Speas (2001), on Economy of Projection. Assumption 1: Each language has a Master Tree that includes all possible projections occurring in the language. Assumption 2: All and only those projections occur in the Master Tree for which there is evidence in the language. Assumption 3: Universal Grammar provides the tools for acquiring the Master Tree, based on input. Assumption 4: The Master Tree is acquired from the bottom up. Assumption 5: The Acquisition-Syntax Correspondence: syntax mirrors acquisition. Assumption 6: Actual instantiations of the tree are projected from the bottom up, based on the Master Tree. Assumption 7: Partial trees may be projected for constructions which do not involve the full Master Tree structure. Assumption 8: Lexical and functional projections differ in terms of how they are represented in the grammar. Assumption 9: Cross-categorial generalizations about structure are possible. Assumption 10: Only as much adjunction is posited as necessary. Assumptions 3, 4 and 5 are relevant to our discussion of acquisition in general. Rather than invoke any language-external influence (such as processing as in



Organic Grammar and verb raising 

Processability; see Vainikka and Young-Scholten 2005) on morpho-syntactic development, it is the syntax itself that drives acquisition. With respect to adverbs, these assumptions a priori rule out an approach such as Cinque’s. First, the idea of economy rules out the proliferation of scores of projections in a given language. Second, the prospect of language-specific syntax structures (i.e., Master Trees) obviates the need for including all syntactic possibilities. And finally, assumption 10 notes the undesirability of adding adjoined projections. In the spirit of Emonds’ (1985) Structure Preserving Hypothesis, the Master Tree excludes adjoined projections as these are costly for the grammar to posit. 4.3

Adverbs in Organic Syntax

Under Organic Syntax, adverbs do not occur in adjoined positions, but rather in various specifier positions, depending on the availability of each position. The approach to adverbs introduced here makes it completely plausible to suggest, following Iatridou’s (1991) original proposal, that there is cross-linguistic variation in terms of the specific adverbs that occur in each specifier position. However, it turns out that given Vainikka’s (2007) analysis, the adverb diagnostic for verb raising still holds for certain adverbs. The most reliable class of adverbs for this diagnostic are those adverbs that may raise out of the VP,5 but that cannot raise any higher than the Spec(TP) position. Sentential adverbs, in particular, may occupy positions high in the tree, and such adverbs then precede the verb regardless of whether it remains in the VP or is raised to a functional head such as Agr or T. Adverbs that can be used as a diagnostic are (under Vainikka 2007) Class II adverbs of degree and manner – such as completely or easily – as well as Class III adverbs of time or frequency such as frequently or often, as used in White’s study. Given the idea of a Master Tree, we can posit that the adverb frequently and its French counterpart occupy different syntactic positions. This is an option available under Organic Grammar. Since adverbs are a lexical category (being an open class derivable from the lexical class of adjectives), rather than a functional one, the adverb itself would carry information about its syntactic position, and this lexical information can transfer under an approach such as ours that allows the transfer of lexical, but not functional features. This sort of account would also explain why French learners of English do not raise verbs in their L2 English with negative or

5. Class I adverbs – short adverbs such as well and fast – cannot be raised out of the VP and therefore will not work as a diagnostic for verb raising since the verb would precede the adverb whether raised or not (see Vainikka 2007). While the analysis of the short adverbs is straightforward, it currently remains unclear why they cannot raise.



Anne Vainikka and Martha Young-Scholten

interrogative constructions. As we note above, this would be expected if Frenchtype verb raising were indeed transferred. Having suggested this as a possible scenario, we must rule it out. We do not think this is a fruitful approach; it is highly unlikely that two closely related languages such as English and French would differ in terms of regular adverb placement. Thus, despite the new developments in the theory of adverbs, the central problem therefore remains: what is the syntactic analysis of *John reads frequently linguistics books in French speakers’ L2 English? We think that the answer lies in a theory of movement currently under further development: the Full House Principle (FHP). 4.4

The Full House Principle

According to this principle, there is a tendency on the learner’s part to fill an empty syntactic position by moving an element to it. The earliest version of the idea that syntactic positions need to be filled – preferably with overt material – is from Vainikka and Young-Scholten (1994); see also Vainikka and Levy’s (1999) POOL, Principle of Obligatory Occupant Licensing, which is useful in accounting for the patterns of empty subjects in Finnish and Hebrew. The FHP was proposed to account for movement into a functional projection without syntactic features (the FP), evident in data from learners who in their acquisition had moved beyond their projection of a bare VP. Consider a typical situation during the L2 acquisition of English where a TP (or FP, as in previous publications by Vainikka and YoungScholten) has been projected, but an AgrP projection has not yet been posited. If the speaker – perhaps based on an overt element in the specifier position – somewhat prematurely posits an AgrP position without having yet acquired the full agreement paradigm, the grammar would attempt to fill that empty (i.e., devoid of features) Agr position by moving a verb from below into it, regardless of whether the L1 or the L2 have verb raising. That is, verb raising would be a readily available option in UG that would surface especially during acquisition prior to the acquisition of the appropriate morphology. Example (5), repeated here as (13), illustrates this, where the adverb jetzt ‘now’ fills an empty position rather than the required subject. (13)

Jetzt brau Wohnungsamt fragen. now need-0/1SG housing.authority ask-INF [Jetzt brauche ich das Wohnungsamt fragen.] ‘Now (I) need to ask (the) housing authority’

(Sevinc/Turkish L1)

Returning to English, once the agreement morphology has been acquired, the learner is in a position to also acquire the universally exceptional feature lowering of English (from Agr and T to the non-raised main verb). With respect to White’s data, an experimental situation (a grammaticality judgment task) such as hers may



Organic Grammar and verb raising 

be particularly likely to involve positing of projections that learners have not yet fully acquired, thus resulting in filling an empty higher position by verb movement. In negative sentences, negative morphology would occupy NegP, and verb raising would not be required to fill a head position, and similarly for question morphology and the CP projection. Thus, rather than providing evidence for transfer of verb raising, White’s data may provide evidence for a stage at which all functional projections have not yet been acquired. 5. Conclusion: Ramifications for the Failed Functional Features Hypothesis If, as under the Failed Functional Features Hypothesis, uninterpretable features not present in the learner’s L1 cannot be activated by adult second language learners, it will matter whether verb raising is always driven by such features (as assumed under Minimalism, Chomsky 1995), or whether there is a mechanism such as the Full House Principle, whereby an element may raise in order to fill a syntactic position. If verb raising during L2 acquisition is motivated by a lack of features/ phonological material in a functional projection, then – despite any possible nonacquisition of uninterpretable features – verb raising could nonetheless occur during the development of a second language. If, however, such features are required in order for verb raising to take place, the FFFH would predict that no non-L1type verb raising should be found during adult second language acquisition. White’s original data would then also present a problem for the FFFH, since it cannot be shown that the adverb examples in L2 English involve French-type verb raising, given the lack of such raising in negative sentences and questions. While we remain agnostic about a critical period for uninterpretable features which results in the impossibility of acquiring such features during adult L2 acquisition, our solution to the adverb dilemma allows (non-L1) verb raising to occur during L2 acquisition, even assuming the Failed Functional Features Hypothesis. References Adger, D. 2003. Core syntax. A minimalist approach. Oxford: OUP. Alexiadou, A. 1997. Adverb placement. A case study in antisymmetric syntax. Amsterdam: John Benjamins. Ayoun, D. 2005. Verb movement in the L2 acquisition of English by adult native speakers of French. In Eurosla Yearbook 5, S. H. Foster-Cohen, M. del Pilar Garcia-Mayo & J. Cenoz (eds), 35–76. Amsterdam: John Benjamins.



Anne Vainikka and Martha Young-Scholten Baker, M. 1985. Incorporation: A theory of grammatical function changing. PhD dissertation, MIT. Bayley, R. 1996. Competing constraints on the variation of the speech of adult Chinese learners of English. In Second language acquisition and linguistic variation, R. Bayley & D. R. Preston (eds), 97–120. Amsterdam: John Benjamins. Bley-Vroman, R. (1990) The logical problem of foreign language learning. Linguistic Analysis 20: 3–49. Boser, K., Lust, B., Santelmann, L. & Whitman, J. 1992. The syntax of CP and V-2 in early child German (ECG): The Strong Continuity Hypothesis. In Proceedings of NELS 23, K. Broderick (ed.), 51–65. Amherst MA: GLSA. Chomsky, N. 1995. The minimalist program. Cambridge MA: The MIT Press. Cinque, G. 1999. Adverbs and functional heads: A cross-linguistic perspective. Oxford: OUP. Clahsen, H. & Muysken, P. 1986. The availability of universal grammar to adult and child learners: A study of the acquisition of German word order. Second Language Research 2: 93–119. Emonds, J. 1978. The verbal complex V’-V in French. Linguistic Inquiry 9: 151–175. Emonds, J. 1985. A unified theory of syntactic categories. Dordrecht: Foris. Eubank, L. 1994. On the transfer of parametric values in L2 development. Language Acquisition 3: 183–208. Fukui, N. & Sakai, H. 2003. The visibility guideline for functional categories: Verb raising in Japanese and related issues. Lingua 113: 321–375. Gerbault, J. 1978. The acquisition of English by a five-year-old French speaker. MA thesis, UCLA. Giorgio, A. & Pianesi, F. 1997. Tense and aspect. From semantics to morphosyntax. Oxford: OUP. Goad, H., White, L. & Steele, J. 2003. Missing inflection in L2 acquisition. Defective syntax or L1-constrained prosodic representations? Canadian Journal of Linguistics 48: 243–263. Grimshaw, J. 1997. Projection heads and optimality. Linguistic Inquiry 28: 373–422. Grondin, N. & White, L. 1993. Functional categories in child L2 acquisition of French. McGill Working Papers in Linguistics, 121–145. Håkansson, G. 2001. Against full transfer: Evidence from Swedish learners of German. Lund University Department of Linguistics Working Papers 48: 67–86. Hawkins, R., Towell, R. & Bazergui, N. 1993. Universal Grammar and the acquisition of French verb movement by native speakers. Second Language Research 9: 189–233. Hawkins, R. 2001. Second language syntax: A generative introduction. Oxford: Blackwell. Hawkins, R. 2005. Revisiting wh-movement: The availability of an uninterpretable [wh] feature in interlanguage grammars. In Proceedings of the 27th Generative Approaches to Second Language Acquisition Conference, B. Beachley, A. Brown & F. Conlin (eds.), 124–137. Somerville MA: Cascadilla Press. Hawkins, R. & Chan, C. Y-H. 1997. The partial availability of Universal Grammar in second language acquisition. The Failed Functional Features Hypothesis. Second Language Research 13: 187–226. Hawkins, R. & Hattori, H. 2006. Interpretation of English multiple wh-questions by Japanese speakers: A missing uninterpretable feature account. Second Language Research 22(3): 269–301. Hawkins, R., Casillas, G., Hattori, H., Hawthorne, J., Husted, R., Lozano, C., Okamoto, A., Thomas, E. & Yamada, K. 2008. The semantic effects of verb raising and its consequences in



Organic Grammar and verb raising  second language grammars. In The role of features in second language acquisition, J. Liceras, H. Zobl & H. Goodluck (eds), 328–351. New York, NY: LEA.   Hyams, N. 1992. The genesis of clausal structure. In The acquisition of verb placement, J. M. Meisel (ed.), 371–400. Dordrecht: Kluwer. Iatridou, S. 1991. About AgrP. Linguistic Inquiry 21: 551–577. Kayne, R. 1994. The antisymmetry of syntax. Cambridge MA: The MIT Press. Laenzlinger, C. 1998. Comparative studies in word order variation. Amsterdam: John Benjamins. Poeppel, D. & Wexler, K. 1993. The full competence hypothesis of clause structure in early German. Language 69: 1–33. Pollock, J.-Y. 1989. Verb movement, Universal Grammar and the structure of IP. Linguistic Inquiry 20: 365–424. Prévost, P. & White, L. 2000a. Accounting for morphological variation in L2 acquisition: Truncation or missing inflection? In The Acquisition of Syntax, M.-A. Friedemann & L. Rizzi (eds.), 202–235. London: Longman. Prévost, P. & White, L. 2000b. Finiteness and variability in SLA: More evidence for missing surface inflection. In Proceedings of the 23rd annual Boston University Conference on Language Development, A. Greenhill, H. Littlefield & C. Tano (eds.), 439–450. Somerville MA: Cascadilla Press. Prévost, P. & White, L. 2000c. Missing surface inflection or impairment in second language acquisition? Evidence from tense and agreement. Second Language Research 16: 103–134. Rizzi, L. 1993/4. Some notes on linguistic theory and language development: The case of Root Infinitives. Language Acquisition 3: 371–393. Sayehli, S. 2001. Transfer and Syntax. A study on the acquisition of German word order by Swedish speakers. MA Thesis, University of Lund. Schwartz, B. D. 1996. On two hypotheses of ‘transfer’ in L2A: Minimal trees and absolute L1 influence. In The generative study of second language acquisition, S. G. Flynn, G. Martohardjono & W. O’Neil (eds.), 17–34. Hillsdale NJ: Lawrence Erlbaum. Schwartz, B. D. & Sprouse, R. A. 1996. L2 cognitive states and the full transfer/full access model. Second Language Research 12: 40–77. Speas, M. 2001. Constraints on null pronouns. In Optimality-Theoretic Syntax, G. Legendre, J. Grimshaw & S. Vikner (eds.), 393–425. Cambridge MA: The MIT Press. Tiphine, U. 1983. The acquisition of English statements and interrogatives by French-speaking children. PhD dissertation, University of Kiel. Tsimpli, I.-M. & Dimitrakopoulou, M. 2007. The interpretability hypothesis: Evidence from whinterrogatives in L2 acquisition. Second Language Research 23(2): 215–242. Vainikka, A. 2007. Adverb movement in organic syntax. Ms. (submitted for publication in Syntax). Vainikka, A. & Levy, Y. 1999. Empty subjects in Hebrew and Finnish. Natural Language and Linguistic Theory 17: 613–671. Vainikka, A. & Young-Scholten, M. 1994. Direct access to X’-theory: Evidence from Korean and Turkish adults learning German. In Language acquisition studies in generative grammar, T. Hoekstra & B.D. Schwartz (eds.), 265–316. Amsterdam: John Benjamins. Vainikka, A. & Young-Scholten, M. 1996a. Gradual development of L2 phrase structure. Second Language Research 12: 7–39. Vainikka, A. & Young-Scholten, M. 1996b. The early stages in adult L2 syntax: Additional evidence from Romance speakers. Second Language Research 12: 140–176.



Anne Vainikka and Martha Young-Scholten Vainikka, A. & Young-Scholten, M. 1998. Morphosyntactic triggers in adult SLA. In Morphology and its interfaces, M-L. Beck (ed), 89–113. Amsterdam: John Benjamins. Vainikka, A. & Young-Scholten, M. 2002. Restructuring the CP in L2 German. In Proceedings of the 26th BUCLD, B. Skarabela, S. Fish & A. H.-J. Do (eds.), 712–722. Somerville MA: Cascadilla Press. Vainikka, A. & Young-Scholten, M. 2005. The roots of syntax and how they grow. Organic grammar, the basic variety and processability theory. In Paths of development in L1 and L2 acquisition, S. Unsworth, A. Sorace, T. Parodi & M. Young-Scholten (eds.), 77–106. Amsterdam: John Benjamins. Vainikka, A. & Young-Scholten, M. 2007. Minimalism vs. organic syntax. In Clausal and phrasal architecture: Syntactic derivation and interpretation. Papers in honour of Joseph Emonds, S. Karimi, V. Samiian & W. Wilkins (eds.) Amsterdam: John Benjamins. Weissenborn, J. 1990. Functional categories and verb movement: The acquisition of German syntax reconsidered. Linguistische Berichte 3: 190–224 (Special Issue). Wexler, K. 1994. Optional infinitives, head movement and the economy of derivations in child grammar. In Verb movement, D. Lightfoot & N. Hornstein (eds.), 305–350. Cambridge: CUP. White, L. 1991a. Adverb placement in second language acquisition: Some effects of positive and negative evidence in the classroom. Second Language Research 7: 133–160. White, L. 1991b.: The verb-movement parameter in second language acquisition. Language Acquisition 1: 337–360. White, L. 1992. Long and short verb movement in second language acquisition. Canadian Journal of Linguistics 37: 273–286.

Non-permanent representational deficit and apparent target-likeness in second language Evidence from wh-words used as universal quantifiers in English and Japanese speakers’ L2 Chinese Boping Yuan

University of Cambridge This paper reports on an empirical study examining L2 acquisition of Chinese wh-words used as universal quantifiers (UQs) by Japanese- and Englishspeaking learners. Results from a sentence acceptability judgment test and a sentence interpretation test indicate that wh-words can be acquired as UQs in English and Japanese speakers’ L2 Chinese grammars. However, the acquisition takes place rather late and there is evidence for representational deficit in beginner, intermediate and post-intermediate Japanese learners’ L2 Chinese grammars, i.e., wh-words in subject position can have target-like behaviours, but those in object position cannot. We argue that the underlying L2 Chinese representations in these learners’ L2 Chinese grammars are still divergent from that of the native Chinese grammar in spite of some apparently native-like behaviours.

1. Introduction In English, wh-words, such as who, what, when, where, how and why, are mainly used as interrogatives in wh-questions.1 In contrast, Chinese wh-words, apart from being used as interrogatives, can also be used as universal quantifiers, as in (1a) and (2a). However, there are constraints on the use of a wh-word as a universal quantifier in Chinese (Cheng 1991, 1995; Lee 1986; Xu 1997). When used as a universal quantifier, the wh-word has to be licensed by dou, which is considered the head of a functional category in this paper. Without the use of dou, the sentence 1.

Of course, English wh-words can be used as relative pronouns as well.



Boping Yuan

would be ungrammatical, as in (1b) and (2c). This is believed to be due to the fact that the wh-word in Chinese has an uninterpretable distributing feature and does not have any quantificational force, and that it is dou which values the wh-word and contributes the universal quantificational force to it. Another constraint is that the wh-word has to appear adjacent to the left of dou. This is because the head dou carries a movement-forcing interpretable feature, which forces the wh-word to move to its specifier for feature checking. This can be clearly seen from the contrast in (2a,b). The wh-word shenme “what” in (2) is the object of the verb chi “eat”; the wh-word has to raise to the left of dou for checking and valuing, as in (2a), and the wh-word staying in situ would lead to ungrammaticality, as in (2b). (1) a. Shei dou xihuan ta. who all2 like her ‘Everyone likes her.’ b. * Shei xihuan ta. who like her ‘Everyone likes her.’ (2) a. Ta shenmei dou chi ti. he what all eat ‘He eats everything.’ b. *Ta dou chi shenme. he all eat what ‘He eats everything.’ c. *Ta chi shenme. he eat what ‘He eats everything.’ Similar to wh-words in Chinese, Japanese wh-words can also be used as universal quantifiers. In this case they have to be combined with the particle –mo. This can be seen in the list in (3) and example sentences in (4) and (5).3

2. Strictly speaking, it is inaccurate to translate dou as “all”. As we will see below, dou is the head of a functional category in Chinese while “all” can be a floating quantifier in English although both are very close in their meanings. 3. According to Nishigauchi (1990), a wh-word combined with the particle –mo is a universal quantifier only when it is used in a non-negative environment. If it is used in a negative environment, it is a negative existential polarity word.



Wh-words as universal quantifiers 



(3) A list of Japanese wh-words used as interrogatives and universal quantifiers (adapted from Watanabe 1992: 48–49) As interrogatives dare “who” nani “what” doko “where” itsu “when” naze “why”

As universal quantifiers (-mo) dare-mo “everyone” #nani-mo “everything”4 doko-mo “everywhere” itsu-mo “whenever” *naze-mo ---------

(4) dare-mo-ga ringo-o tabeta. everyone-Nom apple-Acc ate ‘Everyone ate an apple’

(from Watanabe 2001: (26b))

(5) Sensei-ga dare-mo-o hometa. teacher-Nom everyone praised ‘The teacher praised everyone.’

(Kuribara, p.c )

As we can see from (4) and (5), wh-words in Japanese behave as universal quantifiers in combination with the quantificational element –mo. There is no need to use something equivalent to the Chinese dou in a Japanese sentence with a wh-word used as a universal quantifier, and unlike the case in Chinese, there is no requirement for movement or raising in such a sentence.5 In this paper, we will follow Nishigauchi (1990), Watanabe (1992, 2001) and Tsai (1994, 1999) in assuming that the wh-word is combined with the particle –mo in order to function as a universal quantifier in Japanese sentences like (4) and (5). In these sentences, the wh-word itself does not have a fixed semantic force, and only when a relation is established between the wh-word and the particle –mo and when the former is licensed by the latter will the wh-word be identified as a universal quantifier. In this sense, whwords in Japanese are variables and they have to be bound by their licensers.6 4. According to Watanabe (1992), this combination is only possible in a fixed expression like (i). (i) Nani-mo-ga iya-ni natta. -Nom hateful-be became

‘Everything has become hateful/ I have become sick with everything.’

5. The fact that the objects ringo-o (=apple) and dare-mo-o (=everybody) in (4) and (5) are in a preverbal position is not a result of movement. This is due to the fact that Japanese has a head final VP. 6. An anonymous reviewer of the paper suggests that wh-words like whoever, whatever, whenever, wherever, etc. can function as universal quantifiers in English. However, it should be pointed out that they are actually different from Chinese and Japanese wh-words used as universal quantifiers both syntactically and semantically. Syntactically, these English wh-words always



Boping Yuan

It has been proposed by some second language (L2) researchers (e.g., Smith and Tsimpli 1995, Hawkins and Chan 1997) that features that are not used in L2 learners’ first language (L1) but are required in the L2 will become inaccessible in their L2 acquisition after a critical period. This has been referred to as the Representational Deficit Hypothesis (cf. Hawkins 2003). More recently, this group of researchers (Tsimpli 2003; Tsimpli and Dimitrakopoulou 2007; Hawkins 2004; Hawkins and Hattori 2006) propose that in adult L2 acquisition, uninterpretable features, but not interpretable features, will be inaccessible if they are not selected in the construction of the L1. This has been called the Interpretability Hypothesis. According to this hypothesis, the inaccessibility of the uninterpretable feature will lead to a representational deficit in underlying L2 grammars although L2 learners’ performance may still appear native-like (cf. Hawkins and Hattori 2006). As neither the functional projection headed by dou nor the uninterpretable distributing feature of the Chinese wh-word is available in English and Japanese, both the Representational Deficit Hypothesis and the Interpretability Hypothesis would predict a deficit in this aspect of English and Japanese speakers’ L2 Chinese grammars although it would not rule out the possibility that some behaviours of the wh-word and dou may appear native-like in the learners’ L2 Chinese. This paper reports on an empirical study examining whether dou in Englishand Japanese-speaking learners’ L2 Chinese grammars can project a functional category with a movement-forcing interpretable distributing feature and whether the uninterpretable distributing feature of the wh-word used as a universal quantifier can be acquired in adult L2 acquisition of Chinese. The paper is organized as follows. The next section provides an analysis of dou and how Chinese wh-words are licensed and valued as universal quantifiers. Section 3 presents some data from studies reported in the L2 literature which indicate that adult L2 learners are unable to acquire functional categories and the related features which are required in the L2 and not available in their L1. In Section 4, we describe our empirical study of the Chinese wh-word used as a universal quantifier in English and Japanese speakers’ L2 Chinese, and we will also present the data collected from the study.

have to introduce an embedded clause, as in “Whoever wins the game will get an award of 2,000 pounds.” or “She will accept whatever you would give to her.” and these English wh-words cannot behave like the Chinese and Japanese wh-words in (1a), (2a), (4) and (5), where no embedded clause has to follow these Chinese and Japanese wh-words. Semantically, the English wh-words, whoever, whatever, etc. do not share the meanings of universal quantifiers, such as each, every, etc. and they seem to be close in meaning to the negative polarity item any, as in “Anyone who wins the game will get an award of 2,000 pounds.” or “She will accept anything you would give to her.” Based on this analysis, I would argue that it is indeed the case that English does not have wh-words which can be used as universal quantifiers.



Wh-words as universal quantifiers 

Section 5 will discuss the findings in the empirical study and their implications for the Interpretability Hypothesis, and Section 6 is a conclusion. 2. Analyses of dou and the wh-word used as a universal quantifier in Chinese 2.1

Characteristics of dou

It is generally agreed in the literature (Lee 1986; Cheng 1991, 1995; J. Li 1995; Xu 1997; among others) that a main function of dou is that of a distributor and that the sentence with dou is grammatical only when the NP quantified by dou is interpreted distributively.7 This can be clearly seen in the examples in (6) (from Lee 1986: 58) and (7) (from Cheng 1995: 212). (6) a. Zhangsan he Mali mingtian jiehun. Zhangsan and Mary tomorrow marry ‘Zhangsan and Mary will marry tomorrow’

(collectively)

b. Zhangsan he Mali dou mingtian jiehun. (distributively) Zhangsan and Mary all tomorrow marry ‘Zhangsan and Mary will both marry (with someone else) tomorrow.’ (7) a. Women heyong yi-ge chufang. We share one-CL kitchen ‘We share a kitchen.’ b. Women dou heyong yi-ge chufang. We all share one-CL kitchen ‘We each share a kitchen (with someone else).’

(collectively)

(distributively)

In (6b), due to the distributive use of dou, the sentence does not have a meaning that Zhangsan and Mary are getting married with each other tomorrow. However, the sentence in (6a), which does not have the word dou, has such a meaning. The same applies to the sentences in (7a) and (7b). The sentence in (7a) implies that there is only one kitchen which is shared by us. In contrast, the sentence in (7b) can only have the interpretation that each of us shares a kitchen with someone else. All these contrasts can be accounted for by the distributive character of dou. A second characteristic of dou is that it requires the preceding NP it quantifies over to have a referentially plural interpretation (cf. J. Li 1995; Xu 1997). This re7. Dou can also be used (often together with the word lian) to provide a meaning of emphasis. Lü (1980) points out that in this case, dou is toneless. Also in this usage, the NP preceding dou can be singular, unlike the distributor dou, which requires the preceding NP to be referentially plural. As our focus is on the distributive dou, we will not discuss the lian … dou structure here.



Boping Yuan

quirement actually follows directly from the distributive characteristic of dou. This can be seen in the contrast between (8a) and (8b). While the sentence in (8a) containing a third-person singular pronoun is ruled out, the sentence in (8b) is grammatical due to the fact that a plural pronoun is used. (8) a. *Ta dou chi-le yi-ge pingguo. he all eat-PFV one-CL apple ‘*He each ate an apple.’ b. Tamen dou chi-le yi-ge pingguo. they all eat-PFV one-CL apple ‘They each ate an apple.’ Syntactically, dou only occurs in a preverbal position and is never allowed postverbally. This can be seen in the contrast between (8b) and (9). In addition, when the NP quantified by dou is an object, the object has to raise to a preverbal position preceding dou. It can move adjacent to the left of dou, as in (10a) or further to a topic position, as in (10b). However, dou cannot occur before a subject even though it is adjacent to the topic it quantifies over, as in (10c). (9) a. *Tamen chi-le dou yi-ge pingguo. they eat-PFV all one-CL apple ‘They each ate an apple.’ b. *Tamen chi-le yi-ge pingguo dou. they eat-PFV one-CL apple all ‘They each ate an apple.’ (10) a. Ta naxie shui dou kan-guo ti. he those books all read-EXP ‘He has read all of those books.’ b. Naxie shui ta ti dou kan-guo ti. those books he all read-EXP ‘He has read all of those books.’ c. *Naxie shui dou ta kan-guo ti. those books all he read-EXP ‘He has read all of those books.’



Wh-words as universal quantifiers 

2.2

What is the status of dou when it functions as a distributor8?

There has been a vigorous debate with regard to the status of dou in Chinese, which has generated different analyses of dou and its associated structure. Lee (1986) and Cheng (1995) both argue that dou should be regarded as an adverb with quantificational force. As we see above, dou has to occur preverbally, and so do all adverbs in Chinese. This might suggest that dou should indeed be categorised as an adverb. However, as pointed out by Xu (1997), apart from the adverb, there are many non-adverbial elements in Chinese which cannot occur postverbally. So we cannot rule out the possibility that dou is a non-adverbial element just on the basis of its preverbal position in a sentence. In Lee (1986), it is argued that the relation between dou and the NP it quantifies over is established through a co-indexing mechanism in which dou co-indexes with any left constituent it c-commands. However, apart from some empirical problems9, Lee fails to explain why the NP that dou co-indexes with cannot occur to the right of dou. Cheng’s (1995) proposal is that dou has a dual status. On the one hand, when quantifying over plural NPs, dou is a quantifier with a particular quantificational requirement. On the other hand, when used with wh-words, it is

8. Dou has a number of different meanings. Apart from the usage mentioned in Note 1, it can also mean ‘already’, as in (i), or be used in a question with a wh-interrogative word to indicate that the expected answer should provide a list of things or people, as in (ii). I follow Lü (1980) in assuming that dou used in these sentences is an adverb. As dou in these situations does not have a quantificational force, it should be treated differently from the use of dou we are concerned with here. (i) Dou san dian le. already 3 o’clock Part ‘It is already 3 o’clock. (ii) a. Dou shei hui shuo Fayu? all who can speak French ‘Who are all the people who can speak French?’ b. Ni dou qu-guo shenme difang? you all go EXP what place ‘What are all the places you have been to?’ 9. Chiu (1990) points out that Lee’s co-indexing mechanism cannot rule out the ungrammatical sentences like (i) and (ii), where the plural NPs in the embedded clauses which can potentially be associated with dou are all c-commanded by dou. (i) (ii)

* [Akiu du zhexie shu] dou zui heshi. Akiu read these book each most suitable * [tamen renshi] de nei-ge ren dou lai le. they know DE that-CL person each come SFP



Boping Yuan

both a trigger and a binder, providing quantificational force to the wh-word.10 Cheng rejects Lee’s co-indexing mechanism and argues for an analysis in which dou is adjoined at LF to the NP it quantifies over. The adjunction of dou to the NP turns the complex [NP + dou] into a quantificational structure, and this complex structure then undergoes QR, creating a variable for the complex structure to bind. As dou is considered an adverb, Cheng argues that the movement of dou is a type of adjunct movement. However, Cheng does not provide a clear explanation as to what specific feature triggers the movement of dou to the NP at LF. Another problem is the treatment of dou as an adverb. Adverbs are not obligatory constituents of a syntactic structure and the absence of an adverb should not lead to ungrammaticality of the sentence. However, the contrasts in (1) and (2) demonstrate that dou is compulsory in a sentence with a universal quantifier, and it would be difficult to account for the ungrammaticality caused by the absence of dou if dou is treated as an adverb.11 2.3

Dou as the head of a functional category

In this paper, we follow Hsieh (1995), X. Li (1996), Lin (1998) and Wu (1999) in assuming that dou is the head of a functional category. Within the framework of Chomsky’s (1995) Minimalist Program, these researchers have argued that dou projects a functional category with a movement-forcing distributing feature, which forces the distributed universally quantified NP to move to its specifier for feature checking. Let’s call this functional category “Distributional Projection” (DisP). This can account for the requirement in Chinese that universally quantified NPs, such as meige ren (=everyone), suoyou de ren (=all the people), and renhe ren (=any­one) must occupy a preverbal position and must co-occur with dou, as in (11). It is also not possible to put dou before the universal quantifier, as in (11c). (11) a. Meige ren (suoyou de ren / renhe ren) dou xue Zhongwen. Everyone (all people anyone ) DOU study Chinese ‘Everyone (all the people / anyone) studies Chinese.’

10. We will adopt in our analysis Cheng’s (1995) idea about the triggering effect when dou is used with wh-words in Chinese. 11. Another proposal is made by Chiu (1990), who, following Sportiche’s (1988) analysis of floating quantifiers in French, regards dou as a floating quantifier which heads its own maximal projection, DouP. Apart from other problems with Chiu’s proposal, she does not consider the interaction between dou and wh-words used as a universal quantifier, which further shows limitations of her account as a general analysis of dou in Chinese. So, we will not discuss her proposal here.



Wh-words as universal quantifiers 

b. *Meige ren (suoyou de ren / renhe ren) xue Zhongwen. Everyone (all people anyone ) study Chinese ‘Everyone (all the people / anyone) studies Chinese.’ c. *Dou meige ren (suoyou de ren / renhe ren) xue Zhongwen. DOU everyone (all people anyone ) study Chinese ‘Everyone (all the people / anyone) studies Chinese.’ (12)

DisP NP Meige reni

DisP’ Dis

vP

dou <Meige reni>

v’ v

VP

In this paper, we use angled brackets to indicate the position from which a constituent has moved. As we can see in (12), the universal quantifier meige ren in (11a) is base-generated in the specifier of vP and moves to specifier of DisP to check the distributing feature. This analysis provides an account for the requirement that the NP dou is associated with must be plural or the entity represented by the NP can be divided, as in (8). This is because only plural NPs can be distributed. Following Heim, Lasnik and May (1991), Hsieh (1995) argues that there is a distinction between plural NPs with a collective feature and plural NPs with a distributing feature. She points out that the NP Naxie xuesheng in (13a) has a collective feature while the NP Naxie xuesheng in (13b) has a distributing feature, which explains the contrast between (13a) and (13b). (13) a. Naxie xuesheng mai-le yi tai dianshiji. (Naxie xuesheng=collective) Those students buy PFV a CL TV set ‘Those students bought a TV set.’ b. Naxie xuesheng dou mai-le yi tai dianshiji. (Naxie xuesheng=distributive) Those students DOU buy-PFV a CL TV set ‘Those students each bought a TV set.’ It seems natural to assume that DisP is dominated by TP and that the subject meige ren in (11a) and (12) moves further from the specifier of dou to the specifier of TP



Boping Yuan

although there is no overt morphological marking of tense in Chinese.12 This analysis provides an account for the sentence in (10b), where the dou-associated NP Naxie shu (=those books) appears at the sentence initial position without being adjacent to the left of dou. In (10b), the NP Naxie shu moves to Spec of DisP from the object position to check the [dis] feature before it is topicalized to the sentence initial position. The subject ta (=he) in (10b) is base-generated in the specifier of vP and moves to the specifier of TP as a result of the EPP, an abbreviation for the Extended Projection Principle (EPP), which originally required that every clause must have a subject (Chomsky 1981). In his more recent work, Chomsky (2001) has extended the EPP to mean that functional categories like C and T carry an [EPP] feature requiring them to project a specifier. We have seen that not only preverbal universal quantifiers must co-occur with dou, as in (11), this co-occurrence requirement also applies to Chinese wh-words when they are used as universal quantifiers, as shown in the contrast between (1a) and (1b). In this case, dou can be considered a licenser or a valuer.13 It has been widely accepted that for a Chinese wh-word to have a universal quantifier reading, it has to be licensed or valued by dou (cf. Cheng 1991, 1995; Hsieh 1995; X. Li 1996; Lin 1998; Wu 1999). In this case, we can then assume that the wh-word in (1a) is base-generated in Specifier of vP and moves to the specifier of DisP in order to be licensed and valued by dou as a universal quantifier, as shown in (14).

12. Hsieh (1995) assumes that the functional projection headed by dou is higher than AspP because dou must precede the progressive marker zai, as in (i) and (ii). (i) Tamen dou zai changge. They all ASP sing ‘They are all singing.’ (ii) *Tamen zai dou changge. They ASP all sing 13. Apart from a universal quantifier, a Chinese wh-word can be interpreted as an existential polarity word or as a wh-interrogative. This depends on what licenses/values the wh-word. See Yuan (2007a,b,c) for discussions of Chinese wh-words used as interrogatives, existential polarity words and their behaviours in L2 Chinese.



Wh-words as universal quantifiers 

(14)

TP NPi Sheii

T’ T

DisP Dis’ <Sheii>

vP

Dis dou <Sheii>

v’ v

VP

It seems feasible to apply Adger’s (2003) mechanism of feature checking and valuing of Tense to our analysis of DisP. According to Adger, the verb has an uninterpretable tense feature and the head of TP an interpretable tense feature, and the uninterpretable feature on the verb has to get a value from a matching interpretable feature on T. Likewise, we can assume that the head dou carries an interpretable [dis] feature and the wh-word used as a universal quantifier an uninterpretable [udis] feature. Adger uses an italic “u” to mark an uninterpretable feature. When the uninterpretable feature of the wh-word is checked against the interpretable feature of dou, it gets valued by dou. In other words, what the checking operation does is value and check the [udis] feature of the wh-word because the [udis] feature of the wh-word is unvalued and uninterpretable. This valuation operation is done by Agree, whose function is to value and delete the uninterpretable feature. Adger uses the configuration in (15) to illustrate the operation of Agree (2003: 169). (15) Agree In a configuration X [F: value] … Y[uF:] where … represents c-command, the F checks and values uF, resulting in: X [F: value] … Y[uF: value] The approach of checking by valuing, as illustrated in (15), shows that the uninterpretable feature can get a value from a matching interpretable feature. As mentioned above, dou has been considered to have a movement-forcing [dis] feature by many researchers (e.g., Hsieh 1995; X. Li 1996; Lin 1998; Wu 1999; among others). The movement-forcing feature on dou requires that the checking and valuing have to take place locally, rather than at a distance. In other words, the wh-word is



Boping Yuan

forced to move to the specifier of DisP for feature checking. We can follow Adger (2003) by using an asterisk “*” to represent the movement-forcing feature. The movement of the wh-word triggered by the movement-forcing feature on dou is illustrated in (16). (16) dou[dis*] … wh-word[udis] → wh-word[udis] dou[dis*] … < wh-word[udis] >

When the wh-word moves to the specifier of DisP, as in (16), the uninterpretable [udis] feature can be checked and get deleted. This feature checking relationship between the wh-word and dou is considered as the cause for the co-occurrence requirement between the wh-word and dou and for the pre-dou position of the wh-word. To sum up, dou is considered in this paper as a functional category heading a distributional projection (DisP) with an interpretable [dis*] feature, which forces the wh-word with the uninterpretable [udis] feature to move to its specifier for feature checking. In other words, for the wh-word to be used as a universal quantifier in Chinese, it has to move to the specifier of DisP in order to be licensed and valued by dou. Neither English nor Japanese has DisP, and in these two languages, there is no counterpart of the Chinese dou. The Japanese particle –mo, as illustrated in (3–5), is lexical in nature and does not head a functional projection. 3. L2 studies of the Representational Deficit Hypothesis and the Interpretability Hypothesis Hawkins and Chan (1997) propose that any functional categories and their features which are required by the target language but not available in their L1 can no longer be acquired by adult L2 learners. This is referred to as the Failed Functional Features Hypothesis. Hawkins and Chan base their hypothesis on a study of L2 acquisition of English restrictive relative clauses by adult Chinese speakers. English restrictive relative clauses involve wh-operator movement while their Chinese counterparts do not. Based on the data from their study, Hawkins and Chan argue that the gap in Chinese speakers’ L2 English restrictive relative clauses like “The girl [who I like __ ] is here” is a null pronominal rather than a trace created by wh-operator movement, and that the wh-operator is a base-generated topic in Chinese speakers’ L2 English restrictive relative clauses. This is believed to be due to the fact that Chinese speakers’ L2 English is constrained by the feature specification of functional categories in their L1 Chinese and the wh-operator, which is not available in Chinese, cannot be acquired by these adult learners. On the basis of these data, Hawkins and Chan propose that adult L2 learners no longer have access



Wh-words as universal quantifiers 

to the full range of functional features made available by UG because features not selected by their L1 disappear after a critical period. Hawkins and Chan also propose that where the L2 specification of a functional feature is different or absent in the L1, L2 grammars can make a local adjustment and approximate to the target language grammar, the result of which is that L2 learners can produce target-like sentences. This accounts for the Chinese learners’ analysis of the wh-operator in their English restrictive relative clauses as a topic, because Chinese is a topic-prominent language and the topic can be basegenerated at the sentence initial position co-indexed with an overt or covert constituent in the sentence. Hawkins (2004) calls this a case of feature missing despite appearances. That is, in spite of the observation of target-like performance by L2 learners, certain features are unavailable in their L2 grammars, as a result of the critical period. Hawkins and Hattori (2006) make a more fine-tuned and more explicit claim about the representational deficit in L2 grammars. They make a distinction between interpretable features and uninterpretable features, with the former being used in syntactic computation by the semantic component in determining the meaning of syntactic expressions and the latter being formal and grammatical features and not usable by the semantic component. They follow Tsimpli (2003) and Tsimpli and Dimitrakopoulou (2007) by claiming that uninterpretable features disappear in adult L2 grammars if they are not selected from the feature inventory of UG during the acquisition of primary grammar(s) within the critical period, the result of which is that L2 speakers’ mental representations of the target language will permanently diverge from that of native speakers. Interpretable features remain available, even those not selected during the acquisition of the L1. To investigate this hypothesis, Hawkins and Hattori (2006) conduct a study of adult Japanese speakers’ L2 acquisition of the uninterpretable feature that forces wh-movement in interrogatives in English. Japanese is a wh-in-situ language which lacks the movement-forcing feature. In the study, highly proficient Japanese speakers of English who were first exposed to English relevant input after the critical period were asked to interpret English bi-clausal multiple wh-questions like “Who did John say bought what?” and *“What did John say who bought?”. The results show that while the Japanese speakers are not significantly different from native English speakers in judging grammatical sentences, they are significantly different from the native speakers in judging ungrammatical sentences. The Japanese speakers are also found to be not significantly different in their judgment of grammatical and ungrammatical sentences. Based on these results, Hawkins and Hattori argue that Japanese speakers are unable to acquire the uninterpretable feature forcing wh-movement in English interrogatives, which is present in the native English grammar but has not been selected by Japanese learners because it has



Boping Yuan

disappeared from the UG feature inventory of these Japanese learners. Hawkins and Hattori refer to this as a permanent loss of capacity to acquire in this domain. Hawkins and Hattori’s account for the Japanese speakers’ successful interpretation of English wh-interrogatives is that Japanese speakers represent the wh-movement in English as cases of wh-scrambling in Japanese, which obligatorily moves a whphrase leftwards because of a Focus requirement rather than the uninterpretable movement-forcing feature. Hawkins and Hattori use this as an example to argue that L2 learners will construct representations for the relevant L2 structures with alternative resources made available by UG if a certain uninterpretable feature is no longer available in their UG feature inventory. Tsimpli and Dimitrakopoulou (2007) study the use of a resumptive strategy in wh-subject and object extraction by intermediate and advanced Greek learners of English, and find that, while interpretable features of animacy and discourse-linking can be acquired by Greek-speaking learners of English, their L1 specification of resumptive pronouns as clusters of uninterpretable Case and Agreement features resists resetting. Based on these findings, Tsimpli and Dimitrakopoulou argue that interpretable features are acquirable in L2 acquisition, but uninterpretable features available in L1 but not in L2, or vice versa, lead to persistent problems in L2 grammars, even at the final state of L2 acquisition. 4. Empirical Study 4.1

Research questions

a. As we argued above, English and Japanese do not have DisP, nor do they have a counterpart of the Chinese dou. Given these facts, will English and Japanese speakers be able to project a new functional projection DisP in their L2 Chinese grammars? b. Can Chinese wh-words be interpreted as universal quantifiers by both English and Japanese speakers given that English does not allow wh-words to be used this way? A positive answer to this can implicate that the wh-word can be valued by the interpretable feature [dis] on dou in these learners’ L2 Chinese grammars. c. Will the wh-word move to the left of dou in English and Japanese speakers’ L2 Chinese? This should happen if the interpretable [dis] feature on dou has the movement-forcing feature [*] in their L2 Chinese grammars, which triggers the wh-word to move to the specifier of DisP to have its [udis] feature checked against the [dis*] feature of dou locally.



Wh-words as universal quantifiers 

d. If we follow Hawkins and Hattori (2006), Tsimpli (2003), and Tsimpli and Dimitrakopolou (2007) in assuming that uninterpretable syntactic features unselected in adult L2 learners’ L1s disappear at the end of the critical period, neither adult English nor adult Japanese learners of Chinese are expected to acquire the wh-word as a universal quantifier in their L2 grammars. If this happens, what alternative resources will these learners use to represent the relevant structure? e. As illustrated in (3–5), wh-words in Japanese can be used as universal quantifiers when lexically licensed by the particle –mo. In acquiring Chinese whwords as universal quantifiers, will this give Japanese speakers any advantage or disadvantage over English speakers whose L1 does not allow its wh-words to be used as universal quantifiers? 4.2

Informants

To answer the research questions above, an empirical study was conducted, which included 102 English speakers and 106 Japanese speakers as informants. It also included 20 native speakers of Chinese as controls. The English and Japanese informants were mainly undergraduate students, post-graduate students from universities in the U.K. and Japan as well as English- and Japanese-speaking students attending Chinese language programmes in China. We also included English- and Japanese-speaking lecturers and professors teaching Chinese or Chinese-related courses in universities in the U.K and Japan. All informants were over 17 when they started learning Chinese.14 The native Chinese speakers were university students or office workers in China. On the basis of their performance in a Chinese cloze test, the English and Japanese speakers were divided into five Chinese proficiency groups respectively: Beginner Group, Post-Beginner Group, Intermediate Group, Post-intermediate Group and Advanced Group. The native Chinese speakers are in the Native Chinese Group. Information about each of the 11 groups is given in Table 1. We use “E-” and “J-” to represent English and Japanese groups respectively. From Table 1, we can see that E-Advanced and J-Advanced groups have a history of learning Chinese for 17.5 and 16 years, which suggests that they are very advanced learners of Chinese and can probably represent the final state of L2 Chinese grammars. An ANOVA shows that there is a significant difference between all groups in their performance in the cloze test, (F=(10, 227)798.226, p<0.001) and the following-up Scheffé tests indicate that except for the E-Advanced and J-Advanced groups, each 14. Eight learners started learning Chinese in evening classes or language schools and the rest started learning Chinese after entering universities.



Boping Yuan

Table 1  Information about each group Groups

No. of Average Average months informAge of ants studying Chinese

Average months in China/Taiwan

Mean scores in the cloze test (total=40) (ranges in brackets)

E-Beginner

20

19

4

0.5



3

(1–6)

E-Post-beginner

20

22

9

2



10

(7–15)

E-Intermediate

25

22

28

5



21 (16–25)

E-Post-intermediate

22

28

81

19



30 (26–34)

E-Advanced

15

37

211

40



37 (35–39)

J-Beginner

22

20

8

0.5



J-Post-beginner

24

21

16

3



J-Intermediate

22

23

31

9



22 (16–25)

J-Post-intermediate

20

29

42

20



31 (26–34)

J-Advanced

18

32

196

31



37 (35–39)

Native Chinese

20

25

N/A

N/A



39 (38–40)

3

(1–6)

8

(7–15)

of the learner groups is significantly different from the native Chinese group. It is also found that there is no significant difference between any two groups of the same proficiency level, that is, between E-Beginners and J-Beginners, between EPost-beginners and J-Post-beginners, Between E-Intermediate and J-Intermediate, etc. This implicates that every English group is compatible with the corresponding Japanese group in terms of their Chinese language proficiency. Materials and procedures

4.3

Each of the informants had to complete two tasks, an acceptability judgment task and a sentence interpretation task. In the acceptability judgment task, which preceded the sentence interpretation task, the subject was asked to judge the acceptability of each sentence by circling a number on a scale as given in (17). On the basis of the scale, we treat any score of “+1” or above that the informant assigned to a particular sentence as a sign of accepting the sentence, and conversely any score of “-1” or below as a sign of rejecting that sentence. The score of “0” is treated as a sign that the informant is not sure. (17)

-2

-1

0

+1

+2

completely unacceptable

probably unacceptable

“I don’t know”

probably acceptable

completely acceptable



Wh-words as universal quantifiers 

The acceptability judgment task included, apart from distracters, 8 types of test sentences and each type had four tokens. Examples of the 8 types are given in (18).15 (18) Types 1: shenmo (=what) in subject position and adjacent to the left of dou: (In that country,) shenmo dou hen gui. what dou very expensive ‘(In that country,) everything is expensive.’ Type 2: shenmo (=what) in subject position but adjacent to the right of dou: *(In that country,) dou shenme hen gui. dou what very expensive ‘(In that country,) everything is expensive.’ Type 3: shei (=who) in subject position and adjacent to the left of dou: (In our school,) shei dou xihuan ta. who dou like him ‘(In our school,) everyone likes him.’ Type 4: shei (=who) in subject position but adjacent to the right of dou: *(In our school,) dou shei xihuan ta. dou who like him ‘(In our school,) everyone likes him.’

Type 5: shenme (=what) moves from object position to the left of dou: Wo shenme dou chi. I what dou eat ‘I eat everything.’

Type 6: shenme (=what) remains in situ in object position in a sentence with dou: *Wo dou chi shenme. I dou eat what ‘I eat everything.’ Type 7: shei (=who) moves from object position to the left of dou: (He knows Zhang Hong, he knows Li Ming, he knows Wang Ping …….)

15. The glossary in the example sentences is provided for the reader of the paper but was not provided in the test.



Boping Yuan

ta shei dou renshi. he who dou know ‘(He knows Zhang Hong, he knows Li Ming, he knows Wang Ping …….) he knows everybody.’ Type 8: shei (=who) remains in situ in object position in a sentence with dou: *(He knows Zhang Hong, he knows Li Ming, he knows Wang Ping …….) ta dou renshi shei. he dou know who ‘(He knows Zhang Hong, he knows Li Ming, he knows Wang Ping …….) he knows everybody.’ Types 1–4 are designed to investigate whether wh-words in subject position will raise to specifier of DisP to have their [udis] feature checked and valued by the interpretable [dis*] feature on dou in L2 Chinese grammars. If this does happen in L2 Chinese grammars, subjects should accept sentences in Types 1 and 3 and reject sentences in Types 2 and 4. The same applies to Types 5–8. If the checking and valuing take place locally between the wh-word and dou in L2 Chinese grammars, informants should accept Types 5 and 7 and reject Types 6 and 8. As can be seen in (18), there is a correspondence between Types 1 and 2, between Types 3 and 4, between Types 5 and 6, and between Types 7 and 8. In each of these pairs, everything is the same except for the position of the wh-word in the sentence. If the informant accepts one sentence but rejects the other, we can attribute the difference in the informant’s judgment to the different position of the wh-word in the sentence. In the sentence interpretation task, apart from distracters, there are 4 types of sentences, and in each type there are 6 tokens (with 3 tokens using shenme (=what) and 3 using shei (=who)). In this task, the informant was asked to make an interpretation of each sentence by choosing one from five options given16. Examples of the 4 types are provided in (19), where “√” is used to indicate the correct interpretation for the convenience of the reader of the paper, but this was not available to the informant in the experiment.

