•
UFERSA
Marcos Silva 08/10/2024
Prévia do material em texto
<p>http://www.cambridge.org/9780521792059</p><p>This page intentionally left blank</p><p>Second Language Acquisition and Universal Grammar</p><p>This authoritative textbook provides an overview and analysis of current second</p><p>language acquisition research conducted within the generative linguistic frame-</p><p>work. LydiaWhite argues that second language acquisition is constrained by prin-</p><p>ciples and parameters of Universal Grammar. The book focuses on characterizing</p><p>and explaining the underlying linguistic competence of second language learners</p><p>in terms of these contraints. Theories as to the role of Universal Grammar and the</p><p>extent of mother-tongue influence are presented and discussed, with particular</p><p>consideration given to the nature of the interlanguage grammar at different points</p><p>in development, from the initial state to the ultimate attainment. Throughout the</p><p>book, hypotheses maintaining that second language grammars are constrained</p><p>by universal principles are contrasted with claims that Universal Grammar is not</p><p>implicated; relevant empirical research is presented from both sides of the debate.</p><p>This textbook is essential reading for those studying second language acquisition</p><p>from a linguistic perspective.</p><p>lydia white is Professor of Linguistics at McGill University, Montréal, and</p><p>Chair of the Linguistics Department. She is internationally known as a leading</p><p>expert on second language acquisition. She is the author of Universal Grammar</p><p>and Second Language Acquisition (John Benjamins, 1989) and publishes regu-</p><p>larly in major international journals on language acquisition.</p><p>CAMBRIDGE TEXTBOOKS IN L INGUIST ICS</p><p>General editors: p. austin, j. bresnan, b. comrie,</p><p>w. dressler, c. j. ewen, r. huddleston, r. lass ,</p><p>d. lightfoot, i . roberts, s . romaine, n. v. smith,</p><p>n. vincent</p><p>Second Language Acquisition</p><p>and Universal Grammar</p><p>In this series</p><p>p. h. matthews Morphology Second edition</p><p>b. comrie Aspect</p><p>r. m. kempson Semantic Theory</p><p>t. bynon Historical Linguistics</p><p>j. allwood, l.-g. anderson and ö . dahl Logic in Linguistics</p><p>d. b. fry The Physics of Speech</p><p>r. a. hudson Sociolinguistics Second edition</p><p>a. j. elliott Child Language</p><p>p. h. matthews Syntax</p><p>a. radford Transformational Syntax</p><p>l. bauer English Word-Formation</p><p>s . c . levinson Pragmatics</p><p>g. brown and g. yule Discourse Analysis</p><p>r. huddleston Introduction to the Grammar of English</p><p>r. lass Phonology</p><p>b. comrie Tense</p><p>w. klein Second Language Acquisition</p><p>a. j. woods, p. fletcher and a. hughes Statistics in Language Studies</p><p>d. a. cruse Lexical Semantics</p><p>a. radford Transformational Grammar</p><p>m. garman Psycholinguistics</p><p>w. croft Typology and Universals</p><p>g. g. corbett Gender</p><p>h. j. giegerich English Phonology</p><p>r. cann Formal Semantics</p><p>p. j. hopper and e. c. traugott Grammaticalization</p><p>j. laver Principles of Phonetics</p><p>f. r. palmer Grammatical Roles and Relations</p><p>m. a. jones Foundations of French Syntax</p><p>a. radford Syntactic Theory and the Structure of English: A Minimalist</p><p>Approach</p><p>r. d. van valin, jr, and r. j. lapolla Syntax: Structure, Meaning</p><p>and Function</p><p>a. duranti Linguistic Anthropology</p><p>a. cruttenden Intonation Second edition</p><p>j. k. chambers and p. trudgill Dialectology Second edition</p><p>c. lyons Definiteness</p><p>r. kager Optimality Theory</p><p>j. a. holm An Introduction to Pidgins and Creoles</p><p>c. g. corbett Number</p><p>c. j. ewen and h. van der hulst The Phonological Structure of Words</p><p>f. r. palmer Mood and Modality Second edition</p><p>b. j. blake Case Second edition</p><p>e. gussman