16. In the interpretation task, the sentence to be interpreted was given in Chinese, but the interpretations for the informant to choose were given in the informant’s L1.



Wh-words as universal quantifiers 

(19) Interpretation Type 1: the wh-word in subject position adjacent to the left of dou: (In his room,) shenme dou hen ganjing. what dou very clean a. This sentence means: ‘something is clean.’ b. √ This sentence means: ‘everything is clean.’ c. This sentence means: ‘what is clean?’ d. I don’t understand this sentence. e. This sentence is incorrect. Interpretation Type 2: the wh-word in subject position of a sentence without dou: *(In his room,) shenme hen ganjing. what very clean a. This sentence means: ‘something is clean.’ b. This sentence means: ‘everything is clean.’ c. This sentence means: ‘what is clean?’ d. I don’t understand this sentence. e. √ This sentence is incorrect. Interpretation Type 3: the wh-word moves from the object position to the left of dou: (Of the students in our class,) ta shei dou xihuan. he who dou like a. This sentence means: ‘he likes someone.’ b. √ This sentence means: ‘he likes everyone.’ c. This sentence means: ‘who does he like?’ d. I don’t understand this sentence. e. This sentence is incorrect. Interpretation Type 4: the wh-word remains in situ in object position of a sentence without dou: *(Of the students in our class,) ta xihuan shei. he like who a. This sentence means: ‘he likes someone.’ b. This sentence means: ‘he likes everyone.’ c. This sentence means: ‘who does he like?’ d. I don’t understand this sentence. e. √ This sentence is incorrect.



Boping Yuan

The sentence interpretation task is designed to examine whether the wh-word is correctly interpreted as a universal quantifier in L2 Chinese grammars when it is positioned in the specifier of dou, as in Interpretation Types 1 and 3, and whether the wh-word is incorrectly interpreted as a universal quantifier when dou is absent in the sentence. As we can see in (19), the difference between Interpretation Types 1 and 2 is the presence or absence of dou in the sentence, and the difference between Interpretation Types 3 and 4 is that the wh-word raises from object position to the specifier of dou in the former and but remains in situ with dou absent in the latter. The data from the interpretation task is to provide information of whether the wh-word is interpreted as a universal quantifier and whether this interpretation depends on the checking and valuation of dou. All test sentences were randomized and they were presented in Chinese characters. However, instructions were given to the informants in their L1s respectively, i.e., English, Japanese and Chinese. Contexts like those in (19) were provided for some test sentences and they were given in the informants’ L1s as well. In order to minimize any possible effect of vocabulary on the judgments, efforts were made to include only those basic words of daily life. In addition, English/ Japanese translation and the Chinese pinyin (a Chinese phonetic system) were provided for some words potentially unfamiliar to the informants. Example sentences of irrelevant types were given to informants at the start of each task to help them understand how there were expected to do the task. Informants were also asked to fill in a questionnaire asking for information about their history of learning of Chinese. 4.4

Results

4.4.1 Results from the acceptability judgment test Tables 2 and 3 present the mean scores of the groups’ judgment of sentences in Types 1–4, i.e., sentences with the wh-word shenmo (=what) or shei (=who) in subject position adjacent to the left or to the right of dou. As we can see from these two tables, English and Japanese learners at intermediate, post-intermediate and advanced levels correctly accept grammatical sentences with the wh-word in the subject position adjacent to the left of dou, and they also correctly reject ungrammatical sentences with the wh-word in subject position adjacent to the right of dou as their mean scores of the correct sentences are above +1 (except for the E-Intermediate group’s judgment of shei (=who) to the left of dou, which is 0.76.) and their mean scores for the incorrect ones are all below –1. Results of paired-samples t-tests show that except for E-beginners and J-Beginners groups, there is a significant difference in every group’s judgment between the grammatical sentences with the wh-word to the left of dou and the ungrammatical sentences with the wh-word



Wh-words as universal quantifiers 

Table 2.  Mean scores of the judgment of sentences with shenmo (=what) in subject position adjacent to the left and *right of dou (i.e., Types 1 and 2)

E-Beginners E-Post-beginners E-Intermediate E-Post-intermediate E-Advanced J-Beginners J-Post-beginners J-Intermediate J-Post-intermediate J-Advanced Chinese

subj-what + dou

*dou + subj-what

–0.9 0.34 1.13 1.83 1.95 –0.5 0.1 1.25 1.67 1.81 1.58

–0.75 –0.76† –1.52† –1.91† –1.93† –0.06 –0.62† –1.09† –1.72† –1.83† –1.7†

Note: “†” = significantly different from the group’s judgment of the grammatical sentences of “subj-what + dou”.

to the right of dou.17 This seems to suggest that in these groups’ L2 Chinese grammars, whether English-speaking learners or Japanese-speaking learners, the whword has to move to the specifier of dou in order to have its [udis] feature checked and valued. The data in Tables 2 and 3 also show that English and Japanese beginner and post-beginner groups are indeterminate about the acceptability of the grammatical sentences with the wh-word adjacent to the left of dou as their mean scores are between +1 and –1. This suggests that the wh-word is not required to raise to the specifier of dou to have its [udis] feature checked and valued in these learners’ L2 Chinese. Probably DisP has not been projected in these beginning learners’ L2 Chinese grammars. It is also found in independent-samples t-tests that in judging any of these grammatical or ungrammatical sentences, there is no significant difference (p>0.05) between a particular English group and its corresponding Japanese group, i.e., between E-Beginners and J-Beginners, between E-Post-beginners and J-Post-beginners, between E-Intermediate and J-Intermediate, etc. This seems to suggest that L2 Chinese grammars of both English and Japanese speakers have similar developmental patterns as far as this aspect of their L2 Chinese grammars is concerned.

17. The differences are all at p<0.001 except for E-Post-beginners’ and J-Post-beginners’ judgments of “subj-what +dou” versus “*dou + subj-what”, which are at p=0.007 and at p=0.013 respectively, and except for the two groups’ judgments of “subj-who +dou” versus “*dou + subjwho”, which are at p=0.006 and p=0.027 respectively.

 Boping Yuan

Table 3.  Mean scores of the judgment of sentences with shei (=who) in subject position adjacent to the left and *right of dou (i.e., Types 3 and 4)

E-Beginners E-Post-beginners E-Intermediate E-Post-intermediate E-Advanced J-Beginners J-Post-beginners J-Intermediate J-Post-intermediate J-Advanced Chinese

Subj-who + dou

*dou + subj-who

–0.76 0.18 0.76 1.87 1.91 –0.14 –0.1 1.1 1.72 1.85 1.6

–0.03 –1.03† –1.65† –1.94† –1.86† –0.39 –1.06† –1.32† –1.5† –1.65† –1.56†

Note: “†” = significantly different from the group’s judgment of the grammatical sentences of “subj-who + dou”.

However, the developmental patterns suggested by the groups’ judgment of sentences with the wh-word in subject position are not found in their judgment of sentences with the wh-word in object position. As we can see from Tables 4 and 5, although the J-Advanced group, like the E-Advanced group, correctly accept sentences with the wh-word raised from the object position to the position adjacent to the left of dou, the other Japanese groups seem to have problems in accepting these grammatical sentences. Unlike the mean scores of the E-Intermediate and E-Post-intermediate groups, the corresponding J-Intermediate and J-Post-intermediate groups’ (as well as J-Post-beginner and J-Intermediate groups’) mean scores of these sentences indicate that raising the wh-word from the object position to the specifier of dou is not acceptable in these groups’ L2 Chinese grammars. In judging the grammatical sentences of “moved obj-what + dou”, E-Intermediate and E-Post-intermediate groups’ mean scores are 1.18 and 1.85 while J-Intermediate and J-Post-Intermediate groups’ are only 0.08 and 0.43; in judging the grammatical sentences of “moved obj-who + dou”, J-Intermediate and J-Post-intermediate groups’ mean scores are only 0.35 and 0.5, which are in contrast with the mean scores 1.04 and 1.69 of E-Intermediate and E-Post-intermediate groups. These Japanese groups seem indeterminate in this aspect of their L2 Chinese grammars as examinations of individual data show that informants in these groups sometimes accept and sometimes reject these grammatical sentences. Results of independent-samples t-tests indicate that there is a significant difference between E-Post-beginner and J-Post-beginner, between E-Intermediate and J-Intermediate,



Wh-words as universal quantifiers 

and between E-Post-intermediate and J-Post-intermediate in judging the grammatical sentences of “moved obj-what + dou” and “moved obj-who + dou”.18 It is not until the final stage, represented by the J-Advanced group here, do Japanese speakers’ L2 Chinese grammars accept grammatical sentences with the wh-word raised from the object position to the position adjacent to the left of dou. This implicates that this aspect of Chinese grammar is acquired much later in Japanese speakers’ L2 Chinese grammars than in English speakers’, and this also suggests that Japanese speakers’ L2 Chinese grammars and English speakers’ L2 Chinese grammars may follow different development paths and have different rates in acquiring this aspect of Chinese grammar. Table 4.  Mean scores of judgment of sentences with shenme (=what) moved from object position to the position adjacent to the left of dou (i.e., Type 5) and *sentences with shenme (=what) remaining in situ (Type 6) moved obj-what + dou E-Beginners E-Post-beginners E-Intermediate E-Post-intermediate E-Advanced J-Beginners J-Post-beginners J-Intermediate J-Post-intermediate J-Advanced Chinese

–1.19 0.36 1.18 1.85 1.8 –0.69 –0.81 0.08 0.43 1.77 1.64

* dou + obj-what in situ –1.09 –0.84† –1.46† –1.88† –1.5† –0.54 –1.06 –1.08† –1.05† –1.56† –1.56†

Notes: (a) “↔” = a significant difference between the two groups in an independent-samples ttest; (b) “†” = significantly different from the group’s judgment of the grammatical sentences of “moved obj-what + dou” in a paired-samples t-test.

18. In judging “moved obj-what + dou”, the significance level is at p=0.009 between E-Postbeginners and J-Post-beginners, at p<0.001 between E-Intermediate and J-Intermediate, at p=0.002 between E-Post-beginners and J-Post-beginners; and in judging “moved obj-who + dou”, the significance level is at p=0.016 between E-Post-beginners and J-Post-beginners, at p<0.001 between E-Intermediate and J-Intermediate, at p=0.004 between E-Post-intermediate and J-Post-intermediate.



Boping Yuan

Table 5.  Mean scores of judgment of sentences with shei (=who) moved from object position to the position adjacent to the left of dou (i.e., Type 7) and *sentences with shei (=who) remaining in situ (Type 8)

E-Beginners E-Post-beginners E-Intermediate E-Post-intermediate E-Advanced J-Beginners J-Post-beginners J-Intermediate J-Post-intermediate J-Advanced Chinese

obj-who + dou

* dou + obj-who in situ

–1 0.35 1.04 1.69 1.77 –0.92 –0.72 0.35 0.5 1.38 1.58

–1.41 –0.78† –1.53† –1.88† –1.75† –0.85 –1.24 –1.42† –1.05† –1.57† –1.81†

Notes: (a) “↔” = a significant difference between the two groups in an independent-samples ttest; (b) “†” = significantly different from the group’s judgment of the grammatical sentences of “moved obj-who + dou” in a paired-samples t-test.

It is also found in paired samples t-tests that all Japanese and English groups, except for J-Beginner and E-Beginner groups, are able to make a distinction between the grammatical and ungrammatical sentences concerning the raising of the whword from the object position to the specifier of dou as there is a significant difference in each of these groups’ judgment between the grammatical sentences in Types 5 and 7 and the ungrammatical ones in Types 6 and 8.19 Japanese and English groups do not seem to have major problems in rejecting the ungrammatical sentences of “* dou + obj-what in situ” and “* dou + obj-who in situ” as their mean scores for the two types of sentences are all below -1, except for the mean scores of E-Post-beginners (-0.84 and -0.78) and J-Beginners (-0.54 and -0.85). 4.4.2 Results from the sentence interpretation test Results from the acceptability judgment test only inform us whether learners accept or reject grammatical or ungrammatical sentences, and they do not provide us with information of whether the wh-word in the test sentence is interpreted as 19. The significance is all at p<0.001, except for the E-Post-beginners’, J-Intermediate’s and JPost-intermediate’s judgments of “moved obj-what + dou” vs. “* dou + obj-what in situ”, which are at p=0.002, p=0.003 and p=0.008, and except for the E-Beginners’, J-intermediate’s and JPost-intermediate’s judgments of “obj-who + dou” vs. “* dou + obj-who in situ”, which are at p=0.006, p=0.008 and p=0.026 respectively.



Wh-words as universal quantifiers 

a universal quantifier. To get information of this type, we have to look at the data in the sentence interpretation test, which tell us not only whether the wh-word is interpreted as a universal quantifier in L2 grammars when the wh-word is adjacent to the left of dou, but also whether such an interpretation depends on the presence of dou in the sentence in L2 Chinese grammars. As we can see in Table 6, the percentage of correctly interpreting the wh-word as a universal quantifier is very high in English and Japanese intermediate, postintermediate and advanced groups (86.9%-98.8%) when the wh-word used as a subject appears to the left of dou. Recall that in the acceptability judgment test, these English and Japanese groups correctly accept sentences with the wh-word in the subject position adjacent to the left of dou (see Tables 2 and 3). With the data in Table 6, we can confidently argue that when informants in these groups accept the grammatical sentences in the judgment, they indeed interpret the wh-word as a universal quantifier. Table 6 also shows that the percentage of the English and Japanese Beginner and Post-beginner groups’ correct interpretation of the wh-word as a universal quantifier is not very high (28.3% and 60.8% for the E-Beginners and E-Post-beginners, and 67.6% and 58.8% for the J-Beginners and J-Post-beginners), which seems to confirm the indeterminacy found in these groups’ performance in the acceptability judgement of sentences with the wh-word in the subject position adjacent to the left of dou (see Tables 2–3). Table 6.  Percentage of correct/incorrect interpretations of the wh-word (i.e., what or who) as a universal quantifier in subject position adjacent to the left of dou and *when the whword is in subject position without dou (i.e., Interpretation Types 1 and 2) subj-wh + dou E-Beginners E-Post-beginners E-Intermediate E-Post-intermediate E-Advanced J-Beginners J-Post-beginners J-Intermediate J-Post-intermediate J-Advanced Chinese

28.3% 60.8% 86.9% 96.0% 98.8% 67.6% 58.8% 89.3% 96.6% 97.7% 100%

*subj-wh (without dou) 3.3% 8.3% 7.1% 7.3% 2.4% 17.6% 26.5% 28.7% 17.3% 3.0% 1.7%

Note: “↔” = a significant difference between the two groups in an independent-samples t-test.



Boping Yuan

When dou is not present in the sentence, the percentage of the incorrect interpretation of the subject wh-word as a universal quantifier is low across all the groups, as shown in the third column of Table 6. This suggests that the interpretation of the whword as a universal quantifier does depend on the availability of dou, and this is particularly the case in post beginner stages of L2 Chinese acquisition. However, the table also shows that the Japanese groups, except for the J-Advanced group, make more errors than their corresponding English groups in incorrectly interpreting the whword as a universal quantifier in these ungrammatical sentences as the percentage of their incorrect interpretation is higher than that of their English counterparts. To find out whether any Japanese group is significantly less accurate than its corresponding English group in interpreting the wh-word as a universal quantifier, we use the number of correct and incorrect interpretations of the wh-word by the informants in each group as numerical data to conduct independent-samples t-tests.20 As indicated in the third column of Table 6, the J-Beginners, J-Post-beginners, J-Intermediate and J-Post-intermediate groups are found to have significantly more incorrect interpretations of the subject wh-word than their English counterparts,21 which suggests that the wh-word is more likely to be incorrectly interpreted as a universal quantifier in the absence of dou in these Japanese groups’ L2 Chinese grammars than in their English counterparts’.22 When interpreting sentences with the wh-word raised from the object position to the left of dou, the E-Intermediate, E-Post-intermediate as well as E-Advanced groups are quite accurate, as can be seen in the second column of Table 7, and the percentage of these groups’ correct interpretations of these grammatical sentences are 82.9%, 91.8% and 96.5%. This implicates that these groups’ L2 Chinese grammars are able to interpret the wh-word as a universal quantifier when it raises to the specifier of dou. In contrast, the Japanese groups’ interpretations of these grammatical sentences are rather inaccurate, except for the J-Advanced group. As we can see in the second column of Table 7, unlike the E-Intermediate and E-Post-intermediate groups, whose accuracy rates in correctly interpreting the raised object wh-word as 20. Recall that there are 6 tokens in each Interpretation Type. Therefore, 6 is the maximum score an informant can get for correctly or incorrectly interpreting the wh-word as a universal quantifier, and the minimum score is 0. 21. In interpreting “*subj-wh in situ (without dou)”, the significance level for the difference between E-Beginners and J-Beginners is at p=0.002, between E-Post-beginners and J-Post-beginners and between E-Intermediate and J-Intermediate are both at p<0.001, and between EPost-intermediate and J-Post-intermediate is at p=0.012. 22. It is not clear why the J-Beginners group is significantly more accurate than its English counterpart when the subject wh-word appears adjacent to the left of dou, i.e., in the grammatical “subj-wh + dou” type , as shown in the second column of Table 6.



Wh-words as universal quantifiers 

a universal quantifier are 82.9% and 91.8%, the J-Intermediate and J-Post-intermediate groups’ accuracy rates are only 39.8% and 56.6%. In independent-samples ttests, which again use the number of correct interpretations of the six tokens by the informants in each group as a dependent variable, the J-Intermediate and J-Postintermediate are found to be significantly less accurate than their English counterparts in interpreting the raised object wh-word as a universal quantifier. The same is also found in the groups’ interpretations of ungrammatical sentences where the object wh-word remains in situ and there is no dou present; as indicated in the third column of Table 7, the J-Post-beginners, J-Intermediate and J-Post-intermediate groups are found to have significantly more incorrect interpretations than their English counterparts of the object wh-word as a universal quantifier in ungrammatical sentences where the wh-word remains in situ and where dou is absent.23 Table 7 also shows that although Japanese learners at beginner, intermediate and even post-intermediate levels have problems in correctly interpreting the raised object wh-word as a universal quantifier, the J-Advanced group, like its English counterpart, can correctly interpret the raised object wh-word as a universal quantifier; its accuracy rate is 97.8%. The data in Table 7 implicate that in both English and Japanese speakers’ L2 Chinese grammars, there is a requirement that Table 7.  Percentage of correct/incorrect interpretation of the wh-word (i.e., what or who) as a universal quantifier when the wh-word raises from the object position to the left of dou and *when it remains in situ in object position in a sentence without dou (Interpretation Types 3–4) raised obj-wh + dou E-Beginners E-Post-beginners E-Intermediate E-Post-intermediate E-Advanced J-Beginners J-Post-beginners J-Intermediate J-Post-intermediate J-Advanced Chinese

16.6% 55.0% 82.9% 91.8% 96.5% 24.1% 41.4% 39.8% 56.6% 97.8% 100%

*obj-wh in situ (without dou) 15.6% 7.2% 6.3% 4.6% 3.8% 22% 28.7% 30.6% 20.1% 6.8% 2.2%

Notes: “↔” = a significant difference between the two groups in an independent-samples t-test.

23. The significance is all at p<0.001, except for the difference between E-Post-intermediate and J-Post-intermediate in “*obj-wh in situ (without dou)”, which is at p=0.003.



Boping Yuan

the wh-word must move to the specifier of dou in order to have its [udis] feature checked and valued. However, this requirement seems to become available in Japanese speakers’ L2 Chinese grammars much later than in English speakers’. 5. Discussion Let us have a summary of our findings: a. Data from English and Japanese groups indicate that in the end state of their L2 Chinese grammars, the functional category DisP can be projected, the whword with the uninterpretable [udis] feature can raise to the specifier of dou to have its [udis] feature checked and valued, and the wh-word can be interpreted as a universal quantifier, although these seem to take place significantly later in Japanese speakers’ L2 Chinese than in English speakers’; b. There seems to be an asymmetry in beginner, intermediate and post-intermediate Japanese learners’ L2 Chinese wh-words as universal quantifiers; unlike English speakers who have no major problems in acquiring either the subject wh-word or the object wh-word as a universal quantifier, Japanese learners at beginner, intermediate and even post-intermediate levels have problems accepting sentences with the wh-word raised from the object position to the specifier of dou, and they are also unable to accurately interpret the raised object wh-word as a universal quantifier. They do not seem to have major problems accepting and interpreting the subject wh-word as a universal quantifier though; c. Japanese speakers are found to make significantly more incorrect interpretations of the wh-word as a universal quantifier than English speakers when the wh-word is in an ungrammatical sentence without dou. The Failed Functional Features Hypothesis (Hawkins and Chan 1997) would predict that the functional category DisP headed by dou would not be projected in adult English and Japanese speakers’ L2 Chinese grammars and the Interpretability Hypothesis (Tsimpli 2003; Tsimpli and Dimitrakopoulou 2007; Hawkins and Hattori 2006) would predict that the uninterpretable [udis] feature is not acquirable in adult English or Japanese speakers’ L2 Chinese grammars because the functional category and the uninterpretable feature are not selected in the construction of their L1 English and Japanese grammars and have disappeared from their feature inventory of UG at the end of the critical period. They would predict that there should be a permanent loss of capacity to acquire in this domain. However, our data clearly show that learners from both language groups, particularly those in advanced and end state of L2 Chinese, are able to acquire the functional



Wh-words as universal quantifiers 

category DisP and the uninterpretable [udis] feature. In these groups’ L2 Chinese, the wh-word is attracted to the specifier of dou to have its uninterpretable [udis] feature checked and valued. In these learners’ L2 Chinese, the interpretation of the wh-word as a universal quantifier depends on the availability of dou, without which the wh-word cannot be interpreted as a universal quantifier in the sentence. Our data here suggest that there is no permanent representational deficit in this aspect of L2 Chinese grammars in spite of the fact that the functional category DisP and the uninterpretable [udis] feature are not selected in the learners’ L1 English and Japanese. However, our data do provide evidence for a non-permanent representational deficit in Japanese speakers’ L2 Chinese grammars. Japanese speakers at beginner, intermediate and even post-intermediate levels have problems in accepting Chinese sentences with the wh-word raised from the object position to the specifier of dou although they accept sentences with the subject wh-word to the left of dou. And when interpreting the wh-word appearing to the left of dou, they are also found less accurate in interpreting the object wh-word as a universal quantifier than the subject wh-word. Their judgment and interpretation of the object whword seem to be rather variable and indeterminate. This asymmetry suggests that what is available to the subject wh-word is different from what is available to the object wh-word in these Japanese speakers’ L2 Chinese grammars. We would argue that in these Japanese speakers’ L2 Chinese grammars, dou projects the DisP with an interpretable [dis] feature, and the wh-word, whether used as a subject universal quantifier or an object universal quantifier, has an uninterpretable [udis] feature. However, unlike the native Chinese grammar which has an asterisked [dis*] feature on dou which attracts the wh-word to the specifier of dou to have its [udis] feature checked and valued, the interpretable feature on dou in the Japanese speakers’ L2 Chinese grammars is a non-asterisked [dis] feature which is unable to trigger the raising of the wh-word to the specifier of dou for feature checking and valuing. As a result, the [udis] feature on the wh-word cannot be checked and valued, and the wh-word cannot be accurately interpreted as a universal quantifier. This can account for the problems that the Japanese speakers have with the object wh-word in their L2 Chinese. Obviously we have to provide an explanation as to why the subject wh-word is accepted by these Japanese speakers when it appears to the left of dou and why they can accurately interpret the subject wh-word as a universal quantifier. Let us continue to assume that dou has a non-asterisked [dis] feature in these Japanese speakers’ L2 Chinese grammars which is unable to trigger the raising of either the subject wh-word or the object wh-word to the specifier of dou for feature checking and valuing. However, we would argue that the target-like behaviour of the subject wh-word in these Japanese speakers’ L2 Chinese is due to the [uepp*] feature of TP



Boping Yuan

which requires the subject to raise to the specifier of TP. On the way of raising to the specifier of TP, the subject wh-word stops at the specifier of dou, where it has its [udis] feature checked and valued by the interpretable [dis] feature of dou. Metaphorically speaking, we can say that the [uepp*] feature of T provides the subject wh-word with a free ride which enables it to check and value its [udis] feature on its way to the specifier of TP, hence the target-like behaviour of the subject whword in these Japanese speakers’ L2 Chinese. It is well-known that there is no morphological marking for tense in Chinese. However, the absence of overt tense marking in Chinese does not necessarily implicate that there is no TP in the Chinese sentence. Many researchers (e.g., Ernst 1994; Gasde and Paul 1996; Huang 1982, 1984, 1987, 1989; A. Y.-H. Li 1985, 1990; Ting 1998; Tang 2000; among others) argue that there does exist a distinction between finiteness and non-finiteness in Chinese, hence the existence of TP in Chinese. According to these researchers, the overt tense marking is not the only way to identify finiteness in Chinese, and some syntactic phenomena would not be adequately accounted for if the distinction between finiteness and non-finiteness is not assumed in Chinese. An example frequently cited by these researchers in support of their argument is that in Chinese the subject position of a non-finite clause is obligatorily null and that only a finite clause can have a lexical NP in its subject position. This in fact is in conformity with our assumption here that the [uepp*] feature of TP requires that its specifier be filled. This requirement triggers the subject wh-word in the sentence to raise to the specifier of TP to check the [epp] feature. On its way to the specifier of TP, the subject wh-word passes through the specifier of DisP where it has its [udis] feature checked and valued by the [dis] feature of dou. It is not of much relevance here whether the TP and its [uepp*] feature in the Japanese speakers’ L2 Chinese grammars are transferred from their L1 or acquired in the process of L2 acquisition of Chinese. What is important is that it is the free ride made possible by the checking of the [uepp*] feature of T which enables the subject wh-word to have its [udis] feature checked and valued at the specifier of DisP in the L2 grammars of the Japanese beginner, intermediate and post-intermediate learners of Chinese. The diagram in (20) provides an illustration of the Japanese learners’ underlying representation. The asterisked feature [*] is missing on the [dis] of dou in their L2 Chinese grammars, which results in lack of trigger for the subject wh-word shei to raise to the specifier of DisP in order to check its uninterpretable [udis] feature. However, the specifier of dou provides a landing site for the subject wh-word shei on its way to the specifier of T to check its [uepp*] feature, and at this landing site, the [udis] feature of the subject wh-word shei is checked and valued by the interpretable [dis] feature of dou.



Wh-words as universal quantifiers 

(20) (In our school,) shei dou xihuan ta. TP Shei [EPP] T[uEPP*]

T’ DisP

<Shei[udis][EPP]> Dis

Dis’ vP

dou[dis] <Shei[udis][EPP]>

v’

v xihuan

VP V

NP

<xihuan> ta

The diagram in (20) provides an account for the target-like behaviour of the subject wh-word in the Japanese speakers’ L2 Chinese. However, it should be noted that the Japanese speakers’ underlying L2 Chinese representation is still deficient in this aspect of their L2 Chinese grammars. In the native Chinese grammar, the head of DisP, dou, has an asterisked feature [dis*], which attracts the wh-word to the specifier of DisP from the specifier of vP, and the [uepp*] of T then triggers the wh-word to raise further to the specifier of TP to check the [epp] feature. Unlike the native Chinese grammar, the Japanese speakers’ L2 Chinese grammars have a [dis] with the asterisked feature [*] missing. As a result, dou is unable to attract the wh-word to its specifier for feature checking and valuing. It is the [uepp*] of T which triggers the raising of the subject wh-word shei from the specifier of vP to the specifier of TP via the specifier of DisP, where the [udis] feature on the whword is valued and then gets deleted. This suggests that the checking and valuing of the subject wh-word is a result of a free ride in the Japanese speakers’ L2 Chinese. In this sense, the native-like behaviour of these learners is only apparent and their underlying representation concerning this part of their L2 Chinese grammars is still divergent from that of native Chinese speakers. As the object wh-word does not have this kind of free ride available to it, it remains in situ and its [udis] feature cannot be properly checked and valued, hence the non-target behaviour of the object wh-word in the Japanese speakers’ performance.

 Boping Yuan

Recall that in the final state of Japanese speakers’ L2 Chinese grammars, as represented by the Japanese Advanced Group, the asterisked [dis*] feature is acquired, which triggers both the subject wh-word and the object wh-word to raise to the specifier of DisP to have their [udis] feature checked and valued. However, it is not clear what positive evidence in the input data helps Japanese speakers’ L2 Chinese grammars acquire the asterisked [dis*] feature and why the positive evidence does not have much effect on Japanese learners’ earlier stages of L2 Chinese. We will leave these for future research. Another question one may ask is why the asymmetry found in Japanese beginner, intermediate and post-intermediate learners’ L2 Chinese is absent in English learners’ L2 Chinese. In English, wh-questions involve A’-movement which is triggered by the requirement for checking the [wh] feature. However, wh-movement is not required in Japanese, and when the wh-word does move to the sentence initial position in Japanese wh-questions, it is not obligatory and is due to wh-scrambling (cf. Kobayashi 2000, Bailyn 2001) rather than to wh-feature checking. In other words, the obligatory wh-feature checking in English wh-questions is likely to facilitate English learners’ L2 Chinese acquisition of the checking between the [dis*] feature on dou and the [udis] feature on the wh-word used as a universal quantifier. The data in our empirical study also show that beginner, intermediate and post-intermediate Japanese learners make significantly more incorrect interpretations of the Chinese wh-word as a universal quantifier than English learners when the wh-word is in an ungrammatical sentence without dou. We can attribute this to L1 transfer. Unlike Chinese where the wh-word has to undergo a syntactic feature-checking operation in order to be interpreted as a universal quantifier, the Japanese wh-word, when used as a universal quantifier, is lexically licensed by the particle –mo. It is likely that the Japanese learners sometimes treat the Chinese wh-universal-quantifier as a counterpart of the Japanese wh-universal-quantifier; when the Chinese wh-word appears in an ungrammatical sentence without dou, their L2 grammars will interpret it as a universal quantifier with the particle missing. As the English wh-word cannot be used as a universal quantifier and as it is generally used as an interrogative word in a wh-question and has to undergo a syntactic feature-checking operation, English learners are able to reject the ungrammatical sentences with the wh-word but without dou. In short, the possibility to use the lexically licensed wh-word as a universal quantifier in Japanese, but not in English, is likely to be the source of Japanese learners’ inaccurate interpretation of the Chinese wh-word as a universal quantifier when it appears in an ungrammatical sentence without dou.



Wh-words as universal quantifiers 

6. Conclusion Our data indicate that the functional category DisP and its [udis] feature can be properly represented in English and Japanese speakers’ L2 Chinese grammars even though they are not selected in their L1s. This suggests that there is no permanent loss of capacity to acquire, at least in this aspect of L2 Chinese grammars. However, we do find evidence for a non-permanent representational deficit in beginner, intermediate and post-intermediate Japanese learners’ L2 Chinese grammars. As we can see in (20) above, although the subject wh-word in these Japanese learners’ L2 Chinese is able to check and value its [udis] feature and can have an apparently target-like behaviour, the underlying representation of their L2 Chinese grammars is divergent from that of the native Chinese grammar as the interpretable feature of dou in their L2 Chinese grammars is a non-asterisked [dis] which cannot attract the wh-word in its c-commanding domain to its specifier for feature checking and valuing. However, it should be pointed out that unlike the phenomenon of a missing feature despite target-like appearance as suggested in Hawkins (2004), which is a result of an effort in L2 grammars to make local adjustments or use alternative resources to accommodate the input data of the target language, the apparent target-likeness found in our study is believed to be due to the “free ride” made possible by the requirement of checking the [epp] feature of TP, and it is coincidental in nature. Nevertheless, we believe that our study has provided data in support of Hawkins and Hattori’s call that “caution is required in interpreting apparent targetlike L2 performance as evidence for the acquisition of underlying properties of grammar assumed to be present in the grammars of native speakers.” (2006: 298) Acknowledgements The research reported in this paper is funded by a grant from the Arts and Humanities Research Council in the UK (the grant reference number: AH/E005322/1). I gratefully acknowledge the financial support for the research project from the AHRC. I am in debt to students and teaching staff at Waseda, Keio, Nihon and Osaka Universities in Japan, to Japanese, British and American students studying at Beijing Foreign Studies and Peking Universities in China, and to students and teaching staff of Oxford, London, Westminster, Leeds, Edinburgh and Cambridge Universities in the U.K for their participation in my empirical study. Without their help, this research project would have been impossible. I would thank Chieko Kuribara for providing me relevant Japanese sentences. Parts of the paper were presented at the 2007 annual conference of the British Chinese Language Teaching Society and at PaCSLRF 2008, and I am grateful to the audience for their questions

 Boping Yuan

and comments. I would also like to thank Ingrid Leung and Neal Snape and two anonymous reviewers for their valuable and helpful comments on earlier versions of the paper. Of course, all remaining errors are my own. References Adger, D. 2003. Core syntax: A minimalist approach. Oxford: OUP. Bailyn, J. F. 2001. On scrambling: A reply to Bošković and Takahashi. Linguistic Inquiry 32: 635–657. Cheng, L.-S. L. 1991. On the typology of Wh-questions. PhD dissertation, MIT. Cheng, L.-S. L.1995. On dou-quantification. Journal of East Asian Linguistics 4: 197–234. Chiu, B. 1990. A case of quantifier floating in Mandarin Chinese. Paper presented at Northeast Conference on Chinese Linguistics, Cornell University Chomsky, N. 1981. Lectures on government and binding. Dordrecht: Foris. Chomsky, N. 1995. The minimalist program. Cambridge MA: The MIT Press. Chomsky, N. 2001. Derivation by phase. In Ken Hale: A life in language, M. Kenstowicz, (ed.), 1–52, Cambridge MA: The MIT Press. Ernst, T. 1994. Functional categories and the Chinese Infl. Linguistics 32: 191–212. Gasde, H.-D. & Paul, W. 1996. Functional categories, topic prominence, and complex sentences in Mandarin Chinese. Linguistics 34: 263–294. Hawkins, R. 2003. ‘Representational deficit’ theories of (adult) SLA: Evidence, counterevidence and implications. Paper presented at the 13th EUROSLA Conference, University of Edinburgh. Hawkins, R. 2004. Revisit wh-movement: The Availability of an uninterpretable [wh] feature in interlanguage grammars. In Proceedings of the 7th Generative Approaches to Second Language Acquisition Conference (GASLA 2004), L. Dekydtspotter, R. Sprouse & A. Liljestrand (eds), 124–137, Somerville MA: Cascadilla Press. Hawkins, R. & Chan, C. 1997. The partial availability of Universal Grammar in second language acquisition: The ‘failed functional features hypothesis’. Second Language Research 13: 187–226. Hawkins, R. & Hattori, H. 2006. Interpretation of English multiple wh-questions by Japanese speakers: A missing uninterpretable feature account. Second Language Research 22: 269–301. Heim, I., Lasnik, H. & May, R. 1991. Reciprocity and plurality. Linguistic Inquiry 22: 63–102. Hsieh, R. 1995. Dou and universal quantification in Chinese. In Proceedings of 6th North American Conference on Chinese Linguistics, J. Camacho & L. Choueiri (eds), 85–99. University of Southern California: Graduate Students in Linguistics. Huang, C.-T. J. 1982. Logical relations in Chinese and the theory of grammar. PhD. dissertation, MIT. Huang, C.-T. J.1984. On the distribution and reference of empty pronouns. Linguistic Inquiry 15: 531–574. Huang, C.-T. J.1987. Remarks on empty categories. Linguistic Inquiry 18: 321–337. Huang, C.-T. J. 1989. Pro-drop in Chinese: A generalized control theory. In The Null Subject parameter, O. Jeaggli & K. Safir (eds.), 185–214. Dordrecht: Kluwer.



Wh-words as universal quantifiers  Kobayashi, A. 2000. The third position for a wh-phrase. Linguistic Analysis 30: 177–215. Lee, H.-T. T. 1986. Studies on quantification in Chinese. PhD. dissertation, University of California, Los Angeles. Li, J. 1995. Dou and wh-question in Mandarin Chinese. Journal of East Asian Linguistics 4: 313–323. Li, Y.-H. A. 1985. Abstract case in Chinese. PhD dissertation, University of Southern California. Li, Y.-H. A. 1990. Order and constituency in Mandarin Chinese. Dordrecht: Kluwer. Li, X. 1996. Deriving distributivity in Mandarin Chinese. PhD. dissertation, University of California, Irvine. Lin, J.-W. 1998. Distributivity in Chinese and its implications. Natural Language Semantics 6: 201–243. Lü, S.-X. 1980. Xiandai Hanyu Babaici. Beijing: Commercial Press. Nishigauchi, T. 1990. Quantification in the Theory of Grammar. Dordrecht: Kluwer. Smith, N.V. & Tsimpli, I.-M. 1995. The mind of a savant: Language learning and modularity. Oxford: Blackwell. Sportiche, D, 1988. A theory of floating quantifiers and its corollaries for constituent structure. Linguistic Inquiry 19: 425–449. Tang, T.-C. C. 2000. Finite and non-finite clauses in Chinese. Language and Linguistics 1: 191–214. Ting, J. 1998. Deriving the bei-construction in Mandarin Chinese. Journal of East Asian Linguistics 7: 319–354. Tsai, W.-T. D. 1994. On econmizing the theory of A-bar dependencies. PhD. dissertation, MIT. Tsai, W.-T. D. 1999. On lexical courtesy. Journal of East Asian Linguistics 8: 39–73. Tsimpli, I.-M. 2003. Features in language development. Paper presented at the 13th Eurosla Conference, University of Edinburgh. Tsimpli, I.-M. & Dimitrakopoulou, M. 2007. The interpretability hypothesis: Evidence from whinterrogatives in L2 acquisition. Second Language Research 23: 215–242. Watanabe, A. 1992. Subjacency and S-structure movement of wh-in-situ. Journal of East Asian Languages 1: 255–291. Watanabe, A. 2001. Wh-in-situ languages. In The handbook of contemporary syntactic theory, M. Baltin & C. Collins (eds.), 203–225. Oxford: Blackwell. Wu, J. 1999. Syntax and semantics of quantification in Chinese. PhD dissertation, University of Maryland, College Park. Xu, D. 1997. Functional categories in Mandarin chinese. The Hague: HAG. Yuan, B. 2007a. Superficial native-likeness and fossilization in Japanese speakers’ L2 Chinese wh-questions: A lexical morphological feature defect account. Second Language Research 23: 329–357. Yuan, B. 2007b. Behaviours of wh-words in English speakers’ L2 Chinese wh-questions: Evidence of no variability, temporary variability and persistent variability in L2 grammars. Bilingualism: Language and Cognition 10(3): 277–298. Yuan, B. 2007c. Discrepancy in English speakers’ L2 acquisition of Chinese wh-words as existential polarity words: The L1 dependent interface hypothesis. Paper presented at 2007 GASLA Conference, University of Iowa.

Acquisition of the local binding characteristics of English reflexives and the obligatory status of English objects by Chinese-speaking learners Lin Jiang

Guangdong University of Foreign Studies This study investigates whether Chinese speakers can acquire the local binding characteristics of English reflexives and the obligatory status of English objects. Chinese allows both the long-distance (LD) and the local binding of reflexives, whereas English only allows the local binding of reflexives. Similarly, Chinese allows both overt and null objects, whereas English only allows overt objects. Even though the two properties involve different types of antecedent-anaphora and operator-variable binding relations, both of them are likely to cause poverty of stimulus for Chinese learners of English. The results of the experiment suggest that the learners’ interpretation of binding of reflexives is more or less native-like because of the availability of the triggering data but the obligatory status of English objects is not acquirable because of the lack of the relevant triggering data.

1. Introduction Two differences between English and Chinese are that English reflexives require their antecedents to be within the same clause (i.e., local), as shown in (1), whereas the Chinese reflexive ziji (self) permits the antecedent to be in a different clause (i.e., LD) as well as in the same clause (i.e., local), as shown in (2).

(1) Johni thinks [Billj trusts himself *i/j].

(2) Zhangsani renwei [Wangwuj xiangxin ziji i/j]. Zhangsan think Wangwu trust self ‘Zhangsan thinks that Wangwu trusts him/himself.’ In addition, English does not allow null objects. However, null objects occur rather freely in Chinese, which can be seen in (3) and (4).

 Lin Jiang



(3) Mary’s bikei has been stolen. *The police found Øi yesterday.

(4) A: Shei kanjian le Zhangsani? Who see ASP Zhangsan ‘Who saw Zhangsan?’ B: [Top øi] Lisi shuo Wangwu kanjian le ei Lisi say Wangwu see ASP ‘Lisi says that Wangwu saw Zhangsan.’ Notice that reflexives may only be bound by antecedents in an argument (A) position, not in a non-argument (A’) position, as can be seen by comparing (5) and (6):

(5) Theyi took pictures of themselvesi.



(6) *Which boysi did pictures of themselvesi convince Mary that she shouldn’t date ti. (Harbert 1995: 235)

On the other hand, null objects in Chinese cannot be bound by a c-commanding nominal occurring in an A position. As illustrated in (4), the null object is not bound by Lisi or Wangwu. Rather, it is bound by a null topic in an A’ position. The relationship between a topic and its variable is the A’ equivalent of the relationship between an argument and a reflexive that it binds (i.e. A’-binding vs. A-binding). Hence, investigating these two linguistic properties (reflexive interpretation vs. object realization) together allows one to determine whether learners can establish binding properties of a second language (L2) which are different from the first language (L1), in both A-binding and A’-binding domains. As far as reflexives are concerned, English input may provide ample evidence showing that local binding is possible, but this by no means suggests that LD binding is impossible in English. For example, a Chinese speaker may encounter English sentences like:

(7) Mary talked about herself.



(8) Lisa wants Mary to paint herself.



(9) Lisa believed that Mary was painting herself.

where Mary can be an antecedent for herself, but there is nothing to tell that Chinese speaker that in sentences like the following Mary cannot be an antecedent for herself: (10) * Mary liked John’s pictures of herself. (11) * Mary wanted Tom to like herself. Cases like (10)-(11) could only be determined as impossible if a learner already knows that English reflexives can only be locally bound. Furthermore, it is quite



Reflexive interpretation and object realization 

rare that learners are explicitly told or taught about the locality constraints on the binding of English reflexives. The impossibility of null objects in English is also underdetermined by input for native speakers of Chinese in the following sense. A Chinese speaker will not hear a native speaker of English using null objects in cases like Mary’s bike has been stolen. *The police found yesterday. A native speaker of English uses the pronoun it in cases like these. However, what this kind of fact triggers and demonstrates to a Chinese speaker may be merely that in English objects can be overt; they are possible just like overt objects in their L1, and this does not imply that null objects could not exist in English. Furthermore, some verbs that are traditionally classified as transitive can regularly appear without an overt object, such as read, eat, write, draw, drink, hit, cut, build, etc (Yip and Matthews 2007 for a discussion of null objects and the learnability issue). (12) I bought a book but did not read. (=did not read anything) (13) I bought a book but did not read it. (=did not read that book but could have read something else) Use of read without an object in (12) is a meaningful, interpretable choice for native English speakers (i.e., these are interpreted as activities, without a link to the potential referent). However, for Chinese learners of English, this could be an ambiguous structure, referring to the activities alone, or involving specific reference to the potential antecedent, as would be the case in their L1, for example: (14) Wo mai le yi ben shui dan meiyou du ei I buy ASP one CL book but not read I bought a book but did not read it. So, these types of sentences might encourage a Chinese speaker to assume that English allows null objects more generally. As noted above, both of these linguistic properties are likely to cause a poverty of stimulus for Chinese learners of English. No previous study has considered both knowledge of reflexive binding and knowledge of the (im)possibility of null objects in the same population of advanced L2 learners of English. Previous studies of reflexive binding have typically involved L2 speakers with lower intermediate or intermediate proficiency. Although there have been studies of null objects in advanced speakers, these have not simultaneously addressed reflexive binding. The present study aims to compare Chinese learners’ interpretation of binding of English reflexives and their knowledge of the ungrammaticality of null objects in English. The research question that will be asked is: whether the grammar of Chinese advanced learners of English might converge on the grammar of native

 Lin Jiang

speakers, being identical in both respects, or diverge to a greater or lesser extent. This question is closely linked to the issues of the extent to which L1 influences L2 development, and whether Universal Grammar (UG) is available (and, if so, to what an extent) in second language acquisition (SLA). According to the Representational Deficit Hypothesis (RDH; Hawkins 2003), only features and categories instantiated in the L1 are available to the L2 learner, and syntactic features absent from the L1 will not be acquirable. As we will see in Section 2, tense is implicated for reflexives and a [Top] feature for objects. The [Top] feature is available in the L1 Chinese while the [Tense] feature is not. According to the RDH, Chinese learners would be predicted to acquire the obligatory status of English objects but not the local binding characteristics of English reflexives. As part of the purpose of this study, the RDH will be tested, in order to see whether this prediction holds. I begin with the theoretical assumptions on both linguistic properties, followed by a review of the literature on the interpretation of reflexives and the acquisition of the ungrammaticality of null objects in SLA of English. Next, I describe the methodology. After presenting the results, I return to the research question mentioned above. 2. Theoretical accounts 2.1

Reflexive binding

According to Chomsky (1981), anaphors, which include reflexives and reciprocals, are subject to Principle A of the Binding Theory, which states that an anaphor must be bound in its binding domain. Principle A is an invariant property of human language. However, there is considerable cross-linguistic variation in the domains in which anaphors must be bound. In English, the reflexives like himself, herself can be bound up to, but not beyond, a local subject; binding to subjects or nonsubjects is allowed. (15) Johni said that [Billj gave Tomk a picture of himself *i/j/k]. In Russian, the reflexive sebja (self) can be bound across the PRO subject of a nonfinite clause, but not across the nominative subject of a finite clause (examples from Bailyn 1992); it requires the antecedent to be a subject. (16) Sašai poprosila Marinuj [PROj narisovat’ sebjai/j]. Sasha requested Marina-ACC draw self ‘Sasha asked Marina to draw her/herself.’



Reflexive interpretation and object realization 

(17) Sašai preset [čtoby Marinaj narisovala sebja*i/j]. Sasha requests that Marina draw-PAST self ‘Sasha requests that Marina draw herself.’ In Chinese, the reflexive ziji can even be bound across the nominative subject of a finite clause, and shows strict subject orientation.1 (18) Zhangsani gaosu Lisij [Wangwuk xiangxin ziji i/*j/k]. Zhangsan tell Lisi Wangwu trust self ‘Zhangsan tells Lisi that Wangwu trusts him/himself.’ In addition to the LD reflexive ziji, Chinese has a local reflexive taziji (himself/ herself), which shares the same morphological structure and behaviour as English reflexives. (19) Zhangsani gei Lisij yi zhang tazijii/j de zhaopian. Zhangsan give Lisi one CL himself DE photograph ‘Zhangsan gives Lisi a photograph of himself.’ There are many proposals in the literature that are concerned with capturing the facts of the syntactic binding of reflexives, for example, the parameterized approach (Manzini and Wexler 1987; Wexler and Manzini 1987), the Move-to-INFL approach (e.g., Battistella 1989; Pica 1987), and the Relativized subject approach (Progovac 1992; 1993). In particular, the discussion focuses on the Move-to-INFL approach, which is adopted as the framework for the present study. According to the Move-to-INFL approach (e.g., Battistella 1989; Pica 1987), anaphors are defective and thus must move at Logical Form in order to get licensed. Monomorphemic reflexives like ziji in Chinese and sebja in Russian are heads and do not have features of person, number and gender. Hence, they can raise out of VP into INFL of the same clause. After the monomorphemic reflexive raises into INFL, it will be c-commanded by the subject NP only, but not object, explaining why monomorphemic reflexives cannot be bound by any nonsubject antecedent. We also assume, following Reinhart and Reuland (1991), that head movement of reflexives involves incorporation. When the reflexive head X adjoins to another head Y, excorporation is blocked. As a result, subsequent movement of 1. Although tense and agreement are not systematically marked in Chinese, there are still ways to make a distinction between finite and nonfinite clauses in the language. Huang (1984, 1987, 1989, 1998 [1982]) argues that Chinese uses AUX to encode finiteness, the occurrence of a lexical subject in the language is systematically licensed by an AUX even though AUX may not always have its overt realization. According to Huang, there are mainly two types of verbs in Chinese: one including verbs like shuo (say) and xiangxin (believe), which can be followed by finite clauses, and the other including the so-called “control verbs” like zhunbei (prepare), shefa (try), quan (persuade), and bi (force), which can only be followed by nonfinite clauses.