Phonology: Analysis and Theory</p><p>m. yip Tone</p><p>w. croft Typology and Universals</p><p>f. coulmas Writing Systems: An Introduction to Their Linguistic Analysis</p><p>l. white Second Language Acquisition and Universal Grammar</p><p>Second Language</p><p>Acquisition and</p><p>Universal Grammar</p><p>LYDIA WHITE</p><p>McGill University, Montréal</p><p> </p><p>Cambridge, New York, Melbourne, Madrid, Cape Town, Singapore, São Paulo</p><p>Cambridge University Press</p><p>The Edinburgh Building, Cambridge , United Kingdom</p><p>First published in print format</p><p>- ----</p><p>- ----</p><p>- ----</p><p>© Lydia White 2003</p><p>2003</p><p>Information on this title: www.cambridge.org/9780521792059</p><p>This book is in copyright. Subject to statutory exception and to the provision of</p><p>relevant collective licensing agreements, no reproduction of any part may take place</p><p>without the written permission of Cambridge University Press.</p><p>- ---</p><p>- ---</p><p>- ---</p><p>Cambridge University Press has no responsibility for the persistence or accuracy of</p><p>s for external or third-party internet websites referred to in this book, and does not</p><p>guarantee that any content on such websites is, or will remain, accurate or appropriate.</p><p>Published in the United States of America by Cambridge University Press, New York</p><p>www.cambridge.org</p><p>hardback</p><p>paperback</p><p>paperback</p><p>eBook (NetLibrary)</p><p>eBook (NetLibrary)</p><p>hardback</p><p>http://www.cambridge.org</p><p>http://www.cambridge.org/9780521792059</p><p>Contents</p><p>Preface page xi</p><p>Abbreviations xiv</p><p>1. Universal Grammar and language acquisition 1</p><p>1.1 Introduction 1</p><p>1.2 Universal Grammar in L1 acquisition 2</p><p>1.3 Why UG? The logical problem of language acquisition 3</p><p>1.3.1 An example: the Overt Pronoun Constraint 4</p><p>1.4 Parameters of Universal Grammar 9</p><p>1.4.1 An example: feature strength and movement 10</p><p>1.5 UG access: earlier approaches to UG and SLA 15</p><p>1.6 Methodological issues: ‘tapping’ linguistic competence 17</p><p>1.7 Conclusion 19</p><p>Topics for discussion 19</p><p>Suggestions for additional reading 20</p><p>2. Principles of Universal Grammar in L2 acquisition 22</p><p>2.1 UG and the logical problem of L2 acquisition 22</p><p>2.1.1 The Overt Pronoun Constraint in L2 23</p><p>2.1.2 Process versus result nominals in L2 French 30</p><p>2.1.3 Principles of UG in early interlanguage grammars:</p><p>the ECP 35</p><p>2.2 The logical problem of L2 revisited: alternative accounts 39</p><p>2.2.1 L2 input 40</p><p>2.2.2 The L1 grammar as the source of knowledge of UG</p><p>principles 41</p><p>2.3 Problems for the UG claim: wild interlanguage grammars 42</p><p>2.3.1 Reflexive binding 43</p><p>2.3.2 Null prep 51</p><p>2.4 Methodological issues 54</p><p>2.5 Conclusion 56</p><p>Topics for discussion 57</p><p>Suggestions for additional reading 57</p><p>vii</p><p>viii Contents</p><p>3. The initial state 58</p><p>3.1 What is the initial state? 58</p><p>3.2 A grammar as the initial state 61</p><p>3.2.1 The Full Transfer Full Access</p><p>Hypothesis 61</p><p>3.2.2 The Minimal Trees Hypothesis 68</p><p>3.2.3 The Valueless Features Hypothesis 78</p><p>3.3 UG as the initial state 87</p><p>3.3.1 The Initial Hypothesis of Syntax 87</p><p>3.3.2 Full Access (without Transfer) 88</p><p>3.4 Assessing initial-state hypotheses: similarities and</p><p>differences 95</p><p>3.5 Interlanguage representation: defective or not? 96</p><p>3.6 Conclusion 98</p><p>Topics for discussion 98</p><p>Suggestions for additional reading 99</p><p>4. Grammars beyond the initial state: parameters and</p><p>functional categories 100</p><p>4.1 Introduction 100</p><p>4.2 Parameters in interlanguage grammars 100</p><p>4.3 Global impairment 102</p><p>4.3.1 Breakdown of the Null Subject Parameter 102</p><p>4.3.2 Breakdown of a word-order