 Lin Jiang

X requires the pied piping of Y. Movement of Y is possible only when the position of Y is not crucial for the interpretation of sentence. The movement of tensed INFL from its own clause would be prohibited because it would result in a structure in which tense would have scope over a higher clause. Thus, the interclausal movement of monomorphemic reflexives would be possible only in (a) languages lacking tensed INFL and (b) nonfinite structures. In Chinese, the total absence of tense morphemes enables the interclausal movement of ziji, which decides the property of LD binding of ziji. In Russian, tense is realized morphologically in finite clauses and therefore prohibiting the interclausal movement of sebja. Consequently, sebja must be bound locally in a finite clause. In a nonfinite clause, tense is morphologically null, thus, enabling the interclausal movement of sebja. In consequence, sebja can have LD binding out of nonfinite clauses. Polymorphemic reflexives like himself in English and taziji in Chinese, on the other hand, are maximal projections, which can adjoin only to the nearest maximal projection, namely the VP in which they originate. There they remain in the binding domain of either a local subject or a local object. Thus, the polymorphemic reflexives will always require local antecedents, which are not necessarily subjects. To summarize thus far, the cross-linguistic variation in the grammars of reflexives can be explained in terms of two parameters, a reflexive parameter (following MacLaughlin 1998) and a tense parameter: (20) a. Reflexive parameter: a reflexive is monomorphemic or polymorphemic. b. Tense parameter: tense is null or morphological. The reflexive parameter distinguishes between different types of reflexives. The tense parameter, on the other hand, distinguishes between null or overt tense morphology. These two parameters yield the three possible binding patterns, as shown in Table 1. Type I is the binding pattern in which only local binding is allowed, as in English (himself) and Chinese (taziji); Type II is the binding pattern in which LD binding is permitted out of nonfinite clauses only, as in Russian (sebja); Type III is the binding pattern in which LD binding is permitted out of finite clauses, as in Chinese (ziji). Table 1.  Three types of binding patterns

Type I Type II Type III

Reflexive

Tense

Example

poly poly mono mono

overt null overt null

English himself Chinese taziji Russian sebja Chinese ziji



Reflexive interpretation and object realization 

2.2

Null objects

Some linguists argue that the possibility of null objects in Chinese is directly connected with the widespread presence of topic constructions in this language. Topic constructions are certainly possible in English, for example: (21) Those musselsi, I wouldn’t touch ei with a barge pole! (Hawkins 2001: 211) Topics are assumed to locate in the specifier position of CP (Rizzi 1997), where the head C bears an uninterpretable topic feature [uTop] which checks with a constituent with an interpretable [Top] feature under c-command. For illustrative purposes, assume that the constituent the head C c-commands with [Top] feature is the object. Since [uTop] on C is strong, the object will have to move into the specifier of CP to get the feature checked against the [uTop] feature in C, as depicted by the tree diagram in (22) below (the position from which the movement took place is marked by enclosing a copy of the moved element in brackets). (22) CP

CP

DP [Top] Zhangsan Zhangsan

C' C [uTop]

IP NP Lisi Lisi

I' I

VP kanjian guo see ASP

(Zhangsan) (Zhangsan)

‘As for Zhangsan, Lisi saw him’

English shares with Chinese the property of having a strong optional [uTop] feature in C, which motivates the object to move to the Spec of CP for feature checking purposes. What, then, makes English different from Chinese with respect to objectdrop is that Chinese C licenses a null specifier, whereas, English does not. Chinese C licensing a null specifier means that topic-drop is possible, which decides the possibility of null objects in this language. The topic-drop is made possible by the feature of discourse-orientation in Chinese, in which a topic NP can be deleted if it is identified with a topic in the preceding sentence. According to Tsao (1977), this is a Topic NP Deletion Rule. The result of such a deleting process is formally a topic

 Lin Jiang

chain in which a topic extends its domain to a sequence of several CPs. English, in contrast to Chinese, is assumed to disallow a null specifier, that is, topic-drop is not allowed, which determines the impossibility of null objects in this language. To summarize, the difference between English and Chinese with respect to the possibility of null objects can be captured by the Topic-drop parameter (e.g., Huang 1984; Yuan 1997); Chinese has a setting of [+ topic-drop] and English [– topic-drop].2 3. Previous research on reflexive binding and null objects in SLA of English There has been considerable research which investigates the acquisition of reflexives in L2 English (e.g., Akiyama 2002; Cook 1990; Eckman 1994; Finer 1991; Finer and Broselow 1986; Hirakawa 1990; MacLaughlin 1998; Thomas 1989; 1991; Wakabayashi 1996; Wells 1998; White 1995; White et al. 1997; Yip and Tang 1998). Several distinct findings emerged from those studies. 1. Some learners showed only local binding in the L2 English. 2. Some learners showed LD and local binding in the L2 English. 3. A significantly higher percentage of LD binding occurs in nonfinite clauses than in finite clauses. As for these findings, most researchers have claimed that UG is available and resetting of parameters is possible in L2 acquisition (e.g., Hirakawa 1990), while some have argued that the L2 binding results can be accounted for solely in terms of transfer from the native language and thus do not provide any evidence for parameter resetting or access to UG (e.g., Yuan 1994). Some have further argued that experimental data of the acquisition of English reflexives suggest that UG does not 2. An anonymous reviewer pointed out that ‘matching of an interpretable [Top] feature and an uninterpretable one on C’ cannot account for the multi-topic drop phenomenon in Chinese, as in (i). (i) A: Ni du guo zheben shu You read ASP this book ‘Have you read this book?’

ma? Q?

B: e du guo e [topØi Øj [ei read ASP ej]] ‘I have read it’ We assume, following Yuan (1997), that absence of agreement inflections in INFL in Chinese licenses pro in the specifier of IP, rather than a trace. Thus the status of null subjects and null objects is different in Chinese: subjects are pro, objects are traces. This proposal also means that sentences like (i) don’t involve multi-topic drop.



Reflexive interpretation and object realization 

operate in adult L2 acquisition and the nature of the L2 learner’s behavior is no more than general problem-solving (e.g., Schachter 1996). As a matter of fact, it is difficult to draw firm conclusions about how the performance data in the case of reflexive binding might relate to underlying knowledge of binding and to what extent the ability to reset parameters is involved. This is because of (a) methods of collection of experimental data: in early studies which used a multiple-choice task or a picture-identification task to test L2 learners’ knowledge of reflexive binding, researchers have found it extremely difficult to establish whether their responses reflect the full range of their competence rather than preferred interpretations of an anaphor in a given context (e.g., Akiyama 2002; Wakabayashi 1996; White et al. 1997); (b) methods of analysis of experimental data: the percentage of the responses for each sentence type aggregated by groups of subjects has been examined in most of the previous studies, however, this kind of data may not provide a precise description of interlanguage grammars (ILGs), although they may reveal some general tendencies of L2 acquisition (e.g., Akiyama 2002; Wakabayashi 1996); (c) subject pool: previous studies of reflexive binding have typically involved L2 speakers with lower intermediate or intermediate proficiency, hence it is difficult to determine their ability to reset parameters. These problems were kept in mind when the test for the present study was devised. In contrast with the number of studies on reflexives in SLA of English, not much L2 research has been done on null objects: Gundel and Tarone (1992) document the use of null objects by L1 Spanish speakers; Fuller and Gundel (1987), Zobl (1994) and Yuan (1997) report that Chinese learners of English allow null objects at relatively high rates in both written tasks of acceptability and oral production tasks (in contrast to more accurate performance with null subjects); Park (2004) confirmed the same pattern for L1 Korean speakers. The study by Yuan (1997) was the first systematic study of the null object phenomenon in L2 English. In order to account for the difficulty in unlearning of null objects by Chinese learners of English, Yuan adopted Huang’s (1984) proposal on the topic chain in his explanation of null objects along the following lines: Chinese is a [+ topicdrop] language and the null object is a result of moving the object to the topic position and then having it deleted by the Topic NP Deletion Rule. English is a [– topic-drop] language, and the object can be topicalized and appear in the sentence initial position, but it cannot be dropped, so null objects are not allowed is this language. The difficulty the Chinese learners had in the unlearning of null objects lies in the fact that they were unable to abandon the [+ topic-drop] setting of their L1, possibly due to the lack of positive evidence in the input data to help them be aware of the [– topic-drop] setting of English.

 Lin Jiang

4. The present study 4.1

Subjects

The subjects were 48 adult EFL learners and 8 native-speaker English language teachers. The learners were studying for bachelor degrees in English at a large public university in southern China. They all spoke Chinese as their L1. Based on the results of their language placement test (Quick Placement Test),3 they were considered to be advanced EFL learners. There were 38 females and 10 males and their ages ranged from 20 to 23 years (M=21.7). They had started learning English at or after the age of 11, with an average period of learning English of 10.1 years. 4.2

Methodology

The Chinese learners’ interpretation of English reflexives and their knowledge of the ungrammaticality of null objects in English were tested separately. A storybased truth-value judgement task based on that in White et al. (1997) was developed for this study to test Chinese speakers’ knowledge of reflexive binding in L2 English. The individual test items consisted of a context in the form of a story in Chinese and two comments about the story in English. The controls were tested on the same stories and comments in English, as shown in (23). (23) Lisa just broke up with her boyfriend. She was down in the dumps, so she had a crew cut. Now she looks like her brother. Lisa thinks that her mother will mistake her for her brother. Comments: A. Lisa thinks that her mother will not recognize herself. (True/ False) B. Lisa was lovelorn. (True/ False) Of the two comments for each story, one was a stimulus sentence containing a reflexive, the other was a distractor. Four types of stimulus sentences were used in this experiment, as illustrated in Table 2. There were four tokens for each sentence type. Types 1 and 2 investigate whether subjects allow coreference between himself/herself and a local subject; Types 3 and 4 between himself/herself and a LD subject. The participants read the stories and had to decide for each subsequent comment whether it was a natural sentence that followed appropriately from the story by circling either ‘True’ or ‘False’ printed at the end of the comment. They were told 3. The Quick Placement Test (Oxford University Press 2001) is a multiple-choice task, covering a wide range of morpho-syntactic, lexical and pragmatic properties in English. The maximum score for the test is 60. Scores in the range of 48-60 are equivalent to ‘advanced proficiency’.



Reflexive interpretation and object realization 

Table 2.  Sentence types used to test EFL learners’ interpretation of reflexives Type 1: Biclausal sentences with finite embedded clauses; context suggests binding to the local subject (true): Mary hoped that Janei would paint herselfi Type 2: Biclausal sentences with nonfinite embedded clauses; context suggests binding to the local subject of the infinitive (true): Mary asked Janei to paint herselfi Type 3: Biclausal sentences with finite embedded clauses; context suggests binding to the LD subject (false): *Maryi hoped that Jane would paint herselfi Type 4: Biclausal sentences with nonfinite embedded clauses; context suggests binding to the LD subject (false): *Maryi asked Jane to paint herselfi

that sometimes in the test both comments would follow naturally, but sometimes only one or even none of the comments would follow from the story. The task consisted of 36 stories; 16 stories (those testing the sentence types 1 to 4) related to the issues being investigated here, and are randomized. The order of the comments is randomized across test items. The participants were given as much time as needed to complete the task. We presented the context in Chinese and the test comments in English for the following reasons. First, presenting the context in Chinese ensures that all of the subjects in the experimental group clearly and unambiguously understand the context, hence reducing the possibility that choices of answers result from misunderstandings of stories. Second, presenting only the test comment in the target language (English) eliminates any possible attempts by the subjects to somehow use the surface grammatical form of the stories as a source of aid in judging the comments. Therefore, Chinese contexts seem preferable. After the truth-value judgment task had been completed, a grammaticality judgment task was administrated to test whether Chinese learners of English were aware that English does not allow null objects. The task consisted of 24 target sentences and 24 distractors, with an even number of grammatical and ungrammatical sentences. Participants were asked to evaluate the sentences on a three-point scale (correct, incorrect or ‘not sure’) and correct the perceived errors in sentences that they judged to be incorrect. The participants were given as much time as needed to complete the task. The target sentences included two types of structures: coordinate structures and short question-answer dialogues. The ungrammatical target sentences all involved null objects with specific referents. Examples of grammatical and ungrammatical target sentences are given in Table 3.

 Lin Jiang

Table 3.  Grammaticality judgment task target structures

Coordinate structures (n=12)

– Grammatical sentences (n=6) Mary lost her bike last week, but her boy says the police have found it for her. – Ungrammatical sentences (n=6) *I want to borrow that book from her, but she says she is using at the moment.

Question-answer dialogues (n=12)

– Grammatical sentences (n=6) A: What did you do with the apples? B: I put them in the fridge. – Ungrammatical sentences (n=6) A: Can I use your computer? B: *The computer has gone wrong and you won’t be able to use now.

5. Results 5.1

Reflexive binding

One point was given when a participant answered “True” to a sentence where the context set up the local subject as the appropriate antecedent or “False” to a sentence where the context set up the LD subject as the appropriate antecedent (impossible in English). No point was given for incorrect answers. Table 4 presents the mean accuracy scores, by sentence type (Maximum 4), for EFL learners and controls. For Type 1 sentences, the learner group did not differ significantly from the controls; For Type 2 sentences, again no significant differences were found between the two groups. The learner group’s performance on Type 1 versus Type 2 sentences showed no significant differences either. For Type 3 sentences, the differences between the groups were not significant. For Type 4 sentences, the differences between the groups were again not significant. The learner group’s performance on the Type 3 versus Type 4 sentences showed no significant differences. Table 4.  Mean accuracy scores of the EFL learners and controls (SD in parentheses)

Type 1 (local/finite) Type 2 (local/nonfinite) Type 3 (LD/finite) Type 4 (LD/nonfinite)

EFL learners

Native controls

3.92 (0.28) 3.88 (0.33) 3.60 (0.54) 3.52 (0.58)

4.00 (0) 4.00 (0) 3.88 (0.35) 3.88 (0.35)



Reflexive interpretation and object realization 

Table 5.  Classification of respondents according to the consistency criterion

EFL learners (n = 48) Native controls (n = 8)

Type I (English)

Type II (Russian)

Type III (Chinese)

Inconsistent judgments

46 96% 8 100%

0

0

0

0

2 4% 0

The analysis of individual data was based on the consistency of responses each subject exhibited. Following MacLaughlin (1998), we defined consistency as 3 or 4 acceptances or rejections out of 4. We first evaluated the subjects’ response patterns to each sentence type separately, then combined the results in order to classify the individual subject as exhibiting either a particular type of binding (Type I, Type II, or Type III) or being inconsistent in their judgments. For example, a subject would be classified as exhibiting the binding pattern of Type I, if he or she accepted local binding at the “3 or 4 out of 4” criterion to Types 1 and 2 sentences respectively, meanwhile rejected LD binding at the “3 or 4 out of 4” criterion to Types 3 and 4 sentences respectively. As illustrated in Table 5, a majority of the EFL learners (46 out of 48) showed the binding pattern of Type I; no EFL learners exhibited a Type II or a Type III binding pattern; two EFL learners showed inconsistency in their judgments. 5.2

Null objects

One point was given when a subject correctly accepted a grammatical sentence or rejected an ungrammatical sentence with the appropriate correction. No point was given for the ‘not sure’ responses or incorrect responses. Table 6 presents the mean accuracy scores, by sentence type (Maximum 6), for EFL learners and native controls. The EFL learners, like the native controls, readily accepted the grammatical sentences in both the coordinate structures and question-answer dialogues. However, in judging the ungrammatical sentences, the EFL learners failed to detect the ungrammaticality of English sentences with null objects in coordinate structures, t (54) = 9.805, p<.000, or in question-answer dialogues, t (54) = 9.212, p<.000. To sum up the results in terms of reflexive binding and null objects, the Chinese EFL learners gave native-like responses in interpretation of binding of English reflexives, but have typically not recognized the ungrammaticality of English sentences involving null objects.

 Lin Jiang

Table 6.  Mean accuracy scores of the EFL learners and controls (SD in parentheses) Coordinate structures

EFL learners Native controls

Question-answer dialogues

Grammatical items

Ungrammatical items

Grammatical items

Ungrammatical items

5.60 (0.54) 5.88 (0.35)

2.88 (0.76) 5.63 (0.52)

5.56 (0.58) 5.75 (0.46)

2.75 (0.81) 5.50 (0.54)

6. Discussion The results from this study provided clear evidence that it was much easier for the Chinese speakers to acquire the local binding characteristics of English reflexives than the obligatory status of objects in English, even though Chinese differs from English in both cases. In what follows, two possibilities are proposed to account for this asymmetry in the L2 acquisition of English by Chinese speakers. 6.1

Native language transfer in L2 acquisition

For a speaker whose native language allows LD binding and null objects, the acquisition of an L2 allowing only local binding and overt objects may present a logical problem, because the learner presumably progresses from a more permissive to a more restricted grammar. However, L2 learners might have a way out of this dilemma. One possibility is that they might take advantage of certain similarities between their native language and the L2. Therefore, the first proposal to account for the asymmetry in question is derived from the presence of certain ‘similarities’ between English and Chinese in the case of reflexive binding. Given that the polymorphemic reflexive taziji in Chinese has similar structure and properties as its English counterparts himself/herself, successful acquisition of English local binding by Chinese speakers can be accounted for by assuming that target knowledge of locality restrictions is transferred from the learners’ native language. In fact, this line of analysis is pursued by Yuan (1994) and Lakshmanan and Teranishi (1994), who argued that successful acquisition of L2 binding properties might be explained in terms of similar properties instantiated in the learners’ native lan-



Reflexive interpretation and object realization 

guage.4 The antecedent for taziji is limited to [3rd person, singular, masculine/ feminine, human], but the antecedent of ziji can be any person, number, and gender as long as it is human, so ziji can often replace taziji in sentences with a masculine/feminine, third-person singular antecedent. Otherwise the sentence with taziji may sound very unnatural. In the Current Mandarin Chinese Corpus,5 ziji is much more frequent than taziji of both types: ziji occurs 223298 times, which accounted for 86.1% of all reflexives used; taziji (himself) occurs 12286 times, constituting 4.7%; taziji (herself) 4153 times, constituting 1.6%. Since the polymorphemic reflexive taziji has weaker influence on Chinese speakers, the syntactic properties of the LD reflexive ziji might interfere with their acquisition of English reflexives, making them incorrectly assume that English reflexives serve a similar function as the ziji in Chinese. This line of analysis accords well with Ying (1999), who argued that the high frequency usage of zibun (self) would lead Japanesespeaking learners of English to initially assume that English has a monomorphemic reflexive even though Japanese also has polymorphemic reflexives like karezisin (himself). As a result, the transfer proposal seems insufficient to explain the asymmetry in question. 6.2

A new approach involving an interaction of the L1 grammar and the L2 input

L2 speakers of a language must parse (or process) the input; that is, ‘they must assign a structure to each utterance’ (White 2003: 153). Following the Full Transfer/ Full Access hypothesis (Schwartz and Sprouse 1994, 1996), it will be assumed that the L2 learners adopt the L1 grammar as the initial ILG. Hence, they attempt to parse the input on the basis of the existing L1 grammar. ‘If the parse is unsuccessful, or if the parse suggests the need for an analysis inconsistent with the current grammar, this signals that the grammar is in some sense inadequate, motivating 4. Yip and Tang (1998) investigated the positive transfer hypothesis of Yuan (1994) and Lakshmanan and Teranishi (1994) in L2 acquisition of English reflexives by native speakers of Cantonese. Like Chinese, Cantonese has both a LD reflexive zigei (self) and a local reflexive keoizigei (himself/herself). The results showed that the subjects’ judgments on the L1 did not match the L2 data; that is, of those who consistently observed local binding in English, many allowed Cantonese keoizigei to be LD-bound. The data seemed to undermine the positive transfer hypothesis, according to which Cantonese learners were predicted to acquire English local binding simply by reanalyzing the target reflexives as having the same properties as their respective native language forms. In fact, those Cantonese learners must have set English and Cantonese polymorphemic reflexives apart and cannot simply have treated them as equivalents. 5. The Current Mandarin Chinese Corpus is an on-line corpus compiled by Center for Chinese Linguistics at Peking University. The Corpus consists of about 264,444,436 words drawn from both written and spoken current Mandarin Chinese.

 Lin Jiang

restructuring’ (White 2003:  153). In the current study, the Chinese learners of English parse the input on the basis of the Chinese grammar. The Chinese reflexive ziji allows LD as well as local binding and requires the antecedent to be a subject, as illustrated in the early example (18), whereas the English reflexive allows only local binding and lacks strict subject orientation. Since LD binding subsumes local binding and is fully consistent with it, the Chinese speakers’ current grammar is able to parse the English sentences involving local binding to subjects, consequently, no parsing failure occurs at this point. However, when their ILG is faced with English sentences like (24), an inconsistency between the linguistic input and the current grammar will arise. (24) John talked to Amyi about herselfi. Sentences like (24) indicate that English reflexives allow binding to nonsubjects. However, the Chinese speakers’ current grammar requires the antecedent to be a subject. Once the learners detect this mismatch, restructuring may be initiated. In fact, this line of analysis is pursed by many researchers (Berwick and Weinberg 1984; Carroll 2001; Gibson and Wexler 1994; Schwartz and Sprouse 1994, 1996; White 1987), who have proposed that grammar change in L2 acquisition, as in L1 acquisition, is driven by parsing failure and have further argued that structural properties of the L2 input act as cues to parameter setting. According to this proposal, a function of the input data is to fix one of the possible settings and some property of the input triggers a particular setting. In the context of the L2 acquisition of English local binding, what properties in the input could potentially trigger local binding for Chinese speakers? A number of possibilities suggest themselves. Firstly, the L2 input provides fairly transparent evidence about the morphological form of the English reflexive. The English reflexive has the structure as: a (genitive) pronoun + self (although there are irregularities, e.g., himself (not hisself). Occurrences such as her own self (Progovac and Connell 1991) and self-derived words (self-help, self-centered, self-image) provide evidence that self in English is a distinct morpheme. Hence, the polymorphemic nature of English reflexives is evident in their morphological composition. Once the learners recognize that the English reflexive is morphologically complex, then the English reflexive in their ILGs can adjoin only to the nearest maximal projection so as to comply with the universal principle (namely, the Move-to-INFL approach). There the reflexive remains in the binding domain of either a local subject or a local object. Consequently, the Chinese learners of English would realize that LD binding is prohibited. Secondly, each reflexive type has a cluster of properties associated with it: monomorphemic reflexives allow LD antecedents (as well as local), and require the antecedent to be a subject, e.g., ziji in Chinese; polymorphemic reflexives require local antecedents and allow both subjects and nonsubjects as antecedents,



Reflexive interpretation and object realization 

e.g., himself, herself in English. In general, then, current accounts of reflexives agree on the following cluster of properties of UG: (25) a. LD reflexives must be subject-oriented. b. Reflexives which allow nonsubject antecedents must be local. In the present study, observation of positive data in the input (e.g., John talked to Amyi about herselfi.) can, of course, inform the learner that English allows binding to nonsubjects. According to the cluster of binding properties of UG, nonsubject binding can inform the Chinese learners that the English reflexive must be local. Hence, LD binding would be prohibited in their ILGs.6 Thirdly, the verbal inflections realising tense (e.g. –ed, and the forms had, was, were) might help Chinese speakers become aware that tense in English is morphologically overt, hence prohibiting the interclausal movement of the English reflexive. Consequently, the English reflexive must be bound locally in Chinese learners’ L2 grammar of English. In summary, ‘if parameter setting depends on predetermined, structural cues, then, where such cues are available in the positive L2 input, parameter resetting should in principle be possible’ (White 2003: 166). As we have seen, in the case of the acquisition of English local binding by Chinese speakers, there have been parsing inconsistencies between an analysis based on the L1 grammar and properties of the L2 input, which suggest the need for parameter resetting. Evidence in the input in the form of the polymorphemic nature of English reflexives, nonsubject binding, and overt tense morphology triggers an awareness of local binding of English reflexives. However, the situation regarding English obligatory overt objects in the present study is quite different from the situation regarding reflexives, where results suggest problems with parameter resetting. We turn now to the issue that Chinese advanced learners of English had difficulty recognizing the ungrammaticality of English sentences involving null objects. Chinese has the setting of [+ topic-drop] allowing objects to be null as well as overt, whereas English has the setting of [– topic-drop] allowing overt objects 6. The original study also investigated the interpretation of orientation of English reflexives by Chinese speakers. Two sentence types were used and there were four tokens for each sentence type: –

Monoclausal sentences; context suggests binding to the subject Maryi sold Lisa a picture of herselfi.

–

Monoclausal sentences; context suggests binding to the object Mary sold Lisai a picture of herselfi.

The results show that for both subject and object antecedents, there are no significant differences between the EFL learners and the native controls. Therefore, the Chinese EFL learners gave native-like responses in interpretation of orientation of English reflexives.

 Lin Jiang

only. If the ILG of Chinese-speaking learners of English has the value of [+ topicdrop], this value must be reset to [– topic-drop], if the relevant property of English is to be acquired. Sentences in the L2 English input, such as (26) which involves topicalization but does not involve topic-drop, can successfully be parsed in terms of the L1 grammar where topics are allowed to occur overtly, hence are unlikely to lead to parameter resetting. (26) Those musselsi, I wouldn’t touch ei with a barge pole! ( Hawkins 2001: 211) Similarly, sentences, such as (27) which does not involve topicalization, can be parsed in terms of the L1 grammar where objects are not obligatorily topicalized, hence cannot serve as a cue to resetting. (27) I wouldn’t touch those mussels with a barge pole. Therefore, the absence of suitable triggers in the positive evidence eliminates any motivation for the Chinese speakers to reset the Topic-drop parameter from the [+ topic-drop] setting to the [– topic-drop] setting of the L2. In addition, some transitive verbs can be used intransitively, as in (28) and (29). (28) He didn’t eat ___ today. (29) Yesterday, she presented ___ at a conference. A Chinese speaker may mistakenly take sentences like (28) and (29) as positive evidence indicating that English allows null objects more generally. Such positive evidence may also discourage Chinese-speaking learners to reset the Topic-drop parameter from the [+ topic-drop] setting to the [– topic-drop] setting of the L2. As a result, their L1 setting may persist even into the advanced stage, leading to the observed result in the present study, that is, advanced Chinese learners of English have typically not recognized the ungrammaticality of English sentences involving null objects. One other consideration is whether or not indirect negative evidence might be effective in this case. According to White (2003), indirect negative evidence is an alternative kind of negative evidence and works in the context of a cue-based theory. The rationale is that ‘in the context of predetermined parameter settings and cues, it is conceivable that the grammar might be ‘searching’ for the cue for one setting to some parameter and that, in the absence of the relevant cue, the other setting would be motivated’ (White 2003: 165). If the setting of [+ topic-drop] is instantiated in the ILG of a Chinese-speaking learner of English, there will in fact be few instances of topic drop in the L2 input to confirm this. In some sense, then, failure of the L2 input to reach the necessary threshold of topic drop might consti-



Reflexive interpretation and object realization 

tute a form of indirect negative evidence, indicating the need to reset the setting of the parameter from [+ topic-drop] to [– topic-drop]. However, a number of acquisition researchers have argued against the efficacy of indirect negative evidence in language acquisition because of its vagueness. White (1989), for example, argues that learners make many correct generalizations which go beyond the data they are exposed to, and this finding is problematic for the idea of indirect negative evidence, which would predict that non-occurrence of grammatical sentences would lead to them being treated as ungrammatical. Therefore, the issue of whether indirect negative evidence can trigger the appropriate English value of Topicdrop parameter for Chinese speakers is still open. In conclusion, what the problem with null objects reflects is indeed the presence of non-target like syntactic representations; that is, the Chinese learners have empty categories where native speakers of English would use overt objects. The obligatory status of English objects appears not to be acquirable, considering that these learners are advanced. Given that an uninterpretable [Top] feature is available in both Chinese and English, – hence, available for transfer from the L1 grammar – this result is unexpected according to the RDH. However, as suggested above, despite sharing this feature, there are parametric differences between the two languages such that triggers for parameter resetting appear to be unavailable. On the other hand, the advanced Chinese speakers’ target-like performance on reflexive interpretation shows no syntactic deficit in their L2 grammars. This result goes against the RDH, according to which Chinese learners are expected not to acquire the uninterpretable [Tense] feature in the L2 because it is lacking in the L1. In other words, the results from this study are not consistent with the RDH. Acknowledgements This study was supported by an MOE Project (No. 07JJD 740067) of the Centre for Linguistics and Applied Linguistics of Guangdong University of Foreign Studies, China. This study was based on my PhD dissertation research. I wish to thank Prof. Roger Hawkins for his supervision as well as two anonymous reviewers. Any errors, of course, are my own. References Akiyama, Y. 2002. Japanese adult learners’ development of the locality condition on English reflexives. Studies in Second Language Acquisition 24: 27–54.

 Lin Jiang Bailyn, J. 1992. LF movement of anaphors and the acquisition of embedded clauses in Russian. Language Acquisition 2: 307–335. Battistella, E. 1989. Chinese reflexivisation: a movement to INFL approach. Linguistics 27: 987–1012. Berwick, R. & Weinberg, A. 1984. The grammatical basis of linguistic performance: Language use and acquisition. Cambridge MA: The MIT Press. Carroll, S. 2001. Input and evidence: The raw material of second language acquisition. Amsterdam: John Benjamins. Chomsky, N. 1981. Lectures on government and binding. Dordrecht: Foris. Cook, V. J. 1990. Timed comprehension of binding in advanced L2 learners of English. Language Learning 40: 557–599. Eckman, F. 1994. Local and long-distance anaphora in second language acquisition. In Research methodology in second language acquisition, E. Tarone, S. Gass & A. Cohen (eds), 207–225. Hillsdale NJ: Lawrence Erlbaum. Finer, D. L. 1991. Binding parameters in second language acquisition. In Point counterpoint, L. Eubank (ed), 351–374. Amsterdam: John Benjamins. Finer, D. L. & Broselow, E. I. 1986. Second language acquisition of reflexive binding. Proceedings of NELS 16: 154–168. Fuller, J. & Gundel, J. 1987. Topic prominence in interlanguage. Language Learning 37: 1–18. Gibson, E. & Wexler, K. 1994. Triggers. Linguistic Inquiry 25: 407–454. Gundel, J. & Tarone, E. 1992. Language transfer and the acquisition of pronouns. In Language transfer in language learning, S. Gass & L. Selinker (eds), 87–100. Amsterdam: John Benjamins. Harbert, W. 1995. Binding Theory, control, and pro. In Government and binding theory and the minimalist program, G. Webelhuth (ed.), 177–212. Oxford: Blackwell. Hawkins, R. 2001. Second language syntax: A generative introduction. Oxford: Blackwell. Hawkins, R. 2003. ‘Representational deficit’ theories of (adult) SLA: Evidence, counterevidence and implications. Paper presented at European Second Language Association annual conference (EUROSLA 13), Edinburgh, September. Hirakawa, M. 1990. A study of the L2 acquisition of English reflexives. Second Language Research 6: 60–85. Huang, C.–T. J. 1984. On the distribution and reference of empty pronouns. Linguistic Inquiry 15: 531–574. Huang, C.–T. J. 1987. Remarks on empty categories. Linguistic Inquiry 18: 321–337. Huang, C.–T. J. 1989. Pro-drop in Chinese: A generalized control theory. In The null subject parameter, O. Jaeggli & K. Safir (eds.), 185–214. Dordrecht: Kluwer. Huang, C.–T. J. 1998 [1982]. Logical relations in Chinese and the theory of grammar. New York NY: Garland. Lakshmanan, U. & Teranishi, K. 1994. Preferences versus grammaticality judgments: Some methodological issues concerning the Governing Category Parameter in SLA. In Research methodology in second language acquisition, E. Tarone, S. Gass & A. Cohen (eds.), 185–206. Hillsdale NJ: Lawrence Erlbaum. MacLaughlin, D. 1998. The acquisition of the morphosyntax of English reflexives by non-native speakers. In Morphology and its interfaces in second language knowledge, M. L. Beck (ed), 195–226. Amsterdam: John Benjamins. Manzini, M. R. & Wexler, K. 1987. Parameters, binding theory, and learnability. Linguistic Inquiry 18: 413–444.



Reflexive interpretation and object realization  Park, H. 2004. A minimalist approach to null subjects and objects in second language acquisition. Second Language Research 20: 1–32. Pica, P. 1987. On the nature of the reflexivisation cycle. Proceedings of the North Eastern Linguistics Society 17: 483–499. Progovac, L. 1992. Relativised SUBJECT: Long-distance reflexives without movement. Linguistic Inquiry 23: 671–680. Progovac, L. 1993. Long-distance reflexives: Movement-to-Infl vs. relativised SUBJECT. Linguistic Inquiry 24: 755–772. Progovac, L. & Connell, P. 1991. Long-distance reflexives, Agr-subjects, and acquisition. Paper at the 2nd meeting of FLSM, University of Michigan, Ann Arbor MI. Reinhart, T. & Reuland, E. 1991. Anaphors and logophors: An argument structure perspective. In Long-distance anaphor, J. Koster & E. Reuland (eds), 283–321. Cambridge: CUP. Rizzi, L. 1997. A parametric approach to comparative syntax: Properties of the pronominal system. In Elements of grammar: Handbook in generative syntax, L. Haegeman (ed.), 281–337. Dordrecht: Kluwer. Schachter, J. 1996. Maturation and the issue of Universal Grammar in L2 acquisition. In Handbook of language acquisition, W. Ritchie & T. Bhatia (eds), 159–193. New York NY: Academic Press. Schwartz, B. D. & Sprouse, R. 1994. Word order and nominative case in nonnative language acquisition: A longitudinal study of (L1 Turkish) German interlanguage. In Language acquisition studies in generative grammar, T. Hoekstra & B.D. Schwartz (eds), 317–368. Amsterdam: John Benjamins. Schwartz, B. D. & Sprouse, R. 1996. L2 cognitive states and the full transfer/ full access model. Second Language Research 12: 40–72. Thomas, M. 1989. The interpretation of English reflexive pronouns by non-native speakers. Studies in Second Language Acquisition 11: 281–303. Thomas, M. 1991. Universal Grammar and the interpretation of reflexives in a second language. Language 67: 211–239. Tsao, F. 1977. A functional study of topic in Chinese: The first step towards discourse analysis. PhD dissertation, University of Southern California. Wakabayashi, S. 1996. The nature of interlanguage: SLA of English reflexives. Second Language Research 12: 266–303. Wells, T.L. 1998. L2 acquisition of English binding domain. In Morphology and its interfaces in second language knowledge, M. L. Beck (ed), 227–256. Amsterdam: John Benjamins. Wexler, K. & Manzini, R. 1987. Parameters and learnability in binding theory. In Parameter setting, T. Roeper & E. Williams (eds), 41–76. Dordrecht: Reidel. White, L. 1987. Against comprehensible input: The input hypothesis and the development of L2 competence. Applied Linguistics 8: 95–110. White, L. 1989. Universal grammar and second language acquisition. Amsterdam: John Benjamins. White, L. 1995. Input, triggers and second language acquisition: Can binding be taught? In Second language acquisition theory and pedagogy, F. Eckman, D. Highland, P. Lee, J. Mileman & R. Rutkowski Weber (eds.), 63–78. Mahwah NJ: Lawrence Erlbaum. White, L. 2003. Second Language Acquisition and Universal Grammar. Cambridge: CUP. White, L., Bruhn-Garavito, J., Kawasaki, T., Pater, J. & Prevost, P. 1997. The researcher gave the subject a test about himself: problems of ambiguity and preference in the investigation of reflexive binding. Language Learning 47: 145–172.

 Lin Jiang Ying, H. G. 1999. Access to UG and language transfer: A study of L2 learners’ interpretation of reconstruction in Chinese. Second Language Research 15: 41–72. Yip, V. & Matthews, S. 2007. The bilingual child: Early development and language contact. Cambridge: CUP. Yip, V. & Tang, G. 1998. Acquisition of English reflexive binding by Cantonese learners. In Morphology and its interfaces in second language knowledge, M. L. Beck (ed), 165–194. Amsterdam: John Benjamins. Yuan, B. 1994. Second language acquisition of reflexives revisited. Language 70: 539–545. Yuan, B. 1997. Asymmetry of null subjects and null objects in Chinese speakers’ L2 English. Studies in Second Language Acquisition 19: 467–497. Zobl, H. 1994. Prior linguistic knowledge and conservatism of the learning procedure: Grammaticality judgments of unilingual and multilingual learners. In Language transfer in language learning, S. Gass & L. Selinker (eds), 176–196. Amsterdam: John Benjamins.

Selective deficits at the syntaxdiscourse interface Evidence from the CEDEL2 corpus Cristóbal Lozano

University of Granada Previous research shows that English-speaking learners of Spanish show (i) early sensitivity to the syntactic mechanisms licensing overt and null pronominal subjects, yet (i) persistent and long-lasting deficits when pronominal distribution is constrained by topic/focus at the syntax-discourse interface. It has been assumed that such vulnerability affects the whole set of phi-features of the pronominal paradigm, but I will use near-native corpus evidence to show that the observed deficits are selective, i.e., they do not affect the whole set of phi-features in the pronominal paradigm but rather a subset: due to their representational nature (which is constrained by Universal Grammar), only third person singular animate pronouns are targets for vulnerability, while the rest of the paradigm remains rather stable.

1. Introduction In the second language literature over the past two decades, researchers have mainly focused on the role of formal (i.e., morphosyntactic) features in L2 acquisition. This is reflected in the publication of monographs on this issue (e.g., Liceras et al. 2007), textbooks (e.g., Hawkins 2001, White 1989, 2003) as well as innumerable articles. Several proposals have been put forth trying to account for how some features can lead to representational deficits but others lead to native-like knowledge (e.g., Hawkins and Chan’s 1997 Failed Functional Features Hypothesis, Prévost and White’s 2000 Missing Surface Inflection Hypothesis, Beck’s 1998 Local Impairment Hypothesis, just to name a few). By contrast, relatively little is known about the role of features operating at the interfaces outside narrow syntax. Recent studies have started to address the issue of why features at the syntax-discourse interface can be problematic for L2 learners even at end-states (e.g., Sorace 2004, 2005,

 Cristóbal Lozano

2006). Further distinctions have been made about syntax-semantics vs. syntaxdiscourse features, with different predictions for vulnerability (Tsimpli and Sorace 2006). In the context of the first property of the pro-drop parameter (or null-subject parameter) it is well known since the 80’s that English-speaking learners of L2 Spanish acquire from early stages the formal features licensing null referential pronominal subjects in L2 Spanish (e.g., Liceras 1989, Lozano 2002a, Phinney 1987), yet studies from the late 90s report that such learners show deficits with the distribution of overt and null referential pronominals when constrained outside syntax, i.e., when regulated by discursive features like [Topic] and [Focus] (e.g., Lozano 2002b, Montrul and Rodríguez-Louro 2006, Pérez-Leroux and Glass 1997, 1999). These studies claim that the acquisition of pronominal subjects results in (i) native-like knowledge of formal features operating at narrow syntax from early states, yet (ii) divergent knowledge and deficits when features operate at the syntax-discourse interface, which appear to be persistently problematic even at end-states. Crucially, this observed ‘syntax-before-discourse’ phenomenon (i.e., the claim that the formal features licensing null subjects are acquired early and effortlessly while the discursive features are persistently problematic) comes from studies presenting evidence mostly from third singular pronominal subjects, but the claim has been made about the whole pronominal paradigm (singular: 1st, 2nd and 3rd person; plural: 1st, 2nd and 3rd). In this study, evidence from the whole paradigm will be presented and it will be shown that deficits with pronominal subject at the syntax-discourse interface are selective, since not all persons in the paradigm are equally vulnerable. It will be proposed that such selectiveness is a result of how Universal Grammar organises pronominal features according to a pronominal ‘Feature Geometry’ (Harley and Ritter 2002a). This study is structured as follows. Section 2 presents the theoretical background on the distribution of pronominal subjects at the syntax-discourse interface in native Spanish and the so-called pronominal ‘Feature Geometry Analysis’. Section 3 reports on previous L2 Spanish studies on the acquisition of the distribution of overt and null pronominal subjects. In Section 4 the relevant predictions and hypotheses are presented. Section 5 describes the methodology used and Section 6 presents the results. In Section 7 the results are discussed and a conclusion is reached in Section 8. 2. Theoretical background In this section, I will discuss first how the distribution of overt and null pronominal subjects is constrained in native Spanish by discursive properties like Topic



Selective deficits at the syntax-discourse interface 

and Topic-Shift. Later, I will show an analysis of the hierarchy of pronominal subject features, which is constrained by Universal Grammar (UG). 2.1

Distribution of pronominal subjects at the syntax-discourse interface

Since the formulation of the null-subject parameter (Jaeggli 1982, Rizzi 1982), it is well known that in null-subject languages like Spanish, overt and null personal pronominal subjects can alternate. In (1) the overt pronoun él and the null pronoun pro are in free alternation (e.g., Fernández-Soriano 1989, 1993, 1999, Luján 1999, Picallo 1994, Rigau 1986). Notice that a full Noun Phrase (NP) subject like Pedro ‘Peter’ can also alternate in the same position.1 By contrast, in non-null subject languages like English a null pronoun is not possible, (2). The licensing of null subjects in languages like Spanish has been attributed to formal syntactic features. In particular, the formal [+D] and [+AGR] features of the T(ense) head in Spanish is a proper licensor of pro (Rizzi 1997).2



(1)

Pedro tiene mucho dinero. él pro

Peter/he/Ø have.3sg much money ‘Pedro/he has a lof of money’

(2)

Peter has a lot of money. he *pro

This apparently free alternation of overt and null pronominal subjects is constrained by discursive factors in null-subject languages like Spanish (see, inter alia, Alonso-Ovalle et al. 2002, Fernández-Soriano 1989, Pérez-Leroux and Glass 1997, 1999). It is well known that at the syntax-discourse interface information packaging is crosslinguistically articulated into topic and focus and that different languages 1. In current generative work, what was traditionally analysed as Noun Phrase (NP) is now analysed as Determiner Phrase (DP). For simplicity reasons, I will choose the terminology NP throughout, as the precise syntactic analysis (whether NP or DP) is irrelevant in this work. 2. In the generative literature, there are many theoretical explanations of why null-subject languages allow pro, while non-null subject languages cannot. While the precise technical details of the mechanisms licensing pro vary in these studies, what is common to all of them is that formal features are responsible for the licensing of pro. For different views, see, e.g., Alexiadou and Anagnostopoulou (1998), Rizzi (1997).

 Cristóbal Lozano

can use different mechanisms to encode it, such as morphology, syntax, prosody or a combination of these (e.g., Casielles-Suarez 2004, Rochemont 1998, Zubizarreta 1998, 1999, Vallduví 1992, Vallduví and Engdahl 1996). For the purposes of this study, I will focus on two different types of contexts that regulate the distribution of overt and null pronominal subjects: topic contexts and topic-shift contexts. Topic information represents discourse-old, known information which has been previously evoked in the preceding discourse. Topic continuity is marked in Spanish via a null pronoun. Consider Spanish native data from the CEDEL2 corpus, (3), where the informant is talking about el protagonista ‘the main character’ of the film Escondido ‘Hidden’. The first instance of el protagonista ([3] [sing] [masc]) is clearly focus (new information), as it has not been mentioned in the preceding context. The following references to el protagonista can, at least theoretically, be realised in three possible ways: as a full NP again (el protagonista), as an overt pronoun (él ‘he’) and as a null pronoun (pro). Due to economy reasons, a null pronoun (pro) is the pragmatically felicitous option since it marks topic continuity in the discourse, though the other options (full NP or overt pronoun) would be grammatically possible but pragmatically infelicitous in native Spanish. (3) (Previous context: the informant is talking about the main character of a film) En la película “Escondido” el protagonista tiene una familia y pro trabaja en un programa de televisión. Un día pro empieza a recibir videos anónimos... [RSZ, Spanish native, CEDEL2 corpus] ‘In the film “Hidden”, the main character has a family. He works on a TV program. One day he starts receiving anonymous videos...’

Obviously, pro can mark topic-continuity for any other person and number of the pronominal paradigm. For example, in (4) the informant is talking about herself and her holidays, so she uses the null pronoun pro (phi-features: [1] [sing] [±masc], discursive features: [Topic]). (4) (Previous context: the informant is talking about her holidays) A mediados de julio, pro estaba muy cansada y con mucha tensión emocional, porque pro tuve que dejar casi terminado un trabajo de cierta envergadura. pro pasé unos días en la playa de Guardamar... [MCL, Spanish native, CEDEL2 corpus] ‘Around mid July, I was very tired and with a lot of emotional stress, because I had to finish off an important assignment. I spent a few days on Guardarmar beach...’ Topic-shift (also referred to in the literature as contrastive focus) requires a change of referent, which is typically realised via an overt pronoun. In (5), the native



Selective deficits at the syntax-discourse interface 

informant is talking about los protagonistas ‘the main characters’ of a film, a man and a woman. When the informant wants to refer to the man, the overt pronoun él ‘he’ (phi-features: [3] [sing] [masc], discursive features: [Topic-shift]) is required. Likewise, a change of reference to the woman requires the overt pronoun ella ‘she’ (phi-features: [3] [sing] [fem], discursive features: [Topic Shift]). While it would be grammatically possible to use a null pronoun pro to refer to either the man or the woman (since the null pronoun is unspecified for gender and it could then refer to either of the referents), its use would cause ambiguity, hence an overt pronoun is pragmatically preferred. (5) (Previous context: the informant is talking about the main characters of the film ‘The Illusionist’: a man and a woman) La última película que he visto es la de “El Ilusionista”... Los protagonistas son dos jóvenes que se conocen y se enamoran. Él es de clase baja, mientras que ella es de familia noble... [SPH, Spanish native, CEDEL2 corpus] ‘The last film I saw is “The illusionist”... The main characters are two young people who meet and fall in love. He comes from a working-class family, whereas she comes from a noble family...’

Note that in topic-shift contexts it is also possible to use a full NP (instead of an overt pronoun). This fact has been often overlooked in the literature, where it has been standardly assumed that an overt pronoun is the default option. An analysis of the native data in the CEDEL2 corpus (see Table 4 and Figure 4, Section 6.1) reveals that in native Spanish (as well as in non-native Spanish) topic-shift is significantly realised via a full NP more frequently than via an overt pronoun. In (5), the informant is talking about the main characters of the film ‘The Illusionist’: el príncipe ‘the prince’ ([3] [sing] [masc]), su prometida/la chica ‘his fiancé/the girl’ ([3] [sing] [fem]) and el ilusionista ‘the illusionist’ ([3] [sing] [masc]). Given that two masculine referents are brought into the discourse (el principe and el ilusionista), the native informant chooses full NPs (which can be specified as [Topic-shift], as well as [Focus]) to mark topic shift, instead of overt pronouns (which would cause certain ambiguity, since él ‘he’ could refer to either masculine antecedent). (6) (Previous context: the informant is talking about the main characters of the film ‘The Illusionist’: the Prince, his fiancé, the illusionist) Un día el príncipe y su prometida acuden a ver el espectáculo... El príncipe sospecha de que su prometida le es infiel y pro manda a uno de sus secuaces a perseguirla... Al final, el ilusionista y la chica preparan su huida... El príncipe termina suicidándose y, al final, el ilusionista y la chica consiguen estar juntos [SPH, Spanish native, CEDEL2 corpus]

 Cristóbal Lozano

‘One day, the Prince and his fiancé go to see the show... The Prince suspects that his fiancé is cheating on him and he orders one of his henchmen to chase her... At the end, the illusionist and the girl prepare their escape... The Prince ends up committing suicide and, at the end, the illusionist and the girl get together’. From the data in (6) it appears that in topic-shift contexts the higher the number of potential antecedents, the higher the probability of ambiguity if overt pronouns are used. Full NPs are favoured in these contexts to avoid potential ambiguity, as (7) illustrates. (7) (Previous context: the informant is talking about the characters of the film ‘Miss Sunshine’,which consist of a family: the grandfather, the parents, two children and an uncle) En cuanto al argumento, trata de una familia de clase media estadounidense formada por el abuelo paterno, los padres, dos niños (una niña de 7 años y otro de 16) y un tío materno... En el viaje han de lidiar con problemas personales muy importantes, aquellos que han marcado su vida hasta el momento: el tío se encuentra con el amante que lo traicionó, el padre se da cuenta de que su socio lo ha engañado y lo ha dejado en la bancarrota, el hijo mayor descubre que es daltónico y ello le impedirá ser piloto, su sueño, y el abuelo, el peor parado de la historia, muere por sobredosis de cocaína... [CMM, Spanish native, CEDEL2 corpus] ‘The plot is about a middle-class American family that consists of the paternal grandfather, the parents, two kids (a 7 year-old girl and a 16 yearold boy) and a maternal uncle... During their trip they face important personal problems that have marked their lives: the uncle meets the lover that betrayed him, the father realises that his partner has lied and bankrupted him, the elder son finds out he is colour-blind, which will prevent him from becoming a pilot, his dream, and the grandfather, the loser in the story, dies from a cocaine overdose...’

To summarise, topic continuity is realised in native Spanish via a null pronoun, while topic shift is realised via overt material (either a full NP or an overt pronoun). 2.2

Pronominal feature geometry

In the generative second language acquisition (SLA) literature dealing with the acquisition of pronominal subjects, the inventory of pronominal features has been typically treated as an unordered ‘bundle’. To illustrate the idea of the unordered bundle, consider two pronominal subjects, one specified for [3] [sing] and the other for [1] [sing]. In the literature there is no principled reason to assume that



Selective deficits at the syntax-discourse interface 

the first set of features is easier (or more difficult) to acquire than the second set. This entails that the relative order of acquisition of each pronominal feature would vary little between learners, who would therefore acquire pronominal features simultaneously in one ‘bundle’. In other words, there has been no systematic prediction as to which pronominal feature(s), if any, are acquired first and which later. This is what has been implicitly assumed in the L2 literature since, as it was mentioned earlier, studies have focused mostly on third person singular pronouns and not on a systematic analysis of the full pronominal paradigm (e.g., Al-Kasey and Pérez-Leroux 1998, Davies 1996, Escutia 2002, Diaconescu and Goodluck 2002, Helland 2004, Kanno 1997, Lafond et al.  2001, Liceras 1989, Liceras and Díaz 1999, López-Ortega 2006, Lozano 2002a, 2002b, Montrul 2004, Montrul and Rodríguez-Lourou 2006, Pérez-Leroux and Glass 1997, 1999, Pérez-Leroux et al. 1999, Phinney 1987, Polio 1995, Sorace and Filiaci 2006, White 1985). Harley and Ritter (2002a, 2002b) have proposed the so-called ‘Feature Geometry Analysis’ for pronouns. Drawing from a wide range of typologically unrelated languages, they show that UG provides a constrained set of pronominal features which are systematically and hierarchically organized (Figure 1). The root node is termed referring expression, which corresponds to the traditional idea of a pronoun. There are three main sets of features: participant, individuation and class.3 Referring Expression (= Pronoun)

PARTICIPANT (= person) Speaker (=1st person)

Addressee (=2nd person)

INDIVIDUATION (= number) Group Minimal (=non-sing) (=sing)

CLASS (=gender)

Animate Fem

Inaminate/Neuter

Masc…

(taken from Harley and Ritter 2002a: 508) Figure 1.  Pronominal feature geometry analysis 3. In the Feature Geometry structure shown in Figure 1 the expressions in brackets have been added for clarification purposes, e.g., ‘participant (=person)’, since what Harley and Ritter term participant has been traditionally termed person.

 Cristóbal Lozano

The participant node and its dependents, Speaker and Addressee represent 1st and 2nd person respectively. The individuation node and its dependents, Group and Minimal correspond to non-singular (plural and dual) and singular number respectively. The class node encodes gender and other class information. Note that the underlined nodes Speaker (i.e., 1st person), Minimal (i.e., singular) and Inanimate (i.e., neuter) represent the default interpretation of the node. Importantly, the participant node encodes two features: Speaker (1st person) and Addressee (2nd person), while 3rd person is unmarked. It has been traditionally recognised in linguistics that there is a difference between 1st and 2nd person (which correspond to the speech-act participants, i.e., to a deictic use of the pronoun) vs. 3rd person (which does not correspond to the speech-act participants but rather to an anaphoric use of the pronoun) (e.g., Benveniste 1971, Bloomfield 1933, Forchheimer 1953, Jespersen 1924; see also Cornish 2006 and Saxena 2006 for more updated overviews). According to Harley and Ritter (2002a: 488) ‘The geometry... captures the intuition that so-called 3rd person is in fact not a true personal form... When the Participant node is absent, the underspecified Referring Expression node receives a so-called 3rd person interpretation’. In other words, the authors claim that 3rd person is the absence of grammatical person (=participant) (see also Bianchi 2005, Kayne 2000). Evidence that UG constrains the Feature Geometry analysis in language development comes from child L1 acquisition (e.g., Harley and Ritter 2002a, Hanson 2000). The uniformity and variability in the order of acquisition of pronominal subjects in different L1s shows that in the path of development 1st person singular and 3rd person singular neuter are normally the first pronouns to be acquired, while 3rd person animate comes later. Similar findings are reported for child Specific Language Impairment (Mastropavlou 2006), where 3rd person is selectively impaired or delayed, compared to 1st and 2nd person. Singular is also acquired before plural in normally developing children (Harley and Ritter 2002a, Hanson 2000). It seems then that in L1 acquisition defaults are acquired first (i.e., Speaker, Minimal and Inanimate). As we will see, this has implications for L2 acquisition since generative L2 research has not addressed the issue of emergence of the different pronominal features, i.e., all pronouns have been treated as a ‘bundle’ and no specific predictions have been made about the difficulty or emergence of each of them. Recent corpus SLA research from a functional perspective has shown that learners of L2 English (with L1 Swedish) treat 1st/2nd person pronouns differently from 3rd person (Ädel 2003). As we will see in the experimental section, the distinction will be crucial to account for the L1 English – L2 Spanish data.



Selective deficits at the syntax-discourse interface 

3. Previous L2 studies In the context of the pro-drop parameter, the distribution of overt and null subjects in L2 Spanish has been extensively researched. It is well known since the late 80’s that English-speaking learners of L2 Spanish acquire from early stages and with relatively little effort the formal properties that license a null subject, pro (inter alia, Lafond et al. 2001, Liceras 1989, Lozano 2002a, Phinney 1987). In particular, learners know that an overt and a null pronoun are in free alternation in Spanish, as in (8).

(8)

Él pro

come pasta.

He/pro eat.3sg pasta ‘He eats pasta’ More recent studies have shown that, while the observation that formal licensing features are acquired early is correct, learners do show some deficits with the discursive features that constrain the distribution of overt and null pronouns in the discourse. Pérez-Leroux and Glass (1997) found that very advanced and near-native learners of L2 Spanish (with L1 English) showed native-like knowledge of formal constraints, yet non-native behaviour at the syntax-discourse interface. In one of the tasks, learners were shown topic-shift contexts like (9), where there are several referents (Sampras, Edberg and Agassi) and one of them has to be chosen for contrastive purposes. Learners were required to translate a sentence into Spanish. The context is manipulated such that the expected target translated sentence should contain an overt pronoun to express a shift of referent (i.e., to contrast Sampras against the rest of players). A null pronoun would be unpragmatic since it would cause ambiguity and could refer to any of them.4 Learners showed a low percentage of overt pronouns, but a high percentage of null pronouns, which leads to an unpragmatic overgeneralisation of null pronouns for contrastive purposes. This deficit persists even at end-states.

(9) [Context] My friends are all excited about the US Open Tournament. The player that is most on their mind is Pete Samprasi. They’ve barely mentioned Stefan Edbergj and Andre Agassik. [To translate] Everybody thinks that he will win. [Expected target sentence] Todo el mundom cree que éli ganará. [Unexprected unpragmatic sentence] Todo el mundom cree que #prom ganará. 4. Unpragmatic constructions are indicated with the hash mark (#). Note that constructions can be pragmatically infelicitous while being grammatically correct, i.e., lack of pragmaticality does not entail agrammaticality.

 Cristóbal Lozano

In a later study testing several proficiency levels (elementary, intermediate and advanced), Pérez-Leroux and Glass (1999) found developmental evidence for the early mastery of formal constraints regulating the overt/null distribution, yet late and persistently problematic mastery of such distribution when constrained by topic and contrastive focus. Unpragmatic overuse of null pronouns in topic-shift contexts was found again in the advanced group. The results showed that ‘knowledge of the marking of the topic/focus distinction is acquired over time and experience’ (p. 242), though residual deficits remain. Similar results for contrastive focus contexts have been reported in other studies (e.g., Lozano 2002b, 2003). In topic contexts, it has been reported that learners of Spanish overuse overt pronouns, which results in redundancy (e.g., Lozano 2006a, Montrul and Rodríguez-Louro 2006). In particular, in contexts when the topic is clearly set (Profesor Antonio ‘Professor Antonio’) and there are no potential referents that could lead to ambiguity, as in (10), learners’ tolerance of the (a) sentence, where the overt pronoun él ‘he’ is pragmatically redundant, is significantly higher than the Spanish natives’ tolerance. This tolerance persists even at end-states (Lozano 2006a). (10)

Aunque el profesor Antonioi parece pobre... a. los estudiantes dicen que #éli tiene mucho dinero b. los estudiantes dicen que proi tiene mucho dinero ‘Even though professor Antonio seems poor...

a. students say that he has a lot of money b. students say that has a lot of money’ Other studies testing several aspects of the pronominal paradigm on L2 Spanish support the finding that formal constraints are in place early, yet discursive constrains are acquired over time and tend to result in residual deficits (e.g., Al-Kasey and Pérez-Leroux 1998, Liceras and Díaz 1999, Pérez-Leroux et al. 1999). This ‘syntax-before-discourse’ observation on pronominal subjects does not represent an isolated phenomenon in Spanish L2 acquisition, as it has been attested in other acquisition contexts, as in L1 English–L2 Chinese (Polio 1995), L1 Croatian–L2 Italian (Kras 2006), L1 English–L2 Italian (Sorace and Filiaci 2006), L2 Italian with learners of several L1s (Belletti and Leonini 2004), English-Italian bilingual children (Serratrice 2004, Serratrice et al. 2004), Italian-Dutch bilingual children (Pinto 2006), L1 Greek and Italian attrition under the influence of L2 English (Tsimpli et al. 2004), L1 Spanish attrition under the influence of L2 English (Satterfield 2003), L1 Spanish heritage speakers with dominant English (Montrul 2004) and Spanish L1 acquisition (Grinstead 2004). Additionally, it has been also reported that for the second property of the pro-drop parameter (SubjectVerb inversion), learners of L2 Spanish also show early knowledge of the formal



Selective deficits at the syntax-discourse interface 

properties licensing inversion, but persistent problems with the discursive properties that constrain inversion in the discourse (Hertel 2003, Lozano 2006b, 2006c). Similar results are reported in Spanish heritage speakers with dominant English (Valenzuela and McIlwraith 2007), L2 Portuguese (Fruit 2007) and attrition in L1 Catalan with dominant L2 English (Helland 2004). In this context, Sorace (2004) observes that ‘aspects of grammar at the syntaxdiscourse interface are more vulnerable... than purely syntactic ones’ (p. 143) and that ‘interfaces, because they are more complex than narrow syntax, are inherently more difficult to acquire.’ (p. 144). So, while the ‘syntax-before-discourse’ phenomenon is beyond dispute, its causes are still unclear. Two main proposals have appeared recently in the literature. First, the representational deficit account postulates that underspecification of [+interpretable] features like [Topic] and [Topic-Shift] become underspecified at the syntax-discourse interface, thus triggering the observed deficits (e.g., Montrul 2004, Sorace 2004), but it may be also the case that it is [–interpretable] features that are responsible for such deficits (Lozano 2006b, 2006c). Second, the processing deficit account postulates that the language processor cannot process efficiently properties at the syntax-discourse interface, which results in shallow processing (Sorace 2005, 2006, Sorace and Filiaci 2006). Importantly, in the context of pronominal subjects, most studies mentioned above present evidence from 3rd person singular, but the claims are made (implicitly or explicitly) about the whole pronominal paradigm. In other words, the ‘syntax-before-discourse’ phenomenon has been assumed to affect the whole pronominal paradigm (all three persons and the two numbers). In this study, I will present evidence from the whole pronominal paradigm. As will be shown in the next section (Predictions), deficits at the syntax-discourse interface are selective, since not all persons are equally vulnerable at the syntax-discourse interface. 4. Predictions Based on existing previous L2 studies (Section 3) it was predicted that advanced and highly advanced learners of L2 Spanish (with L1 English) would show deficits at the syntax-discourse interface, in particular, overproduction of overt pronouns in topic contexts where a null pronoun is required. Additionally, following the theoretical assumptions of Harley and Ritter’s Feature Geometry (Section 2.2), it was predicted that not all pronominal features are equally complex/simple to acquire. The hypothesis in (11) was then formulated. (11) H1: Deficits at the syntax-discourse interface do not affect the whole pronominal paradigm, but are rather selective.

 Cristóbal Lozano

In particular, we expect (i) robustness and native-like knowledge with speech-act participants (1st and 2nd person) and with neutrals (3rd person inanimate), but (ii) vulnerability with 3rd person animate. In short, syntax-discourse deficits are selective as they affect certain persons in the pronominal paradigm and not all persons (as previously assumed in the literature). 5. Method In this Section I will detail the properties of the corpus used in the study, as well as the subjects of the corpus. Then, I will describe the concordance software used to analyse the corpus and how the corpus raw data were treated statistically. 5.1

Corpus

CEDEL2 (Corpus Escrito Del Español L2 ‘L2 Spanish Written Corpus’) is a learner corpus that is being developed at the Universidad Autónoma de Madrid (Spain).5 Currently the corpus consists of an L1 English–L2 Spanish subcorpus plus a native Spanish subcorpus used for comparative purposes. At present, the corpus has reached approximately 500,000 words, though it is envisaged that by the end of the project data collection will have reached over half a million words. Data consist of written compositions in Spanish. In particular, learners must fill in several online forms via the internet at: http://www.uam.es/woslac/start.htm i. a learning background questionnaire which provides crucial information such as the learners’ chronological age, age of first exposure to Spanish, years of instruction in Spanish, length of stay in Spanish-speaking countries, their father’s and mother’s native language, language(s) spoken at home, etc. As we will see in Section 5.2 (Table 1), this type of data are useful to filter out learners that do not meet the researcher’s requirements. ii. a composition in Spanish form, where participants can choose between twelve different topics that vary in difficulty and can elicit different types of linguistic structures. iii. a Spanish placement test (only for learners), which is an independent and standardized measure of grammatical proficiency in Spanish (University of Winconsin College-level Placement Test: Spanish grammar, form 96M). Spanish natives must fill in two online forms only: (i) a formación académica form, which is similar to the learners’ learning background form, but it obviously 5.

More information on CEDEL2 is available at: http://www.uam.es/woslac/cedel2.htm



Selective deficits at the syntax-discourse interface 

excludes questions that are not applicable to natives, such as length of instruction in Spanish, age of first exposure, etc; (ii) a redacción en español, which is identical to the learners’ composition form. Given the nature of the data for this study, I selected only those compositions where topic continuity and topic shift would be more likely, i.e., compositions where one character would be activated during the narration (topic continuity) and those where several characters appeared and a change of character would be likely (topic shift). The composition titles were mostly of the type Resume una película que hayas visto recientemente ‘Summarise a film you have seen recently’, Describe un viaje que hayas hecho recientemente ‘Describe a trip you have made recently’, ¿Qué hiciste el año pasado durante tus vacaciones? ‘What did you do last year during your holidays?’ 5.2

Subjects

For this study I selected a small group of CEDEL2 texts that met certain proficiency criteria (see below). Learners were classified into two learner groups (N=10 each group) according to proficiency level. An independent proficiency measure was administered to check learners’ grammatical competence (University of Winconsin College-Level Placement Test – Spanish, 1998). A summary of the learners’ and natives’ bio-data is shown in Table 1 (full bio-data details of each participant are presented in the tables in the Appendix section). As can be seen in Table 1, the proficiency score was higher in the upper-advanced group (mean=99%; range=98%-100%) than in the lower-advanced group (mean=93%; range=95%-91%). Overall, the upper-advanced group was first exposed to Spanish earlier (mean=14 years; median=14; range=3–20 years) than the lower-advanced group (mean=15 years; median=14; range=6–26). Overall, the upper-advanced group received more years of instruction (mean=8; range=5–11) than the lower-advanced (mean=6; range=3–15). Regarding the learners’ stay in a Spanish-speaking country, the upper-advanced group stayed overall longer (mean=29 months; range=3–18 months; outlier=204 months) than the lower-advanced group (mean=27 months; range=0–12 months; outlier=228 months). These variables confirm that the upper-advanced group can be safely regarded as a highly proficient group, reaching near-native grammatical competence, while the lower-advanced group can be considered a very proficient group. The rationale behind the creation of two advanced groups instead of only one was the following: given that the SLA debate on deficits at the syntax-discourse interface has focused on whether learners can overcome discursive deficits at end-states (see sections 1 and 3), it was necessary to discriminate between those learners who had indeed achieved a near-native degree of competence (upper-advanced group) and those who had not but still showed a high level of competence (lower-advanced group).

 Cristóbal Lozano

Table 1.  Summary of subjects’ bio-data (see Appendix for full details) Group

N

Mean proficiency

Mean chronol age

Upper-advanced Lower-advanced Spanish natives

10 10 12

99% 93% n/a

35 32 37

5.3

Mean age Mean years first exposure instruct 14 15 n/a

8 6 n/a

Mean stay (months) 29 27 n/a

Software (concordancer)

The UAM Corpus Tool (version 1.0) is a stand-off XML, freely-downloadable software for the annotation of linguistic corpora developed by Michael O’Donnell.6 Some of its main features, which were used in the current study, are as follows: i. Annotation of segment using an annotation scheme of your design. The annotation of each segment can be made at multiple levels (e.g., NP, Clause, Sentence, whole document). In this study, I annotated each grammatical subject in finite sentences (null pronoun, overt pronoun or full NP) according to the annotation scheme in Section 5.4, Figure 2. ii. Searching for instances of a feature (or combination of features), e.g., any instance of the feature topic-shift, or any instances containing an overt pronoun which is a topic-shift, or any instances of third-person singular subjects realised as a full NP which is topic-shift. iii. Comparative statistics across subsets, e.g., contrasting pragmatic and unpragmatic production of third person singular subjects. 5.4

Data analysis

Using UAM Corpus Tool I designed an annotation scheme (Figure 2). The annotation scheme was implemented taking into account Harley and Ritter’s (2002a) pronominal Feature Geometry and previous literature on pronominal subjects. Each subject (whether a full NP, an overt pronoun or a null pronoun) in finite sentences was tagged for the following properties: i. Syntax: each subject was coded as a full NP, an overt pronoun or a null pronoun. ii. Number: singular (1st, 2nd or 3rd person) or plural (1st, 2nd or 3rd person). iii. Animacy: animate or inanimate. 6. UAM Corpus Tools can be freely downloaded. More information at: http://www.wagsoft. com/CorpusTool/index.html



Selective deficits at the syntax-discourse interface 

iv. Information status: topic (which encodes topic-continuity in the discourse), focus (which encodes discourse-new information), and topic-shift (which encodes a change of topic in the discourse). In the current study, only topic and topic-shift will be analysed. v. Pragmaticality: whether the subject was pragmatically appropriate or rather inappropriate, in which case I coded the type of lack of pragmaticality. This can be of two types: underproduction (i.e., using a null pronoun in topic-shift contexts which require the use of overt material) or overproduction (i.e., using an overt pronoun or a full NP when a null pronoun is required in topic-continuity contexts). The annotation scheme was applied to the following corpora (Table 2). The upperadvanced corpus consisted of 10 texts (i.e., 10 learners, 1 text per learner) containing a total of 8,188 words and 453 tags (i.e., 453 annotated subjects which were statically analysed at a later stage). The ratio of pronominal subjects used in this group was 5.5% (i.e., on average there were 5.5 tagged pronominal subjects per 100 _np SYNTAX p _ _ronoun _ null

subjects _

_1sing _SINGULAR_singular T 2_sing _ YPE _3sing NUMBER _ _1plu _PLURAL_2plu _plural _TYPE _3plu _ NIMACY _animate A _inanimate t_opic _focus _topic-shift _pragmatic PRAGMATICS _ INFO _

_apragmatic

_ PRAGMATICA _TYPE

_underproduction

_overproduction _OVERPRODUCTION _TYPE

_overt-when-null-required

_np-when-null-required

Figure 2.  Annotation scheme used in the software UAM Corpus Tool

 Cristóbal Lozano

Table 2.  Summary of linguistic data (see Appendix for full details) Corpus Upper-advanced Lower-advanced Spanish natives

N texts

Total # words

Total # tags

Mean % pronominal subjects

10 10 12

8188 8521 5954

453 528 299

5.5% 6.2% 5.1%

words). The lower-advanced corpus also consisted of 10 texts with a total of 8,521 words and 528 tags, corresponding to a 6.2% words/tags ratio. Finally, the Spanish native corpus consisted of 12 texts with a total of 5,954 words and 299 tags, the percentage of pronominal subjects being 5.1%. It is important to highlight that, although the total number of words is lower in the Spanish corpus than in the learner corpora (due to the natives’ smaller amount of words per composition), the crucial fact is that the percentage of pronominal subjects used is similar in all three corpora, ranging from 5.1% to 6.1%, which indicates that the three corpora are comparable in terms of the ratio of syntactic subjects produced. As explained briefly above, UAM Corpus Tool outputs raw frequency statistics for each feature (i.e., for each tag type) and compares them via inferential statistics based on the t-test. Since most statistical analyses presented in this study are based on the raw frequencies produced by our participants, I coded the output of raw frequencies into SPSS and treated them to chi-square analyses to check for significance, as is the standard practice in learner corpus studies using frequency data. 6. Results In this section we will see three types of results regarding the syntactic subject: (i) its syntax and information status, (ii) its person and number, and (iii) its information status and pragmaticality. While result (i) is rather descriptive, it gives us an idea of how learners use subjects from a formal and functional point of view, a fact that has been often overlooked in the literature. Results (ii-iii) address our hypothesis and will reveal learners’ deficits at the syntax-discourse interface. 6.1

Syntax and information status of the subject

In this section we will explore whether, according to the literature, a null pronoun is preferred over an overt pronoun in topic-continuity contexts, yet an overt pronoun is preferred to a null pronoun in topic-shift contexts.



Selective deficits at the syntax-discourse interface 

Table 3.  Group * Syntax of topic SYNTAX OF TOPIC GROUP

NP

PRN

NULL

LOW-ADV

Count % within GROUP

10 2.9%

34 9.9%

298 87.1%

342 100.0%

UPP-ADV

Count % within GROUP

13 4.1%

20 6.3%

285 89.6%

318 100.0%

SPANISH

Count % within GROUP

2 1.2%

3 1.8%

159 97.0%

164 100.0%

100.0%

97.0% 89.6% 87.1%

80.0% Percentage of production

Total

GROUP LOW-ADV UPP-ADV SPANISH

60.0%

40.0%

20.0% 2.9% 4.1% 0.0% NP

9.9% 1.2%

6.3%

1.8%

PRN

NULL

SYNTAX of TOPIC

Figure 3.  Syntax of topic

In topic contexts, most subjects are null pronouns in both the Spanish native corpus (97%) and the learner corpora (89.6% and 87.1% for the upper and lower advanced groups respectively), as expected (see Table 3, Figure 3). Overt pronouns

 Cristóbal Lozano

are marginally used in topic contexts by natives (1.8%) and, to a larger extent, by learners (6.3% and 9.9%). Full NPs are clearly disfavoured to encode topic-continuity (1.2% natives; 4.1% and 2.9% learners). In short, in topic contexts all groups prefer mostly a null pronoun to encode topic-continuity, though overt pronouns and full NPs are marginally used in these cases, as can be visually appreciated in Figure 3. The learners’ slightly higher (and pragmatically inappropriate) production of overt pronouns and full NPs in these contexts is significantly different from the Spanish natives’ production (upperadvanced vs. natives: χ2=8.002, df=2, p=0.018; lower-advanced vs. natives: χ2=12.516, df=2, p=0.002). In Section 6.3 we will come back to the issue of incorrect production of overt pronouns in topic contexts. Regarding topic-shift, Table 4 shows that a change of topic is (i) normally encoded via a full NP (87.2% Spanish, 79.4% upper-advanced, 92.3% lower-advanced), (ii) occasionally encoded via an overt pronoun (12.8% Spanish, 11.1% upper-advanced and 4.4% lower-advanced), and (iii) never by a null pronoun in the native group (0%) and occasionally or seldom by a null pronoun (9.5% upperadvanced and 3.3% lower advanced). The fact that topic-shift is encoded typically via a full NP can be better appreciated visually in Figure 4. The corpus data in this study show that natives as well as learners prefer producing an NP (rather than an overt pronoun) to encode topic-shift. There are no significant differences between the learner groups and the natives (upper-advanced vs. natives: χ2=4.740, df=2, p=0.93; lower-advanced vs. natives: χ2=4.634, df=2, p=0.99). This is a fact that has been largely overlooked both in the theoretical and L2 literature, where it has been assumed that in nullsubject languages topic-shift is normally encoded via an overt pronoun. While the causes for this syntactic preference need serious investigation, they are beyond the scope of this study (see Blackwell 1998 and Reinhart 1995). Table 4.  Group * Syntax of topic-shift SYNTAX OF TOPICSHIFT GROUP

NP

PRN

NULL

Total

LOW-ADV

Count % within GROUP

84 92.3%

4 4.4%

3 3.3%

91 100.0%

UPP-ADV

Count % within GROUP

50 79.4%

7 11.1%

6 9.5%

63 100.0%

SPANISH

Count % within GROUP

41 87.2%

6 12.8%

0 0.0%

47 100.0%



Selective deficits at the syntax-discourse interface 

GROUP LOW-ADV

100.0% 92.3%

Percentage of production

80.0%

UPP-ADV

87.2%

SPANISH

79.4%

60.0%

40.0%

20.0% 11.1% 12.8% 4.4%

9.5% 3.3%

0.0%

0.0% NP

PRN SYNTAX of TOPIC-SHIFT

NULL

Figure 4.  Syntax of topic-shift

6.2

Person and number of the subject

In this section I analyse each person in the pronominal paradigm and check whether its production is pragmatically (un)acceptable. Regarding 1st person singular (yo/pro ‘I’) results are shown in Table 5. All groups’ (natives and learners) production of 1st singular was pragmatically correct (98.1% natives, 100% upper-advanced and 98.2% lower-advanced), there being no significant differences between any of the groups (χ2=2.234, df=2, p=0.327). These results clearly show that learners show a native-like production of 1st singular, i.e., 1st singular does not lead to deficits at the syntax-discourse interface. This lends support to H1 in (11) above. As for 1st person plural (nosotros/nosotras/pro ‘we’), the results are similar to those of 1st singular. Production of 1st plural was pragmatically correct (100% for natives and learners), as Table 6 shows. There are obviously no significant differences between groups, since their production rates are identical (hence a χ2 cannot be performed because there is no variation in the production rates). These results confirm again hypothesis H1, since learners’ native-like production rates clearly indicate that they do not show any pragmatic deficits with 1st plural.

 Cristóbal Lozano

Table 5.  Group * Pragmaticality of 1st singular PRAGMATICALITY 1st sing

Total

PRAGMAT- UNPRAGIC MATIC GROUP

LOW-ADV

Count % within GROUP

108 98.2%

2 1.8%

110 100.0%

UPP-ADV

Count % within GROUP

120 100.0%

0 0.0%

120 100.0%

SPANISH

Count % within GROUP

52 98.1%

1 1.9%

53 100.0%

Table 6.  Group * Pragmaticality of 1st plural PRAGMATICALITY 1st plural

Total

PRAGMAT- UNPRAGIC MATIC GROUP

LOW-ADV

Count % within GROUP

42 100.0%

0 0.0%

42 100.0%

UPP-ADV

Count % within GROUP

27 100.0%

0 0.0%

27 100.0%

SPANISH

Count % within GROUP

22 100.0%

0 0.0%

22 100.0%

Regarding 2nd person singular (tú/pro ‘you’) and 2nd person plural (vosotros/vosotras/pro ‘you all’) all groups’ production is pragmatically correct, though the frequencies are so low in our corpus that little else can be said about 2nd person. In particular, there were only two productions of 2nd singular (n=1 upper-advanced group and n=1 lower-advanced group) and only two of 2nd plural (n=1 upperadvanced group and n=1 lower-advanced group). Spanish natives did not produce any 2nd person at all. Regarding 3rd person singular animate (NP/él/ella/pro ‘NP/he/she’), learners produced a considerable amount of pragmatically incorrect forms (14.9% upperadvanced, 16.7% lower-advanced), compared to the negligible pragmatically incorrect production of the Spanish native group (1%, which corresponds to just 1 token), as Table 7 shows. Given that the observed frequencies in the Spanish group



Selective deficits at the syntax-discourse interface 

Table 7.  Group * Pragmaticality of 3rd singular animate PRAGMATICALITY 3rd sing animate

Total

PRAGMAT- UNPRAGIC MATIC GROUP

LOW-ADV

Count % within GROUP

175 83.3%

35 16.7%

210 100.0%

UPP-ADV

Count % within GROUP

120 85.1%

21 14.9%

141 100.0%

SPANISH

Count % within GROUP

97 99.0%

1 1.0%

98 100.0%

are smaller than 5, a Fisher’s Exact test was performed (instead of Chi-square), which reveals that each learner group significantly differs from the Spanish group (upper-advanced vs. natives: p≤0.001; lower-advanced vs. natives: p≤0.001). By contrast, the upper-advanced group does not differ from the lower-advanced group (p=0.386). These results support H1, since learners clearly show deficits with 3rd singular animate (i.e., they produce a considerably high number of non-pragmatic instances), while the learner groups do not differ from each other. Examples (12) and (13) illustrate overproduction deficits with 3rd person singular animate. In (12) the upper-advanced learner produces an overt pronoun (él ‘he’, [3] [sing] [masc] [anim]) to refer to the previous antecedent mi novio de EEUU ‘my boyfriend from USA’. The overt pronoun is pragmatically redundant, since there is no ambiguity of referent in this situation. Similarly, in (13) another upperadvanced learner produces a full NP (Penelope, [3] [sing] [fem] [anim]) to refer to the previous antecedent (Penelope). The second NP is pragmatically redundant since there is no ambiguity in the discourse. In both (12) and (13) a null pronoun would have been the pragmatically desirable option. (12) [Context: The informant is talking about a recent trip: she went to Seville (Spain) on holiday and her boyfriend came over from the USA to visit her] Cuando me faltaban dos semanas, mi novio de EEUU me visitó unos días para ver la ciudad que me encantaba tanto. Era la primera vez que #él salió de su pais, por eso era un viaje importante. [CPB, upper-advanced, CEDEL2 corpus] ‘Two weeks before I went back, my boyfriend from the USA paid me a visit for a few days to see the city I liked so much. It was the first time that he left his country, that is why it was an important trip.’

 Cristóbal Lozano

(13) [Context: The informant is talking about the film ‘Volver’: Penelope Cruz kills her father and she and her mum hide the corpse in a fridge in their restaurant] ...Penelope limpia todo el sangre y lo ocultan en la nevera del restaurante abajo, que esta cerrado y por venta. #Penélope tiene las llaves por si acaso alguien quiera echar un vistazo para comprarlo. [JEL, upper-advanced, CEDEL2 corpus] ‘Penelope cleans all the blood and they hide it [the corpse] in the fridge in the restaurant below, which is shut down and for sale. Penelope has the restaurant keys in case anyone wants to have a look and buy it.’ Additionally, deficits with 3rd singular animate can be of the underproduction-type. In (14), the production of a null pronoun (pro) is pragmatically ambiguous, as it could refer to two potential antecedents in the immediately preceding context: either mi amiga ‘my (girl) friend’ or su novio-prometido ‘her fiancé’. As it stands, the sentence can have two possible readings, (i) it is the girl friend who wanted to check whether her fiance would like Canada, or (ii) it is the fiancé who wanted to check whether he himself would like Canada. In this context, an overt pronoun would be pragmatically desirable (either él ‘he’ or ella ‘she’, depending on the informant’s intention). (14) [Context: The informant is talking about her best girl friend and her fiancé] Mi amiga me contó que iban a visitar a Canadá este mes para que su novio-prometido conociera a su familia aquí y sus amigos también. Su novio tampoco había estado en Canadá así que #pro quería ver si a él le gustara y cosas así para ver si querían volver a Canadá después de casarse. [KEM, upper-advanced, CEDEL2 corpus] ‘My best friend told me they were going to visit Canada this month so that her fiancé could meet her family here [in Canada] and her friends too. Her fiancé hadn’t been to Canada either so that pro [=he/she] wanted to check whether he like it and stuff, to check whether they wanted to come back to Canada after their marriage.’ By contrast, learners show native-like behaviour with 3rd person singular inanimate (NP/ello/pro ‘NP/it’), Table 8. Learners’ production is pragmatic (97.7% upper-advanced, 100% lower-advanced), similar to Spanish natives’ production (98.7%, except for 1 residual token which represent 1.3% of lack of pragmaticality). As the raw data reveal, the learner groups are not significantly different from the Spanish group. Fisher’s exact test indeed confirms this: upper-advanced vs. natives, p=0.589; lower-advanced vs. natives, p=0.467). These results confirm H1, since learners do not show any pragmatic deficits with 3rd singular inanimate.



Selective deficits at the syntax-discourse interface 

Table 8.  Group * Pragmaticality of 3rd singular inanimate PRAGMATICALITY of 3rd sing inanim

Total

PRAGMAT- UNPRAGIC MATIC GROUP

LOW-ADV

Count % within GROUP

90 100.0%

0 0.0%

90 100.0%

UPP-ADV

Count % within GROUP

43 97.7%

1 2.3%

44 100.0%

SPANISH

Count % within GROUP

78 98.7%

1 1.3%

79 100.0%

Sentence (15) illustrates how an upper-advanced learner produces a pragmatically felicitous sentence with 3rd singular inanimate. There are two potential inanimate antecedents in the preceding context (Sevilla ‘Seville’ and autobús ‘bus’). The informant wants to say that the city was impressive, hence he/she uses the overt NP la ciudad ‘the city’ to avoid ambiguity (otherwise, the use of a null pronoun could be interpreted as (i) it is Seville that was impressive, or (ii) it is the bus that was impressive). Sentence (16) shows how the null pronoun (pro) is used pragmatically to refer to the 3rd person inanimate antecedent El paseo de los ingleses ‘The English avenue’. (15) [Context: The informant is talking about her arrival to Seville (Spain)] Cuando llegué a Sevilla, supe que había un autobús que iba al centro de la ciudad. No pensaba que hubiera ninguno, y por eso, había planeado ir en taxi. La ciudad era muy impresionante. [AK, upper-advanced, CEDEL2 corpus] ‘When I arrived in Seville, I know that there was a bus going to the city centre. I didn’t think there was going to be any, that’s why I had planned on going by taxi. The city was very impressive.’ (16) [Context: The informant is talking about her trip to a university course in France and how she used to walk from her apartment to the university every day] Cada día caminaba de mi apartamento a la universidad por “El paseo de los ingleses”. pro era un camino muy lindo con vistas de hoteles y también el mar azul y claro del Mediterráneo. [ARGL, upper-advanced, CEDEL2 corpus] ‘Every day I used to walk from my apartment to the university via the “The English avenue”. It was a beautiful walk with sights overlooking the hotels and the sky-blue Mediterranean’.

 Cristóbal Lozano

Table 9.  Group * Pragmaticality of 3rd plural animate PRAGMATICALITY of 3rd plu animate

Total

PRAGMAT- UNPRAGIC MATIC GROUP

LOW-ADV

Count % within GROUP

47 90.4%

5 9.6%

52 100.0%

UPP-ADV

Count % within GROUP

102 90.3%

11 9.7%

113 100.0%

SPANISH

Count % within GROUP

34 94.4%

2 5.6%

36 100.0%

Regarding 3rd person plural animate (NP/ellos/ellas/pro ‘NP/they’), Table 9 shows that learners’ production of unpragmatic forms (9.7% upper-advanced, 9.6% lower advanced) is higher than natives’ production (5.6%), though these differences are non-significant (Fisher’s Exact Test: upper-advanced vs. natives: p=0.348; loweradvanced vs. natives: p=0.394). These results indicate that production of 3rd plural animate is largely pragmatic for all groups but learners show a higher (but nonsignificant) percentage of unpragmatic production than natives do. While differences between learners and natives are non-significant, the results lend support to H1 since learners do show some pragmatic deficits with 3rd plural animate. Finally, regarding 3rd plural inanimate (NP/pro ‘they’), learners show nativelike behaviour (Table 10) since they do not produce unpragmatic tokens (0% upper-advanced, 4.5% lower-advanced, which corresponds to only one residual token, which inflates the percentage due to the low number of productions, i.e., 22). Spanish natives do not produce any unpragmatic tokens either (0%). These results confirm that with 3rd plural inanimate, learners’ behaviour is native-like as their production is pragmatic and they do not significantly differ from Spanish natives (since both the Spanish native group and the upper-advanced group produced 100% of pragmatic cases, there are no significant differences; regarding the loweradvanced group, they produced only 1 token which corresponds to 4.5%, but this difference is non-significant when compared to the natives, p=0.710 with Fisher’s Exact Test). These results support H1 as learners do not show any pragmatic deficits with 3rd plural inanimate.



Selective deficits at the syntax-discourse interface 

Table 10.  Group * Pragmaticality of 3rd plural inanimate PRAGMATICALITY of 3rd plu inanim

Total

PRAGMAT- UNPRAGIC MATIC GROUP

LOW-ADV

Count % within GROUP

21 95.5%

1 4.5%

22 100.0%

UPP-ADV

Count % within GROUP

9 100.0%

0 0.0%

9 100.0%

SPANISH

Count % within GROUP

14 100.0%

0 0.0%

14 100.0%

To summarise (Table 11), learners show deficits only with 3rd person animate (in particular in the singular number), as predicted by H1. By contrast, the rest of the pronominal paradigm is intact in our learners’ representation. This suggests that deficits at the syntax-discourse interface are selective and do not affect the whole pronominal paradigm, as stipulated in H1. Table 11.  Summary of learners’ results on person and number

1st 2nd

3rd

SINGULAR

PLURAL

No deficits: Learners behave statistically like Spanish. No deficits: Learners behave statistically like Spanish. [BUT more research needed due to low frequencies]

No deficits: Learners behave statistically like Spanish. No deficits: Learners behave statistically like Spanish. [BUT more research needed due to low frequencies]

Deficits: Animate Learners behave differently from Spanish.

No deficits: Animate Learners behave statistically like Spanish. BUT learners produce more unpragmatic tokens than Spanish (non sig.) No deficits: Inanimate Learners behave statistically like Spanish.

No deficits: Inanimate Learners behave statistically like Spanish.

 Cristóbal Lozano

6.3

Information status and pragmaticality of the subject

In this section I focus on the pragmaticality of the information of status of the subject, i.e., whether the use of the full NP/overt pronoun/null pronoun is pragmatically correct. First, consider topic contexts, where a null pronoun is expected to express topic continuity (Table 12, shown graphically in Figure 5A). Both learner groups (12.3% lower-advanced, 8.8% upper-advanced) produce a considerable amount of unpragmatic topic, which is significantly higher than those of Spanish natives (3%): upper-advanced vs. natives: χ2=5.621, df=2, p=0.018; lower-advanced vs. natives: χ2=11.269, df=2, p=0.001. Notice that the learner groups do not significantly differ from each other (χ2=2.137, df=2, p=0.144). Example (17) illustrates unpragmatic topic: the upper-advanced learner produces an overt pronoun (ellos ‘they’) to refer to the immediate antecedent (los chicos ‘the boys’) in a topic-continuity context which would require a null pronoun, as there is no possible ambiguity. Notice that the learner subsequently uses a pragmatic null pronoun. Similarly, in (18) the lower-advanced learner is talking about la madre ‘the mother’. A null pronoun is expected, as used in the first instance (...pero pro es muy trabajadora ‘but she is hard-working’), but the learner later uses two overt pronouns (ella ‘she’), which are pragmatically redundant. (17) [Context: The informant is talking about a group of teenagers in Ecuador. She was her teacher during her stay there] Cuando me integré en el grupo, en realidad los chicos no podían cantar ni tocar muy bien. Sin embargo, poco a poco a lo largo del año, #ellos se mejoraron bastante y no sólo pro desarrollaron su grupo y sus talentos musicales, sino también pro crecieron como individuos. [ELS, Upper-Advanced, CEDEL2 corpus] ‘When I got into the group, the boys could not really sing or play well. However, little by little during that year, they improved a lot, pro developed their group and their musical skills, and pro grew up as individuals.’ (18) [Context: The informant is talking about the main character of the film “Spanglish”] La madre no puede hablar inglés pero pro es muy trabajadora. #Ella empieza a trabajar... #Ella no puede comunicar[se] con esta familia …. [SMM, Lower-Advanced, CEDEL2 corpus] ‘The mother cannot speak English but pro is very hard-working. She starts working... She cannot communicate with the family...’



Selective deficits at the syntax-discourse interface 

Table 12.  Group * Pragmaticality of Topic PRAGMATICALITY

Total

PRAG TOP UNPRAG TOP GROUP

LOW-ADV

Count % within GROUP

299 87.7%

42 12.3%

341 100.0%

UPP-ADV

Count % within GROUP

290 91.2%

28 8.8%

318 100.0%

SPANISH

Count % within GROUP

159 97.0%

5 3.0%

164 100.0%

Table 13.  Group * Overproduction type with Topic OVERPRODUCTION TYPE

Total

OVERT NP instead of instead of NULL NULL GROUP

LOW-ADV

Count % within GROUP

33 78.6%

9 21.4%

42 100.0%

UPP-ADV

Count % within GROUP

19 67.9%

9 32.1%

28 100.0%

SPANISH

Count % within GROUP

3 60.0%

2 40.0%

5 100.0%

To summarise, while the production of topic subjects is largely correct for all groups, the two learner groups produced a large proportion of unpragmatic topic subjects, which is significantly higher than the Spanish natives’ production. This corroborates H1 in that learners at advanced and very advanced levels of competence show deficits with discursive features like [Topic]. Importantly, recall that the unpragmatic errors with topic can be of two types, namely, (i) the production of an overt pronoun when a null pronoun is required, and (ii) the production of a full NP when a null pronoun is required. Figure 5B (which corresponds to Table 13), shows the percentage of overproduction types with topic, out of the percentages of unpragmatic topic (Figure 5A). As can be appreciated in Figure 5B, all groups overproduce more overt pronouns than full NPs when a null pronoun is required, although the trend of overproduction of overt pronouns decreases towards the native norm, and the trend of overproduction of full NPs increases towards the native norm.

0.0%

20.0%

40.0%

60.0%

80.0%

100.0%

8.8% 3.0%

UNPRAG TOP

12.3%

PRAGMATICALITY of TOPIC

PRAG TOP

87.7%

91.2%

97.0%

Figure 5.  Pragmaticality of Topic

Percentage of production

Figure A

67.9% 60.0%

40.0%

NP instead of NULL

21.4%

32.1%

OVERPROD TYPE with TOPIC

OVERT instead or NULL

78.6%

Figure B

SPANISH

UPP-ADV

GROUP LOW-ADV

 Cristóbal Lozano



Selective deficits at the syntax-discourse interface 

Instances of overproduction of overt pronouns are shown in (19), where the use of an overt pronoun (ellos ‘they’) to refer to the antecedent los chicos ‘the boys’ is pragmatically redundant. A null pronoun would be more adequate, as the learner correctly produces in the next clauses. Additionally, overproduction of full NPs is shown in (19) as well, where the final instance of los chicos ‘the boys’ (in los chicos habían decidido...) is redundant, since los chicos is the topic in the discourse and, therefore, requires a null pronoun (pro), as the learner correctly produces in the previous lines. (19) [Context: The informant is talking about a group of teenagers in Ecuador. She was her teacher during her stay there] Cuando me integré en el grupo, en realidad los chicos no podían cantar ni tocar muy bien. Sin embargo, poco a poco a lo largo del año, #ellos se mejoraron bastante y no sólo pro desarrollaron su grupo y sus talentos musicales, sino también pro crecieron como individuos. Tuvimos un retiro en que hablamos sobre las razones por las cuales #los chicos habían decidido participar en el grupo...[ELS, Upper-Advanced, CEDEL2 corpus] ‘When I got into the group, the boys could not really sing or play well. However, little by little during that year, they improved a lot, pro developed their group and their musical skills, and pro grew up as individuals. We retreated to a quiet place where we talked about the reasons why the boys had decided to participate in the group...’ To summarise, the data on the pragmaticality of topic and the types of overproduction errors with topic indicate that learners significantly overproduce more overt material (mainly overt pronouns but also full NPs) than Spanish natives do in topic continuity contexts. This indicates that advanced and even end-state learners show overproduction deficits at the syntax-discourse interface, as predicted by H1. Consider now the pragmaticality of topic-shift. Recall that to mark a change of topic, overt material is required (either an overt pronoun or a full NP) in native Spanish, which prevents ambiguity between referents. The type of unpragmatic errors with topic-shift relate to underproduction, i.e., production of a null pronoun when overt material is required (cf. sentence (14)). Table 14 (shown visually in Figure 6) indicates that learners produce relatively low percentages of underproduction (7.9% upper-advanced, 3.3% lower-advanced), which corresponds to a few tokens, while Spanish natives never underproduce (0%). Technically, learners behave like natives in topic-shift contexts, as there are no significant differences, but learners do still produce some residual null subjects when an overt pronoun is required (Fisher’s Exact Test: upper-advanced vs. natives p=0.57, just non-significant; lower-advanced vs. natives p=0.283). This lends support to our H1, as learners show underproduction deficits at the syntax-discourse interface in topic-shift contexts, though these deficits are not significantly different from Spanish natives’ zero underproduction rate.

 Cristóbal Lozano

Table 14.  Group * Pragmaticality of Topic-shift PRAGMATICALITY

GROUP

100.0%

PRAG TOPICSHIFT

UNPRAG TOPICSHIFT

LOW-ADV

Count % within GROUP

88 96.7%

3 3.3%

91 100.0%

UPP-ADV

Count % within GROUP

58 92.1%

5 7.9%

63 100.0%

SPANISH

Count % within GROUP

47 100.0%

0 0.0%

47 100.0%

96.7%

GROUP LOW-ADV

100% 92.1%

UPP-ADV SPANISH

80.0% Percentage of production

Total

60.0%

40.0%

20.0% 3.3%

0.0% PRAG TOPSHIFT

7.9%

0.0%

UNPRAG TOPSHIFT

PRAGMATICALITY of TOPIC-SHIFT

Figure 6.  Pragmaticality of Topic-shift



Selective deficits at the syntax-discourse interface 

7. Discussion Results on the non-pragmaticality of topic and topic-shift confirm the general hypothesis that advanced and end-state L2 learners show deficits at the syntax-discourse interface with pronominal subjects. In particular, English-speaking learners of L2 Spanish produce a considerable proportion of unpragmatic subjects in topic contexts, i.e., they produce overt material (mostly overt pronouns, but also full NPs) in topic-continuity contexts where a null pronoun would be pragmatically acceptable. They also show a residual but non-significant amount of underproduction, i.e., production of a null subject in topic-shift contexts requiring overt material. It may be argued that these results are based on written production data from the CEDEL2 corpus and therefore do not reflect directly learners’ spontaneous intuitions as in oral corpora or in experimental work. However, the general findings on overproduction and underproduction with pronominal subjects clearly replicate those of previous research. This indicates that written corpora are a legitimate source for exploring learners’ competence (see Granger, Dagneaux and Meunier 2002, Granger, Hung and Petch-Tyson 2002, Granger and Petch-Tyson 2003). Most L2 Spanish studies reviewed above report on unidirectionality in the type of production, i.e., overproduction (and not underproduction) is the norm, though bidirectionality has been also attested (Montrul and Rodríguez-Louro 2006). This entails that errors of both overproduction and underproduction are found, but there is an asymmetry, since overproduction errors are significantly greater than underproduction errors. The results of the current study show bidirectionality, though statistically significant differences between learners and natives were found only with overproduction (and not with underproduction). Sorace (2006) argues that deficits at the syntax-discourse interface in Italian L1 attrition are unidirectional (overproduction only) as a result of processing deficits at the interfaces. In particular, overt pronominal subjects can be used as the ‘default’ processing option when the language processor is overloaded as it tries to integrate both syntactic and interface information. Sorace and Filiaci (2006) also report unidirectionality in L2 Italian, which is claimed to be a result of representation deficits. In particular, [+interpretable] features like [Topic] at the syntax-discourse interface are vulnerable. Whatever the causes (whether processing deficits or incomplete representations), unidirectionality has been also observed in native Spanish speakers, who tend to produce some overt pronominal subjects in topic-continuity contexts (Alonso-Ovalle et al. 2002). Our data also show that Spanish natives do produce some unpragmatic overt material in topic-continuity contexts (see Table 12 and Table 13). This is surprising, as the theory would predict that a null pronoun is the pragmatically preferred choice to mark topic continuity. Further research will

 Cristóbal Lozano

need to elucidate whether L2 learners overproduce overt subjects as a consequence of a supposedly universal default mechanism. Results on phi-features have shown that learners’ behaviour is statistically similar to Spanish natives regarding 1st person (singular and plural), 2nd person (singular and plural) and 3rd person inanimate. By contrast, learners behave significantly different from natives as regards 3rd person animate singular (and, to a lesser extent, plural). In other words, it seems that learners show robust knowledge at the syntax-discourse interface with the deictic use of pronouns (i.e., when these refer to speech-act participants: 1st and 2nd person) and with inanimacy (3rd neutral). Vulnerability at the interface is observed only with 3rd person animate, i.e., with the anaphoric use of the pronouns. This indicates that deficits with discursive features like [Topic] and [Topic-shift] at the syntax-discourse interface are selective and do not affect the whole pronominal paradigm, as certain features (3rd animate) are vulnerable, while the rest appear to be relatively robust. Montrul (2006) compares incomplete acquisition in heritage speakers with other cases of language change in progress in Brazilian Portuguese (BP). Interestingly, BP is shifting towards the English value of the null subject parameter (i.e., a non-null subject language) and the most affected person is 3rd singular, as reported in this study. Additionally, the observed syntax-discourse deficits are mainly a matter of overproduction, since overt pronouns, which are specified as [Topic-shift], can be occasionally specified for [Topic]. Such deficits interact with participant features of the Feature Geometry, as deficits affect the [non-participant] feature. Table 15 shows a likely mental representation of discursive pronominal features for both Spanish natives and L2 Spanish learners based on the results. The discursive feature [Topic] is realised in Spanish via a null pronoun (pro) unspecified for person, number and animacy, i.e., the null pronoun can refer to any person, number and animate or inanimate entities. English-speaking learners of L2 Spanish also realise the [Topic] feature mainly via a null pronoun for the whole pronominal paradigm, but they show a residual deficit in the sense that such discursive [Topic] can be realised via an overt pronoun when it is specified for [3] and [+animate]. In topicshift contexts, Spanish natives as well as learners realise the discursive feature [Topic-shift] as an overt pronoun for the whole pronominal paradigm. As indicated by the round brackets, learners may exceptionally encode [Topic-shift] via a null pronoun with a certain combination of phi-features: [3] and [+animate]. This last unpragmatic choice shows a exceptionally low frequency in the learners and it is not significantly different from Spanish natives (who never produce it).