parameter 108</p><p>4.3.3 Global impairment: assessment 113</p><p>4.4 Local impairment 114</p><p>4.4.1 Local impairment: evidence 115</p><p>4.4.2 Local Impairment: assessment 117</p><p>4.5 UG-constrained grammars and parameter setting 118</p><p>4.6 No parameter resetting 119</p><p>4.6.1 No parameter setting: evidence 120</p><p>4.6.2 No parameter resetting: assessment 125</p><p>4.7 Parameter setting and resetting 127</p><p>4.7.1 The Verb Movement Parameter: acquiring new feature</p><p>strength 128</p><p>4.7.2 Nominal projections: feature strength, features and</p><p>categories 132</p><p>4.8 Settings of neither L1 nor L2 141</p><p>4.8.1 Settings of neither L1 nor L2: reflexives 143</p><p>4.8.2 Settings of neither L1 nor L2: case checking 146</p><p>4.9 Parameter setting and resetting: assessment 148</p><p>4.10 Conclusion 149</p><p>Topics for discussion 149</p><p>Suggestions for additional reading 150</p><p>Contents ix</p><p>5. The transition problem, triggering and input 151</p><p>5.1 Introduction 151</p><p>5.2 Parsing 153</p><p>5.3 The filtering effects of grammars 153</p><p>5.4 Parameter setting: triggers and cues 157</p><p>5.4.1 Morphological triggers: a digression 160</p><p>5.5 Triggers</p><p>o katta to] itta no?</p><p>Whoi nom (hei) car acc bought that said Q</p><p>‘Who said that (he) bought a car?’</p><p>Is there a logical problem for the L2 learner of a null subject language like</p><p>Spanish or Japanese with respect to this interpretive restriction on embedded</p><p>24 Principles of Universal Grammar in L2 acquisition</p><p>pronominal subjects? Consider first the L2 input. Discovering the restriction on</p><p>the basis of input represents as much of a challenge for the L2 learner as it does</p><p>for the L1 learner. In general, overt and null pronouns appear in similar or over-</p><p>lapping contexts (see chapter 1, table 1.1), they take similar antecedents, and there</p><p>appears to be nothing in the L2 input to signal the difference between them as</p><p>far as bound variable status is concerned. Thus, frequency of occurrence in the</p><p>input is unlikely to provide any useful clue as to when pronouns may occur and</p><p>under which interpretations. The L2 learner, like the L1 learner, somehow has to</p><p>discover that, in a restricted and rather uncommon set of sentence types, an overt</p><p>pronoun cannot appear with a particular interpretation (as a bound variable), even</p><p>though it can appear with another interpretation (with a referential antecedent). In</p><p>addition, classroom input does not appear to be helpful in this respect. Accord-</p><p>ing to Pérez-Leroux and Glass (1997), this issue is not discussed in L2 Spanish</p><p>textbooks or taught in L2 classrooms; Kanno (1997) makes the same observations</p><p>for L2 Japanese. In other words, knowledge of the interpretative restriction on</p><p>overt pronouns is underdetermined by the L2 input, satisfying the first of the two</p><p>conditions described in section 2.1.</p><p>As for attaining such knowledge on the basis of themother tongue, if L2 learners</p><p>are native speakers of a [−null subject] language, such as English, nothing in the</p><p>L1 grammar would allow them to arrive at the appropriate distinction, since overt</p><p>pronouns in English are not restricted in the same way. Thus, an investigation of</p><p>the acquisition of Spanish or Japanese by native speakers of English would also</p><p>meet the second requirement outlined in section 2.1.</p><p>Acquiring the interpretive constraint on overt pronouns in L2, then, constitutes</p><p>a classic learnability problem. If L2 learners are successful in this domain, it</p><p>would strongly support the claim that interlanguage grammars areUG-constrained.</p><p>Recently, researchers have conducted experiments to investigate whether adult</p><p>learners observe the Overt Pronoun Constraint. Pérez-Leroux and Glass (1997,</p><p>1999) have examined L2 Spanish, while Kanno (1997, 1998b) has investigated L2</p><p>Japanese. In both cases, the L2 learners are native speakers of English. Thus, both</p><p>conditions for investigatingwhether the interlanguage grammar is UG-constrained</p><p>are met: the L2 input underdetermines the phenomenon being investigated and the</p><p>L1 is not a potential source of information about restrictions on overt pronouns.</p><p>The issue, then, is whether L2 learners in fact behave in ways that are consistent</p><p>with this constraint. If they do, UG is implicated since other potential sources of</p><p>such behaviour have effectively been eliminated.</p><p>In Pérez-Leroux and Glass (1999) (see box 2.1), L2 learners of Spanish at</p><p>different levels of proficiency were tested by means of a task which involved</p><p>translating biclausal sentences from the L1 English into the L2 Spanish, following</p><p>written contexts (in the L1) which strongly favoured either a quantified antecedent</p><p>2.1 UG and the logical problem of L2 acquisition 25</p><p>Box 2.1 Overt Pronoun Constraint (Perez-Léroux and Glass 1999)</p><p>Languages: L1 = English, L2 = Spanish.</p><p>Task: Translation from English into Spanish. Each sentence preceded by a para-</p><p>graph (in English) to provide a context.</p><p>Sample stimuli:</p><p>Bound variable context Referential context</p><p>The court charged that some journalists</p><p>had been in contact with the jurors. Several</p><p>of them were questioned by the judge.</p><p>In the O.J. Simpson trial, it is clear that</p><p>the press has a negative bias against the</p><p>defendant in their reporting. Some</p><p>journalist said that he was a wife-beater.</p><p>To translate: To translate:</p><p>No journalist admitted that he had talked to</p><p>the jurors.</p><p>But no journalist said that he is guilty.</p><p>Results:</p><p>Table 2.1.1 Production of null and overt pronouns (in %)</p><p>Bound variable stories Referential stories</p><p>(# = 4) (# = 4)</p><p>Null Overt Null Overt</p><p>L2 groups Elementary (n = 39) 57.7 34 21.2 67.9</p><p>Intermediate (n = 21) 73.8 26.2 35.7 59.5</p><p>Advanced (n = 18) 93.1 0 58.3 31.9</p><p>Native speakers (n = 20) 85 13.7 31.3 67.5</p><p>n = number of subjects</p><p># = number of stimuli</p><p>(within the same sentence) or a discourse-based referential antecedent (external to</p><p>the sentence) for the embedded subject pronoun. If the interlanguage grammar is</p><p>constrained by the Overt Pronoun Constraint, translations should disfavour overt</p><p>pronouns and favour null subjects where the antecedent is quantified, since overt</p><p>pronouns are prohibited in this context. In the case of the referential stories, on the</p><p>other hand, either kind of pronoun is grammatical. As can be seen in table 2.1.1,</p><p>production of overt pronouns is significantly lower in bound-variable contexts</p><p>than in referential contexts and this is true of all groups. Correspondingly, the use</p><p>of null pronouns is significantly higher following bound variable contexts than</p><p>referential contexts.1 These results suggest that L2 learners, like native speakers,</p><p>26 Principles of Universal Grammar in L2 acquisition</p><p>distinguish between referential and bound variable interpretations of pronouns,</p><p>largely disallowing overt pronouns in the latter context, compared with their use</p><p>in other contexts. Thus, the results are supportive of the claim that interlanguage</p><p>grammars are constrained by the Overt Pronoun Constraint, even at the elementary</p><p>level.</p><p>However, ideally, it would seem that the incidence of overt pronouns in bound</p><p>variable contexts ought to be 0%, a result achieved only by the advanced group.