Selective deficits at the syntax-discourse interface 

Table 15.  Likely mental representation of discursive pronominal features

Spanish natives Spanish L2ers

Topic contexts

Topic-shift contexts

Discurs. Feature [Topic]

Realisation

Discurs. Feature [Topic-shift]

Realisation

[Topic]

pro [upers][unum][uanim] overt [3][+ anim]

[Topic-shift]

overt [upers][unum][uanim] pro [3][+ anim]

pro [upers][unum][uanim]

overt [upers][unum][uanim]

Recall that in L1 acquisition the relative order of acquisition is typically 1st and 3rd inanimate, then 3rd animate. This path of development is claimed to be guided by the Feature Geometry, which is ultimately constrained by UG. Our L2 data suggest that English-speaking adult learners’ knowledge of L2 Spanish may be also guided by such Feature Geometry in the sense that (i) 1st and 3rd inanimate must have been acquired before end-states as they do not appear to lead to deficits at the syntax-discourse interface in our learners, while (ii) 3rd animate is acquired later and can lead to residual deficits at the interface, as observed. As it stands, this proposal needs further empirical corroboration by means of L2 developmental studies testing (i) how the Feature Geometry interacts with discursive features like topic and topic-shift in L2, and (ii) whether the observed path of development in child L1 acquisition is also observed in adult L2 acquisition. In the context of general theories of SLA, recall from the introduction that most proposals have focused on formal (morphosyntactic) features operating at narrow syntax, and not on discursive features like [Topic] and [Focus] operating at the syntax-discourse interface. Despite their original aim, these theories centre around two opposing views, as the source of deficits can be of two general types: representational vs. mapping deficits. Let us see in turn how these theories can account for our data. The representational approach stipulates that learners’ linguistic representation of the relevant features show deficits of various kinds. Beck’s (1998) Local Impairment Hypothesis states that strong (i.e., uninterpretable) features become unspecified, which leads to a permanent state of unconstrained optionality. If this were the case, we would expect our learners to show a more random behaviour and produce optionally both overt and null pronominals in topic contexts and topic-shift contexts, which is contrary to our results. Hawkins and Chan’s (1997) Failed Functional Feature Hypothesis stipulates that L2 uninterpretable formal features become defective if they have not been acquired before the Critical Period.

 Cristóbal Lozano

This proposal is difficult to test in the current study since the learner sample contains both pre- and post-pubescent learners (see Table 16 and Table 17) The mapping approach claims that learners’ linguistic representation of features is intact and that the observed deficits are a result of performance when mapping linguistic knowledge onto the relevant morphophonological forms, i.e., the learner cannot retrieve the required morphological item (Prévost and White 2000). If this were the case, we would expect our learners to show the inverse deficit, i.e., they would underproduce null subjects in topic-shift contexts significantly more than they do now since they would supposedly have surface problems when mapping the relevant features onto the correct pronominal form and thus retrieve a null form (pro). Additionally, we would expect our learners not to overproduce overt material in topic-continuity contexts, which is contrary to fact. Current approaches to the problem of the interfaces in language acquisition (e.g., Sorace 2004, 2005, 2006) postulate that it is interpretable discursive features like [Topic] that are vulnerable since they operate at the interface between the computational system and other systems of thought. This is a plausible explanation for the current data but (i) it cannot account for why deficits are selective, i.e., they are observed only with 3rd person animate subjects and not with the rest of the paradigm, and (ii) it would entail that the problem is discursive (i.e., learners are unable to acquire the discourse-related features [Topic] and [Topic-shift] to a native-like extent) rather than a problem with the realisation of these discursive features onto the correct pronominal form. There is no principled reason to believe that advanced learners are unable to interpret the information status of [Topic] as representing old information and [Topic-shift] as a change of discourse referent, since information status is a cross-linguistic universal (e.g., Casielles-Suarez 2004, Vallduví 1992, Vallduví and Engdahl 1996). It seems more reasonable to assume that the learners’ deficits are a result of the constraints UG imposes on the Pronominal Feature Geometry. Further research will need to investigate more closely the locus of deficits with discursive features and how these interact with the Feature Geometry. Finally, the conclusion that deficits at the syntax-discourse interface with pronominal subjects are selective in L2 Spanish must be taken cautiously since the evidence presented here comes from corpus data. Two of the well-known limitations of data-driven approaches to language acquisition is that (i) performance data does not guarantee a genuine reflection of learners’ competence and (ii) biases in the corpus sample can skew the data (as is the case in this study with 2nd person). Corpus results must be complemented with detailed experimental work. Future research needs to address whether the observed selective impairment in production data is also attested in comprehension data and, if so, what is the ultimate source of the observed deficits.



Selective deficits at the syntax-discourse interface 

8. Conclusion This study has shown that English-speaking learners of L2 Spanish at advanced and very advanced levels of competence show deficits with the properties that constrain the distribution of overt and null pronominal subjects at the syntaxdiscourse interface, as previous SLA research has shown. Unlike previous SLA research, it has been shown that such deficits are selective, as they do not affect equally all phi-features in the pronominal paradigm, but rather a subset of them. It is concluded that the observed deficits stems from the way Universal Grammar constrains pronominal features. References Ädel, A. 2006. Metadiscourse in L1 and L2 English. Amsterdam: John Benjamins. Alexiadou, A. & Anagnostopoulou, E. 1998. Parametrizing Agr: word order, verb- movement and EPP-checking. Natural Language and Linguistic Theory 16(3): 491–539 Al-Kasey, T. & Pérez-Leroux, A.-T. 1998. Second language acquisition of Spanish null subjects. In The generative study of second language acquisition, S. Flynn, G. Matohardjono & W. O’Neil (eds.), 161–185. Hillsdale NJ: Lawrence Erlbaum. Alonso-Ovalle, L., Fernández-Solera, S., Frazier, L. & Clifton, C. 2002. Null vs. overt pronouns and the topic-focus articulation in Spanish. Journal of Italian Linguistics 14: 151–169. Beck, M.-L. 1998. L2 acquisition and obligatory head movement: English-speaking learners of German and the local impairment hypothesis. Studies in Second Language Acquisition 20: 311–348. Belletti, A. & Leonini, C. 2004. Subject inversion in L2 Italian. In EUROSLA Yearbook 4, S. Foster-Cohen, M.A. Sharwood Smith, A. Sorace & M. Ota (eds.), 95–118. Amsterdam: John Benjamins. Benveniste, E. 1971. Problems in general linguistics. In The nature of pronouns, E. Benveniste (ed.). Coral Gables FL: University of Miami Press. Bianchi, V. 2005. The person asymmetry: Underspecification of person and number features? Paper presented at the workshop ‘Underspecification in morphology and syntax’, University of Cologne, Germany. Blackwell, S. E. 1998. Constraints on Spanish NP anaphora: The syntactic versus the pragmatic domain. Hispania 81: 606–618. Bloomfield, L. 1933. Language. New York NY: Holt, Rinehart and Winston. Casielles-Suarez, E. 2004. The syntax-information structure interface: Evidence from Spanish and English. Oxford: Routledge. Cornish, F. 2006. Discourse anaphora. In Encyclopedia of language and linguistics, 2nd edn, K. Brown (ed.), 631–638. Oxford: Elsevier. Davies, W. D. 1996. Morphological uniformity and the null subject parameter in adult SLA. Studies in Second Language Acquisition 18: 475–493. Diaconescu, C. R. & Goodluck, H. 2002. Structural and discourse factors in Romanian L2 English Learners’ interpretation of pronouns. In Proceedings of the 6th Generative Approaches to

 Cristóbal Lozano Second Language Acquisition Conference (GASLA 2002), J. M. Liceras, H. Zobl & H. Goodluck (eds.), 71–75. Somerville MA: Cascadilla Press. Escutia, M. 2002. Universal Grammar, transfer and variability: A case study. Estudios Ingleses de la Universidad Complutense 10: 67–85. Fernández-Soriano, O. 1989. Strong pronouns in null-subject languages and the Avoid Pronoun Principle. MIT Working Papers in Linguistics 11: 228–239. Fernández-Soriano, O. 1993. La visión paramétrica del lenguaje: Más sobre los sujetos y objetos nulos. In La lingüística y el análisis de los sistemas no natives, J. M. Liceras, (ed.), Ottawa: Dovehouse. Fernández Soriano, O. 1999. El pronombre personal. Formas y distribuciones. In Pronombres átonos y tónicos, V. Demonte & I. Bosque (eds.), 1209–1273. Madrid: Espasa-Calpe. Forchheimer, P. 1953. The category of person in language. Berlin: Walter de Gruyter. Fruit, M.N. 2007. Divergence at the syntax-discourse interface: Evidence from the L2 acquisition of contrastive focus in European Portuguese. Paper presented at ISB6 (6th International Symposium on Bilingualism), University of Hamburg. Grinstead, J. 2004. Subjects and interface delay in child Spanish and Catalan. Language 80: 40–72. Granger, S., Dagneaux, E. & Meunier, F. (eds.) 2002. International corpus of learner English [inc. CD ver 1.1]. Louvain: UCL Presses Universitaires de Louvain. Granger, S., Hung, J. and Petch-Tyson, S. (eds.) 2002. Computer learner corpora, Second language acquisition and foreign language teaching. Amsterdam: John Benjamins. Granger, S. & Petch-Tyson, S. (eds.) 2003. Extending the scope of corpus-based research: New applications, new challenges. Amsterdam: Rodopi. Hanson, R. 2000. Pronoun acquisition and the morphological feature geometry. Calgary Working Papers in Linguistics 22: 1–14. Harley, H. & Ritter, E. 2002a. Person and number in pronouns: A feature-geometric analysis. Language 78: 482–526. Harley, H. & Ritter, E. 2002b. Structuring the bundle: A universal morphosyntactic feature geometry. In Pronouns: Grammar and representation, H. J. Simon & H. Weise, (eds.), 23–39, Amsterdam: John Benjamins. Hawkins, R. 2001. Second language syntax: A generative introduction. Oxford: Blackwell. Hawkins, R. & Chan, C. Y. 1997. The partial availability of Universal Grammar in second language acquisition: The ‘failed functional features hypothesis’. Second Language Research 13: 187–226. Helland, C. 2004. Attrition and syntactic subjects in Catalan. Paper presented at The Romance Turn, Universidad Nacional de Educación a Distancia (Madrid, September). Hertel, T.J. 2003. Lexical and discourse factors in the second language acquisition of Spanish word order. Second Language Research 19: 273–304. Jaeggli, O. 1982. Topics in Romance syntax. Dordrecht: Foris. Jespersen, O. 1924. The philosophy of grammar. London: George Allen and Unwin. Kayne, R.S. 2000. Person morphemes and reflexives in Italian, French, and related languages. In Parameters and universals, R. S. Kayne (ed.), 131–162. Oxford: OUP. Kanno, K. 1997. The acquisition of null and overt pronominals in Japanese by English speakers. Second Language Research 13: 265–287. Kras, T. 2006. Interface instability in bilingual language acquisition: In search of the causes. Paper presented in EUROSLA 2006, 13–16 September, Turkey.



Selective deficits at the syntax-discourse interface  Lafond, L., Hayes, R. & Bhatt, R. 2001. Constraint demotion and null-subjects in Spanish L2 acquisition. In Romance syntax, semantics and L2 acquisition, J. Camps & C. Wiltshire, (eds.), 137–153, Amsterdam: John Benjamins. Liceras, J. M. 1989. On some properties of the “pro-drop” parameter: looking for missing subjects in non-native Spanish. In Linguistic perspectives on second language acquisition, S. M. Gass & J. Schachter (eds.), 109–133. Cambridge: CUP. Liceras, J. M. & Díaz, L. 1999. Topic drop versus pro-drop: Null subjects and pronominal subjects in the Spanish L2 of Chinese, English, French, German and Japanese speakers. Second Language Research 15: 1–40. Liceras, J. M, Zobl, H. & Goodluck, H. 2007. The role of formal features in second language acquisition. Mahwaw NJ: Lawrence Erlbaum Associates. López Ortega, N. R. 2006. Presencia y ausencia de sujeto en el discurso oral de español no-nativo: Un estudio cualitativo longitudinal en contexto de inmersión. In Perspectivas interdisciplinares de la lingüística aplicada, Tomo 1, M.L. Carrió Pastor (ed.), 175–186. Valencia: Universidad Politécnica de Valencia. Lozano, C. 2002a. Knowledge of expletive and pronominal subjects by learners of Spanish. ITL Review of Applied Linguistics 135/6: 37–60. Lozano, C. 2002b. The interpretation of overt and null pronouns in non-native Spanish. Durham Working Papers in Linguistics 8: 53–66. Lozano, C. 2003. Universal Grammar and focus constraints: The acquisition of pronouns and word order in non-native Spanish. PhD dissertation, University of Essex. Lozano, C. 2006a. Explaining the “syntax-before-discourse” phenomenon: Pronominal subject distribution in L1 Greek-L2 Spanish. Paper presented at The Romance Turn II, University of Utrecht, 7th September. Lozano, C. 2006b. Focus and split intransitivity: The acquisition of word order alternations in non-native Spanish. Second Language Research 22: 1–43. Lozano, C. 2006c. The development of the syntax-discourse interface: Greek learners of Spanish. In The acquisition of syntax in Romance languages, V. Torrens & L. Escobar (eds.), 371–399. Amsterdam: John Benjamins. Luján, M., 1999. Expresión y omisión del pronombre personal. In Gramática descriptiva de la lengua española, I. Bosque & V. Demonte (eds.), 1275–1315. Madrid: Espasa-Calpe. Mastropavlou, M. 2006. The role of phonological salience and feature interpretability in the grammar of typically developing and language impaired children. PhD dissertation, Aristotle University of Thessaloniki. Montrul, S. 2004. Subject and object expression in Spanish heritage speakers: A case of morphosyntactic convergence. Bilingualism: Language and Cognition 7: 125–142. Montrul, S. 2006. Bilingualism, incomplete acquisition and language change. In L2 Acquisition and creole genesis. Dialogues, C. Lefebvre, L. White & C. Jourdan (eds.), 379–400. Amsterdam: John Benjamins. Montrul, S. & Rodríguez-Louro, C. 2006. Beyond the syntax of the Null Subject Parameter: A look at the discourse-pragmatic distribution of null and overt subjects by L2 learners of Spanish. In The acquisition of syntax in Romance languages, V. Torrens & L. Escobar (eds.), Amsterdam: John Benjamins. Pérez-Leroux, A. T. & Glass, W. R. 1997. OPC effects on the L2 acquisition of Spanish. In Contemporary perspectives on the acquisition of Spanish, Vol. 1: Developing grammars, A. T. Pérez-Leroux & W. R. Glass, (eds.), 149–165. Sommerville MA: Cascadilla Press.

 Cristóbal Lozano Pérez-Leroux, A. T. & Glass, W. R., 1999. Null anaphora in Spanish second language acquisition: Probabilistic versus generative approaches. Second Language Research 15: 220–249. Pérez-Leroux, A.T., Scott, A.M., Hertel, T., Kellar, V. & Glass, W.R. 1999. Sources of knowledge in second language acquisition. Spanish Applied Linguistics 3: 33–63. Phinney, M. 1987. The pro-drop parameter in second language acquisition. In Parameter setting, T. Roeper & E. Williams (eds.), 221–238. Dordrecht: Reidel. Picallo, M. C. 1994. Catalan possessive pronouns: The avoid pronoun principle revisited. Natural Language and Linguistic Theory 12: 259–299. Pinto, M. 2006. Subject pronouns in bilinguals: Interference or maturation? In The acquisition of syntax in Romance languages, V. Torrens & L. Escobar, (eds.), 331–350. Amsterdam: John Benjamins. Polio, C. 1995. Acquiring nothing? The use of zero pronouns by nonnative speakers of Chinese and the implications for the acquisition of nominal reference. Studies in Second Language Acquisition 17: 353–377. Prévost, P. & White, L. 2000. Missing surface inflection or impairment? Evidence from tense and agreement. Second Language Research, 16(2): 103–133. Reinhart, T. 1995. Interface strategies. OTS Working Papers, TL-95–002. Rigau, G. 1986. Some remarks on the nature of strong pronouns in null-subject languages. In Generative studies in Spanish syntax, I. Bordelois, H. Contreras & K. Zagona (eds.), Dordrecht: Foris. Rizzi, L. 1982. Issues in Italian syntax. Dordrecht: Foris. Rizzi, L. 1997. A parametric approach to comparative syntax: properties of the pronominal system. In The new comparative syntax, L. Haegeman (ed.), 268–285. London: Longman. Rochemont, M.S. 1998. Phonological focus and structural focus. In The limits of syntax [Syntax and Semantics 29], P.W. Culicover & L. McNally (eds.), 337–363. London: Academic Press. Satterfield, T. 2003. Economy of interpretation: Patterns of pronoun selection in transitional bilinguals. In Effects of the second language on the first, V. Cook, (ed.), 214–233. Clevedon: Multilingual Matters. Saxena, A. 2006. Pronouns. In Encyclopedia of language and linguistics, 2nd edn, K. Brown (ed.), 131–133. Oxford: Elsevier. Serratrice, L. 2004. Anaphoric interpretation of null and overt pronominal subjects in Italian. Paper presented at The Romance Turn, Universidad Nacional de Educación a Distancia, Madrid, September. Serratrice, L., Sorace, A. & Paoli, S. 2004. Transfer at the syntax-pragmatics interface: Subjects and objects in Italian-English bilingual and monolingual acquisition. Bilingualism: Language and Cognition 2: 169–186. Sorace, A. 2004. Native language attrition and developmental instability at the syntax-discourse interface: Data, interpretations and methods. Bilingualism: Language and Cognition 7: 143–145. Sorace, A. 2005. Selective optionality in language development. In Syntax and variation: Reconciling the biological and the social, L. Cornips & K. P. Corrigan (eds.), 55–80. Amsterdam: John Benjamins. Sorace, A. 2006. Possible manifestations of shallow processing in advanced second language speakers. Applied Psycholinguistics 27: 88–91. Sorace, A. & Filiaci, F. 2006. Anaphora resolution in near-native speakers of Italian. Second Language Research 22: 339–368.



Selective deficits at the syntax-discourse interface  Tsimpli, I.-M. & Sorace, A. 2006. Differentiating interfaces: L2 performance in syntax-semantics and syntax-discourse phenomena. BUCLD Proceedings 30: 653–664. Tsimpli, I.-M., Sorace, A., Heycock, C. & Filiaci, F. 2004. First language attrition and syntactic subjects: A study of Greek and Italian near-native speakers of English. International Journal of Bilingualism 8: 257–277. University of Wisconsin, 1998. The University of Wisconsin college-level placement test: Spanish (grammar) form 96M. Madison WI: University of Wisconsin Press. Valenzuela, E. & McIlwraith, E. 2007. Topic and focus in early bilingual grammars. Paper presented at ISB6, University of Hamburg. Vallduví, E. 1992. The informational component. New York NY: Garland. Vallduví, E. & Engdahl, E. 1996. The linguistic realization of information packaging. Linguistics 34: 459–519. White, L. 1985. The “pro-drop” parameter in adult second language acquisition. Language Learning 35: 47–62. White, L. 1989. Universal grammar and second language acquisition. Amsterdam: John Benjamins. White, L. 2003. Second language acquisition and Universal Grammar. Cambridge: CUP. Zubizarreta, M. L. 1998. Prosody, focus, and word order. Cambridge MA: The MIT Press. Zubizarreta, M. L. 1999. Las funciones informativas: tema y foco. In Gramática descriptiva de la lengua española, Vol.3, I. Bosque & V. Demonte (eds.). Madrid: Espasa.

Appendix Table 16.  Upper-advanced group’s data Initials

Profic

Age

ELS JSD KEM KEM2 OPE CPB MEA JGP LP JEL

100% 100% 100% 100% 98% 98% 98% 98% 98% 98%

25 20 28 28 48 21 52 58 51 20

99% 35 (mean) (mean)

Age first Years Stay # words # tags exposure instruct (yrs;mths) 15 12 20 20 15 11 3 13 16 13

5 9 8 8 11 11 ? 7 6 8

18 13 10 10 17 3 204 6 4 7

647 819 1032 866 798 1010 865 554 781 816

14 (mean)

(mean)

29 (mean)

818 (total)

40 41 65 50 43 45 38 27 46 58 453 (total)

% pronom subjects 6.2% 5.0% 6.3% 5.8% 5.4% 4.5% 4.4% 4.9% 5.9% 7.1% 5.5% (mean)

 Cristóbal Lozano

Table 17.  Lower-advanced group’s data Initials

Profic

Age

ARGL SAR DKH AK SMG KMH CCO SMM CN ACC

95% 95% 95% 95% 93% 93% 93% 91% 91% 91%

49 58 28 23 19 21 20 18 55 25

Age first Years Stay # words # tags exposure instruct (yrs;mths)

93% 32 (mean) (mean)

16 20 6 13 16 10 14 13 26 12

? 3 15 10 3 11 5 6 3 2

9 12 3 1 1 4 0 0 228 7

15 (mean)

6 (mean)

27 (mean)

981 858 870 907 840 834 849 705 871 806 8521 (total)

63 46 65 63 65 47 44 50 42 43 528 (total)

% pronom subjects 6.4% 5.4% 7.5% 6.9% 7.7% 5.6% 5.2% 7.1% 4.8% 5.3% 6.2% (mean)

Table 18.  Spanish native group’s data Initials

Age

# words

# tags

MSCL GDC CAC MCL AHN ENB MHMR CLB RSZ MDD SPH CMM

33 51 33 40 33 39 56 32 26 32 30 33

642 392 377 667 470 442 335 456 538 422 372 841

17 15 17 38 30 21 17 17 27 28 32 40

37 (mean)

5954 (total)

299 (total)

% pronom subjects 2.6% 3.8% 4.5% 5.7% 6.4% 4.8% 5.1% 3.7% 5.0% 6.6% 8.6% 4.8% 5.1% (mean)

Clitic doubling and clitic left dislocation in Spanish and Greek L2 grammars Teresa Parodi

University of Cambridge Research Centre for English and Applied Linguistics Morphological properties in L2 acquisition have a prominent role in the current research agenda, particularly with respect to the optionality observed and what this tells us about learners’ representations. Some authors (Smith and Tsimpli 1995, Hawkins and Chan 1997, Hawkins and Franceschina 2004, among others) argue for representational deficits in learners’ systems, particularly with respect to uninterpretable features. Clitics, as manifestations of agreement and bearers of uninterpretable features, offer a promising testing ground. This paper contributes to the debate by observing clitic doubling and clitic left dislocation in L2 Spanish and Greek when English is the L1. Differences in morphological richness and in the properties of their pronominal systems allow us to test how learners interpret clitics in the L2.

1. On L2 acquisition: optionality, learnability and features There is a long research tradition on variation in learners’ systems. Within the generative framework, however, it has become a focus of research relatively recently and has been placed within the more general question of how to account for optionality. In this framework grammars are assumed to offer categorical options, with only one candidate out of a set being grammatical and excluding others. Faced with optionality in learners’ systems the question arises as to whether this optionality is ‘real’ or apparent. In this latter case the learners may have a systematic grammar, but based on criteria which differ from those valid in the L2. The former case, on the other hand, would represent true randomness. Optionality may have different sources, related on the one hand to the mechanism of feature checking (Chomsky 1995 and later work) and, on the other, to the nature of the L2 acquisition process. With respect to the former, it is important to bear in mind the distinction between interpretable (i.e., visible at LF, with semantic

 Teresa Parodi

content) and uninterpretable features (i.e., not visible at LF). Checking is a syntactic operation: features of lexical entries are checked against those of terminal nodes. As a result, uninterpretable features are erased from the derivation to LF. In a developmental perspective, it has been claimed that at early acquisitional stages features may not be fully specified or may be specified with values that do not correspond to those of the steady state. Consider, for instance, Hyams’s (1996) use of underspecification to account for root infinitives in child language, or Hawkins and Chan’s (1997) Failed Functional Features Hypothesis for L2 acquisition. Hawkins and Chan (1997) and, in the same line, Hawkins and Franceschina (2004) claim that L2 grammars suffer from representational deficits, particularly with respect to uninterpretable features. Hawkins and Franceschina (2004) study the acquisition of gender concord in French and Spanish, contrasting L1 and L2 acquisition. Following Carstens (2000), they view gender concord as a morphophonological reflex of feature checking, specifically of uninterpretable gender features. Carstens (2000) offers an account of feature checking based on Chomsky (1995). In this account, D and A have uninterpretable gender features, [ugender], which are checked against the interpretable [+/-fem] feature of N and erased if they match. This is reflected in the phonological forms inserted. On the basis of existing studies of L1 acquisition of French and Spanish (Karmiloff-Smith 1979 for French, Müller 1998 for bilingual German-French, Pérez-Pereira 1991), Hawkins and Franceschina (2004) observe that, although children are able to make consistent gender choices of determiner for nonce Ns, these choices are based on the phonological form of the N and not on checking of uninterpretable features on D. They claim that uninterpretable [ugender] on D develop at a later stage in L1 acquisition. According to these authors, in L2 acquisition all post-childhood learners go through an initial phase during which uninterpretable [ugender] is not part of their representation. Learners whose L1 does not have these features are unable to establish them in the L2; on the other hand, those whose L1 does have them will be able to rely on their L1 and become target like in the relevant domain. Hawkins and Franceschina (2004) present data from English and Italian learners of Spanish and show that the latter become targetlike with respect to gender concord. The English learners, on the other hand, seem to rely on phonological information for one of the gender forms and overgeneralise the other. Prévost and White (2000) and White (2003) put forward a different view of the source of optionality and mismatches between learners’ systems on the one hand and both their L1 and the L2 on the other. As opposed to the previous view, in this perspective the problem lies in the mapping between lexicon and overt realisation, in other words, a PF-phenomenon. Prévost and White (2000) base their claim on evidence from tense and agreement: in the learners’ production, non-finite forms occur in finite contexts but not vice versa. Furthermore finite forms are



Clitics in Spanish and Greek 

used accurately, both in terms of form and of distribution. This is taken as an indication that the learners have knowledge of the relevant tense and agreement features and that their representation is not impaired. The hypothesis predicts that underspecified forms will occur instead of fully specified ones, but not vice versa. On the background of this discussion, the current study examines the L2 acquisition of clitics in Spanish and Greek by speakers of English. Spanish and Greek, on the one hand, and English, on the other, differ in aspects which make them suitable for the acquisitional debate mentioned above. In the first place, Spanish and Greek have more overt morphology than English. They also differ in their pronominal systems: Spanish and Greek, unlike English, have object clitic pronouns which are, arguably, functional heads consisting of Case and AGR-O features, on a V-related projection (Uriagereka 1995, Sportiche 1996, Parodi 1998). That clitics are verbal rather than nominal categories, distinguishes them from English pronouns in a variety of morphosyntactic and semantic ways (Cardinaletti and Starke 1994). Differences in feature specification of pronouns versus clitics can lead to the misanalysis of clitics as pronouns and have consequences on their distribution, given that different conditions hold for them. Parodi and Tsimpli (2005) studied clitics, pronouns and gaps in Null Operator Structures in L2 Spanish and Greek on the one hand and L2 English on the other, taking into account finiteness (finite and non-finite contexts). Relevant for the current study is their section on L2 Spanish and Greek. Parodi and Tsimpli (2005) found true optionality only at lower proficiency levels for both groups. For more advanced learners, however, there was a systematicity, although not targetlike: speakers of English learning Spanish or Greek seemed to assign to clitics an interpretation compatible with that of English strong pronouns. In what follows I will present aspects of the (morpho)syntax and semantics of Spanish and Greek clitics relevant for the current discussion. Particularly, I will deal with cases in which Spanish and Greek allow for a clitic to co-occur with a coreferent NP, namely clitic doubling and clitic left dislocation. 2. On clitics There is extensive research on the status of clitics in languages which, as Spanish and Greek, allow for clitic doubling. In a nutshell, the question is whether clitics are arguments of the verb, given that they appear preverbally, i.e., not with the normal distribution of arguments. This is the line taken in studies such as Kayne (1975) and Strozer (1976), among other more recent ones. An alternative is to consider them agreement markers, specifically markers of object agreement (Suñer

 Teresa Parodi

1988, Franco 1993, 2000, Torrego 1995, Sportiche 1996, Parodi 1998, Anagnostopoulou 1994, 1999). I will follow the latter line, analysing Spanish and Greek clitics as instances of object agreeement in a continuum with inflectional affixes at one end and pronouns on the other. This assumption leads to related questions. One of them is that of the place of clitics in the sentence structure. Although the analyses vary, many authors suggest that clitics head their own projections: that is the case in Uriagereka (1995) and in Sportiche’s (1996) proposal of Clitic Voices. Analyses such as Torrego (1995), Tsimpli (1999), Franco (2000), among others, also assume clitics to be heads of a functional projection, but in this case the assumption is that they merged in AgrO. I will follow the latter line which seems more economical in that there is no reduplication of functional projections.1 Another aspect of clitics discussed in this context is whether they are independent syntactic objects or affixes. Following Franco (2000) and Parodi (1998) I will claim that, as part of the agreement morphology, they are part of a continuum with pronouns at one end and inflectional affixes at the other. Clitics may occupy different places in this continuum in different languages or dialects, or at different diachronic stages. It is worth noting that the existence of clitic doubling in Spanish and Greek does not mean that it is obligatory. Several factors play a role here: the syntactic category of the double (pronoun or full NP), case (accusative, dative, genitive), properties of the NP such as definiteness and specificity. In what follows I will give an overview of the conditions under which clitics co-occur with a coreferent NP in Spanish and Greek. The aim of this overview is to illustrate the claim that they are instances of object agreement. 2.1

Clitic doubling in Spanish

Clitic doubling is obligatory when the coreferent element is a pronoun, both in accusative and dative contexts, as shown in the Spanish examples in (1) and (2). (1) accusative *(la) veo a ella acc.3fs see.1s P her ´I see her´ (2) dative *(le) doy la carta a ella dat.3s give.1s the letter P her ´I give the letter to her´

1. For the purposes of the current paper not much hinges on the exact analysis, as long as they are compatible with the claim that we are dealing with agreement markers.



Clitics in Spanish and Greek 

With Spanish full NPs, clitic doubling is strongly preferred, if not obligatory, in the dative (independently of the properties of the NP, e.g., whether the NP is definite or specific or animate). (3) le doy la carta al vecino ´I give the letter to the neighbour´ (4) …a un vecino P a neighbour

+definite, +specific -definite, -specific

In the accusative, however, dialects of Spanish vary in their acceptance of clitic doubling. While it is disallowed in European Spanish, it is possible, for example, in Río de la Plata Spanish, provided that the NP is definite as in (5); otherwise it is disallowed as in (6)2. (5) la veo a la mujer 3accfs see.1s P the woman ´I see the woman´

+def

(6) *…a una mujer P a woman ´…a woman´

-def

2.2

Clitic doubling in Greek

In Greek the use of clitic doubling is unrestricted in the genitive, but again restricted to definites in the accusative, as shown in the examples under (7) and (8). (7)

genitive, +definite o Pétros tis chárise éna portofóli tis Elénis the Peter-nom 3fs-gen gave a purse the Helen-gen ´Peter gave a purse as a present to Helen´

(8) a.

accusative, +definite i Eléni ta diábase ta biblía pou tis ícha dósi the Helen 3accnpl read.past the books that 3genfs had given.1sg ´Helen read the books I gave her´

2. This is a simplification, since specificity also plays a role. Clitic doubling is impossible with a definite, but non-specific NP as in (i). (i) la veo a la mujer que sepa turco 3accfs see.1s P the woman who know.subj Turkish ´I see the woman who might know Turkish´

 Teresa Parodi

b.

accusative, -definite *O Kóstas to agórase chthes ena avtokíneto the Costas 3ms-acc bought yesterday a car ´Costas bought a car yesterday´

The conditions are summarised in Table 1 below. The conditions described above can be interpreted as the interaction of a case hierarchy on the one hand, and, possibly, a specificity hierarchy on the other. The hierarchy in (9) modelled on Larson (1988), shows the mapping of theta roles onto case. The one under (10) is based on Silverstein (1976) and Comrie (1981).3 This is basically an animacy hierarchy. Comrie (1981) explains that, specifically for full NPs, issues like definiteness and specificity play a role in addition to animacy, and that these are actually means of identifying an entity. This view is in line with Anagnostopoulou’s (1999: 783) referentiality scale, represented under (11). (9) agent > recipient/goal > patient nominative > dative > accusative (10) pronoun

> +anim NP > +def NP > +spec NP

> -anim NP > -def NP > -spec NP

(11) anaphoric pronouns > demonstratives > proper names / definite descriptions > novel definites > weak definites > partitives > referential indefinites The question might now arise as to whether the claim that clitics are agreement markers can be maintained, given that it does not always obtain. It is possible to assume that this is a case of language change in which object clitics have been Table 1.  Clitic doubling in Spanish and Greek

obligatory

European Spanish

Río de la Plata Spanish

Greek

accusative and dative pronoun

accusative and dative pronoun

genitive pronoun or NP accusative NP, +def

strongly preferred possible

dative NP

disallowed

accusative NP

3.

dative NP accusative NP, +def accusative NP, -def

See also Anagnostopoulou (1999) and Aissen (2003).

accusative NP, -def



Clitics in Spanish and Greek 

developing into agreement markers. Agreement relations are known to be subject to prominence hierarchies. This is where case on the one hand and properties of the NP such as definiteness and specificity on the other play a role. Object agreement is established in Spanish for the dative and in Greek for the genitive, but it is less grammaticalised for the accusative. As argued by Leonetti (2007), the definiteness feature becomes less relevant as more advanced stages of grammaticalisation are reached. This is why the definiteness effect is no longer relevant for dative/ genitive clitics. The above discussion aims at explaining why clitics are considered instances of agreement, by showing which factors play a role and it will be complemented by observations about clitic left dislocation in the next section. 2.3

Clitic left dislocation (CLLD)

A related case of cooccurrence of a clitic and a coreferential NP, namely clitic left dislocation (CLLD) in which topicality plays a role (Anagnostopoulou 1997, Cinque 1997, Escobar 1997). If we compare English and Spanish, a distinction can be drawn between contrastive left dislocation (CLD) and clitic left dislocation (CLLD), as illustrated in (12) and (13) vs (14) (from Valenzuela 2006: 284). (12) water,

I drink Op every day

(13) agua, tomo Op todo el día water, drink.pres.1s the whole day ´water, I drink all day´ (14) a Juan lo vi ayer. P Juan 3accsm see.past.1s yesterday. ´Juan, I saw yesterday´ In both cases the left dislocated element is a topic, and it is linked to the clause by a resumptive element. In (12) and (13) the topic is assumed to be linked by a null operator Op, while in (14) the resumptive element is a clitic. English always has the (null anaphoric) operator Op. Spanish (and Greek) can have a resumptive clitic. Specific topics, like the one in (14) must take a clitic. Non-specific (15) or generic (13) topics do not. Following the debate mentioned above for clitic doubling, it could be argued that specificity is construed as d-linking. (15) un vaso de agua, (*lo) tomé a glass of water, (3accsm) drank.past.1s ´a glass of water I drank´

 Teresa Parodi

There has been some debate about similarities and differences between clitic doubling and clitic left dislocation (among others, Agouraki 1992). I will assume that the clitic is a marker of agreement in both cases. In CLLD the clitic is also part of a chain which licenses the relation with the topicalised NP (Cinque 1990). As for the NP, following Rizzi’s (1997) analysis of the left periphery, it can be assumed that the clitic has moved from the argument position to SpecTopP (see (16)). (16) [ TopP el auto [IP no loi necesito hasta tarde t i / *a él ] the car not 3accsm need.pres.1s until late P it ´the car I don´t need until late´ The displaced NP is doubled by a clitic if it is construed as specific, independently of case. Consider the contrast between (17) and (18). (17) el auto no lo necesito hasta tarde the car not 3accms need until late ´the car I don´t need until late´

+spec

(18) * un alumno nuevo no lo he visto P a student new not 3accms have.1s see ´a new student I haven´t seen´

-spec

Similarly, in Greek clitic left dislocation is possible provided that the dislocated NP is specific (as in (19–20). Non-specific indefinite NPs cannot be topicalised, as in (21). Case is not a crucial aspect. As shown by the English translation left-dislocation structures are impossible with indefinite NPs in English too. (19)

accusative, +definite to Gianni ton ida sto párti to sábbato the Iánnis.acc 3accms saw at the party on Saturday John I saw at the party on Saturday

(20)

genitive, +definite tis Elénis o Pétros tis chárise éna portofóli the Helen-gen the Peter-nom 3fs-gen gave a purse ´to Helen Peter gave a purse as a present´

(21)

accusative, -definite *ena koritsi, to sinantisa xtes one girl it met.1s yesterday ´*a girl, I met her yesterday´

The conditions for clitic left dislocation are summarised in Table 2 below.



Clitics in Spanish and Greek 

Table 2.  Clitic left dislocation in Spanish and Greek Spanish (both varieties)

Greek

obligatory

accusative or dative pronoun or NP

accusative or genitive pronoun or NP

disallowed

accusative or dative, -def,-spec NP

accusative or genitive -def, -spec NP

Recall that English, the L1 of the learners studied, has strong pronouns (in Cardinaletti and Starke’s (1994) classification), but no clitics. English pronouns are used in definite contexts. As opposed to Spanish and Greek, case plays no role in English. Some pronominal forms (me, her, him, us, them) are used in non-nominative contexts. There is nothing equivalent to the accusative, dative or genitive distinctions carried out in Spanish and Greek. In English subject-verb agreement is marked morphologically, but not object agreement. To sum up, the current study is based on the following assumptions: a. Clitics in Spanish and Greek are agreement markers, spell-out of agreement features, head of a functional projection (a.o. Suñer 1988, Franco 1993, 2000, Torrego 1995, Sportiche 1996, Parodi 1998, Tsimpli 1999). b. Case differences are very important: clitic doubling is (nearly) obligatory in the dative and genitive in Spanish and Greek respectively, but possible only under certain conditions in the accusative in both languages. c. A uniform analysis is adopted for clitic doubling and for clitic left dislocation in terms of the relations between clitic and double (following, among others, Agouraki 1992, Sportiche 1996). CLLD is the result of raising a doubled NP to a Topic position in the left periphery of the clause, i.e., a case of topicalisation. Case plays a secondary role in CLLD. The analysis just presented will constitute the background for the research questions to be developed in the next section. 3. Research questions The study of clitics in clitic doubling and left dislocation structures aims at shedding light on issues of optionality, parallel to Parodi and Tsimpli (2005): i. is there optionality at earlier and later developmental stages? ii. if so, is this ‘real’ or apparent optionality? iii. if apparent, what are the regularities found?

 Teresa Parodi

Parodi and Tsimpli (2005) found evidence of optionality in the data of both lower and higher proficiency English learners of Spanish and Greek who were tested on their preferences for gaps or resumption (i.e., pronominal clitics) in Null Operator structures. The lower proficiency group did not display clear preferences for gaps or resumption, which was taken as evidence of true optionality. The results of the higher proficiency group, on the other hand, indicate that the subjects fail to recognise the agreement properties of Spanish and Greek pronominal clitics (which gives an impression of optionality), but tend to accept them in contexts in which a strong pronoun would have been felicitous in their native English. The authors interpret this result as a case of apparent optionality, as it provides evidence of systematic, albeit non-targetlike, choices. A further aspect to be discussed is whether the results give any insight into the nature of the learners’ representation. Under the assumption that clitics are agreement markers, they carry uninterpretable features of person, number and, for the 3rd person, also gender. Interpretable features are involved in definiteness, specificity and topicality. These aspects tie in with the discussion about a potential asymmetry in the acquisition of interpretable and uninterpretable features (Smith and Tsimpli 1995, Hawkins and Chan (1997), Hawkins and Franceschina 2004). If these researchers are right interpretable features such as definiteness and topicality should not be problematic: learners should be able to deal with those properties of clitics related to definiteness and topicality. They should, on the other hand, have difficulties with agreement, as realised by object clitics. An asymmetry between interpretable and uninterpretable features is not expected under Prévost and White’s (2000) and White’s (2003) approach. The test factors were case (accusative and dative in Spanish, accusative and genitive in Greek) and definiteness. An observation is here in point as to the role of definiteness and specificity in the test. Notice that the specificity contrast is particularly relevant for definite NPs in some dialects of Spanish, such that doubling is not possible with non-specific definite NPs. Doubling with indefinites, whether specific or not, is not possible in any Spanish dialect. Given this situation, it seemed possible to simplify the test by using definiteness as a factor, while controlling for specificity in the following way. No context was given, but an effort was made to bias a specific reading of the definite NPs in absence of context. Consider (22), where the preferred interpretation of ‘the boy’ is a specific boy the speaker is in charge of, or (23), where ‘the post office’s woman’ is likely to be the one who works at the local post office. (22) al niño lo llevo a la escuela en bicicleta. the boy 3acc.m.s take.1s to the school on bicycle. ´the boy I take to school on the bike.´



Clitics in Spanish and Greek 

(23) todos los días la veo a la mujer del correo. all the days 3acc.f.s see.1s P the woman of the post ´every day I see the post office´s woman.´ If the test participants are sensitive to definiteness, they are more likely to accept a clitic and a coreferent NP, if this NP is definite. If the test participants are also sensitive to case, they are more likely to accept a clitic and a coreferent NP in a dative (for Spanish) or genitive (for Greek) context than in an accusative one. Sensitivity to both case and definiteness is compatible with the interpretation of clitics as agreement markers. If the learners are transferring the properties of English, then they are not expected to show sensitivity to case (which plays no role in English); sensitiveness to only definiteness is compatible with the interpretation of clitics as pronouns. It is therefore expected that the English learners will be more likely to accept a clitic if it is coindexed with a definite NP. Notice finally that under the assumption of straightforward transfer from English the learners are expected to find resumption in the form of clitic doubling or clitic left dislocation unacceptable, since English has empty categories in all these contexts. 4. Methodology The research questions were tested in a study which involved 56 participants as illustrated in Table 3. Different levels of L2 proficiency were distinguished according to the results of a cloze test. Thus, the learners of Greek were divided into three groups, namely, a Lower Intermediate, an Upper Intermediate and an Advanced group. The learners of Spanish were divided into two groups (Intermediate and Advanced).4 The subjects were language students at local schools and language schools as well as university students in Cambridge. The native speakers of Greek were recruited in Greece. Those of Spanish were South Americans and they were recruited in Cambridge. Table 3.  Participants included in the study native speakers Spanish Greek

9 14

L2 learners 18 (9 upper, 9 lower proficiency) 15 (5 upper, 5 middle, 5 lower)

4. Notice that the test does not include beginners: a certain proficiency level is required to test the relevant questions involving pronouns and clitics.

 Teresa Parodi

The test was an acceptability judgment as described in Bard, Robertson and Sorace (1996). According to this methodology the subjects have to rank the test items on a scale they establish themselves. The participants are explicitly instructed not to use zero or negative numbers. In order to make the different scales comparable, while maintaining the rankings, the results are converted into logarithms. They are validated in terms of self-consistency. The stimuli were presented visually and aurally, and in a randomised order. The test items were grammatical and ungrammatical sentences involving clitic doubling and left dislocation structures. They test case – accusative vs dative (in Spanish) and genitive (in Greek) – a contrast which does not exist in English. They also test the contrast between definite and indefinite NPs, which does exist in English.5 Examples of test items are presented in (22) and (23) above and in (24) – (26) below. (24) *¿cuándo la vas a comprar una mesa nueva? when 3acc.f.s buy.2s.fut a table new? ´when are you going to buy a new table?´ (25) to Gianni ton ida sto party to savato Giannis.acc 3acc.m.s saw.1s at the party on Saturday ´Giannis I saw at the party on Saturday´

Spanish

Greek

(26) *mia istoria mu tin ipe o pateras mu a story me.gen 3acc.f.s told.3s the father me.gen ´a story my father told me´ 5. The test results 5.1

The clitic doubling test

The results are presented in graphs, which display on the x-axis the categories tested, i.e., accusative definite and indefinite NPs as well as dative definite and indefinite NPs. On the y-axis we can read the ranking assigned to the different items in logarithms: 0.5, 1, etc are log values. What is taken into account is the ranking of the test items within each group: e.g., how the native speakers of Spanish rank accusative NPs which are definite with respect to those which are indefinite. This pattern is compared with the one found in the learner groups.6 5. Notice that the focus is on the factors which guide the acceptance of a clitic under different conditions of case and definiteness and not on testing correct or incorrect morphology on clitics. 6. Not taken into account is the absolute ranking, ie. whether an item reaches 1 or 2 on the scale.



Clitics in Spanish and Greek  2.5 2 1.5 native speakers L1E upper group L1E lower group

1 0.5 0

acc, +def

acc, -def

dat, +def

dat, -def

Figure 1.  Clitic doubling in English learners’ L2 Spanish

Consider Figure 1 corresponding to clitic doubling in Spanish. The graph shows a preference for clitic doubling with the dative over the accusative for native Spanish speakers as well as a preference of doubling of definites in the accusative, while definiteness does not make a difference in the dative. An analysis of variance was conducted in which case and definiteness were within-subjects factors. There was a main effect of case (F (1,8)= 12.377, p <.01) and a marginal effect of definiteness (F (1,8) = 4.496, p =.06), as well as a significant interaction between case and definiteness (F (1,8) = 7.855, p <.05). These effects cannot be observed in the English learners of Spanish: they rate clitic doubling worse than the native speakers of Spanish but do not show an effect of case or definiteness on full NPs. We do find a significant interaction between case, definiteness and level (F (2,24) = 5.073, p <.02), indicating that the upper group are closer to the native speakers. Turning to Greek (Figure 2) and focussing first on the native speakers, the graph reflects the asymmetry between accusative and genitive and also, in the accusative, between definite and indefinite NPs. An analysis of variance showed significant effects of case (F (1,13)=8.833, p<.02), such that clitic doubling in a genitive context is ranked higher than that in an accusative context. Furthermore, in the accusative native speakers also rank clitic doubling with definite NPs significantly higher than indefinite NPs (F (1,13)=33.182, p<.001). As for the learners, the upper group pattern like the native speakers: they are confident about the distinction in definiteness and the resulting level of acceptability, but the difference between definite and indefinite NPs does not reach significance. No differences between case and definiteness show for the middle and lower groups of learners.

 Teresa Parodi 2.5 2 1.5 native speakers L1E upper group L1E middle group L1E lower group

1 0.5 0

acc, +def

*acc, -def

gen, +def

Figure 2.  Clitic doubling in English learners’ L2 Greek

5.2

The co-occurrence of a clitic and an NP in clitic left dislocation

The results of the Spanish clitic left dislocation tests are shown in Figure 3. As expected, in the native Spanish speakers’ judgments there is no significant difference based on case with left dislocation, which contrasts with clitic doubling. There is, however, a significant effect of definiteness: native speakers of Spanish rank clitic left dislocation with definite NPs higher than indefinite NPs (F (1,8) = 7.782, p<.05), particularly with the accusative. A significant interaction between case and definiteness can also be observed (F (1,8) = 8.956, p<.02), showing that the definiteness effect is stronger in the accusative. The upper group of learners perform similarly to the native speakers; in this group, case differences show no effect, as opposed to definiteness (F (1,8) = 21.528, p<.002): that is, test items in which left dislocated NPs are definite are ranked higher than those with indefinite NPs. The lower group of learners, on the other hand, do not show a different treatment of the clitic based on case or definiteness. Figure 4 presents the results of the Greek test for clitic left dislocation. An analysis of variance with case and definiteness as within-subjects factors showed a highly significant effect of definiteness for native speakers (F (2,26)=22.769, p<.001): test items with definite left-dislocated NPs are ranked significantly higher than those with indefinite NPs. Keeping definiteness constant, the difference between accusative and genitive is significant as well (F (2,26) = 5.556, p<.05). The analysis of variance reveals no effects for the upper and the middle groups of learners. There is, however, significant effects of case (F (1,4)=10.360, p<.05) and of definiteness (F (1,4) = 9.199, p<.05) for the lower group, although they do not pattern like the native speakers.



Clitics in Spanish and Greek  2.5 2 1.5 native speakers L1E-upper group L1E-lower group

1 0.5 0

acc, +def

acc, -def

dat, +def

dat, -def

Figure 3.  Clitic left dislocation in English learners’ L2 Spanish

2.5 2 1.5 native speakers L1E upper group L1E middle group L1E lower group

1 0.5 0

acc, +def

*acc, -def

gen, +def

Figure 4.  Clitic left dislocation in English learners’ L2 Greek

6. Summary and discussion The test results from the native speakers groups are compatible with the analysis of clitics as agreement markers. Their judgments in both language groups point to differences based on case and definiteness. The definiteness effect is also obvious in clitic left dislocation, confirming the line of analysis represented by Cinque (1990, 1997, Escobar 1997, Anagnostopoulou 1997). The results from the learner

 Teresa Parodi

groups, however, show a different picture. In clitic doubling structures the English learners of Spanish and Greek allow for clitics regardless of definiteness and case features of the doubled noun phrase, contrary to the target languages. There are no statistically significant differences between case or definiteness conditions neither for the learners of Spanish nor for those of Greek, although the latter pattern with the native speakers. On the other hand, in clitic left dislocation structures, the definiteness of the doubled noun phrase does make a significant difference for the learners of Spanish in the upper group: clitic left dislocation is preferred with definites, both in the dative and in the accusative. The results of the learners of Greek in the upper group display a pattern which resembles that of the native speakers and of the learners of Spanish, without reaching significance. Recall that one of the questions is whether optionality is attested at different developmental stages. Based on the results of the statistical analysis the question receives an affirmative answer, both for earlier and later developmental stages, with some exceptions. These exceptions allow us to distinguish between ‘real’ and apparent optionality. There is clearly ‘real’ optionality with respect to case, in that none of the learner groups show sensitivity to case differences. There is, however, as pointed out above, some sensitivity to definiteness differences, an indication of regularities in the learners’ perception of clitics which diverge from those attested in the native controls and which can be interpreted as cases of apparent optionality. This sensitivity to definiteness is compatible with a consistent interpretation of clitics as strong pronouns, which is the category these learners have in their L1. In this sense we can observe a systematicity in their choices which confirms comparable results obtained in Parodi and Tsimpli (2005), as mentioned in Section 3 above. A possible explanation for these findings is a transfer effect from English where definite NPs can be left dislocated; obviously no such transfer is possible in clitic doubling structures since English does not have a similar structure. Thus, the definiteness constraints responsible for the availability of a left dislocation structure in English L1 are transferred to the L2 Spanish and Greek grammars. Another linguistic property which appears to affect the process of L2 acquisition of different pronominal systems is overt agreement morphology. Spanish and Greek, in contrast with English, are languages with rich overt morphology both for subject and for object agreement, realised in the form of a verbal suffix for subject agreement and object clitics for object agreement. Learners of Spanish and Greek have to change from a phonetically null category to overt agreement morphology in subject and object position. The data indicate that English learners of Spanish or Greek, especially the upper groups, have a tendency to prefer the clitic option. This can be interpreted as the use of a generalised strategy to accept overt agreement morphology in the form of clitics in these structures. A general preference for overt morphology over gaps was also observed in Parodi and Tsimpli’s



Clitics in Spanish and Greek 

(2005) results on Null Operator structures. We can assume that the saliency of rich agreement morphology in the learners’ input guides the restructuring of their grammar in changing from null to overt elements. Syntactically, the choice of the clitic could be triggered by the requirement to identify the empty category in the object position. Features such as definiteness which rule the choice of a clitic or an empty category in the target language do play a role in these L2 learners’ grammars. However, this role differs from the one they have in native grammars. The overall picture indicates that the learners are sensitive to definiteness, but, interestingly, rather impervious to case, in contrast with the native speakers of Spanish and Greek. Although the saliency of overt morphology in these languages guide the learners towards the acceptance of clitics, they may also be misanalysing the clitic as a pronoun, i.e., not recognising the role of clitics as agreement markers.7 In other words they misinterpret the phonological reflex of the uninterpretable agreement feature. This is compatible with work by Smith and Tsimpli (1995) and Hawkins and Franceschina (2004), among others, according to which the grammar of learners whose L1 does not instantiate a certain uninterpretable feature will not restructure when faced with it in the L2.8 In the current case English learners of Spanish and Greek, whose L1 does not instantiate object agreement, seem not to be able to realise it in an L2 which does. The type of optionality observed in this study seems not to be captured by analyses such as Prévost and White’s (2000) and by White’s (2003), according to which learners can achieve a native-like representation of the L2, but not always map this representation to the right overt realisation, since in the current case the representation of the L2 appears not to be native-like. Further light on these results could be shed by a study which includes learners whose first language have contrasting feature settings, more specifically a language which instantiates object agreement overtly, for example Greek learners of Spanish and vice versa. The results could be stronger if complemented by other tasks, such as production or acceptability tasks. At this stage this remains as a desideratum for further research.

7. As opposed to native speakers, they are not treating clitics as head of functional categories: they are treating them as pronouns. 8. An anonymous reviewer observes that this result may be related to the fact that the subjects´ proficiency was just not advanced enough and more advanced subjects would have performed better. This cannot be excluded and could be tested with near-natives.

 Teresa Parodi

Acknowledgments I would like to thank Francis Eaves-Walton for carrying out the data collection and Ernest Lee for statistical advice. For insightful comments I am grateful to two anonymous reviewers as well as to the audience at the University of Essex, where a preliminary version of the study was presented, and to Ianthi Tsimpli for long and fruitful discussions. References Agouraki, Y. 1992. Clitic-left dislocation and clitic doubling: a unification. UCL Working Papers in Linguistics 4: 45–70. Aissen, J. 2003. Differential object marking: iconicity vs. economy. Natural Language and Linguistic Theory 21: 435–483. Anagnostopoulou, E. 1994. Clitic dependencies in Modern Greek. PhD dissertation, University of Salzburg. Anagnostopoulou, E. 1997. Clitic left dislocation and contrastive left dislocation. In Materials on left dislocation, E. Anagnostopoulou, H. van Riemsdijk & F. Zwarts (eds.), 151–191. Amsterdam: John Benjamins. Anagnostopoulou, E. 1999. Conditions on clitic doubling. In Clitics in the languages of Europe, H. van Riemsdijk (ed.), 759–798. Berlin: Mouton de Gruyter. Bard. E. G., Robertson, D. & Sorace, A. 1996. Magnitude estimation of linguistic acceptability. Language 72(1): 32–68. Cardinaletti, A. & Starke, M. 1994. The typology of structural deficiency. On the three grammatical classes. Working Papers in Linguistics 4(2). Centro Linguistico Interfacoltà, Università degli Studi di Venezia. Carstens, V. 2000. Concord in minimalist theory. Linguistic Inquiry 31: 319–355. Chomsky, N. 1995. The minimalist program. Cambridge MA: The MIT Press. Cinque, G. 1990. Types of A’-dependencies. Cambridge MA: The MIT Press. Cinque, G. 1997. Topic constructions in some European languages and connectedness. In Materials on left dislocation, E. Anagnostopoulou, H. van Riemsdijk & F. Zwarts (eds.), 93–118. Amsterdam: John Benjamins. Comrie, B. 1981. Language universals and linguistic typology. Oxford: Basil Blackwell. Escobar, L. 1997. Clitic left dislocation and other relatives. In Materials on left dislocation, E. Anagnostopoulou, H. van Riemsdijk & F. Zwarts (eds.), 233–273. Amsterdam: John Benjamins. Franco, J. 1993. Conditions on clitic doubling: The agreement hypothesis. Anuarios del Seminario de Filología Vasca “Julio de Urquijo” XXVII(2): 285–298. Franco, J. 2000. Agreement as a continuum. The case of Spanish pronominal clitics. In Clitic phenomena in European languages, F. Beukema & M. den Dikken (eds.), 147–189. Amsterdam: John Benjamins. Hawkins, R. & Chan, C.-Y. 1997. The partial availability of Universal Grammar in second language acquisition: The failed functional features hypothesis. Second Language Research 13: 187–226.



Clitics in Spanish and Greek  Hawkins, R. & Franceschina, F. 2004. Explaining the acquisition and non-acquisition of determiner-noun gender concord in French and Spanish”. In The acquisition of French in different contexts, P. Prévost & J. Paradis (eds.), 175–205. Amsterdam: John Benjamins. Hyams, N. 1996. The underspecification of functional categories in early grammar. In Generative perspectives on language acquisition, H. Clahsen (ed.), 91–127. Amsterdam: John Benjamins. Karmiloff-Smith, A. 1979. A functional approach to child language: A study of determiners and reference. Cambridge: CUP. Kayne, R. 1975. French syntax. The transformational cycle. Cambridge MA: The MIT Press. Larson, R. 1988. On the double object construction. Linguistic Inquiry 19: 335–391. Leonetti, M. 2008. Specificity in clitic doubling and in differential object marking. Probus 20(1): 33–66. Müller, N. 1998. UG access without parameter setting: A longitudinal study of (L1 Italian) German as a second language. In Morphology and its interfaces in L2 knowledge, M.-L. Beck (ed.), 115–163. Amsterdam: John Benjamins. Parodi, T. 1998. Aspects of clitic doubling and clitic clusters in Spanish. In Models of inflection, R. Fabri, A. Ortmann & T. Parodi, Tübingen: Niemeyer. Parodi, T. & Tsimpli, I.-M. 2005. ‘Real’ and apparent optionality in second language grammars: Finiteness and pronouns in null operator structures. Second Language Research 21(3): 250–285. Pérez-Pereira, M. 1991. The acquisition of gender: What Spanish children tell us. Journal of Child Language 18(3): 571–590. Prévost, P. & White, L. 2000. Missing surface inflection or impairment in second language acquisition? Evidence from tense and agreement. Second Language Research 16: 103–133. Rizzi, L. 1997. The fine structure of the left periphery. In Elements of Grammar, L. Haegeman (ed.), 281–337. Dordrecht: Kluwer. Silverstein, M. 1976. Hierarchy of features and ergativity. In Grammatical categories in Australian languages, R. M.W. Dixon (ed.), 112–171. Canberra: Australian Institute of Aboriginal Studies. Smith, N. & Tsimpli, I.-M. 1995. The mind of a savant. Oxford: Blackwell. Sportiche, D. 1996. Clitic constructions. In Phrase structure and the lexicon, J. Rooryck & L. Zaring (eds.), 213–276. Dordrecht: Kluwer. Strozer, J. 1976. Clitics in Spanish. PhD dissertation, UCLA. Suñer, M. 1988. The role of agreement in clitic-doubled constructions. Natural Language and Linguistic Theory 6: 391–434. Torrego, E. 1995. On the nature of clitic doubling. In Evolution and revolution in linguistic theory, H. Campos (ed.), 399–418. Washington DC: Georgetown University Press. Tsimpli, I.M. 1999. Null operators, clitics and identification: A comparison between Greek and English. In Studies in Greek Syntax, A. Alexiadou, G. Horrocks & M. Stavrou (eds.), 241–262. Dordrecht: Kluwer. Uriagereka, J. 1995. Aspects of the syntax of clitic placement in Western Romance. Linguistic Inquiry 26: 79–123. Valenzuela, E. 2006. L2 end state grammars and incomplete acquisition of Spanish CLLD constructions. In Inquiries in linguistic development, R. Slabakova, S. Montrul & P. Prévost (eds.), 283–304. Amsterdam: John Benjamins. White, L. 2003. Second language acquisition and Universal Grammar. Cambridge: CUP.

Aspect and the interpretation of motion verbs in L2 Greek Ianthi Maria Tsimpli and Despina Papadopoulou Aristotle University of Thessaloniki

This study examines the role of aspect in L2 acquisition within the framework of the Interpretability Hypothesis (Tsimpli 2003; Hawkins and Hattori 2006). Aspect in Greek is a grammaticalized, interpretable feature affecting the argument structure and the telic/atelic interpretation of manner-of-motion verbs. As such, aspect is relevant to the syntax-semantics and the syntaxdiscourse interfaces. Native speakers and L2 learners of Greek were tested on the comprehension and production of manner-of-motion verbs. The results show that aspectual distinctions are appropriately used by L2 learners, at both interfaces. However, unlike native speakers, L2 learners rely more on lexical properties of prepositions and verbs than on grammatical aspect to encode (a)telicity. Moreover, L2 grammars of Greek seem to involve a one-to-one correspondence between perfectivity and telicity.

1. Introduction Research on second language acquisition (L2) has concentrated on general issues such as whether UG is or is not involved in the construction of an L2 grammar, but also on more specific issues within a UG-constrained second language acquisition approach such as the role of L1 and the constraints it imposes on the development of the second language. A number of alternative accounts proposed differ with respect to the extent of L1 transfer and its persistence through advanced levels of L2 competence (Schwartz and Sprouse 1996; 2000; White 2003), as well as with respect to the domain of L1 transfer, namely computational syntax or the interfaces: syntax-prosody (Goad and White 2006) and syntax-discourse (Belletti et al. 2005). To begin with, the present paper assumes a UG-constrained approach to L2 acquisition. Furthermore, it adopts the Interpretability Hypothesis (IH), which is closely related to the Failed Functional Features Hypothesis (FFFH), in that both theories assume that adult L2 acquisition of uninterpretable features is problematic

 Ianthi Maria Tsimpli and Despina Papadopoulou

due to critical period constraints (Hawkins and Chan 1997; Tsimpli 2003; Hawkins and Hattori 2006; Tsimpli and Dimitrakopoulou 2007).1 Although in early stages of L2 acquisition L1 transfer of uninterpretable features is more obvious, in more advanced stages L2 learners appear to approximate target performance. According to IH, this is the result of some compensatory mechanism of the L2 grammar which accommodates the input using interpretable features, thus eliminating real optionality from the system (Tsimpli and Mastropavlou 2008). In IH, the distinction between interpretable and uninterpretable features is crucial with respect to learnability. Specifically, it is claimed that interpretable features, even if not grammaticalized similarly in L1 and L2, are accessible to the L2 learner since they are not subject to critical period constraints. Thus, early sensitivity as well as ultimate attainment with respect to interpretable features are predicted. In contrast to the above view on the accessibility of interpretable features, Belletti et al. (2005) and Sorace (2006) have argued that features relevant to the syntax-discourse interface are problematic in L2 acquisition due to L1 interference. In Tsimpli and Sorace (2006) a distinction between syntax-semantics and syntaxdiscourse interfaces is argued for and, hence, interpretable features are further differentiated. Accordingly, linguistic phenomena which involve formal features and operations within syntax and LF, such as focus, topic and quantification belong to the syntax-semantics interface whereas phenomena which involve ‘higher’ levels of pragmatic processing such as contextual appropriateness in the use of overt subject pronouns in a null subject language belong to the syntax-discourse interface. In the latter case, features such as [contrast] and [topic-shift] are involved. Within the framework of FFFH and IH, the morpho-syntactic features investigated thus far are uninterpretable formal features with distinct values in L1 and L2: resumptive pronouns in relatives (Hawkins and Chan 1997), gender (Hawkins and Franceschina 2004), wh-feature in interrogatives (Hawkins and Hattori 2006), clitics and determiners (Tsimpli 2003, Tsimpli and Mastropavlou 2008) and resumptive pronouns in interrogatives (Tsimpli and Dimitrakopoulou 2007)2. In contrast, the present study is concerned with Aspect, a functional category with interpretable feature values (cf. Smith 1991). The FFFH in its current form does not make specific predictions for the L2 acquisition of Aspect even in the case where the L1 differs from the L2 in the encoding of the relevant features. On the 1. In Minimalism interpretability is a notion that pertains to both interfaces, namely LF (Logical Form) and PF (Phonetic Form). In this paper we are concerned with interpretability at LF, i.e., the semantic interface. 2. In some of these works, the acquisition pattern of an interpretable feature is contrasted with the uninterpretable one under investigation, provided both are specified on members of the same paradigm. For example, in Tsimpli and Mastropavlou (2008) the definite article is contrasted with the indefinite and the 3rd person clitic with 1st and 2nd person clitics.



Aspect and motion verbs 

other hand, the IH claims that interpretable features should be accessible to L2 learners, even if their instantiation in L2 is different from L1. This is based on the assumption that interpretable features are not subject to maturation constraints for reasons that have to do with their dual representation in language and cognition (cf. Smith and Tsimpli 1995; Tsimpli 1996). In Greek, Aspect is morphologically expressed in a binary way, namely perfective and imperfective. It has been argued that the semantic features involved in the perfective/imperfective distinction include [+/-bounded], [+/-iterative] and [+/-habitual] (Comrie 1976; Mozer 1994). Specifically, the imperfective can be [-bounded, -iterative, -habitual] or [+bounded, +iterative, +habitual], as shown in examples (1a) and (1b) respectively, whereas the perfective is [+bounded, -iterative, -habitual] and [+bounded, +iterative, -habitual] (cf. (2a) and (2b) respectively). The difference between the imperfective and the perfective in the [+iterative] specification is that iteration in the imperfective is equivalent to the habitual, whereas iteration in the perfective blocks the habitual reading (see the ungrammatical choices in (1b) and (2b)). The examples in (1) and (2) include an activity predicate3: (1) a. Htes oli mera zoghrafize. yesterday all day painted.imp.3s “Yesterday he was painting all day.” b. Persi zoghrafize /*zografise ena portreto kathe Kiriaki. last-year painted.imp.3s / *painted.perf.3s one portrait every Sunday “Last year he used to paint one portrait every Sunday.” (2) a. Htes zoghrafise. yesterday painted.perf.3s “Yesterday he painted.” b. Persi pije / *pijene tris fores sinema. last-year went.perf.3s / went.imp.3s three times cinema “Last year he went to the cinema three times.” The interpretability of Aspect is assumed for all languages with some aspectual marking. English marks aspect with the use of auxiliaries (3a), Russian with the use of lexical means (cf. (3b) from Babko-Malaya 1999) and Greek with morphological aspect (3c): 3. It is well known that the viewpoint aspect interacts with the situation type of the verb and in some cases leads to aspectual shifts (Smith 1991). Since motion verbs are activity predicates, we restrict our discussion to them and the aspectual shift they involve, namely from activity to accomplishment.

 Ianthi Maria Tsimpli and Despina Papadopoulou



(3) a. John is writing / writes / has written a letter.

b. Ivan pisal / napisal pis’mo. Ivan wrote.imp / na-wrote.perf letter c. O Yanis egrafe / egrapse ena grama. the.nom Yanis wrote.imp / wrote.perf a letter “Ivan was writing / wrote a letter.” The central role of Aspect in the clause-structure and in particular in the syntaxsemantics interface via the projection of argument structure has led researchers to suggest that the mapping between syntax and thematic arguments is mediated by Aspect (Tenny 1987; Borer 1994; 2005; Travis 2000; for Greek see Mozer 1994; Chila-Markopoulou and Mozer 2001; Sioupi 2002; 2005; Tsimpli and Papadopoulou 2006). A strong version of this hypothesis is proposed by Borer (1994; 2005), according to which verbs project arguments freely within VP and higher aspectual projections attract internal and external arguments. It is further claimed that the aspectual features of the verb interact with the specificity of the DP-complement in deriving the telic or atelic interpretation: (4) I Maria zoghrafise to portreto. the Maria painted.perf.3s the portrait “Maria painted the portrait.” (5) I Maria zoghrafize to portreto. the Maria painted.imp.3s the portrait “Maria was painting the portrait.” On the basis of this assumption, Tsimpli and Papadopoulou (2006) have argued that the telic/atelic distinction in the interpretation of (4) and (5) is formally captured by a difference in the Merge position of the DP-complement, as shown by the structures in (6) and (7) below4:

(6)

vP Subj

v’ v

AspP Obj

Asp’

Asp VP [+perf] 4. In Tsimpli and Papadopoulou (2006) the phrase intervening between vP and VP is TransP V’ (transitivity phrase) and Aspect is a feature specified on Trans. V

Complement

Subj

v’ v



AspP Obj

Aspect and motion verbs 

Asp’ Asp [+perf]

VP V’ V



(7)

Complement

vP Subj

v’ v

AspP Obj i

Asp’ Asp [+imp]

VP ti

V’ V

Complement

The specific DP-complement of a perfective verb is merged in Spec,AspP, whereas the DP complement of an imperfective verb is merged in Spec,VP and is then moved to Spec,AspP. In both structures, the DP-complement eventually appears in Spec,AspP for Case reasons. Tsimpli and Papadopoulou (2006) report on an empirical study including an adult and a child group of native speakers of Greek, in which perfective verbs were found to strongly favour an overt object as opposed to imperfective verbs, which were preferred as intransitive. We argued that this preference is due, first, to economy restrictions on the derivation of DP complements with perfective vs. imperfective verbs (cf. (6) and (7)) and, second, to an interpretative difference, namely that perfectivity includes an endpoint made visible through an overt object, thus rendering the predicate telic (cf. Horrocks and Stavrou 2003). The implication of the above is that telicity in Greek is relevant to the syntax-semantics interface as it is compositionally derived by the perfective form of the verb and the specificity of the DP complement. In this respect, the interpretation of an activity predicate with an overt DP-complement is unambiguously telic or atelic depending on the perfective/imperfective distinction. Note, however, that, given the role of the DP complement in determining the predicate’s interpretation as being telic or atelic, it is concluded that (a)telicity cannot be exclusively determined by the aspectual form of the verb. This is in contrast to

 Ianthi Maria Tsimpli and Despina Papadopoulou

other languages, in which telicity is morphologically expressed in the verbal projection; for example, verbal particles in Germanic, the clitic se in Spanish (Zagona 1996; Sanz 2000) and verbal prefixes in Slavic (Filip 1999). Our presentation of the encoding of (a)telicity in Greek thus far involves activity predicates with a DP-complement. The linguistic phenomenon addressed in the present paper, however, concerns the structure and the interpretation of a subclass of activity predicates, namely, manner-of-motion verbs. These verbs can be followed by a PP, which can be a complement (GOAL) or an adjunct (PATH) (Zubizarreta and Oh 2007; for Greek cf. Horrocks and Stavrou 2007). (8)

I ghata etrekse/etrehe ston kipo. the cat ran.perf.3s / ran.imp.3s s-the garden a. [IP…[AspP [perf/imp]….[VP V PP ]]] b. [IP…[AspP [perf/imp]….[ VP [VP V …] PP ]]]

PP complement PP adjunct

Notice that example (8) differs crucially from sentences such as (4) and (5), since DP-complements are merged in Spec,AspP or Spec,VP with perfective and imperfective verbs respectively, whereas PP complements are merged directly with the verb (in the complement position of structures (6) and (7) above). In other words, example (8) is ambiguous with respect to the structural position of the PP. In (8a), the PP complement denotes the endpoint of the motion event which can be understood either as reached (telic) or not (directional/atelic). On the other hand, in (8b) the PP adjunct has a locative interpretation and the predicate is atelic. The first question that we address in the present study is whether the choice of perfective/imperfective affects native speakers’ preference for structure (8a) or (8b), a syntax-semantics interface issue. On the basis of our previous findings from native speakers of Greek, structure (8a) will be preferred with perfective verbs and structure (8b) with imperfective verbs. The second question concerns native speakers’ preferences for the telic over the atelic (directional or locative) reading, a syntax-discourse interface issue. Note that the telic reading is unavailable in sentence (8) with an imperfective verb. As far as perfective verbs are concerned, given that, as mentioned above, perfectivity implies telicity, the telic reading is predicted to be the default interpretation with perfective manner-of-motion verbs and thus strongly preferred over the directional. The same questions arise in relation to non-native speakers of Greek. With respect to the first question, the present study investigates whether in L2 grammars aspectual distinctions interact with the complement vs adjunct status of the PPpath, as regulated by the syntax-semantics interface. With respect to the second question, we examine whether the syntax-discourse related preference for the telic over the directional reading with perfective manner-of-motion verbs is native-like. Both questions can be addressed provided L2 learners have mastered the



Aspect and motion verbs 

morphological expression of aspectual distinctions (i.e., [+/-perfective]) at a certain stage of development. According to the IH, L2 acquisition of Aspect being an interpretable feature is predicted to be unproblematic. Any differences between the native and the non-native speakers could then be attributed to L2 problems related to the role of Aspect in the syntax-semantics, the syntax-discourse interface or both (cf. Belletti et al. 2005; Sorace 2006). In the next section we present the structure and interpretation of motion verbs in Greek and compare them with the corresponding structures in the native languages of the L2 learners of this study. In Section 3, previous L2 studies on the same phenomenon are briefly presented. In Section 4, the empirical study is presented, followed by a discussion of the most important similarities and differences between the native and the L2 speakers of Greek. Finally, in Section 5 the implications for IH with respect to the two interface levels to which aspectual distinctions are (in)directly relevant are discussed. 2. Manner-of-motion verbs in Greek Motion verbs are activity predicates which can acquire an accomplishment reading in certain contexts (Talmy 1985). In Greek, as in other languages, motion verbs differ as to whether they are compatible with a PPpath complement, thus denoting directed motion, or not. Accordingly, motion verbs are distinguished between unambiguously locative manner-of-motion verbs (without directed motion) (e.g., (9)), unambiguously non-locative motion verbs (with directed motion) (e.g., (10)) and manner-of-motion verbs which are ambiguous between a locative and a non-locative reading. The last group of verbs will be presented in detail in the following section. Consider the following examples5: (9) I Maria xoreve / xorepse (mesa) sto spiti. PP adjunct the Maria danced. imp.3s / danced.perf.3s (inside) s-the house “Maria was dancing/danced inside/in/*to/*into the house.” (10) a. I Maria pijene / pije (mesa) sto spiti. PP complement the Maria went.imp.3s / went.perf.3s (inside) s-the house “Maria was going/went inside/into/to the house.” 5. Notice that the difference between the perfective and the imperfective verb forms in (9) and (10) has to do with the boundedness of the event: the event is unbounded with the imperfective verb forms. However, in (9) the event is necessarily atelic and the adjunct PP is a path and not a goal. On the other hand, in (10) the PP complement provides the endpoint of the motion event which is reached when the verb is perfective. In this case the interpretation is telic. With the imperfective the interpretation is directional and atelic.

 Ianthi Maria Tsimpli and Despina Papadopoulou

b. I Maria ebene / bike (mesa) sto spiti. the Maria entered.imp.3s / entered.perf.3s (inside) s-the house “Maria was entering/entered inside/into the house.” The unambiguously locative verbs such as xorevo (=dance), parapato (=stagger, stumble), periplanjeme (=wander) lack a directed motion feature. Thus, the PPpath sto spiti in example (9) can only function as a locative modifier of the motion event and, as such, it is a VP-adjunct. Accordingly, the verb structure in (9) is unergative. Crucially, the aspectual choice between perfective and imperfective does not affect the unambiguously locative interpretation of the event. Notice that the corresponding verbs in English can have a directional reading: (11) John wandered to the lake. (adapted from Zubizarreta and Oh 2007) Zubizarreta and Oh (2007) distinguish between two types of manner-of-motion verbs in English: those that can take a ‘distance’ complement and those that cannot (cf. ‘run a mile’ and ‘*wander a mile’). In English, both classes can participate in structures that are ambiguous between a locative and a directional reading6. In contrast, in Greek, verbs which cannot take a ‘distance’ complement (‘*xorevo ena mili’ = dance a mile) are unambiguously locative, and, thus, lack a directional reading (cf. 9). Given that in both languages this class of manner-of-motion verbs cannot have an inherent directed motion feature, the difference must be attributed to the inherent directional meaning of the preposition ‘to’ in English7, and lack thereof in the case of the Greek preposition s(e) (Horrocks and Stavrou 2007). Thus, the

6. According to Zubizarreta and Oh (2007), the ambiguity in English arises depending on the choice of the preposition (cf. ia and b), the choice of the motion verb (cf. iia and b), and the form used in combination with a locative P (cf. the gerund in (iiia and b)). (Examples adapted from Zubizarreta and Oh, op.cit): (i) a. John ran into / out of the room. b. John ran inside / outside / in the house

(only directional) (locative or directional).

(ii) a. John ran / walked in /inside the house. b. John danced in/inside the house.

(locative or directional) (only locative)

(iii) a. John’s running (in)to the house b. John’s running in/inside the house

(directional) (locative)

7. As an anonymous reviewer points out, the [+/-progressive] feature of an English verb affects the possibility of the telic reading of an otherwise directional atelic predicate: (i) He was going into the house. (ii) He went into the house. In both cases the PP into the house is a goal complement but in (ii) the endpoint has been reached and the predicate is telic.



Aspect and motion verbs 

Greek class of unambiguously locative verbs (cf. 9) is co-extensive with the class of manner-of-motion verbs which cannot take a ‘distance’ DP complement. The examples in (10) are similar to those in (9), in that the perfective / imperfective choice of the verbal form does not affect the inherent semantic feature (directed motion, in this case) of the verbs (Zubizarreta and Oh 2007). The motion events in (10) are unambiguously non-locative and as such, the PPpath is a complement of the verb8. 2.1

Aspect and manner of motion verbs

The third class of motion verbs in Greek consists of activity predicates, which can optionally be construed as either involving directed motion or not: (12) a. I Maria etrexe (mesa) sto parko. the Maria ran.imp.3s (inside) s-the park “Maria was running inside/in/to/*into the park.” b. I Maria etrekse (mesa) sto parko. the Maria ran.perf.3s (inside) s-the park “Maria ran inside/in/into/to the park.” The manner-of-motion verbs exemplified in (12) can take a ‘distance’ complement (‘trexo ena mili’ = run a mile)9. As shown by the translation, these sentences are ambiguous between a locative and a non-locative reading in both the perfective and the 8. As noted in the literature, there is a difference between the verb pijeno (=go) in (2a) and beno (=enter) in (2b) (Talmy 2000; Oh 2003). The former is a neutral motion verb whereas the latter is a path verb since it encodes, apart from motion, a path feature. 9. It should be pointed out that some native speakers consider the structures in (12) unambiguously locative (cf. Horrocks and Stavrou (2007)). For these speakers, the directional or telic reading can be expressed unambiguously with structures consisting of a simple motion verb such as beno (=enter) or pijeno (=go) and the manner-of-motion verb as a gerund (e.g., (ia)). Alternatively, an unambiguous structure can include one of these manner-of-motion verbs with a preposition clearly specified for direction (e.g., (ib)): (i) a. I Eleni ebene/bike sto dhomatio trexondas. the Eleni entered imp.3s/perf.3s s-the room running “Eleni was entering/entered the room running.”

b. I Eleni etrexe/etrekse pros to dhomatio. the Eleni ran.imp.3s/perf.3s towards the room “Eleni was running/ran towards the room.”

As shown in Papadopoulou’s (1996) empirical study as well as in the data from the native speakers presented in this study, the ambiguity is indeed attested. We believe that the contrast in native speakers’ judgements may be due to the strong preference for the unambiguous options in (i) above.

 Ianthi Maria Tsimpli and Despina Papadopoulou

imperfective form (cf. Horrocks and Stavrou 2007). However, (12a) is ambiguous between two atelic interpretations, the locative and the directional, whereas (12b) has an additional reading, which is telic and implies that the endpoint has been reached. In the locative reading, the PPpath is a VP-adjunct as in (9), while in the other two readings the PPpath is a complement (‘goal’) as in (10). We return to the syntactic representation of each reading of the ambiguity in the following section. For the moment, it suffices to point out that the ambiguity between locative and non-locative readings is a phenomenon relevant to the syntax-semantics interface, since it involves differences in the predicate’s argument structure. On the other hand, the difference between the directional and the telic reading is perceived at the discourse level. On these grounds one could argue that atelicity – and not telicity – is grammaticalized in Greek, as the ambiguity in the imperfective form involves only atelic, i.e., locative and directional, interpretations of the predicate. We consider this an unwelcome suggestion because it would imply that imperfective aspect entails atelic readings in all cases, contrary to fact. In the example below, although the main verb is imperfective, the temporal reading is [+iterative, +habitual] and the interpretation of the predicate is that of a series of telic events (Iatridou 2000): (13) Otan itan mikros, etrehe sto parko se mia ora. when was young, ran.imp.3s s-the park in an hour “When he was young, he used to run to the park in an hour.” We thus conclude that morphological aspect alone is not specified with respect to the (a)telic interpretation; for the atelic interpretation to hold, the imperfective needs to be [-iterative]. With respect to the ambiguities observed in (12), Papadopoulou (1996) found that adult native speakers of Greek prefer the telic reading with perfective mannerof-motion verbs than the atelic and the directional readings, while with imperfective verbs they prefer the atelic locative interpretation. Moreover, the status of the preposition, simple (se) or complex (mesa se) (cf. (12)) was also shown to affect native speakers’ preferences: the simple preposition is more strongly associated with non-locative readings whereas the complex preposition favours locative readings. This is probably due to the fact that the complex preposition has a richer semantics and lexicalizes the path more clearly. The simple preposition is underspecified for either direction or location and allows the aspectual features of the verb (situation type and viewpoint aspect) to interact more transparently with the noun of the PP (see fn. 6) (Theophanopoulou-Kontou 1992; Terzi 2007). Based on Papadopoulou’s (1996) study, Table 1 summarizes the (un)available interpretative options for the ambiguous class of Greek manner-of-motion verbs exemplified in (12) with the simple preposition se. Preferences are marked with a double tick:



Aspect and motion verbs 

Table 1.  Preferred readings with manner-of-motion verbs in Greek Aspect

Perfective Imperfective

Atelic

Telic

Locative (PP adjunct)

Directional (PP complement)

(PP complement)

√ √√

√ √

√√ X

Notice that the directional reading is not the preferred one either for the perfective or the imperfective form. This is probably due to the lack of an inherent directed motion feature on the simple preposition se in Greek, along the lines suggested in Horrocks and Stavrou (2007)10. However, we expect direction to be more strongly preferred with perfective than imperfective verbs for reasons to do with the overall preference for transitive readings with perfective verb forms (Tsimpli and Papadopoulou 2006). To summarize the discussion so far, morphological aspect on Greek ambiguous manner-of-motion verbs affects (a) the predicate’s representation of arguments and (b) the strength of preference associated with each reading of the ambiguity involved. (a) is a property of the syntax-semantics interface and distinguishes between manner-of-motion verbs with complement vs adjunct PPpath. (b) is a syntax-discourse interface issue, whereby the perfective verbs acquire a pragmatically preferred interpretation for the telic over the directional reading. Before we provide the structure corresponding to the examples in (12), we briefly present how the distinction between directional and locative readings is captured in the L1s of the L2 learners participating in this study. In German, some manner-of-motion verbs are compatible with either a directional or a locative reading, but unambiguously so11: (14) Der Wurm kriecht in die / der Tasse. the worm crawls in the.acc / the.dat cup “The worm crawls into / in the cup.” The directional reading is encoded by the accusative case marking of the DP complement of the preposition in, whereas the locative reading is encoded by the dative case marking.

10. As Horrocks and Stavrou (2007) claim, Greek has a clearly directional preposition pros (=towards) which explicitly marks direction in motion structures. These authors, however, suggest that the directional readings of PPs headed by se may be more readily available with some manner-of-motion verbs such as trexo (=run) due to their lexical / encyclopaedic properties. 11. The examples are from Rothweiler et al. (2007).

 Ianthi Maria Tsimpli and Despina Papadopoulou

Russian motion verbs encode direction or location morphologically12: (15) a. let-e-tj flyimp.dir.inf b. let-a-tj flyimp.ndir.inf Directed motion verbs cannot be used to express repeated or habitual motion events (Romanova 2006: 16–17): (16) Ja begun a zanjatija I runimp.dir.1sg on clases.acc “I am running to the classes.” (17) a. *Ona často letit v Moskvu. She often fliesimp.dir in Moscow.acc “She often flies to Moscow.” b. Ona často letajet v Moskvu. She often fliesimp.ndir in Moscow.acc “She often flies to Moscow.” In addition, directed motion verbs and non-directed motion verbs can combine with PPs headed by the same preposition. However, the object of P is marked for accusative and the PP denotes the goal with directed motion verbs, while the object of P is in the locative case and the PP has a locative reading with non-directed motion verbs13 (Romanova 2006: 131): (18) a. Žuk polz v korobku. imp beetle.nom crept .dir.sg.m in box.acc “The beetle crept into the box.” b. Žuk polzal v korobke. imp beetle.nom crept .ndir.sg.m in box.loc “The beetle crept in the box.” The addition of perfectivizing prefixes, like vo in (19b), to a directed motion verb renders the predicate telic (Romanova 2006: 143): (19) a. idti v magazin imp walk .dir.inf in shop.acc “walking to the shop” 12. Aspect marking in the Russian examples is part of the stem (shown as superscript in the gloss). 13. Notice that non-directed motion verbs can co-occur with directional PPs when they have a repeated or habitual reading (cf. 17b).



Aspect and motion verbs 

b. vojti v magazin perf in-walk .dir.inf in shop.acc “walk into the shop / enter the shop” To summarize, contrary to the Greek class of manner-of-motion verbs which are ambiguous between a locative (adjunct PP) and a directional/telic (complement PP) reading, in English, German and Russian (non)directed motion is unambiguously expressed with manner-of-motion verbs either through the lexical choice of the verb root or the preposition, and/or the case marking of the prepositional complement. Accordingly, the task of English, German and Russian learners of Greek is to acquire the preference effects of the interaction between perfective and imperfective aspect with the complement vs. the adjunct option in the representation of the PPpath as well as with the telic vs. the atelic readings of the predicate. 2.2

The syntax of manner of motion verbs

According to Zubizarreta and Oh (2007), the ‘bare’ motion verb go has the structure in (20) below: (20)

VP DP

V V

PPGOAL

O Yanis pije sto parko the Yanis went.perf.3s s-the park We assume that the structure of unambiguously non-locative verbs (e.g., (10) above), such as vgheno (=exit), beno (=enter), vutao (=dive) is also the one in (20), the difference being that all of these verbs with the exception of pijeno (=go) encode ‘path’ as well as directed motion. The similarity remains, however, that the PP is a goal complement of V and that the reading is necessarily non-locative. The perfective form of these verbs renders the complement PP the endpoint of the motion event and the predicate is interpreted as telic. In the imperfective form, the PP-complement is still the goal but the unboundedness of the event leads to the interpretation of the predicate as directional/atelic.

 Ianthi Maria Tsimpli and Despina Papadopoulou

The second class of manner-of-motion verbs, exemplified in (9) above, is unambiguously locative and thus atelic. We assume that the relevant structure is the one in (21) below, where the PPpath is a VP-adjunct, and the structure is unergative: (21)

VP DP

V’ V’

PPPATH (LOC)

V

O Yanis periplanjotan / periplanithike sto parko the Yanis wandered.imp.3s / wandered.perf.3s s-the park Turning to the ambiguous manner-of-motion verbs, the locative and the non-locative readings should in principle correspond to the structures in (21) and (20) above, respectively. In order to establish the complement vs adjunct difference in each case, we can subject the perfective and imperfective verb forms of these verbs to the usual aspectual PP test (Dowty 1979) and change the position of the PPpath to be adjacent or non-adjacent to the motion verb (Zubizarreta and Oh 2007): (22) a. O Yanis etrekse sto parko se mia ora. the Yanis ran.perf.3s s-the park in an hour b. ??O Yanis etrekse se mia ora sto parko. the Yanis ran.perf.3s in an hour s-the park “Yanis ran to/*in the park in an hour.” (23) a. O Yanis etrekse sto parko ja mia ora. the Yanis ran.perf.3s s-the park for an hour b. O Yanis etrekse ja mia ora sto parko. the Yanis ran.perf.3s for an hour s-the park “Yanis ran in/*to the park for an hour.” The pairs in (22) and (23) involve the verb trexo (= run) in the perfective form. Both sentences in (23) are grammatical: the locative reading induced by the temporal PP allows the path PP sto parko to appear in sentence-final position, thus suggesting that this PP is also an adjunct. In the pair in (22), however, the nonlocative and, in particular, the telic reading is the only one compatible with the temporal PP se mia ora (=in an hour). The reduced acceptability of (22b) in which



Aspect and motion verbs 

the temporal, adjunct PP intervenes between the verb and the PP sto parko supports the complement analysis of the PPpath. Consider the same examples with the verb in the imperfective form: (24) a. O Yanis etrehe sto parko se mia ora. the Yanis ran.imp.3s s-the park in an hour b. ??O Yanis etrehe se mia ora sto parko. the Yanis ran.imp.3s in an hour s-the park “Yanis ran to the park / *in the park in an hour.” (25) a. O Yanis etrehe sto parko ja mia ora. the Yanis ran.imp.3s s-the park for an hour b. O Yanis etrehe ja mia ora sto parko. the Yanis ran.imp.3s for an hour s-the park “Yanis ran in the park / *to the park for an hour.” Although the acceptability of the sentences in pairs (24) and (25) is parallel to the ones in (22)-(23), (24a) is acceptable only if the imperfective aspect is read as [+iterative] (Iatridou 2000). In other words, the translation of (24a) cannot be ‘*Yanis was running to the park in an hour’ but only ‘Yanis used to run to the park in an hour’. In this respect, the event is construed as a series of telic events (see Section 2.1 above). This constraint in the interpretation of (24) is forced by the temporal PP, which only allows a telic reading of the motion event and neither of the atelic ones (directional or locative)14. Thus, the [+iterative] rather than the [-iterative] (durative) reading of imperfective aspect is the only one available for deriving the series of telic events. Finally, the reading of (25a) is either locative or directional whereas in (25b) the directional reading is unavailable given that the aspectual PP intervenes between the verb and its complement sto parko. It thus seems that the class of ambiguous manner-of-motion verbs in Greek can be associated with either a locative reading in which the PPpath is an adjunct or the non-locative reading in which the PPpath is a complement (directional or telic). According to Zubizarreta and Oh (2007), the fact that Germanic languages, but not French or Spanish, allow the directional (or telic) reading with manner-of14. The PPs in an hour and for an hour lead to the interpretation of the situation type of the verb as accomplishment and activity respectively (Dowty 1979; Verkuyl 1972). In turn, whether the accomplishment reading is also telic depends on the specificity of the complement. Given that in the activity reading the manner-of-motion verbs discussed here lack a complement, whereas in the accomplishment reading the PP is a complement with a specific DP object, we assume that the distinction between activity and accomplishment in (22)-(25) is co-extensive with atelic vs. telic. Accordingly, the directional interpretation is excluded from the examples modified by in an hour regardless of the aspectual morphology of the verb.

 Ianthi Maria Tsimpli and Despina Papadopoulou

motion verbs whereby the PPpath is a complement, is associated with the parametrically defined property of languages to allow compound formation in the syntax of the V-V type. Zubizarreta and Oh argue that the compounding in this case involves two Vs the first of which is specified for ‘manner’ and the second for ‘directed motion’. It is the latter, which is phonologically null, that selects the PPpath as a complement and thus it is represented as a complement of the whole V-V compound. The corresponding structure is presented in (26) below: (26)

VP DP

V

V John wobbled

PPPATH (goal) V

to the park

Recall from the previous section that the Greek class of manner-of-motion verbs which lack a ‘distance’ complement (e.g., xorevo ‘dance’) are unambiguously locative and thus cannot participate in (24a). In this respect, Greek patterns with French and Spanish and not with Germanic languages. In French and Spanish, however, the class of manner-of-motion verbs, which are ambiguous in Greek, is unambiguous. Directed motion can be expressed in these languages periphrastically, i.e., with a gerund or with an unambiguously specified preposition. The problem then is, if V-V compounding in (26) is a syntactic option in Greek why is it restricted to a subclass of manner-of-motion verbs only and is not generalized to the other class exemplified in (9) (e.g., dance, stumble, gallop etc)? This problem is addressed by Zubizarreta and Oh (2007) with respect to Italian, a language that seems to be closest to Greek with respect to the classification of manner-of-motion verbs. Folli (2001) suggests that Italian has three classes of manner-of-motion verbs, similar to the ones presented in (9), (10) and (12) in the previous section. The crucial difference between Greek and Italian, however, is that within the ambiguous class, Italian disambiguates between locative and nonlocative readings through auxiliary use (essere for the unaccusative vs avere for the unergative structure), whereas Greek does not disambiguate the two readings in the grammar through morphological aspect (and choice of preposition), but instead renders one of the two readings a strongly preferred choice at the syntaxdiscourse interface. Zubizarreta and Oh (2007) suggest that Italian uses the position of these motion verbs which, in this structure, are semi-functional to compose a structure such



Aspect and motion verbs 

as the one in (26). We would like to suggest that Greek uses the functional projection of Aspect to distinguish between the complement vs adjunct choice for the PPpath. In the empirical study of Tsimpli and Papadopoulou (2006), it is found that optionally transitive verbs in the perfective show a significantly stronger preference for the realization of an overt object than in the imperfective. In present terms, this finding is translated as perfective aspect requiring a complement PPpath more than imperfective aspect. In other words, the non-locative (i.e., the directional or telic) reading should be clearly preferred over the locative with perfective verbs. In line with Zubizarreta and Oh’s analysis regarding the compositional nature of ‘manner’ and ‘directed motion’ features in the V-V compounding structure in (26), we suggest that aspect in Greek manner-of-motion structures hosts the directed motion feature, as in the following representation: (27)

vP v’

Subj v

AspP Spec

Asp’

Asp [perf ]/[imperf]

VP V’

VPP

PATH

In case the perfective form is used, the complement PP is consistent with a directional or telic reading but not with the locative reading in which the directed motion feature is missing. The latter has a VP-adjunct status as in the structure below: (28)

vP v’

Subj

AspP

v Spec

Asp’

Asp [perf]/[imperf] VP

VP PPPATH (LOC)

 Ianthi Maria Tsimpli and Despina Papadopoulou

In (27), if aspect is imperfective the only possible reading is the directional and not the telic. This is due to the fact that imperfective aspect denotes an unbounded event and as such it cannot derive a telic interpretation15. If aspect is perfective, the telic reading is also available. The representation is identical to the directional reading and the difference involves a syntax-discourse enrichment whereby the goal PP is also understood as the reached endpoint of the motion event. In (28), the PP is locative and the interpretation is always atelic. We claim that native speakers of Greek associate the structure in (27) with perfective aspect on the basis of the independently attested property of perfective to favour a complement. On the other hand, we expect that the structure in (28) is strongly associated with imperfective aspect and a locative interpretation. 3. Previous studies The acquisition of motion verbs in the second language has been studied by Inagaki (2001), Matsunaga (2006), Montrul (2001) and Navarro and Nicoladis (2005). Their aim was to test whether the argument structure of the L1 constrains motion expressions in the second language. An additional aim of the studies by Inagaki and Montrul was to investigate whether positive evidence facilitates the acquisition of motion verbs. Inagaki (2001) tested the argument structure of manner-of-motion verbs in the interlanguage of intermediate Japanese learners of English and advanced English learners of Japanese. Notice that Japanese, contrary to English, does not allow goal PP complements to appear with manner of motion verbs (cf. 29b). Instead they use a periphrastic structure (cf. 29c), which includes apart from the main directed motion verb a gerund expressing the manner (examples taken from Inagaki 2001: 155): (29) a. He ran into the house. b. *John-ga gakkoo-ni aruita. John-nom school-at walked c. John-ga gakkoo-ni aruite itta. John-nom school-at walking went In a written grammaticality judgment task with pictures, Inagaki found that the Japanese learners of English accepted sentences such as (29a), which suggests that they were able to recognize the grammaticality of manner-of-motion verbs with 15. Recall from the discussion above that the only possibility for a telic reading with imperfective is if imperfective is [+iterative].



Aspect and motion verbs 

goal PPs, due to the availability of positive evidence in the input (Inagaki 2001: 164). However, L1 effects were also obtained, since the learners accepted constructions which are possible in their L1 and which received low judgments by the native speakers of English. On the other hand, the English learners of Japanese exhibited difficulties in identifying the ungrammaticality of sentences such as (29b), due to the lack of positive evidence. Matsunaga (2006) also tested manner-of-motion verbs with goal PPs in L2 English by German and Japanese speakers. German is similar to English in that it allows goal PPs with manner-of-motion verbs: (30) Er rannte ins Haus. he ran in.acc house “He ran into the house.” In a sentence-combining task, she replicated Inagaki’s finding that advanced Japanese learners of English did produce sentences such as (29a) and actually to the same extent as German speakers. L1 influence from Japanese was observed in less proficient learners of English. Montrul (2001) investigated the (un)availability of transitivity alternations with manner-of-motion verbs in L2 English and Spanish. More specifically, in English unergative manner of motion verbs (31a) undergo a transitivity alternation when the PP denoting the endpoint is present (31b), whereas this construction is not possible in Spanish (31c) and Turkish (31d): (31) a. The soldiers marched. b. The captain marched the soldiers to the tents. c. *El capitán marchó a los soldados hasta el campamento. e. *Asker-ler heyke-l-e yürü-dü. In a picture judgment and a grammaticality judgment task, she found L1 effects in the acquisition of the argument structure for manner-of-motion verbs. Namely, (i) Spanish and Turkish learners of English were reluctant to accept sentences such as (31b) and (ii) English learners of Spanish were more likely to accept sentences such as (31c) than Turkish learners of Spanish. The fact that positive evidence in the input did not facilitate Spanish and Turkish learners of English is attributed by Montrul to the fact that constructions such as (31b) are not very productive in English. Navarro and Nicoladis (2005) investigated the expressions of motion events used by native speakers and advanced English speaking learners of Spanish through a production task, in which the participants described two silent video

 Ianthi Maria Tsimpli and Despina Papadopoulou

excerpts16. Spanish differs from English in that directed motion is expressed with path verbs – and not manner-of-motion verbs – combined with PP goals. The findings indicated that the learners, as the native speakers, produced more path than manner-of-motion verbs to denote directed motion events, even though this difference was not overwhelming. In addition, the L2 learners were less likely to produce bare path verbs than the native speakers. Navarro and Nicoladis (2005: 106) conclude that the L2 learners of their study show a clear trend towards the complete acquisition of constructions denoting directed motion, even though their first language is typologically different from Spanish in the encoding of path and they have not received explicit instruction on this phenomenon. To summarize the main findings from the previous studies presented here show that, although L1 effects are evident, L2 learners can acquire motion structures even when they are differently encoded in their native language. 4. The empirical study The aim of the empirical study was to investigate the role of the aspectual verb form (perfective vs imperfective) on one hand and the preposition type (simple vs complex) on the other in the comprehension and production of potentially ambiguous manner-of-motion verbs. The study compares the proposed role of aspect and prepositions in native and L2 Greek. With respect to the role of aspect at the syntax-semantics interface, the study aims to address the question of whether L2 learners, similarly with native speakers, use the perfective/imperfective distinction to differentiate between complement and adjunct PPs. Furthermore, the role of aspectual distinctions at the syntax-discourse interface involves the interpretative difference between telic and atelic readings of motion events. 4.1

Predictions

With respect to the native speakers of Greek, the prediction is that they will provide more PPgoal readings (directional and telic) with perfective than with imperfective aspect, since they use aspectual distinctions to differentiate between argument structures. This prediction is based on previous empirical evidence presented in Papadopoulou (1996) and on our earlier study (Tsimpli and Papadopoulou 2006), which shows a close link between perfectivity and transitivity. At the syntax-discourse level,

16. For similar studies in L1 acquisition see (Berman and Slobin 1994; Oh 2003; Özçalişkan and Slobin 2000; Slobin 1996; Hickmann and Hendriks 2006).