</p><p>Pérez-Leroux andGlass attribute the incidence of overt pronouns in the elementary</p><p>and intermediate groups to a tendency to overuse overt pronouns in general (based</p><p>on the L1). In many cases of L2 (and native speaker) performance, there will be</p><p>such additional factors that come into play. Indeed, what is important here is not</p><p>the absolute figures but rather the fact that there are significant differences in per-</p><p>formance across different sentence types. The issue is whether the interlanguage</p><p>grammar shows evidence of certain distinctions (in this case in the incidence of</p><p>overt subjects with referential as opposed to quantified antecedents). If learner per-</p><p>formance on one sentence type differs significantly from performance on another,</p><p>this suggests that the interlanguage grammar represents the relevant distinction</p><p>(whatever it may be). If these sentence types were represented in the same way</p><p>in the interlanguage grammar, such differences would be unexpected; instead, the</p><p>sentences should be treated the same. (See Grimshaw and Rosen (1990) for related</p><p>discussion relevant to L1 acquisition and Martohardjono (1993, 1998) for L2.)</p><p>Turning now to L2 Japanese, Kanno (1997, 1998b) investigates whether adult</p><p>learners know the prohibition on quantified andwh antecedents for overt pronouns.</p><p>(See box 2.2.) Her task was a coreference-judgment task, quite different from the</p><p>task used by Pérez-Leroux and Glass. Subjects were presented with biclausal</p><p>sentences with quantified and referential main-clause subjects and overt or null</p><p>embedded pronoun subjects; they had to indicate whether or not the embedded</p><p>pronoun could refer to the subject of the main clause.</p><p>Kanno found that native speakers and L2 learners alike differentiated in their</p><p>treatment of overt pronouns depending on the type of antecedent involved (quan-</p><p>tified or referential) (see table 2.2.1). Native speakers overwhelmingly rejected</p><p>the interpretation where an overt pronoun took a quantified antecedent, respond-</p><p>ing instead that the overt pronoun must take a sentence-external referent. This</p><p>was not</p><p>due to a general prohibition against quantified antecedents, since these</p><p>were accepted in the case of null subjects. Nor was it due to any general dislike</p><p>of sentence-internal antecedents, since referential antecedents within the same</p><p>sentence were accepted. The L2 learners showed a remarkably similar pattern of</p><p>results; their performance was not significantly different from the controls. Both</p><p>native speakers and L2 learners, then, appeared to be following the Overt Pronoun</p><p>Constraint, disallowing quantified antecedents for overt pronouns. Furthermore,</p><p>Kanno took the precaution of including a control group of native speakers of</p><p>2.1 UG and the logical problem of L2 acquisition 27</p><p>Box 2.2 Overt Pronoun Constraint (Kanno 1997)</p><p>Languages: L1 = English, L2 = Japanese.</p><p>Task: Corereference judgments. Japanese biclausal sentences, each followed by</p><p>a question asking who was performing the action described in the embedded</p><p>clause.</p><p>Sample stimuli:</p><p>Null pronoun with quantified/wh</p><p>antecedent</p><p>Null pronoun with referential antecedent</p><p>Dare ga asita uti ni iru to itteiru n desu ka. Tanaka-san wa raisyuu Kyooto e iku to</p><p>itteimasita yo.</p><p>(Who says that (he) is going to stay home</p><p>tomorrow?)</p><p>(Mr Tanaka was saying that (he) is going</p><p>to Kyoto next week.)</p><p>Q: Who do you suppose is going to stay</p><p>home tomorrow?</p><p>Q: Who do you suppose is going to</p><p>Kyoto next week?</p><p>(a) same as dare (b) another person (a) Tanaka (b) someone other than</p><p>Tanaka</p><p>Overt pronoun with quantified/wh</p><p>antecedent</p><p>Overt pronoun with referential antecedent</p><p>Dare ga kyoo kare ga uti ni iru to itteiru n</p><p>desu ka.</p><p>Tanaka-san wa raisyuu kare ga Tokyoo e</p><p>iku to iimasita yo.</p><p>(Who says that he is going to stay home</p><p>today?)</p><p>(Mr. Tanaka said that he would go to</p><p>Tokyo next week.)</p><p>Q: Who do you suppose is going to stay</p><p>home today?</p><p>Q: Who do you suppose will go to Tokyo</p><p>next week?</p><p>(a) same as dare (b) another person (a) Tanaka (b) someone other than</p><p>Tanaka</p><p>Results:</p><p>Table 2.2.1 Acceptances of antecedents for null and overt pronouns (in %)</p><p>Quantified antecedents Referential antecedents</p><p>(# = 10) (# = 10)</p><p>Null Overt Null Overt</p><p>L2 learners (n = 28) 78.5 13 81.5 42</p><p>Native speakers (n = 20) 83 2 100 47</p><p>English who judged equivalent sentences in English (with overt pronouns in</p><p>the lower clause). This group allowed quantified antecedents and referential an-</p><p>tecedents for overt pronouns to an equal (and high) extent, suggesting that the L1</p><p>English is not the source of the L2 learners’ behaviour in Japanese.</p><p>28 Principles of Universal Grammar in L2 acquisition</p><p>There are a number of differences between these two sets of experiments, both</p><p>in the methodology and in the results. In Pérez-Leroux and Glass’s experiments,</p><p>the task was a production task, involving translation; in Kanno’s studies, learners</p><p>had to make coreference judgments. In Pérez-Leroux and Glass’s task, the subject</p><p>of the main clause was always a quantified NP; referential antecedents had to be</p><p>found elsewhere within the discourse provided by the story. In contrast, Kanno</p><p>varied the nature of the main-clause subject (quantified/wh expression versus a</p><p>referential NP), such that the referential antecedent was always within the same</p><p>sentence as the embedded pronoun. There is one obvious difference in the results</p><p>of the two sets of studies which may be attributed to these task differences. In</p><p>the case of most groups, Pérez-Leroux and Glass found that translations of the</p><p>sentences following a referential context tended to disfavour null subjects; overt</p><p>pronouns were used instead (see table 2.1.1). Kanno, on the other hand, found that</p><p>referential antecedentswere accepted for null subjects (see table 2.2.1).Montalbetti</p><p>(1984) suggests that an overt pronoun is preferred when its referential antecedent</p><p>is not within the same sentence; this would account for the preference observed</p><p>by Pérez-Leroux and Glass.</p><p>Despite this difference in the treatment of null pronouns and their antecedents,</p><p>the crucial issue concerns treatment of overt pronouns with quantified antecedents.</p><p>Here, results from the two studies are consistent with each other: L2 learners</p><p>of Spanish and Japanese alike show significantly lower use or acceptance of</p><p>overt pronouns with quantified antecedents, as do native speakers. Thus, taken</p><p>together, the experiments of Pérez-Leroux and Glass (1999) and Kanno (1997)</p><p>provide evidence for Overt Pronoun Constraint effects in different L2s, with</p><p>learners at different levels of proficiency and with different methodologies being</p><p>employed.</p><p>While these results are compelling, we should bear inmind that we are interested</p><p>in the grammars of individual speakers rather than groups. The Overt Pronoun</p><p>Constraint, as a principle of UG, should constrain the grammar of each individual</p><p>learner. Group results that suggest UG-consistent behaviour on the part of L2</p><p>learners may in fact conceal potentially problematic individual variation. While</p><p>Pérez-Leroux andGlass do not provide analyses of results from individuals, Kanno</p><p>does. Kanno found that 100% of the native-speaker controls and 86% of the L2</p><p>learners demonstrated consistent behaviour with respect to overt pronouns with</p><p>quantified antecedents, where consistency is defined as the rejection of quantified</p><p>antecedents for overt pronouns in 4 or 5 cases out of 5. This level of consistent</p><p>rejections by individuals suggests that the group results are indeed an accurate</p><p>reflection of individual linguistic competence.