Aspect and motion verbs 

we expect native speakers to show a preference for the telic over the directional/atelic reading due to the pragmatic link between perfectivity and telicity. With respect to the L2 grammar17, the Interpretability Hypothesis predicts that the interpretable status of Aspect will help L2 learners of Greek master the morphological aspectual distinctions and integrate grammatical and lexical information to derive the “native” choices of (a) argument structure for perfective and imperfective verbs respectively and (b) telic/atelic interpretations. These predictions run counter the Sorace (2006) and Belletti et al. (2006) claims that interpretable features, relevant at the syntax-discourse interface, are problematic for L2 learners even at advanced stages of L2 acquisition. We therefore expect L2 learners to share structural representations with NSs in so far as the complement/adjunct distinction interacts with the aspectual forms (see (27) and (28) in Section 2.2). Regarding the difference between simple and complex prepositions we predict, in line with Papadopoulou (1996), that complex rather than simple prepositions will enhance the PP-adjunct readings in both the native and the L2-learner data. Simple prepositions, being semantically underspecified, are predicted to show a more transparent relation between the aspectual form of the verb and the complement/adjunct distinction. Finally, we do not expect direct L1 transfer since the L1s of the participants do not exhibit the ambiguity attested in Greek structures with manner-of-motion verbs. However, we do expect L2 learners to prefer unambiguous structures using unambiguously locative or directed-motion verbs for two reasons: the first is due to the lack of ambiguity in the corresponding L1 structures and the second is due to the increased processing load usually associated with ambiguous structures and more so when using a non-native language. 4.2

Participants

Two groups of subjects participated in the study: monolingual native speakers and second language learners of Greek. There were ten (six females, four males) L2 learners of Greek who participated in both the comprehension and the production task. At the time of testing they were all attending Greek lessons at the Aristotle University of Thessaloniki and were at the intermediate level. Their level of proficiency in Greek was determined by a non-standardized placement test used in the School for Modern Greek, Aristotle University of Thessaloniki. In terms of their educational background, four of them had a university degree and six of them 17. Recall from the Introduction section that the FFFH does not make any explicit predictions for the acquisition of interpretable features other than the logically implied hypothesis that interpretable features should be unproblematic in L2 development.

 Ianthi Maria Tsimpli and Despina Papadopoulou

Table 2.  Profile of the second language learners of Greek Profile

Mean scores (SD)

Mean Age Length of residence in Greece (months) Length of L2 instruction (months)

22 (6.24) 11 (3.98) 9.10 (4.12)

were still studying at the University when tested. They came from two different language backgrounds: Slavic (four native speakers of Russian and one of Macedonian) and Germanic (three native speakers of German and two of English). When tested, all participants had been attending courses on Greek for at least seven months and at most twenty months. With respect to the length of residence in Greece, all had spent from seven to eighteen months in the country. Information about the profile of the L2 learners is also provided in Table 2: Since the main experiments investigated the role of aspect in the interpretation of motion events, we wanted to ensure that the L2 participants were aware of (a) the morphological aspectual distinction in the Greek verbal system and (b) the semantic features associated with perfective and imperfective aspect. For this reason, L2 learners initially completed a cloze task (Agathopoulou and Papadopoulou 2007), in which participants had to fill the gaps with the appropriate verb form. The choice was determined by an adverbial included in the sentence, which unambiguously rendered the event bounded or unbounded (e.g., (32) and (33)). The cloze task consisted of thirty sentences, all referring to the past. In sixteen of the sentences the gap had to be completed with the perfective and in fourteen with the imperfective past verb form. All the verbs were in active voice and were given at the end of each sentence in the 3rd person, present, indicative: (32) Otan itan nea,.…..…..…..….. tria foremata ti vdhomada. when was young,……………..three dresses the week “When she was young, ……..three dresses per week.” (33) Persi i Mary …..…..…..….. mono mia hristujeniaktiki karta. last-year the Mary………..only one Christmas card “Last year Mary ……….only one Christmas card.”

(ravi) (sew.3s) (ghrafi) (write.3s)

In (32) the target form is erave (=was sewing/ used to sew), which is imperfective, and in (33) it is eghrapse (=wrote), which is perfective. The following table presents the mean accuracy scores per each aspectual verb form:



Aspect and motion verbs 

Table 3.  Pilot task: Mean accuracy scores Aspect

Mean scores (SD)

Perfective Imperfective Total

14.20/16 (1.69) 12.70/14 (1.57) 26.90/30 (3.00)

As shown in Table 3, all L2 learners performed very well. The mean accuracy rate for both aspectual forms was 90%, which suggests that the L2 learners who participated in the study were able to associate the perfective and the imperfective aspect with the appropriate morphological and semantic features18. The main experiments have also been conducted with native speakers of Greek for control purposes. The comprehension task was run with ten monolingual native speakers of Greek (females: 6; mean age: 22.1 years, SD: 1.66). A different group of ten native speakers of Greek participated in the production task (females: 5, mean age: 24.90 years, SD: 3.81). 4.3

Method

Sentence-picture matching (SPM) task 4.3.1 Materials The SPM task consists of forty-six items corresponding to forty-six quartets of pictures related to forty-six sentences19. Among the forty-six items there are four practice, fourteen filler and twenty-eight critical items. The critical sentences always denote a motion event expressed by a manner of motion verb and a PP. Seven motion verbs are used, namely treho (= to run), peto (= to fly), perpato (= to walk), 18. An anonymous reviewer points out that a pilot task should have been used to test our L2 learners’ knowledge of simple and complex prepositions in Greek. We admit that methodologically this is a flaw in our study. However, simple and complex prepositions of the type tested in the comprehension task and expected to be used in the production task are frequent in Greek and part of the syllabus of elementary Greek courses. 19. The pictures used in the SPM task and the short videos in the production task were prepared for the purposes of a different collaborative project (IKYDA ’04) between the University of Hamburg and Aristotle University of Thessaloniki. We are grateful to our colleagues in the German and the Greek research groups (Monika Rothweiler, Annette Fox, Solveig Kroffke, Nadine Stahl, Maria Mastropavlou, Kalliopi Katsika and Agapi Mylonaki) for letting us use these materials for the present study. The pictures of the SPM task were modified to clearly show the difference between the directional and the telic readings. An example of the picture quartets is provided in the Appendix.

 Ianthi Maria Tsimpli and Despina Papadopoulou

pidho (= to jump), sernome (= to crawl), kolibo (= to swim) and odhigho (= to drive). Each verb is incorporated in two different motion events: in one the PP is introduced by the simple preposition se and in the other by a complex preposition (mesa se = into, pano se = onto). In each motion event the verb appears in its two aspectual forms (perfective vs imperfective). Therefore, each verb appeared in four conditions as shown below: Perfective – simple preposition (34) To alogho etrekse sto tsirko. the horse ran.perf.3s s-the circus “The horse ran in(to) the circus.” Imperfective – simple preposition (35) To alogho etrehe sto tsirko. the horse ran.imp.3s s-the circus “The horse was running in(to) the circus.” Perfective – complex preposition (36) To alogho etrekse mesa sto tsirko. the horse ran.perf.3s inside s-the circus “The horse ran in(to) the circus.” Imperfective – complex preposition (37) To alogo etrehe mesa sto tsirko. the horse ran.imp.3s inside s-the circus “The horse was running in(to) the circus.” In each quartet the pictures illustrate four different events, as follows: i. the moving entity (e.g., the horse) is inside a circus running (atelic – locative event), ii. the moving entity is heading towards a circus (atelic – directional event) as indicated by an arrow showing the path and direction of motion, iii. the moving entity ran and arrived at the circus (telic event) as indicated by an arrow showing the endpoint of the motion event, iv. the moving entity is involved in an activity which is not a motion event (distractor). The filler sentences always matched pictures such as (iv) and were included in the task for control purposes.



Aspect and motion verbs 

4.3.2 Procedure The participants were given a booklet that consisted of forty-six quartets of pictures. They were instructed to look at each quartet of pictures and at the same time the experimenter read aloud a sentence, one at a time. The participants’ task was to match the sentence they heard with one of the four pictures they saw. All the participants were tested individually in a quiet room. The forty-six sentences and quartets of pictures were divided into two sessions, so that each participant never saw the same set of pictures more than twice and never heard the same sentence more than once in the same session. Moreover, the same sets of pictures that appeared in the two different sessions or within the same session always included the four pictures in different order. The items were pseudo-randomized and there was a one-week interval between the two sessions. Production task 4.3.3 Materials The production task consisted of twenty-six short videos: two practice, eight filler and sixteen critical videos. The critical videos involved eight motion events presented in two different conditions: in one condition an entity was performing a motion event (i.e., walking) in a certain location (atelic video), whereas in the other condition the same entity was shown to perform the same motion activity and to arrive at a certain endpoint (telic video). For instance, a video showing a woman who was walking in a kitchen represented the atelic condition. On the other hand, a video depicting a woman walking and arriving at the kitchen represented the telic condition. The eight motion activities employed in this task were supposed to be illustrated by manner of motion verbs such as to walk (twice), to run, to fly (twice), to crawl, to drive and to jump. The filler videos described various kinds of non-motion actions. 4.3.4 Procedure The participants were told that they would watch a short video and would then be asked to describe what they saw. The videos were presented on a computer screen. After each video, the experimenter asked the question “What did X … do?”. The participants provided their response, which was recorded on the computer through Windows Media Player. All participants were tested individually in a quiet room.

 Ianthi Maria Tsimpli and Despina Papadopoulou

4.4

Results

Sentence-picture matching (SPM) task The NSs of Greek always matched the filler sentences with the distractor pictures and the L2 learners did so 98.60% of the time. In addition, the native speakers’ data revealed two incorrect matches of critical sentences with distractor pictures and one “no response”, whereas in the L2 data there were five incorrect matches of a critical sentence with a distractor picture. The incorrect responses to the critical sentences have been eliminated from any further analyses. This resulted in the elimination of 1% of the native speakers’ data and 1.80% of the L2 data. We will begin by the presentation of the participants’ interpretation of the sentences they heard depending on the aspectual form of the verb and irrespectively of the preposition that introduced the path or the goal of motion. First, the participants’ responses, i.e., the picture they chose among the four alternates, were coded depending on whether they denoted path (PP adjunct) or goal (PP complement). The two groups’ PP-complement and PP-adjunct readings for the two aspectual verb forms are presented in Figure 1 below. As shown in Figure 1, both the NSs and the L2 learners of Greek relied on the aspectual form of the verb to interpret the sentences. Namely, when the predicate they heard included an imperfective verb, they interpreted it as denoting a locative event (PP-adjunct), whereas, when the verb in the spoken sentence was in the perfective aspect, the motion event was understood as being non-locative (PPcomplement). These observations have been confirmed statistically. There were 100 90 80 70 60

complement adjunct

50 40 30 20 10 0

NS

L2 learners imperfective

NS

L2 learners perfective

Figure 1.  Mean percentages of (non) locative readings for each aspectual form



Aspect and motion verbs 

significantly more PPpath readings with imperfective than with perfective manner of motion verbs (L1-Greek: χ2= 96.353, p<0.001; L2-Greek: χ2= 18.339, p<0.001). In addition, there were significantly more PPpath than PPgoal interpretations with imperfective aspect (L1-Greek: χ2= 58.696, p<0.001, L2-Greek: χ2= 9.993, p<0.01), whereas the reverse effect was obtained for perfective aspect (L1-Greek: χ2= 38.338, p<0.001, L2-Greek: χ2= 8.377, p<0.01). Note, however, that the performance of the L2 learners is significantly different from that of the NSs in both the imperfective (χ2= 12.759, p<0.001) and the perfective (χ2= 6.327, p<0.02) aspect. The data have also been analyzed with respect to telicity. More specifically, the pictures that illustrated locative and directional events were coded as atelic and the ones that depicted an entity moving and reaching an endpoint as telic. Figure 2 displays the percentages of telic and atelic responses for the two aspectual verb forms and the two groups. The data shown in Figure 2 are similar to those of Figure 1 in that the aspectual form of the verb affected the readings the participants imposed on the sentences. This means that there were more atelic interpretations with imperfective than with perfective verb forms (L1-Greek: χ2= 44.863, p<0.001; L2-Greek: χ2= 22.466, p<0.001). In addition, there were significantly more atelic than telic readings with imperfective aspect (L1-Greek: χ2= 84.522, p<0.001; L2-Greek: χ2= 43.277, p<0.001). On the other hand, the difference between telic and atelic interpretations in the perfective aspect was not significant for either group (L1-Greek: χ2= 0.353, p=0.553; L2-Greek: χ2= 0.029, p=0.865). Moreover, even though no 100 90 80 70 60

telic

50

atelic

40 30 20 10 0

NS

L2 learners imperfective

NS

L2 learners perfective

Figure 2.  Mean percentages of (a)telic readings for each aspectual form

 Ianthi Maria Tsimpli and Despina Papadopoulou

35 30 25 20

NS

15

L2ers

10 5 0 imperfective

perfective

Figure 3.  Directional interpretations per aspectual form and participant group

significant differences were found between the native speakers and the L2 learners in the perfective aspect (χ2= 0.291, p=0.589), in the imperfective the L2 learners did differ significantly from the native speakers (χ2= 6.109, p<0.02). A further analysis has been performed with respect to the directional readings of the sentences. Figure 3 illustrates the percentages of directional responses with imperfective and perfective verbs. As Figure 3 shows, the NSs’ directional responses are affected by the aspectual form of the verb; namely, significantly more directional responses have been obtained with perfective than with imperfective verb forms (χ2= 23.556, p<0.001). On the other hand, no such effect was found in the L2 data (χ2= 0.106, p=0.745). We also counted the adjunct and complement interpretations with respect to both the aspectual form of the verb and the preposition type (simple vs complex): Table 4 shows that the NSs’ interpretations depended on the aspectual form of the verb but not on the type of preposition. This means that for both preposition types the NSs show (a) an unambiguous preference for the PPpath reading in the imperfective (simple P: χ2= 8.471, p<0.01, complex P: χ2= 62.229, p<0.001) and (b) a Table 4.  Mean percentages per aspectual verb form and preposition type Groups

NS L2 learners

Imperfective Simple Complex Adjunct Compl. Adjunct Compl. 68 51

32 49

97 76

3 24

Perfective Simple Complex Adjunct Compl. Adjunct Compl. 19 23

81 77

29 52

71 48



Aspect and motion verbs 

preference for the PPgoal interpretation in the perfective (simple P: χ2= 27.657, p<0.001, complex P: χ2= 12.188, p<0.001). However, we have to note that the PP2 path reading with imperfective significantly increases (χ = 20.888, p<0.001) when the preposition is complex (68% with simple vs 97% with complex Ps), which supports previous findings on manner-of-motion verbs (Papadopoulou 1996). On the other hand, the data from the L2 learners indicate that their interpretations were affected by the choice of the preposition. More specifically, they show a preference for the PP-adjunct reading with the imperfective, only when the preposition is complex (χ2= 18.514, p<0.001). By contrast, in the perfective aspect they exhibit a preference for PP-complement only with simple prepositions (χ2= 19.841, p<0.001). To summarize the results from the comprehension task, we found that both participant groups relied on the morphological aspect of the verb to comprehend motion events. More specifically, they both associated perfective aspect with PPgoal readings and imperfective aspect with PPpath readings. In addition, both groups had similar behaviour with respect to the encoding of (a)telicity: imperfective aspect is clearly linked with atelicity, whereas perfective manner-of-motion verbs were not necessarily interpreted as telic. Nevertheless, there is a quantitative difference between the two groups since these tendencies are stronger in the NSs than the L2 learners. Furthermore, the learners’ preference for PPpath and PPgoal readings needs to be reinforced by the distinction between simple and complex prepositions. Native speakers’ preference for adjunct vs complement is associated with the choice of preposition (simple vs complex) only in the imperfective. Finally, even though for both groups the directional reading is not the most preferred one with either imperfective or perfective aspect, only in the NSs’ data are the directional interpretations affected by the aspectual form of the verb. Production task We eliminated any responses that were irrelevant to the purpose of the task, namely responses that did not involve motion events. This resulted in the elimination of 4% (6 out of 160 responses) of the native speakers’ data and 16% (25 out of 160 responses) of the L2 data. The L2 learners produced significantly more irrelevant responses than the NSs in both the atelic (χ2=5.959, p<0.02) and the telic (χ2=6.944, p<0.01) videos. In all subsequent analyses only relevant responses have been counted. First, we present the target and non-target responses per video condition and participant group (see Figure 4). Any utterances that unambiguously denoted telic motion events in the atelic video condition and atelic motion events in the telic video condition were considered as non-target. For example, the predicate in sentence (38) describes an unambiguously telic event, since the verb beno (enter) incorporates the PATH and the PP stin kuzina is necessarily a complement. Such a

 Ianthi Maria Tsimpli and Despina Papadopoulou

response is non-target, when the video describes an atelic motion event, and target, when the video illustrates a telic motion event. (38) I jineka bike stin kuzina. the woman entered.perf.3s s-the kitchen “The woman entered the kitchen.” In addition, sentence (39) denotes an atelic, locative, motion event, because the manner-of-motion verb is in the imperfective. Such an utterance was considered as non-target for the telic videos and as target for the atelic videos. (39) To aeroplanaki petuse se mia ethusa. the airplane flew.imp.3s in a room “The airplane was flying in a room.” Consider, finally, sentence (40): (40) I petaludha petakse mesa sto vazo. the butterfly flew.perf.3s in s-the vase “The butterfly was flying in the vase.” As already discussed in Section 2.1, the predicate in (40) is ambiguous with respect to the description of (a)telic motion events. In other words, as far as the grammatical representation is concerned, there are two possible structures. In one the PP denotes the GOAL and is the complement of the verb, whereas in the other the 100 90 80 70 60

non-target

50

target

40 30 20 10 0

NS

L2 learners atelic videos

NS

L2 learners telic videos

Figure 4.  (Non)Target responses per condition and participant group (%)



Aspect and motion verbs 

PP denotes the PATH and is an adjunct. Therefore, such utterances have been counted as target responses in both the telic and the atelic video conditions. As is obvious from Figure 4, the NSs’ and the L2 learners’ performance in the atelic video condition was very good and there were no statistically significant differences between the two groups (χ2=0.357, p=0.550). In the telic video condition, however, the L2 learners performed significantly worse than the NSs (χ2=11.782, p<0.01). The following graph presents the percentages of ambiguous (cf. (40)) and unambiguous target responses in each video condition. Notice that unambiguous target responses in the telic condition were utterances including a perfective motion verb that inherently denotes PATH, i.e., go, enter (cf. (38)). In the atelic videos, utterances containing imperfective manner-of-motion verbs (cf. (39)) and responses in which lexical means are used to express location (cf. (41) and (42)) were counted as unambiguous: (41) Enas antras ekane voltes stin apothiki. a man did.3s rounds s-the loft “A man was walking back and forth in the utility room.” (42) To koritsi horopidhise mesa stis laspes. the girl bounced.perf.3s in s-the mud “The girl was bouncing in the mud.” 100 90 80 70 60

ambiguous

50

unambiguous

40 30 20 10 0

NS

L2 learners atelic videos

NS

L2 learners telic videos

Figure 5.  (Un)Ambiguous responses per condition and participant group (%)

 Ianthi Maria Tsimpli and Despina Papadopoulou

Both participant groups produced significantly more ambiguous responses in the telic than in the atelic videos (NSs: χ2=11.588, p<0.01; L2 learners: χ2=9.470, p<0.01), which indicates that the use of perfective manner-of-motion verbs with PPs is preferred when describing a telic rather than an atelic motion event. Furthermore, in the atelic videos both groups produced more unambiguous than ambiguous responses (NSs: χ2=27.000, p<0.001; L2 learners: χ2=47.032, p<0.001), which is consistent with the previous finding. In the telic videos, however, this effect was significant only for the L2 learners (χ2=8.696, p<0.01), which suggests that the construction [manner-of-motion Vperf + PP] is not productively used by the L2 learners in this condition; rather, the L2 speakers prefer to use inherently directional verbs to express telic motion events. Nonetheless, the L2 learners used this construction less frequently than the NSs in both the atelic (χ2=5.216, p<0.03) and the telic (χ2=3.883, p<0.05) videos. An additional analysis has been performed for the unambiguous target responses in the atelic video condition, in order to find out which is the preferred construction for the description of an atelic, locative event. Figure 6 presents the frequency of responses with (a) imperfective manner-of-motion verbs (cf. (39)), (b) light verbs with nouns showing atelic locative events (cf. (41)) and (c) inherently locative verbs marked for perfective (cf. (42)). As is obvious from the data in Figure 6, both the NSs and the L2 learners preferred to use a manner-of-motion verb marked for imperfective aspect to denote location (NSs: χ2=11.267, p<0.001; L2: χ2=20.763, p<0.001). Moreover, there are 100 90 80 70 60

perf + lexical

50

light verb

40

imperfective

30 20 10 0 NS

L2 learners

Figure 6.  Analysis of unambiguously target responses in the atelic video condition (%)



Aspect and motion verbs 

Table 5.  Frequency of PP use per condition and participant group Group NS L2 learners

Simple

Atelic Complex

No PP

Simple

Telic Complex

20/75 (27%) 29/63 (46%)

45/75 (60%) 12/63 (19%)

10/75 (13%) 22/63 (35%)

31/72 (43%) 29/42 (69%)

41/72 (57%) 13/42 (31%)

no statistically significant differences between the two groups in the use of imperfective manner-of-motion verbs to express atelic motion events (χ2=1.032, p=0.310). Notice also that 26% of the imperfective verbs used by the NSs were unambiguously locative, whereas the L2 learners never used such verbs in their responses. This effect is probably due to more constrained lexical repertoire of the L2 learners as compared to the NSs. Finally, in this task as in the comprehension experiment the L2 learners distinguished between telic and atelic motion events through the use of PPs as shown in Table 5. In the telic videos, the L2 learners used more simple than complex prepositions (χ2=12.190, p<0.001), which is not the case for the NSs (χ2=2.778, p=0.096). Recall that L2 learners show the same pattern in the comprehension task. A difference between the two tasks is that PP omission is an option that L2 learners in particular employ in the description of atelic motion events. This option is based on the adjunct status of the PP in this condition. In the atelic video condition, the L2 learners omit PPs more often that the NSs (χ2=8.958, p<0.01). To sum up the main results from the production task, we found that the L2 learners behaved in a native-like way when describing atelic motion events, since they consistently used manner-of-motion verbs marked for imperfective aspect. On the other hand, their performance diverged from that of the NSs’ when expressing a telic motion event. In this condition the learners produced significantly more errors than the NSs, i.e., they used more imperfective forms to describe a telic motion event, on one hand, and, on the other, they did not seem to productively use perfective manner-of-motion verbs together with goal PP complements. Instead they relied on the use of inherently directed motion verbs which in their perfective form are necessarily telic in Greek as in other languages too (cf. fn. 6). Finally, preposition choice seems to strengthen the expression of telic vs atelic motion events. 5. Discussion The aim of this study was to investigate whether native and non-native speakers of Greek make use of grammatical aspect in structures with manner-of-motion verbs

 Ianthi Maria Tsimpli and Despina Papadopoulou

at interface levels. In particular, at the syntax-semantics interface where argument realization interacts with aspectual properties and at the syntax-discourse interface where the interpretation of the predicate as telic or atelic is involved. Since Aspect is an interpretable feature, the study also aims to test the IH with respect to the claim that interpretable features should not be problematic for L2 acquisition. The interaction of manner-of-motion verbs with Aspect provides relevant evidence for the role of interpretability at these two interfaces. Our predictions with respect to the performance of the Greek native speakers have been confirmed. The distinction between perfective and imperfective aspect affected (a) the complement vs adjunct choice in the representation of the PP and (b) the telic vs atelic interpretations of manner-of-motion verbs. It should be noted that (b) is a result relevant to both the comprehension and the production tasks. The findings from the NSs provide further support for the independently attested transitivity preference associated with the perfective form of Greek verbs (Tsimpli and Papadopoulou 2006), as perfective manner-of-motion verbs are preferably construed with PP complements. Moreover, the NSs’ data from this study also support the pragmatic link between perfectivity and telicity suggested in previous research (Chila-Markopoulou and Mozer 2001; Giannakidou 2003; Tsimpli and Papadopoulou 2006). With respect to the interaction of Aspect and preposition choice (simple vs complex) in encoding the complement vs adjunct and the telic vs atelic readings, NSs show a preference for complex over simple prepositions to describe atelic motion events. A similar tendency is found in the comprehension task, where the complex preposition significantly strengthens the adjunct reading. With respect to the L2 learners, the control test we used showed that they have mastered the morphological properties of aspectual distinctions in Greek and use the corresponding forms accordingly. Specifically, they are aware of the [+/-bounded] distinction between perfective and imperfective forms as well as of the [+/-iterative/habitual] specification of the imperfective. Turning to the results of our comprehension experiment, the L2 learners displayed a significant preference for adjunct readings with imperfective verbs and complement readings with perfective verbs. This means that they show sensitivity to the effects of aspect on the argument structure of manner-of-motion verbs in Greek, a syntax-semantics interface issue. With respect to the syntax-discourse interface, the L2 data from both tasks show that imperfective verb forms are strongly associated with atelic readings and perfective forms with telic readings. In this respect, the L2 learners behave in a native-like way. However, the L2 learners differed from the NSs in several respects. First, the L2 learners’ preference for PP-adjunct readings with imperfective and for PP-complement readings with perfective aspect is modulated by the choice of complex vs



Aspect and motion verbs 

simple prepositions. In particular, the adjunct interpretation with the imperfective aspect is significantly favored only when the preposition is complex and the complement reading with the perfective is favored only when the preposition is simple. We, thus, suggest that the choice of structure adopted by the L2 learners for the complement and the adjunct reading is sensitive to the nature of the PP. Note that the two types of prepositions differ in that the simple preposition heads a P with a DP complement, whereas the complex preposition involves two PPs a simple one embedded under a higher P (for further discussion on the structure of simple and complex PPs in Greek see Theophanopoulou-Kontou 1992; Terzi 2007). Our data show that whereas both L2 learners and native speakers associate complex prepositions with locative readings, only the L2 learners associate simple prepositions with non-locative readings. NSs treat simple prepositions as equally compatible with locative and non-locative readings. We will come back to the L2 learners’ strong preference for a one-to-one mapping between form and meaning at the end of the section. Secondly, the L2 learners differ from the NSs in that their directional readings were fewer and did not depend on aspect. Recall that direction is one of the two (the other being the telic reading) possible syntax-discourse interpretations when the PP is a complement. It might be the case that direction is only lexically expressed – by the use of an unambiguously directional preposition –in the L2 learners’ interlanguage. Furthermore, this finding suggests that perfective manner-ofmotion verbs with complement PPs are preferably understood by the L2 learners as telic and not as atelic, since perfectivity strengthened only the telic and not the directional responses. Notice that this crucially differs from the NSs’ behavior, where the perfective aspect of the verb resulted in more directional and telic responses than imperfective aspect. We could thus argue that the L2 learners have a narrower range of possible interpretations at the syntax-discourse interface. They choose to associate (a)telicity with aspectual morphology: perfective manner-ofmotion predicates are interpreted as telic and imperfective ones as atelic. This is the preferred option attested in NSs too who, however, also allow for directional readings with the perfective, but to a smaller extent than the telic reading. Furthermore, in the production task, the L2 learners produced more unambiguous than ambiguous responses in both the telic and the atelic conditions. In the telic videos, they preferred to use an inherently directed motion verb in the perfective than an ambiguous manner-of-motion verb, which, in the perfective, can also have a telic interpretation. Hence, the L2 learners used lexical cues, in this case the semantic features of the verb (Aktionsart), in addition to grammatical aspect, to encode telic events. In the atelic condition they produced fewer perfective manner-of-motion verbs with a PPpath, than the NSs. This implies that perfective aspect is not used to encode atelicity, which supports the claim, based on the

 Ianthi Maria Tsimpli and Despina Papadopoulou

data from the SPM task, that the learners interpret the two aspectual forms according to the telic/atelic distinction. Finally, the L2 learners produced significantly more non-target responses than the NSs in the telic condition of the production task: 70% of the learners’ nontarget responses – the equivalent percentage in the NSs data is 33% – involved an imperfective verb showing overuse of the imperfective aspect. We think that this is not due to the incorrect mapping of the morphological and semantic features of aspect, since in the cloze task (cf. Section 4.2) the L2 learners used the two aspectual forms accurately. Rather, the overuse of imperfective forms is probably related to the formation of the perfective aspect in on-line production which presents learners with some difficulty, since the citation form of a verb is in the imperfective and the derivation of the perfective depends on a large set of morpho-phonological properties. Taken together the findings from the comprehension and the production task indicate that the two participant groups effectively use Aspect to arrive at the preferred argument structure as well as the preferred telic/atelic interpretation of predicates with manner-of-motion verbs. In this respect, the interpretable feature of Aspect is unproblematic for L2 learners in the computational component and at the interfaces, which supports the predictions of IH for L2 acquisition. The two groups differ, however, in (a) the sensitivity they show to lexical cues such as the choice of preposition and the inherent aspectual features of the verb, and (b) the use of perfective manner-of-motion verbs, which are structurally ambiguous. With respect to (a), we suggest that L2 learners rely on lexical means, i.e., preposition type, to strengthen the intended meaning since the salience of lexicalized features is higher than that of grammatical morphology in L2 acquisition. With respect to (b), the preference for unambiguous constructions provides a more direct mapping between the syntax and the interfaces. In the phenomenon we study, the syntax-discourse interface restricts the grammatical options favoring the telic (rather than the directional) interpretation of the perfective. What these findings imply is that the L2 learners prefer to have a one-to-one correspondence between form (perfective vs imperfective aspect, simple vs complex Ps) and meaning (telic vs atelic, non-locative vs locative). This is reminiscent of Rizzi’s (2005) Categorial Uniformity principle, which is argued to apply at the syntax-semantics interface and defines the unmarked case of form-meaning mappings. It is plausible to assume that this principle also guides L2 grammars. In effect both differences distinguishing native from L2 speakers boil down to the L2 learners’ tendency to avoid ambiguity through (lexical or grammatical) form-to-meaning mappings. Whether avoiding ambiguities is an inherent property of developing L2 grammars or is an L1-effect in our study, remains an open question. If the Interpretability Hypothesis is correct, then Categorial Uniformity should apply even more strongly in L2



Aspect and motion verbs 

grammars since interpretable features alone will be responsible for regulating L1-L2 differences. 6. Concluding remarks This study of ambiguous manner-of-motion verbs in Greek L2 attempts to address a question that FFFH and IH have not dealt with yet, namely whether interpretable features can be acquired in a target fashion by L2 learners and thus provide a radically distinct pattern of L2 acquisition compared to uninterpretable features. The interpretable feature studied, Aspect, is a grammatical category with morphological expression of (im)perfectivity in Greek, but is also relevant to argument structure and the (a)telic interpretation of manner-of-motion verbs. Our findings support the predictions of IH, since the group of the L2 learners tested (a) has mastered the features associated with the perfective/imperfective distinction and (b) uses this distinction effectively at the syntax-semantics and the syntax-discourse interfaces. However, the aspectual distinctions employed by the non-native speakers of Greek are reinforced through the use of lexical information, an L2 property not attested in the NSs’ data. Overall, we conclude that interpretable features, in contrast to uninterpretable ones, are not vulnerable in L2 acquisition and any difficulties attested may be attributed to the ambiguities and the resulting underspecification of the form (perfective/imperfective) with respect to the possible interpretations of (a)telicity at the syntax-discourse interface. Acknowledgements Part of this study has been funded by IKYDA ’04. We would like to thank Monika Rothweiler, Annette Fox, Solveig Kroffke, Nadine Stahl, Maria Mastropavlou, Kalliopi Katsika and Agapi Mylonaki for fruitful discussions as well as two anonymous reviewers for useful and constructive comments. We would also like to thank Anastasia Giannakidou, Jason Merchant, Anna Roussou and Tasos Tsangalidis for useful discussions on the topic. All errors remain of course ours.

 Ianthi Maria Tsimpli and Despina Papadopoulou

References Agathopoulou, E. & Papadopoulou, D. 2007. Morphological dissociations in the L2 acquisition of an inflectionally rich language. Paper presented at EUROSLA 17, Newcastle, UK, 11–14 September. Babko-Malaya, O. 1999. Zero morphology: A study of aspect, argument structure and case. PhD dissertation, The State University of New Jersey. Belletti, A., Bennatti, E. & Sorace, A. 2005. Theoretical and developmental issues in the syntax of subjects: Evidence from near-native Italian. Ms, University of Siena. Berman, R.A. & Slobin, D.I. 1994. Relating events in narrative: A crosslinguistic developmental study. Hillsdale NJ: Lawrence Erlbaum Associates. Borer, H. 1994. The projection of arguments. In University of Massachussetts Occasional Papers in Linguistics Functional Projections 17, E. Benedicto & J. Runner (eds.), 19–47. Amherst MA: GLSA. Borer, H. 2005. Structuring sense, Vol. 2: The normal course of events. Oxford: OUP. Chila-Markopoulou, D. & Mozer, A. 2001. Telicity and referentiality in the VP of Modern Greek: Aspect and determiner. In Proceedings of the 4th International Conference on Greek Linguistics, Y. Agouraki et al (eds.), 138–145. Nicosia: University Studio Press. Comrie, B. 1976. Aspect: An introduction to the study of verbal aspect and related problems. Cambridge: CUP. Dowty, D. 1979. Word meaning and Montague grammar. Dordrecht: Reidel. Filip, H. 1999. Aspect, eventuality types and nominal reference. New York NY: Garland. Folli, R. 2001. Constructing telicity in English and in Italian. PhD dissertation, Oxford University. Giannakidou, A. 2003. A puzzle about until and the present perfect. In Perfect explorations, A. Alexiadou, M. Rahter & A. von Stechow (eds.), 101–133. Berlin: Mouton de Gruyter. Goad, H. & White, L. 2006. Ultimate attainment in interlanguage grammars: A prosodic approach. Second Language Research 22: 243–268. Hawkins R. & Chan, C-Y. 1997. The partial availability of universal grammar in second language acquisition: The ‘failed functional features hypothesis’. Second Language Research 13: 187–226. Hawkins, R. & Franceschina, F. 2004. Explaining the acquisition and non-acquisition of determiner-noun gender concord in French and Spanish. In The acquisition of French in different contexts, P. Prévost & J. Paradis (eds.), 175–205. Amsterdam: John Benjamins. Hawkins, R. & Hattori, H. 2006. Interpretation of English multiple wh-questions by Japanese speakers: A missing uninterpretable feature account. Second Language Research 22: 269–301. Hickmann, M. & Hendriks, H. 2006. Static and dynamic location in French and in English: Crosslinguistic developmental perspectives. First Language 26: 103–135. Horrocks, G. & Stavrou, M. 2003. Actions and their results in Greek and English: The complementarity of morphologically encoded (viewpoint) aspect and syntactic resultative predication. Journal of Semantics 20: 297–327. Horrocks, G. & Stavrou, M. 2007.Grammaticalized aspect and spatio-temporal culmination. Lingua 117: 605–644. Iatridou, S. 2000. The grammatical ingredients of counterfactuality. Linguistic Inquiry 31: 231–270.



Aspect and motion verbs  Inagaki, S. 2001. Motion verbs with goal PPs in the L2 acquisition of English and Japanese. Studies in Second Language Acquisition 23: 153–170. Matsunaga, K. 2006. L1-constrained motion expressions in the L2 acquisition of English. In Essex graduate student papers in language and linguistics, Vol. 8, F. Nakatsuhara, G. Sarko, C. Jaensch & N. Snape (eds.), 43–65. Colchester: University of Essex. Montrul, S. 2001. Agentive verbs of manner of motion in Spanish and English as second languages. Studies in Second Language Acquisition 23: 171–206. Mozer, A. 1994. Ποιόν και aπόψεις του pήματος [Situational and view point aspect of the verb]. Athens: Parousia. Navarro, S. & Nicoladis E. 2005. Describing motion events in adult L2 Spanish narratives. In Selected proceedings of the 6th Conference on the Acquisition of Spanish and Portuguese as first and second languages, D. Eddington (ed.), 102–107. Somerville MA: Cascadilla Proceedings Project. Oh, K. 2003. Manner and path in motion event descriptions in English and Korean. In Proceedings of the 27th Annual Boston University Conference on Language Development, B. Beachley, A. Brown & F. Conlin (eds.), 580–590. Somerville MA: Cascadilla Press. Özçalişkan, Ş & Slobin D. I. 2000. Climb up vs. ascend climbing: Lexicalization choices in expressing motion events with manner and path components. In Proceedings of the 24th Annual Boston University Conference on Language Development, B. Beachley, A. Brown & F. Conlin (eds.), 558–570. Somverville MA: Cascadilla Press. Papadopoulou, D. 1996. Ρήματα κίνησης στην ελληνική: σημασιοσυντακτική προσέγγιση [Motion verbs in Greek: A syntactic and semantic approach]. MA thesis, University of Athens. Rizzi, L. 2005. On the grammatical basis of language development: A case study. In The Oxford handbook of comparative syntax, G. Cinque & R. Kayne (eds.) 70–109. Oxford: OUP. Romanova, E. 2006. Constructing perfectivity in Russian. PhD Dissertation, University of Tromsø. Rothweiler, M., Tsimpli, I.-M., Papadopoulou, D., Fox, A., Katsika, K., Kroffke, S.M., Mastropavlou, M., Mylonaki, A. & Stahl, N. 2007. Motion verbs in Greek and German: Evidence from typically developing and SLI children. Paper presented at the 18th International Symposium of Theoretical and Applied Linguistics, 4–6 May, Thessaloniki. Sanz, M. 2000. Events and predication. A new approach to syntactic processing in English and Spanish. Amsterdam: John Benjamins. Schwartz, B. & Sprouse, R. A. 1996. L2 cognitive states and the full transfer/full access model. Second Language Research 12: 40–72. Schwartz, B. & Sprouse, R. A. 2000. When syntactic theories evolve: Consequences for second language acquisition research. In Second Language Acquisition and Linguistic Theory, J. Archibald (ed.) 156–186. Oxford: Blackwell. Sioupi, A. 2002. On the syntax and semantics of verb-complement constructions that involve ‘creation’: A comparative study in Greek and German. In Issues in formal German(ic) typology, W. Abraham & J.-W. Zwart (eds.), 263–284. Amsterdam: John Benjamins. Sioupi, A. 2005. Morphological and telicity aspect with accomplishment VPs in Greek. In Crosslinguistic views on tense, aspect and modality [Cahiers Chronos 13], B. Hollebrandse, A. van Hout & C. Vet (eds.) 131–144, Amsterdam: Rodopi. Slobin, D. 1996. Two ways to travel: Verbs of motion in English and Spanish. In Essays in semantics, M. Shibatani & S.A. Thompson (eds.), 195–317. Oxford: OUP. Smith, C.S. 1991. The parameter of aspect. Dordrecht: Kluwer.

 Ianthi Maria Tsimpli and Despina Papadopoulou Smith, N. & Tsimpli, I.-M. 1995. The mind of a savant. Oxford: Blackwell. Sorace, A. 2006. Possible manifestations of shallow processing in advanced second language learners. Applied Psycholinguistics 27: 88–91. Talmy, L. 1985. Lexicalization patterns: Semantic structure in lexical forms. In Language typology and semantic description, Vol. 3: Grammatical categories and the lexicon, T. Shopen (ed.), 36–149. Cambridge: CUP. Talmy. L. 2000. Toward a cognitive semantics, Vol. 2: Typology and process in concept structuring. Cambridge MA: The MIT Press. Tenny, C. 1987. Grammaticalising aspect and affectedness. PhD dissertation, MIT. Terzi, A. 2007. Locative prepositions, predicate inversion and full interpretation. In Selected papers from the 17th international symposium on theoretical and applied linguistics, E. Agathopoulou, M. Dimitrakopoulou & D. Papadopoulou (eds.), 210–219. Thessaloniki: Monochromia. Theophanopoulou-Kontou, D. 1992. Οι σύνθετες προθετικές φράσεις της ΝΕ και η δομή τους (The complex prepositional phrases in modern Greek and their stucture). In Proceedings of the 13th Annual Meeting of the Department of Linguistics, 311–330. University of Thessaloniki. Travis, L. 2000. Event structure in syntax. In Events as grammatical objects: The converging perspectives of lexical semantics and syntax, C. Tenny & J. Pustejovski (eds.), 145–185. Stanford CA: CSLI. Tsimpli, I.-M. 1996. The prefunctional stage of first language acquisition. New York NY: Garland. Tsimpli, I.-M. 2003. Clitics and determiners in L2 Greek. In Proceedings of the 6th Conference on Generative Approaches to Second Language Acqusition, J. Liceras, H. Zobl & H. Goodluck (eds.), 331–339. Somerville MA: Cascadilla Proceedings Project. Tsimpli, I.-M. & Papadopoulou, D.2006. Aspect and argument realization: A study on antecedentless null objects in Greek. Lingua 116: 1595–1615. Tsimpli, I.-M. & Sorace, A. 2006. Differentiating interfaces: L2 performance in syntax-semantics and syntax-discourse phenomena. In Proceedings of the 30th Annual Boston University Conference on Language Development, DBamman, T. Magnitskaia & C. Zaller (eds.), 653–664. Somerville MA: Cascadilla Press. Tsimpli, I.-M & Dimitrakopoulou, M. 2007. The Interpretability Hypothesis: Evidence from whinterrogatives in second language acquisition. Second Language Research 23(2): 215–242. Tsimpli, I.-M. & Mastropavlou, M. 2008. Feature interpretability in L2 acquisition and SLI: Greek clitics and determiners. In The role of formal features in second language acquisition, J. Liceras, H. Zobl & H. Goodluck (eds.), 142–183. London: Routledge. Verkuyl, H. J. 1972. On the compositional nature of the aspect. Dordrecht: Reidel. White, L. 2003. Second language acquisition and Universal Grammar. Cambridge: CUP. Zagona, K. 1996. Compositionality of aspect: Evidence from Spanish aspectual se. In Aspects of Romance linguistics, C. Parodi, C. Quicoli, M. Saltarelli & M.L. Zubizarreta (eds.), 475–488. Washington DC: Georgetown University Press. Zubizarreta, M. L & Oh, E. 2007. On the syntactic composition of manner and motion. Cambridge MA: The MIT Press.



Aspect and motion verbs 

Appendix

Associating meaning to form in advanced L2 speakers An investigation into the acquisition of the English present simple and present progressive Sarah Ann Liszka

University of London Institute in Paris This paper investigates the L2 acquisition of the distributional and interpretational properties of the English present simple (e.g., She works at home) and present progressive (e.g., She is working at home). To test whether advanced L2 learners are successful in assigning target-like meanings to these forms, sixteen advanced L1 French-L2 English and thirteen L1 English informants participated in two oral tasks and a written gap-fill task. Results indicate that these L2 speakers use both forms productively, but show optionality in consistently producing the progressive in appropriate environments. The study considers a permanent L2 deficit arising from L1-L2 parametric differences as a potential source of difficulty in acquiring the target-like interpretations associated with the two forms.

1. Introduction Locating the source of persistent optional use of overt forms in advanced L2 learners remains a central issue within the Generative Grammar framework. Much research in this domain has focussed on the L2 acquisition of functional categories (FCs) and their features, in relation to issues of accessibility and instantiation of L2 properties that are parametrically different in the L1. Whilst a number of hypotheses have been generated from this perspective, they can broadly be divided into two approaches in an attempt to account for optionality. The first approach (adopting the label from Slabakova 2003) is the Full Functional Representation position, which posits that L2 speakers are theoretically able to acquire FCs, their features and associated featural properties, yielding native-like grammatical representations,

 Sarah Ann Liszka

regardless of L1-L2 parametric variation (e.g., Lardiere, 1998a, 1998b, 2000; Prévost and White, 2000; Schwartz and Sprouse 1996). For example, Prévost and White’s (2000) Missing Surface Inflection Hypothesis suggests that the problem is one of lexical access, resulting in speakers sometimes selecting a lexical item comprising a subset of features of the target lexical item, i.e., a default form. The selection of a default form occurs either if the L2 speaker is unsure of the lexical form in a particular syntactic environment, or when a speaker is experiencing processing/communicative pressure. Such vulnerability to performance factors may obscure L2 learners’ underlying knowledge of a form in production, giving rise to optionality. A second example, the ‘Full Transfer/Full Access’ (FT/FA) hypothesis (Schwartz and Sprouse 1996; Schwartz 1999), claims that all of the fully-feature specified functional and lexical categories of the L1 are entirely transferred to the L2 initially, hence ‘full transfer’. Sufficient exposure to target language (TL) input subsequently reveals that some L1 syntactic representations do not generate TL sentences, triggering a UG-constrained restructuring of the grammar, hence full access. Successful parameter resetting requires positive evidence from L2 data for specific L2 properties to be instantiated and restructuring can happen at any point, producing a distinct IL grammar at each stage of development. However, if input is ambiguous, obscure or non-existent, L2 speakers may not always establish unambiguous underlying representations for certain phenomena, potentially leading to optionality in production. Implicit to the Full Functional Representation (FFR) perspective is that with access to the full range of properties made available by the language faculty, learners will acquire the appropriate form-meaning relationships associated with a particular form. Note, however, that this assumption refers to advanced learners; as Slabakova (2003) points out, studies have shown that overt morphological marking often precedes target use of a form (e.g., Bardovi-Harlig 1992; Montrul and Slabakova 2002). The second approach, the Representational Deficit position (Hawkins 2005), in earlier incarnations attributes optionality to a representational deficit (RD) arising from L1-L2 parametric differences, where a parameterised feature of the L2 that is not selected in the L1 is unavailable to post-childhood L2 learners (e.g., Smith and Tsimpli 1995; Hawkins and Chan 1997). More recently, the RD approach has explicitly claimed that the locus of impairment is restricted to parameterised uninterpretable features (i.e., syntactic features without semantic content, in contrast to interpretable syntactic features which affect semantic interpretation), suggesting that the remaining elements of UG are accessible for second language attainment (e.g., Tsimpli 2003, Tsimpli and Dimitrakopoulou 2007; Hawkins and Hattori 2006). In concrete terms, any uninterpretable feature associated with a FC that is encoded in the L1 is available for instantiation in the L2. Thus, development of that feature will not be impaired, eventually yielding syntactically driven reflexes



Associating meaning to form 

in production. Conversely, as L1 uninterpretable features remained fixed, any in the L2 that differ from the L1 are permanently unavailable throughout development in post-childhood SLA. For example, Hawkins and Franceschina (2004) posit that the L2 acquisition of gender and gender concord for adult second language learners will only be established if [ugender] is a feature of the L1. An abstract property associated with formal features is the property of ‘strength’, described as either ‘strong’ or ‘weak’ (the former triggering syntactic movement, such as v-to-T, in order to allow overt feature-checking to take place), which may also give rise to L1-L2 parametric differences. For example, Hawkins and Hattori (2006) found that although adult L1 Japanese-L2 advanced English learners interpreted wh‑questions appropriately, they appeared to not be constrained by the Attract Closest Principle. To account for these results it seems that Japanese speakers of L2 English may fail to establish a strong [uwh] feature on C[Q], which is parametrically opposed to a weak [uwh] feature on C[Q] in Japanese. They also suggest that where learners interpret and use (multiple) wh-questions appropriately, it is the effect of a strong focus feature in the L1 and apparent native-like use is superficial (urging caution more generally in interpreting results that show native-like production). The current study investigates this second approach, specifically considering the acquisition of the distributional and interpretational properties of the (British) English present progressive (be+V-ing) and present simple, whose core temporal meanings are in complementary distribution. Data from L1 French – L2 English advanced learners, who natively encode one representation (le présent) for both the interpretations associated with the two English forms, are used to discuss how parametric differences might account for any optionality evident in associating target-like meanings to forms. 2. Interpretations of the present simple and present progressive in English and the présent in French Starting with English, the underlying core interpretations of the present simple (1) and present progressive (2) are exemplified below (an inappropriate form in a particular context is preceded by #). (1) a. She works at home on Mondays. b. #She is working at home on Mondays. (2) a. She is working at home (at the moment). b. #She works at home (at the moment).

 Sarah Ann Liszka

As these examples demonstrate, the English present simple has a habitual/generic interpretation (1a) and is inappropriate in a context expressing an event/action that is occurring at Speech Time (ST), i.e., now (1b). Conversely, the English present progressive incorporates two interpretive elements: an existential/event in progress at ST, which is simultaneously an event that is ‘ongoing’1 (2a) and a habitual/generic interpretation is inappropriate (2b). Note that in certain contexts a progressive form is licensed for a temporary habitual interpretation, such as, She is working at home on Mondays for the next month, implying that she does not usually work at home on Mondays. However, for the purposes of this study, the emphasis is on the existential/event in progress at ST meaning underlying the present progressive versus the habitual/generic meaning of the present simple, also precluding the investigation of the acquisition of the aspectual meaning of ‘ongoing’ associated with the v-ing grammatical morpheme. Turning now to French, it differs from English as the présent licences both the interpretations associated with the two English forms: (3) Elle travaille chez elle tous les lundis. She works at home on Mondays. (4) Elle travaille chez elle (à présent). She is working at home (at the moment). Following Al-Hamad, Al-Malki, Casillas, Francheschina, Hawkins, Hawthorne, Karadzovska, Kato, Liszka, Lozano, Ojima, Okuwaki, & Thomas (2002) and Hawkins, Casillas, Hattori, Hawthorne, Husted, Lozano, Okamoto, Thomas and Yamada (2008), this paper assumes that the semantic difference between the existential/event in progress and habitual/generic interpretations encoded in the two English forms is “directly connected to whether or not the verb raises overtly to T(ense) in the syntax” (Al-Hamad et al. 2002: 55), where only raised verbs licence an existential/event in progress interpretation.

1. Treating the ‘existential/event in progress’ meaning and ‘ongoing’ meaning as two separate interpretive elements is motivated by the different grammatical properties underlying both meanings. The existential/event in progress interpretive element, overtly realised by raised aux be, requires linking to a reference time: in the present progressive the reference time coincides with speech time ‘now’; in the past progressive the event is in progress in relation to a reference time in the past. On the other hand, ‘ongoingness’ pertains to grammatical aspect. Crucially, aspect alters the internal temporal contour of an event/situation, which is independent of any reference time (see, for example, Comrie 1976). Thus progressive aspect, overtly realised by the v-ing morpheme in English, expresses the ‘ongoingness’ of an event irrespective of the event’s location in time (i.e., the past, present, future), and without making any explicit reference to the event’s beginning, middle or end.



Associating meaning to form 

Hawkins et al. (2008) claim that v-to-T raising is an operation of ‘narrow syntax’ (thus yielding semantic effects) and that verb raising involves the interaction of interpretable and uninterpretable features via ‘Agree’ in the T-vP configuration. Their argument that v-to-T raising has semantic effects is motivated by Déchaine and Manfredi’s (2000) analysis of the interpretations underlying ‘null tense’ in four languages: English, Italian, and Fongbe and Igbo. ‘Null tense’ refers to the present simple form in English, which as noted above yields a habitual/generic interpretation. It also refers to the present simple form in Italian, which similar to the French présent licences habitual/generic and existential/event in progress interpretations. In Fongbe and Igbo, both of which lack overt inflectional marking, ‘null tense’ refers to bare verb forms. The bare forms yield two interpretations in Fongbe, i.e., past and present perfect, but only a past interpretation is licensed in Igbo. To account for these cross-linguistic differences Déchaine and Manfredi posit the involvement of two parameters. The first is whether interpretations associated with T are independent of the inherent aspect2 of the verb and its predicate, as in English and Italian, or whether T can be interpreted according to inherent aspectual properties of the VP as in Fongbe and Igbo (Hawkins et al.  illustrate this with Déchaine and Manfredi’s example of the accomplishment eat the bread giving rise to a perfective interpretation, where the eventive nature of the accomplishment ‘stands in for’ T). The argument is that in Fongbe and Igbo T only has an uninterpretable categorical feature [V], in contrast to English and Italian having an uninterpretable [V] and additionally an uninterpretable [AGR] feature. This feature is said to be responsible for blocking a meaning based on the inherent aspect of the VP, thus yielding a generic/habitual interpretation. The second parametric difference concerns the ‘strong’ feature of T that forces thematic v-to-T raising, which is a feature of Italian and Igbo, but not English and Fongbe. In Igbo verb raising has the effect of licensing only one interpretation, i.e., the past (in comparison to nonraising Fongbe, which has the additional present perfect interpretation available). The consequence of verb raising in Italian (and French), however, is the availability of the additional existential/event in progress interpretation, which contrasts with the restriction of only the generic/habitual reading in English. Table 1 below summarises the main points so far.

2. Inherent aspect is universal and non-grammaticalised. It is defined by the internal semantic properties of a verb or its predicate, which, following Vendler (1967), fall into one of four categories: states, activities, accomplishments and achievements.

 Sarah Ann Liszka

Table 1.  Summary of Déchaine and Manfredi’s (2000) analysis of ‘null tense’ Lang

‘Null tense’ interpretations

Parameter (1) AGR feature

Parameter (2) Strong T feature

English Italian

(a) Habitual/generic (a) Habitual/generic (b) Existential/event in progress (a) Past (b) Present perfect (a) Past

Yes Yes

No Yes

No

No

No

Yes

Fongbe Igbo

Hawkins et al. (2008) then suggest that the semantic component assigns particular interpretations to the structural configurations created by these uninterpretable features. Furthermore, they link Adger’s (2003) syntactic representation of the simple and progressive with the ideas for semantic interpretation proposed by Déchaine and Manfredi (2000) to demonstrate the semantic effects of raised and non-raised verbs. Briefly, the agreement relationship between T and v requires the interpretable tense features [present], [past] to value and delete an uninterpretable feature of v. The agreement operation is formalised as follows (Adger 2003: 169): (5)

Agree In a configuration X[F: val] … Y[uF: ] where … represents c-command, then F checks and values uF, resulting in: X[F: val] … Y[uF: val] In French, tense valued on thematic (and auxiliary) verbs is always strong, triggering Move (v-to-T raising), as “A strong feature must be local to the feature it checks/is checked by” (Adger 2003: 179). Tense valued on thematic verbs in English, on the other hand, is weak and therefore they remain in situ as there is no locality constraint. Turning to the progressive, Adger’s account posits an independent Progressive category, which has a strong uninterpretable feature, requiring be to raise in order to be locally valued by the interpretable T-features [present] or [past]. Furthermore, the Progressive aspectual head has the interpretable feature [Prog] which values the v[uInfl:] feature locally and is overtly realised by v-ing (which as mentioned above means something like ‘ongoing’). In their analysis of English, Hawkins et al. (2008) go on to suggest that behaving like the [AGR] feature of Décahine and Manfredi (2000), the ‘agreement’ operation in Adger’s system (i.e., between interpretable and uninterpretable features) prevents any chance of a temporal interpretation based on the inherent aspectual properties



Associating meaning to form 

of the verb and its predicate. Also, as English thematic verbs do not raise to T, only the habitual/generic interpretation is available. Aux be, however, does raise to T, giving rise to the existential/event in progress interpretation that maps onto the progressive form. They further point out that a habitual/generic interpretation is blocked “simply because interpretable [Prog] has valued the [uInfl:] feature of v as [uInfl: Prog]”. Returning to French, examples (3) and (4) above showed that thematic verbs in the présent yield habitual/generic and existential/event in progress interpretations, suggesting that v has a strong uninterpretable feature. To account for the two interpretations, following Déchaine and Manfredi’s line of argument, they suggest that the necessity for syntactic agreement gives rise to the habitual/generic interpretation, whilst local valuing gives rise to the existential/event in progress interpretation. Having outlined the grammatical properties of the English present simple, English present progressive and the French présent, we are now in a position to consider the research questions applicable to this study. First, as French has an instantiated [uInfl:], it is assumed that French L1 speakers of L2 English will establish the T and v featural properties of English. It is further assumed that learners need to establish that thematic v has a weak [uInfl:] (unlike French), but that a strong [uInfl:] is a property of the Progressive auxiliary be. To test these assumptions the following research questions ask whether advanced L1 French speakers of L2 English know that: a. A strong uninterpretable feature underlies the syntactically independent category Progressive. b. A weak uninterpretable feature underlies English thematic verbs. In light of any evidence suggesting optionality, these potential loci as sources of persistent difficulty in associating meaning-to-form for the English present simple and present progressive by advanced L2 speakers of L1 French will be discussed in relation to the RD hypothesis. 3. The study Sixteen non-native and thirteen native speakers of English participated in the study. The non-native informants were L1 French speakers, who were enrolled in advanced English language classes at a university institution at the time of data collection. Their levels had been determined by an in-house proficiency test, conforming to the common European framework for determining language levels, prior to enrolment. Table 2 below outlines biographical details of the native French speakers at the time of the experiment. The fourth column detailing classroom exposure includes the 60 hours of advanced-level instruction over 15 weeks that each informant had received at the university.

 Sarah Ann Liszka

Table 2.  Biographical details of non-native informants Learner NB CD GM CB EB SC VL BST FG VF VC SB VB LF BSO NA

Age

Age of first exposure

EFL classes (years)

Naturalistic exposure

51–55 41–45 61–65 41–45 16–20 41–45 36–40 36–40 36–40 36–40 41–45 41–45 36–40 21–25 56–60 26–30

11 13 15 11 10 12 10 13 10 11 11 11 13 12 11 11

2 (+ university) 5 1 (+ school) 7 7.5 9 7 6 16 14 9 9 5.5 13 9.5 16

1 year (UK) Short visit (UK) 4.5 years (UK/Nigeria) Short visits (UK) Short visit (UK) Short visits (UK) Short visits (UK/USA) 7 years (UK) Short visits (UK) 7 months (UK) 8 months (UK) Short visits (UK) 4 years (UK) Short visits (UK) Short visits (USA) Short visits (UK/USA)

The experiment comprised three tasks: two oral and one written. The oral tasks both used visual stimuli, a picture (task one) and a video clip (task two), in an attempt to elicit guided ‘spontaneous’ descriptions, producing a subset of verbs in common across the informants for comparison in addition to individual variation. It was essential to prime the informants to provide descriptions, rather than narratives, as states and ongoing events tend to predominate in descriptions compared to events and states in narratives (Smith 2000). This was particularly pertinent for the video clip as it is made up of a distinct sequence of events, which mainly provide foreground information (Dahl 1984) to a narrative and are generally marked with the simple past or present historic in English (Hopper 1979; Dry 1981, 1983). Therefore, clear instructions were required to ensure that native and non-native informants understood the temporal context (i.e., speech time and event time are in the present) and that descriptions of ongoing events in progress, rather than narratives were required. Furthermore, three modes of testing were used (task one allowing for off-line pre-planning, task two requiring on-line production, and the third allowing for off-line pre-planning in a written paradigm/preference task) in order to accommodate potential task-effects. As Ayoun (2005: 59–60) points out, albeit in terms of judgement tasks, procedural and/or stimulus task-related factors can contribute to variability in performance. Thus, the order of the tasks was motivated first by priming considerations, as task one is based on a typical (and there-



Associating meaning to form 

fore potentially familiar) EFL classroom activity for practising the present progressive in descriptions. Second, sandwiching the on-line task between the two off-line tasks allowed for the most difficult task in terms of processing/communicative pressure (cf. Prévost and White 2000) to be followed by the controlled written task, devoid of the demands of ‘spontaneous’ production. Turning to particular details, all sixteen L2 French speakers provided data for the two oral tasks and five of the thirteen native (monolingual) English speakers provided benchmark data. With regards to task one, informants were given a picture of a busy street scene from the viewpoint of looking out of a window. It depicted people engaged in typical daily activities, such as a woman pushing a pram, a baby crying, children fighting, a boy eating an ice-cream, people chatting and a stall-holder selling vegetables. On a one-to-one basis with the researcher, informants were verbally instructed to “imagine that you are looking out of your window and this view of the street is what you can see outside at the moment. Describe what is happening outside”. Having checked that the informants had understood the relevant operating instructions for the recording equipment, the researcher then left the room. Each informant completed the task, with the possibility of stopping and starting the recording equipment in their own time, at the beginning, end and during the task. The second task, the video clip description, took place straight after the picture description. Mirroring task one in the use of the present progressive and the word ‘describe’ in the instructions (with the intention of establishing the context of a description in the present), informants were told “You are going to watch a TV programme and I would like you to describe what is happening on the screen at the same time as you are watching”. It was also explained that the volume would be turned to zero (which did not detract from the plot as it contains minimal dialogue). Similar to task one, informants completed the task alone. However, as this was an on-line task and informants were obliged to report the events on the screen as they were happening, they were not allowed to operate the recording equipment during the nine minutes of footage. The clip was a comedy sketch showing Rowan Atkinson’s comedy character Mr Bean at an elegant, expensive French restaurant. To summarise, at the beginning Mr Bean is sitting alone at his table, writing his own birthday card. Then, after consulting the menu and finding out how expensive the food is, he orders the cheapest dish available. It is not until the food arrives that he discovers that he has ordered a raw meat dish, steak tartare, which at first perplexes and then disgusts him. Realising that he cannot bring himself to eat it, he spends the rest of the sketch trying to surreptitiously dispose of the meat in various ways, such as pushing it into a vase, dropping it into his neighbour’s handbag and hiding it under his plate. It finishes with a waiter being tripped over by Mr Bean,

 Sarah Ann Liszka

spilling food on his table. Mr Bean complains to the manager, who promptly moves him to another table and serves him with another dish of steak tartare. Data from the description tasks were transcribed and re-examined twice at later dates. Verbs were coded for verbal morphology and a token analysis of appropriate and inappropriate use of verb forms in appropriate contexts was carried out. The same procedure for coding and scoring was followed by the researcher on two further occasions in order to only include tokens which received the same classification three times3. For the picture, as a description of events in progress at the moment of speaking, present progressive forms were coded as appropriate for thematic verbs. The same premise was used for the video clip, unless a context was established, for example, with an adverbial (such as then, suddenly, as soon as) licensing the present simple form of the verb and any sequential verbs also bound by the adverbial. Example (6) from a native speaker and (7) from a non-native speaker (describing the same scene) both illustrate such contexts where the simple present is appropriate (verb and adverb in italics): L1 English speaker (6)

and the waiter’s come back (pres perfect) and he’s undoing his napkin for him (pres progressive) and placing it on his lap (pres progressive) and he’s imitating the action (pres progressive) the waiter’s just done with the napkin (pres perfect) and he’s doing it again (pres progressive) he’s flicked a napkin (pres perfect) and he’s placing it on his own lap (pres progressive) and then he accidentally flicks it across to somebody else’s table (pres simple) but pretends it wasn’t him (pres simple) and err instead he’s tucked the tablecloth into his shirt (pres perfect) the waiter’s brought him his food (pres perfect) French L2 English speaker (7)

the waiter is coming back (pres progressive) he is putting his napkin on his knee (pres progressive) Mr Bean imitates him (#pres simple) and the waiter again is doing it (pres progressive)

3. The coding and scoring process was done three times in an attempt to provide reliability and rigour. However, ideally, as an anonymous reviewer points out, such coding and scoring should be done by two different raters, if not three.



Associating meaning to form 

and then he throws err his napkin to the next table (pres simple) the waiter is coming back again (pres progressive) Two further examples describing the same scene are shown in (8) and (9) below to give a further flavour of the descriptions produced: L1 English speaker (8)

And the wine waiter has just returned (pres perfect) And he’s now putting the napkin out for Mr Bean (pres progressive) Mr Bean is happy about that (cop) And he’s rearranging it over his lap several times (pres progressive) He obviously likes the way (state) The wine waiter did that (simple past) Oh, and he’s just flung the napkin onto the, another table (pres perfect) Table with some other people who are eating on it (pres progressive) And err, they’re looking round (pres progressive) As if they don’t know (state) Where it’s come from (pres perfect) Mr Bean’s ignoring them (pres progressive) And Mr Bean’s food has now arrived (pres perfect)

French L2 English speaker (9)

Mr Bean is trying to play with his knife on the glasses (pres progressive) But the waiter is is coming again (pres progressive) And err the waiter I don’t know (pres simple) The waiter gives him the a napkin (#pres simple) And Mr Bean throw, throws it away to his neighbour (#pres simple) The waiter comes again with the dishes (#pres simple) And Mr Bean give, gives him the money for the dish (#pres simple)

For both tasks, correct or incorrect use of verb forms other than the present simple or present progressive were noted but omitted from the calculations (such as the present perfect, future ‘going to +inf ’). Other omissions included modal auxiliaries, reported speech, inaudible utterances, repetitions and self-corrected errors, where only the corrected version was counted. Contextual ambiguities were also excluded, such as in “the man who sells ice-cream is across the street”. Here the selling of ice-cream could be interpreted as either habitual or event in progress. Percentage scores for appropriate and inappropriate use of the present simple and present progressive forms were then calculated, first for individual informants and then conflated for the L1 and L2 groups.

 Sarah Ann Liszka

For the third task, fifteen of the sixteen L2 informants completed the task, which was piloted and confirmed by eight of the thirteen native controls. It took the form of a contextualised dialogue task, requiring informants, in their own time, to complete a set of short dialogues in a given context, by writing an appropriate verb form from a verb stem which was given in brackets. In total, the task comprised of sixty environments, which required a variety of verb forms, of which the present progressive is the most appropriate in seventeen environments and the present simple is most appropriate in nineteen environments (other environments included the use of the simple past, present perfect and present perfect progressive). The set of instructions (10), followed by a sample context and dialogue (11) to give a sample of the task are set out below: (10) Instructions: The following task has 4 sections. Each section is divided into 3 parts taking the form of dialogues, labelled ‘Dialogue A’; ‘Dialogue B’ and ‘Dialogue C’. Before each dialogue the context is given in bold. It is important to read the contextual information before giving your answer. You have to complete the dialogues with the verbs given in brackets. For example, She ________________ (walk) to work every day. She _____walks______ (walk) to work every day. The first dialogue is for you to practice. Thank you for taking part. (11) Sample context and dialogue from the task: (A mother and her teenage son have recently been arguing and they aren’t talking to each other). Later that evening Brian bursts into tears and starts to tell his mum about his problems. Mum: Why _____________________ (you / cry), darling? Brian: Every day, the older boys at school ______________ (threaten) to hit me if I ____________________ (not / give) them my lunch money. Mum: Right. I’m going to speak to your head teacher first thing tomorrow. This has to stop. Brian: And … Sally, the girl I _________________ (love) more than anything ____________ (organize) a party for next week & she hasn’t invited me. Mum: My poor son. I ________________ (try) to deal with my own problems at the moment and I never realised you had so many of your own. The number of uses of forms for the seventeen present progressive and nineteen present simple environments were calculated. Any forms other than these, such as



Associating meaning to form 

the use of the present perfect, were omitted from the scoring. Percentage scores were calculated for distributions of forms for individuals and the group. 4. Results Table 3 summarises the overall frequency scores of present simple and present progressive forms produced by the NS and NNS in the two oral tasks. Observe that the Representational Deficit account predicts that these advanced L2 English speakers should continue to show optionality in associating meanings to forms, resulting from an L1 influence. Results from the picture description task seem to disconfirm the hypothesis, showing that these learners are successful in assigning target-like meanings to present simple forms (100%) and present progressive forms (92.7%). It should be borne in mind, however, that this was a potentially familiar task, particularly as these were instructed learners. And if we consider the video clip task a marked difference emerges. Whilst maintaining a near native-like use of the present simple in present simple environments (94.9%), the L2 English group produces progressive forms in just over half of the progressive environments (54.8%). Such probabilistic marking could arguably be the result of the communicative pressure imposed by a real-time processing task and the simple form is produced as the default, which is suggestive of evidence in favour of the Table 3.  Raw and percentage scores for present simple and present progressive forms produced in the 2 oral tasks Task

Form

Picture

Present Simple habitual #event in progress Present Progressive event in progress #habitual Total contexts

Video

Present Simple habitual #event in progress Present Progressive event in progress #habitual Total contexts

L1 French (n = 16)

L1 English (n=5)

301/301 (100%) 0/301 (0%)

113/113 (100%) 0/113 (0%)

265/286 (92.7%) 21/286 (7.3%) 587

103/103 (100%) 0/103 (0%) 216

499/526 (94.9%) 27/526 (5.1%)

280/281 (99.6%) 1/281 (0.4%)

304/555 (54.8%) 251/555 (45.2%) 1081

347/350 (99.1%) 3/350 (0.9%) 631

 Sarah Ann Liszka

Missing Surface Inflection Hypothesis (Prévost and White 2000). However, if the problem is one of speed of access to forms in production better performance might be expected from the off-line written task, which followed the more demanding video clip task and was without a time limit. Evidence suggests that this is not the case: Table 4 shows the appropriate and inappropriate use of present simple and present progressive forms produced in the contextualised dialogue task. Similar to the two oral tasks, results indicate near-native-like production in present simple environments (94.2%). This is markedly different to the production of progressive forms in existential/event in progress contexts (59.3%) and to native responses (97.8%) in these contexts and similar to the non-native production rate of appropriate progressive forms in the video clip (54.8%). This raises an interesting question in relation to identifying where optionality in event in progress contexts arises, i.e., are these learners overgeneralising the present simple or undergeneralising the present progressive? This is a subtle difference, which might be explained in terms of cause and effect: (i) if undergeneralising of the present progressive is caused by a problem with the progressive, the over-use of the simple is an effect and (ii) overgeneralising the present simple suggests a problem with the present simple (cause) and the effect is an under-use of the present progressive. In this case, it could be argued that the former aligns with the FFR (Full Functional Representation) approach, i.e. performance underdetermines knowledge. For example, if there are problems of lexical access in producing a more syntactically complex utterance containing be+V-ing (cause), the use of the ‘default’ simple form is the effect. On the other hand, over-use of the simple in (ii), arising from an L1 influence which allows the simple to have habitual/generic and existential/ event in progress interpretations, aligns with the RD hypothesis. And to support the assumption that these learners are overusing the simple, where the learners do produce the present progressive, they use it appropriately, i.e., the present progressive is not used in habitual/generic contexts. Table 4.  Raw and percentage scores for present simple and present progressive forms produced in the contextualised dialogue task Form

L1 French (n= 15)

L1 English (n = 8)

habitual #event in progress

262/278 (94.2%) 16/278 (5.8%)

151/151 (100%) 0/151 (0%)

event in progress #habitual

131/221 (59.3%) 90/221 (40.7%)

133/136 (97.8%) 3/136 (2.2%)

Note: L1 French omissions (neither present simple nor present progressive used): 7 simple, 34 progressive environments. L1 English: 1 present simple.



Associating meaning to form 

To sum up so far, the results indicate that these advanced learners of L2 English have persistent difficulty in matching meaning to form for the present progressive in existential/event in progress contexts in two tasks, but consistently produce the present simple in habitual/generic contexts. Furthermore, the results from the potentially more demanding video clip task were not markedly different from the off-line written task, where better performance might be expected without the constraints of on-line processing. Taken together, these results hint at an L1 influence as predicted by the RD hypothesis and we will now consider how to account for these data from that perspective. 5. Discussion and conclusions Recall that the RD hypothesis claims that adult learners of a second language, even at high levels of proficiency, will fail to represent uninterpretable syntactic features that are not present in their L1s. Consequently, a deficit resulting from such an L1 influence will permanently affect the assignment of native-like meanings to surface forms. By the same token, any uninterpretable feature encoded in the L1 that is also a feature of the L2, is theoretically available for instantiation in the L2, leading to native-like syntactic and semantic representations, overtly realised in production. As observed in the last section, where these advanced learners use the present progressive, they do so appropriately. With regards to the first research question, then, they appear to have knowledge that a strong [uInfl:] is associated with the syntactically independent category Progressive which forces the raising of aux ‘be’ to be checked and deleted by T, yielding an existential/event in progress interpretation. This might be expected, given that all verbs (i.e., auxiliary, thematic and modal) are obliged to raise in French and the L1-L2 settings for feature strength coincide. However, the fluctuation seen in the results from the video clip task and the contextualised dialogue task, between the progressive and the simple in existential/event in progress contexts, shows that its use does not extend to all environments in a target-like way. This is assumed to be an over-use of the present simple, allowing the present simple to have an existential/event in progress reading in addition to a habitual/generic reading (as in French). In relation to the second research question, then, in light of the RD hypothesis, this might suggest that a strong v[uInfl:] is maintained (i.e., the weak setting for English v has not been established) and thematic verbs are raised, thus assigning an existential/event in progress interpretation to present simple forms. Interestingly, the results from Hawkins et al’s (2008) acceptability judgement task, designed to test L2 knowledge of the semantic contrasts between raised and non-raised verbs in English, indicate that their L1 verb-raising informants also

 Sarah Ann Liszka

diverge from native speakers in distinguishing meaning contrasts. But contrary to the findings and conclusions in this current study, Hawkins et al.’s informants appear to be able to distinguish between contexts where a simple form is required with thematic verbs and where it is not, which might suggest that the L1 strong feature has been dropped. On the other hand, the results show that their informants are allowing be + ing to have an existential/event in progress and a habitual/ generic interpretation (again in contrast to the results in the present study), which Hawkins et al. (2008) speculate arises from a failure to recognise Progressive as a syntactically independent category. Furthermore, the results in this study suggest compatibility with Schwartz and Sprouse’s (1996) FT/FA hypothesis. Recall that this hypothesis proposes that the complete range of syntactic properties of the L1 is transferred into the L2 at the initial stage of acquisition (i.e., full transfer). Thus, in the case of these L1 French learners of English, the results could simply reflect that the transferral of the strong [uInfl:] is still in operation in the IL for thematic verbs, accounting for both the habitual/generic and existential/event in progress interpretations assigned to the present simple. However, it could be argued that with full access to UG we might expect these learners to have restructured their grammars, giving rise to a switch from strong to weak v[uInfl:]. This expectation is based on the assumptions that the two forms are neither obscure nor non-existent in the TL input and that as advanced learners they will have been exposed to ample relevant input. Also if L1 hindrance from the initial transfer were a factor, we might expect it to be residual at this stage of development shown by a lower frequency of inappropriate use of the present simple in existential/event in progress contexts (rather than over 40% in tasks 2 and 3 in this study). Finally although it has not been the focus of this study, the acquisition of the grammaticalised aspectual meaning of ongoing, which also underlies the progressive and is mapped on to the v–ing morpheme, is a further avenue of research. Nevertheless, from the instances of appropriate use of the progressive form, it could be argued that these L1 French learners appear to have acquired the linguistic properties which yield the meaning of ‘ongoingness’. However, heeding Hawkins and Hattori’s (2006: 273) warning that “caution is required in treating apparent cases of target-like performance by L2 speakers as evidence for underlying targetlike representations”, it is necessary to look again at the L1. French encodes the imperfect [-perfective], which is said to be conceptually similar to the English progressive, (with the difference that it inflects for prototypical states, as well as dynamic verbs). But as White (2003:  255) points out, English lacks the featural distinction [+/-perfective] and (albeit in relation to Spanish) the progressive is not the same as imperfective. Furthermore, the imperfect encodes both habitual and continuous readings in the past, whereas English has two forms: habitual aspect



Associating meaning to form 

realised by ‘used to +inf ’ and past progressive ‘was/were +ing’ (Comrie 1976). Taking these factors together, we might speculate that L1 French learners utilise the (partial) conceptual similarity between the imperfect and progressive, i.e., durativity, to assign meaning to the progressive (which is nevertheless UG-constrained), in contrast to encoding a native-like representation of progressive aspect. What is obvious, however, is that more research is required to try to unravel the complexities involved in acquiring these temporal distinctions. A broader range of tasks would be of great benefit, as would more testing of L2 speakers living in L2-target communities, especially for studies concerned with high levels of proficiency and issues of ultimate attainment. References Adger, D. 2003. Core syntax: A minimalist approach. Oxford: OUP. Al-Hamad, M., Al-Malki, E., Casillas, G., Francheschina, F., Hawkins, R., Hawthorne, J.,K Karadzovska, D., Kato, K., Liszka, S., Lozano, C., Ojima, S., Okuwaki, N. & Thomas, E. 2002. Interpretation of English tense morphophonology by advanced L2 speakers. Eurosla Yearbook 2, S.H. Foster-Cohen, M. Garcia-Mayo & J. Cenoz (eds.), 49–69. Amsterdam: John Benjamins. Ayoun, D. 2005. Verb movement in the L2 acquisition of English by adult native speakers of French. In EUROSLA Yearbook 5, S. H. Foster-Cohen, M. del Pilar Garcia-Mayo & J. Cenoz (eds.), 35–76. Amsterdam: John Benjamins. Bardovi-Harlig, K. 1992. The relationship of form and meaning: A cross-sectional study of tense and aspect in the interlanguage of learners of English as a second language. Applied Psycholinguistics 13: 253–278. Comrie, B. 1976. Aspect: An introduction to the study of verbal aspect and related problems. Cambridge: CUP. Dahl, Ö. 1984. Temporal distance: Remoteness distinctions in tense-aspect systems. In Explanations for language universals, B. Butterworth, B. Comrie & Ö. Dahl (eds), 105–122. The Hague: Mouton. Déchaine, R. & Manfredi, V. 2000. Interpreting null tense. Paper presented at the Round Table on the syntax of tense and aspect, November 2000, University of Paris. Dry, H. 1981. Sentence aspect and the movement of narrative time. Text 1: 233–40. Dry, H. 1983. The movement of narrative time. Journal of Literary Semantics 12: 19–53. Hawkins, R. 2005. Revisiting Wh-movement: The availability of an uninterpretable [wh] feature in interlanguage grammars. In Proceedings of the 7th Generative Approaches to Second Language Acquisition Conference (GASLA 2004), L. Dekdyspotter, R. Sprouse & A. Liljestrand (eds.), 124–137. Somerville MA: Cascadilla. Hawkins, R. & Chan, C. 1997. The partial availability of Universal Grammar in second language acquisition: The ‘failed functional features’ hypothesis. Second Language Research 13(3): 187–226.

 Sarah Ann Liszka Hawkins, R. & Franceschina, F. 2004. Explaining the acquisition and nonacquisition of determine-noun gender concord in French and Spanish. In The acquisition of French in different contexts, P. Prévost & J. Paradis (eds), 175–206. Amsterdam: John Benjamins. Hawkins, R. & Hattori, H. 2006. Interpretation of English multiple wh-questions by Japanese speakers: A missing uninterpretable feature account. Second Language Research 22(3): 269–301. Hawkins, R., Casillas, G., Hattori, H., Hawthorne, J., Husted, R., Lozano, C., Okamoto, A., Thomas, E. & Yamada, K. 2008. The semantic effects of verb raising and its consequences in second language grammars. In The role of features in second language acquisition, J. Liceras H. Zobl & H. Goodluck (eds), 328–351. New York NY: LEA. Hopper, P. 1979. Aspect and foregrounding in discourse. In Syntax and semantics: Discourse and syntax, T. Givón (ed.), 213–41. New York NY: Academic Press. Lardiere, D. 1998a. Case and tense in the `fossilized’ steady state. Second Language Research 14(1): 1–26. Lardiere, D. 1998b. Dissociating syntax from morphology in a divergent L2 end-state grammar. Second Language Research 14(4): 359–375. Lardiere, D. 2000. Mapping features to forms in SLA. In Second language acquisition and linguistic theory, J. Archibald (ed.), 102–29. Oxford: Blackwell. Montrul, S. & Slabakova, R. 2002. Acquiring morphosyntactic and semantic properties of aspectual tenses in L2 Spanish. In The acquisition of Spanish morphosyntax: The L1-L2 connection, A.-T. Perez-Leroux & J. Liceras (eds), 113–149. Dordrecht: Kluwer. Prévost, P. & White, L. 2000. Missing surface inflection or impairment in second language acquisition? Evidence from tense and agreement. Second Language Research 16(2): 103–133. Schwartz, B.D. 1999. The second language instinct. In Language acquisition: Knowledge, representation and processing, A. Sorace, C. Heycock & R. Shillcock (eds.). Dordrecht: Kluwer. Schwartz, B. D. & Sprouse, R. A.1996. L2 cognitive states and the ‘full transfer/full access’ model. Second Language Research 8: 1–38. Slabakova, R. 2003. Semantic evidence for functional categories in interlanguage grammars. Second Language Research 19(1): 42–75. Smith, C. S. 2000. The domain of tense. Colloquium on Tense and Aspect. Paris Smith, N.V. & Tsimpli, I.-M. 1995. The mind of a savant: Language learning and modularity. Oxford: Blackwell. Tsimpli, I.-M. 2003. Features in language development. Paper presented at the Eurosla conference, University of Edinburgh. Tsimpli, I.-M. & Dimitrakopoulou, M. 2007. The interpretability hypothesis: Evidence from whinterrogatives in L2 acquisition. Second Language Research 23(2): 215–242. Vendler, Z. 1967. Verbs and times. In Linguistics in philosophy, Z. Vendler (ed), 97–121. Ithaca NY: Cornell University Press. White, L. 2003. Second language acquisition and Universal Grammar. Cambridge: CUP.

Index of names A Abney, S.  3, 33 Adger, D.  54, 234 Agathopoulou, E.  208 Agouraki, Y.  174–175 Aissen, J.  172 Anagnostopoulou, E.  170, 172–173, 181 Anderssen., M.  15 B Babko-Malaya, O.  189 Bard. E. G.  178 Beck, M.-L.  2, 127, 159 Belletti, A.  187 Berman, R. A.  206 Bickerton, D.  28 Bohnacker, U.  15 Borer, H.  190 Bruhn de Garavito, J.  2 C Cardinaletti, A.  169, 175 Carstens, V.  168 Chan C.Y.-h.  2, 34, 53, 57, 72, 80, 96, 127, 159, 188, 230 Chierchia, G.  34, 36, 47 Chila-Markopoulou, D.  190 Chomsky, N.  35, 65, 108, 167 Cinque, G.  59–60, 173, 174, 181 Comrie, B.  172, 189 D Déchaine, R.  234–235 Dimitrakopoulou, M.  72, 81–83, 96, 188 Dowty, D.  200 Duffield, N.  2 E Emonds, J.  58, 61 Erguvanli, E. E.  3 Escobar, L.  173, 181 Eubank, L.  2, 58, 61

F Filip, H.  192 Fodor, J.  28 Franceschina, F.  2, 167, 168, 176, 183, 188 Franco, J.  170, 175 Fry, D.  15 G Giannakidou, A.  220 Glass, W. R.  128–129, 133, 136 Gleason, J. L.  28 Goad, H.  1, 3, 5, 8, 10, 23, 27, 35–37, 47, 187 H Harley, H.  128, 133 Hattori, H.  54, 72, 81–83, 96, 101, 230–231, 233, 244 Hawkins, J.  29 Hawkins, R.  2, 27–28, 34, 46–47, 53–54, 56–57, 72, 80–83, 96, 101, 108, 111, 122–123, 127, 159, 167–168, 176, 183, 188, 230–231, 233–234, 243–244 Heim, I.  29 Hendriks, H.  206 Hickmann, M.  206 Horrocks, G.  192 Hovdhaugen, E.  30 Huang, C.-T.  109, 112–113 Huang, S.  36–37 Huebner, T.  1, 28 Hulst, H. van der  10 Hyams, N.  168

Ko, H.  1, 27–30, 32–33, 46 Konrot, A.K.  15 Kornfilt, J.  3, 5, 9, 10 Kupisch, T.  15, 34 L Lardiere, D.  2, 27, 47 Larson, R.  172 Lehiste, I.  15 Leonetti, M.  173 Leung, Y.-k. I.  1, 2, 27, 33, 46 Levi, S.V.  15 Liceras, J.  133, 135–136 Liszka, S.  2 Liu, D.  28 Lozano, C.  54, 128, 133, 135–136 Lyons, C.  3, 7, 29–31 M Maclaran, R.  31 MacLaughlin, D.  110, 112, 117 Manfredi, V.  234–235 Mastropavlou, M.  2, 188 Matsunaga, K.  204 Montrul, S.  2, 128, 133, 136–137, 157–158, 204, 230 Mosel, U.  30 Mozer, A.  189 Müller, N.  168 Murphy, S.  28 N Navarro, S.  204 Newell, H.  10 Nicoladis, E.  204

I Iatridou, S.  196 Inagaki, S.  204 Inkelas, S.  4 Ionin, T.  1, 27–33, 46

O Oh, E.  192 Orgun, C.O.  4 Özçalişkan, Ş  206 Öztürk, B.  3, 5, 6, 7

K Kabak, B.  4, 10 Karmiloff-Smith, A.  168 Kayne, R.  169

P Papadopoulou, D.  190 Parodi, T.  169, 170, 175, 176, 182 Parrish, B.  28

 Representational Deficits in SLA Peperkamp, S.  4 Pérez-Leroux, A. T.  128–129, 133, 136 Pérez-Pereira, M.  168 Pollock, J.-Y.  58, 61 Poser, W. J.  5 Prévost, P.  2, 57, 127, 160, 168, 176, 183, 230, 242 Prince, E.  31 R Radford, A.  34 Reid, J.  33 Ritter, E.  128, 133 Rizzi, L.  129, 174, 222 Robertson, D.  1, 14, 28, 47, 178 Rodríguez-Louro, C.  128, 133, 136, 157 Romanova, E.  198 Roussou, A.  2, 27 S Sag, I.  28 Sanz, M.  192 Schwartz, B. D.  2, 23, 53–54, 58, 60, 187, 230, 244 Selkirk, E. O.  4, 24, 35 Sezer, E.  11 Silverstein, M.  172 Sioupi, A.  190 Slabakova, R.  229–230

Slobin, D.  206 Smith, N.  167, 176, 183, 188–189, 230 Snape, N.  1–2 5, 27–28, 33–34, 46–47 Sorace, A.  127, 137, 160, 178, 188 Sportiche, D.  169, 170, 175 Sprouse, R.  2, 23, 53–54, 58, 60, 187, 230, 244 Starke, M.  169, 175 Stavrou, M.  192 Steele, J.  3, 8, 23 Strozer, J.  169 Suñer, M.  169, 175 T Talmy. L.  193 Tenny, C.  190 Terzi, A.  196 Theophanopoulou-Kontou, D.  196 Thomas, M.  1, 28 Torrego, E.  170, 175 Trademan, J. E.  28 Travis, L.  190 Trenkic, D.  2, 3, 7, 12–14, 20–23, 27, 47 Tsimpli, I.-M.  2, 27, 72, 81–83, 96, 167, 169, 170, 175, 176, 182, 183, 188, 230

U Underhill, R.  3 Uriagereka, J.  169, 170 V Vainikka, A.  53–56, 62–64 Valenzuela, E.  173 Verkuyl, H. J.  210 Vogel, I.  4, 10 W Wakabayashi, S.  28, 112–113 Weijer, R. van de.  10 Wexler, K.  1, 27–30, 32–33, 46 White, L.  1, 3, 5, 8, 10, 23, 27, 35–37, 47, 57–58, 112–114, 119–123, 127, 133, 160, 168, 176, 183, 187, 230, 242 Y Young-Scholten, M.  53–56, 62–64 Yuan, B.  112–113, 118–119 Z Zagona, K.  192 Zec, D.  7, 13, 23–24 Zubizarreta, M. L.  33, 46, 130, 192

Index of subjects A adjectives  1–2, 5–9, 12–20, 22–23 adjunct  192 adjunction  59 adverb placement  53 affixal clitic  4, 8, 10, 12, 18, 23 agree  54, 79 agreement  167–170, 173, 175–177, 182–183 Aktionsart  221 asymmetry  12–13, 19, 96–97, 100, 118–119, 157, 176, 179 articles  1–9, 11–23, 27, 38, 47–48 article choice parameter  32 B bidirectionality  157 C Cantonese  2 case  170, 172–175, 177–183 Categorial Uniformity  222 CEDEL2 corpus  138–139 Chinese  2, 27, 33–34, 36–38, 42–48 clitic doubling  167–185, 170–173, 178–180 clitic left dislocation  167–185, 173–175, 180–181 complement  192 co-occurrence requirement between the wh-word and dou  80 corpus data  127, 138, 160 critical period  34, 54, 65, 72, 81, 83, 96, 159, 188 D definiteness  28–29, 34, 41, 170, 172–173, 176–177, 179–183 definiteness in English  29 directional  194 discourse  128–132, 152–156 discursive features  128–132 distributive characteristic of dou  74

distributing feature  70, 72, 76–77 distributional projection  76, 80

GOAL  192, 216 Greek  187, 189–190, 192–197, 203–204, 209, 220, 223

E Economy of Projection  62 English  1–20, 22–23, 27–30, 32, 34, 36–38, 42, 45–46, 48, 194, 235 English articles  23

I Interpretability Hypothesis  72–73, 96, 187–189, 207, 222–223 interpretable features  28, 33–35, 45–48, 54, 72, 167, 176, 187–188, 230, 220, 222–223, 233–234

F Failed Functional Features Hypothesis  53, 80, 96, 187–188, 223 [1st person] feature  145–146 [2nd person] feature  146 [3rd person] feature  127, 134, 146–151 features  229 feature checking  70, 76, 78, 80, 97, 99–101, 111, 166, 168, 230 feature specification  169 final state  82–83, 100 fluctuate  32 Fluctuation Hypothesis  27, 32–33, 41–42, 45, 48 focus  127–129 [focus] feature  127–129 free clitics  4, 7–12, 23, 35, 37 free ride  98–99, 101 French  2 French-English interlanguage  59 Full House Principle  64 Full Functional Representation  229–230, 242 Full Transfer Full Access  23, 48, 53, 230, 244 Full Transfer Partial Access  23 functional categories  229 Fundamental Difference Hypothesis  53 G German  15, 19, 197

J Japanese  5, 33, 204 K kinds  34, 36 L L1 transfer  45, 100, 187–188, 207 L2 German  55 lexical categories  55 lexical projections  54 local adjustment  81, 101 long-distance binding  106, 110, 112, 117–118, 120–121 locative  194 M Mandarin  8 manner-of-motion verbs  187, 200–201, 205, 215, 219, 221–222 Master Tree  62 Minimal Trees/ Structure Building  53 Mirror Principle  62 Missing Surface Inflection Hypothesis  230 Modulated Structure Building  56 Move-to-INFL approach  109, 120 movement-forcing interpretable feature  70

 Representational Deficits in SLA N negation  53 no permanent loss of capacity to acquire  101 nominal mapping parameter  34 nominal phrases  36 non-permanent representational deficit  97, 101 noteworthy property  30 null objects  105–108, 111–118, 121–123 null pronouns  127–131 null-subject parameter  128 O optionality  167–169, 175–176, 182–183, 229 Optional Infinitive Stage  56 Organic Syntax  63 overgeneralising  242 overt morphology  169–170, 182–183 overproduction  153–155, 157 overt pronouns  127–131 P parametric differences  230 parsing  120–121 partial UG availability  53 PATH  192, 217 pefectivity  187 performance  230 phi features  127, 130, 133, 158 phonological phrase  4, 6, 8–11 pragmatics  128–132, 152–156 present simple  229 present progressive  229 Principle A of the Binding Theory  108 Principle of Obligatory Occupant Licensing  64 Processability  63 processing deficit  136, 159 pro-drop parameter  128

pronominal features  128, 133–135 Pronominal Feature Geometry  132–134, 137 pronominal subjects  128, 152–156 pronoun  169–170, 172, 175–177, 182–183 Prosodic Transfer Hypothesis  1, 3, 8–9, 11–14, 22–23, 27, 35, 44, 48 prosodic word  4, 7–12, 22 Q questions  53 R referring expression  133–134 reflexive binding  107–108, 112–118 reflexive parameter  110 representational deficit  136, 159 Representational Deficit Hypothesis  2, 23, 27, 33, 37, 44–45, 48, 54, 72, 108, 123, 230, 235, 243 Root Infinitives  56 Russian  198 S selective deficits  127–128, 134, 137–138, 151, 158, 160 Serbian  1, 7, 12–14, 24 Serbio-Croatian  23 Spanish  2, 33, 127–133, 205 Spanish  L2 127, 138–140 specifier positions  59 specificity  28–29, 41 stress  1–3, 5, 7–9, 11–19, 22–23 Strong Continuity  60 syntax-before-discourse phenomenon  128, 136 syntax-discourse interface  127–129, 135, 137, 142, 145, 157, 187–188, 193, 220, 223

syntax-discourse interface deficits  137, 142, 145, 151, 155 syntactic licensing of count nouns  34, 38 syntax-semantics interface  187, 191–192, 197, 220, 222–223 T (a)telicity  187, 223 target-like behaviour  97–99, 101 telicity  187 tense parameter  110 topic  127–129 topicality  173, 176 topic-continuity  127–132, 142–151, 152–156 [topic-continuity] feature  127–132, 142–151, 152–156 topic-drop parameter  112, 122–123 [topic] feature  127–129 topic-shift  127–132, 142–151, 152–156 [topic-shift] feature  127–132, 142–151, 152–156 topicalization  122 Turkish  1–12, 14–15, 19–20, 22, 205 U UAM Corpus-Tool  140 underproduction  153–155 undergeneralising  242 unidirectionality  157 uninterpretable features  28, 33–35, 45–48, 53, 70, 72, 81, 167–168, 176, 187–188, 223, 230–231, 233–235 universal quantifier  69–72, 76–80, 82–83, 88, 93–94, 96–97, 100 universal quantificational force  70

In the series Language Acquisition and Language Disorders the following titles have been published thus far or are scheduled for publication: A complete list of titles in this series can be found on the publishers’ website, www.benjamins.com 48 Santos, Ana Lúcia: Minimal Answers. Ellipsis, syntax and discourse in the acquisition of European Portuguese. xv, 293 pp. + index. Expected February 2009 47 Snape, Neal, Yan-kit Ingrid Leung and Michael Sharwood Smith (eds.): Representational Deficits in SLA. Studies in honor of Roger Hawkins. 2009. xxv, 250 pp. 46 Haznedar, Belma and Elena Gavruseva (eds.): Current Trends in Child Second Language Acquisition. A generative perspective. 2008. vi, 363 pp. 45 Guijarro-Fuentes, Pedro, María Pilar Larrañaga and John Clibbens (eds.): First Language Acquisition of Morphology and Syntax. Perspectives across languages and learners. 2008. vi, 302 pp. 44 Sekerina, Irina A., Eva M. Fernández and Harald Clahsen (eds.): Developmental Psycholinguistics. On-line methods in children’s language processing. 2008. xviii, 190 pp. 43 Savickienė, Ineta and Wolfgang U. Dressler (eds.): The Acquisition of Diminutives. A crosslinguistic perspective. 2007. vi, 352 pp. 42 Lefebvre, Claire, Lydia White and Christine Jourdan (eds.): L2 Acquisition and Creole Genesis. Dialogues. 2006. viii, 433 pp. 41 Torrens, Vincent and Linda Escobar (eds.): The Acquisition of Syntax in Romance Languages. 2006. viii, 422 pp. 40 Deen, Kamil Ud: The Acquisition of Swahili. 2005. xiv, 241 pp. 39 Unsworth, Sharon, Teresa Parodi, Antonella Sorace and Martha Young-Scholten (eds.): Paths of Development in L1 and L2 acquisition. In honor of Bonnie D. Schwartz. 2006. viii, 222 pp. 38 Franceschina, Florencia: Fossilized Second Language Grammars. The acquisition of grammatical gender. 2005. xxiv, 288 pp. 37 Montrul, Silvina A.: The Acquisition of Spanish. Morphosyntactic development in monolingual and bilingual L1 acquisition and adult L2 acquisition. 2004. xvi, 413 pp. 36 Bartke, Susanne and Julia Siegmüller (eds.): Williams Syndrome across Languages. 2004. xvi, 385 pp. 35 Sánchez, Liliana: Quechua-Spanish Bilingualism. Interference and convergence in functional categories. 2003. x, 189 pp. 34 Ota, Mitsuhiko: The Development of Prosodic Structure in Early Words. Continuity, divergence and change. 2003. xii, 224 pp. 33 Josefsson, Gunlög, Christer Platzack and Gisela Håkansson (eds.): The Acquisition of Swedish Grammar. 2004. vi, 315 pp. 32 Prévost, Philippe and Johanne Paradis (eds.): The Acquisition of French in Different Contexts. Focus on functional categories. 2004. viii, 384 pp. 31 Marinis, Theodoros: The Acquisition of the DP in Modern Greek. 2003. xiv, 261 pp. 30 Hout, Roeland van, Aafke Hulk, Folkert Kuiken and Richard J. Towell (eds.): The Lexicon– Syntax Interface in Second Language Acquisition. 2003. viii, 234 pp. 29 Fernández, Eva M.: Bilingual Sentence Processing. Relative clause attachment in English and Spanish. 2003. xx, 294 pp. 28 Shimron, Joseph (ed.): Language Processing and Acquisition in Languages of Semitic, Root-Based, Morphology. 2003. vi, 394 pp. 27 Salaberry, M. Rafael and Yasuhiro Shirai (eds.): The L2 Acquisition of Tense–Aspect Morphology. 2002. x, 489 pp. 26 Slabakova, Roumyana: Telicity in the Second Language. 2001. xii, 236 pp. 25 Carroll, Susanne E.: Input and Evidence. The raw material of second language acquisition. 2001. xviii, 461 pp. 24 Weissenborn, Jürgen and Barbara Höhle (eds.): Approaches to Bootstrapping. Phonological, lexical, syntactic and neurophysiological aspects of early language acquisition. Volume 2. 2001. viii, 337 pp. 23 Weissenborn, Jürgen and Barbara Höhle (eds.): Approaches to Bootstrapping. Phonological, lexical, syntactic and neurophysiological aspects of early language acquisition. Volume 1. 2001. xviii, 299 pp.

22 Schaeffer, Jeannette C.: The Acquisition of Direct Object Scrambling and Clitic Placement. Syntax and pragmatics. 2000. xii, 187 pp. 21 Herschensohn, Julia: The Second Time Around – Minimalism and L2 Acquisition. 2000. xiv, 287 pp. 20 Kanno, Kazue (ed.): The Acquisition of Japanese as a Second Language. 1999. xii, 180 pp. 19 Beck, Maria-Luise (ed.): Morphology and its Interfaces in Second Language Knowledge. 1998. x, 387 pp. 18 Klein, Elaine C. and Gita Martohardjono (eds.): The Development of Second Language Grammars. A generative approach. 1999. vi, 412 pp. 17 Archibald, John: Second Language Phonology. 1998. xii, 313 pp. 16 Hannahs, S.J. and Martha Young-Scholten (eds.): Focus on Phonological Acquisition. 1997. v, 289 pp. 15 Brinkmann, Ursula: The Locative Alternation in German. Its structure and acquisition. 1997. x, 289 pp. 14 Clahsen, Harald (ed.): Generative Perspectives on Language Acquisition. Empirical findings, theoretical considerations and crosslinguistic comparisons. 1996. xxviii, 499 pp. 13 Allen, Shanley: Aspects of Argument Structure Acquisition in Inuktitut. 1996. xvi, 244 pp. 12 Juffs, Alan: Learnability and the Lexicon. Theories and second language acquisition research. 1996. xi, 277 pp. 11 Yip, Virginia: Interlanguage and Learnability. From Chinese to English. 1995. xvi, 247 pp. 10 Lakshmanan, Usha: Universal Grammar in Child Second Language Acquisition. Null subjects and morphological uniformity. 1994. x, 162 pp. 9 Adone, Dany: The Acquisition of Mauritian Creole. 1994. xii, 167 pp. 8 Hoekstra, Teun and Bonnie D. Schwartz (eds.): Language Acquisition Studies in Generative Grammar. 1994. xii, 401 pp. 7 Meisel, Jürgen M. (ed.): Bilingual First Language Acquisition. French and German grammatical development. 1994. vi, 282 pp. 6 Thomas, Margaret: Knowledge of Reflexives in a Second Language. 1993. x, 234 pp. 5 Gass, Susan M. and Larry Selinker (eds.): Language Transfer in Language Learning. Revised edition. 1992. x, 236 pp. 4 Eckman, Fred R. (ed.): Confluence. Linguistics, L2 acquisition and speech pathology. 1993. xvi, 260 pp. 3 Eubank, Lynn (ed.): Point Counterpoint. Universal Grammar in the second language. 1991. x, 439 pp. 2 Huebner, Thom and Charles A. Ferguson (eds.): Cross Currents in Second Language Acquisition and Linguistic Theory. 1991. viii, 435 pp. 1 White, Lydia: Universal Grammar and Second Language Acquisition. 1989. xii, 198 pp.