</p><p>However, in another study, Kanno (1998a, b) found greater variability in indi-</p><p>vidual results. A different group of English-speaking intermediate-level learners</p><p>2.1 UG and the logical problem of L2 acquisition 29</p><p>of Japanese was tested at two different points in time, with a twelve-week interval</p><p>between the two sessions. The task was the same as the one in the previously de-</p><p>scribed experiment.At both test sessions, therewas a significant difference between</p><p>coreference judgments involving a quantified antecedent, depending on whether</p><p>the embedded pronoun subject was null or overt. In the latter case, acceptance of</p><p>coreference was significantly lower, consistent with the Overt Pronoun Constraint.</p><p>There were no differences in group performance across the two test sessions. At</p><p>the individual level, however, with consistency defined as above, Kanno found that</p><p>only 9 of 29 L2 learners consistently excluded quantified antecedents for overt pro-</p><p>nouns in both test sessions. A further 15 subjects showed consistent rejection in one</p><p>session or the other. Thus, at the individual level, operation of the Overt Pronoun</p><p>Constraint was not consistent across subjects or over time.</p><p>Kanno (1998b) suggests that, while theOvert PronounConstraint does constrain</p><p>interlanguage grammars, it cannot always be accessed in performance, perhaps</p><p>because it has not been activated, due to insufficient exposure to suitable input.</p><p>However, there is something unsatisfactory about this explanation, in that lack of</p><p>relevant input is precisely the motivation for principles like the Overt Pronoun</p><p>Constraint in the first place. Once the L2 learner discovers that Japanese is a</p><p>null-subject language allowing both null and overt pronouns, the Overt Pronoun</p><p>Constraint ought to come into effect.</p><p>For the sake of the argument, however, let us grant that there may be individual</p><p>variability of the kind suggested by Kanno. What would it mean for the claim</p><p>that UG constrains interlanguage grammars if individual learners do not in fact</p><p>observe a particular constraint consistently? Grimshaw and Rosen (1990) consider</p><p>apparently problematic results in L1 acquisition where children appear to accept</p><p>violations of another principle of UG, namely, Principle B of the Binding Theory.</p><p>Grimshaw andRosen point out that a number of performance factorsmay intervene</p><p>to conceal underlying competence. They suggest that L1 acquirers may ‘know’</p><p>UG principles (unconsciously) but nevertheless fail to observe or ‘obey’ them in</p><p>certain circumstances. In other words, competence and performance will some-</p><p>times</p><p>diverge, such that sentences that are ruled out as ungrammatical violations</p><p>of some principle of UG are nevertheless accepted in certain cases. According</p><p>to Grimshaw and Rosen, even when this happens, it is nevertheless possible to</p><p>show that children know the principle in question. If the child’s grammar were not</p><p>constrained by some principle, one would expect grammatical and ungrammati-</p><p>cal sentences to be treated alike, since the principle ruling out the ungrammatical</p><p>sentences would not be available. Therefore, if children treat grammatical and un-</p><p>grammatical sentences differently (accepting significantly more of the former than</p><p>the latter), this is sufficient to show that the two sentence types are not the same</p><p>in the child grammar and that the principle in question must be operating, even</p><p>30 Principles of Universal Grammar in L2 acquisition</p><p>if children do not perform with a high degree of accuracy on the ungrammatical</p><p>sentences. Linguistic competence accounts for the distinction between grammat-</p><p>ical and ungrammatical sentences; performance factors account for the failure to</p><p>observe the distinction absolutely.</p><p>Let us apply this logic to the Overt Pronoun Constraint results from individual</p><p>learners in Kanno’s study. Looking only at the sentences involving overt pronouns</p><p>and quantified antecedents, Kanno found that some subjects, at both testing ses-</p><p>sions, failed to consistently exclude the ungrammatical interpretation, leading her</p><p>to conclude that the Overt Pronoun Constraint (hence, UG in general) is not con-</p><p>sistently accessible. FollowingGrimshaw and Rosen’s proposal, however, it would</p><p>be more appropriate to compare each learner’s acceptance of antecedents for overt</p><p>pronouns, i.e. referential (grammatical) with quantified (ungrammatical). (In other</p><p>words, the comparison that Kanno makes at a group level should also be made</p><p>at the individual level.) If interlanguage grammars are constrained by the Overt</p><p>Pronoun Constraint, then we expect each subject to show a significant difference in</p><p>acceptance of grammatical versus ungrammatical interpretations. This should be</p><p>so even if performance on particular ungrammatical sentences is somewhat vari-</p><p>able. While Kanno does not provide data to allow one to check this point, she does</p><p>show that the L2 learners were much more consistent in their coreference judg-</p><p>ments involving grammatical interpretations, which suggests that, individually,</p><p>they are in fact distinguishing between licit and illicit interpretations.</p><p>To summarize the findings relating to the Overt Pronoun Constraint, Pérez-</p><p>Leroux and Glass (1997, 1999) and Kanno (1997, 1998a, b) have shown that</p><p>learners of different L2s, at different proficiency levels, tested by means of differ-</p><p>ent tasks, show significant differences in their treatment of overt pronouns with</p><p>quantified antecedents (illicit) versus null pronouns with quantified antecedents or</p><p>overt or null pronouns with referential antecedents (licit). Group and individual</p><p>results suggest that L2 learners are making the relevant distinctions, distinctions</p><p>which could not have been derived from the L1 grammar or the L2 input alone,</p><p>supporting the claim that interlanguage grammars are subject to the Overt Pronoun</p><p>Constraint.</p><p>2.1.2 Process versus result nominals in L2 French</p><p>The previous section discussed L2 knowledge of a principle of UGwhich</p><p>restricts the distribution and interpretation of embedded pronominal subjects. We</p><p>turn now to an examination of another aspect of grammar involving an interplay</p><p>between syntactic and interpretive factors, namely the distinction between two</p><p>types of nominals, known as process and result nominals. The research described</p><p>2.1 UG and the logical problem of L2 acquisition 31</p><p>below addresses the question of whether there is a learnability problem relat-</p><p>ing to the syntax–semantics interface, a relatively new field of inquiry in the L2</p><p>context. Again, the underlying assumption is that knowledge of the property in</p><p>question must have its origin in UG, the issue being whether L2 learners reveal</p><p>unconscious knowledge of subtle distinctions between the two types of nominals,</p><p>distinctions which are unlikely to be derivable from the L2 input alone or from</p><p>the L1.</p><p>Dekydtspotter, Sprouse and Anderson (1997) look at whether L2 learners are</p><p>sensitive to differences between process and result nominals in L2 French. A</p><p>process nominal describes an event or something ongoing; a result nominal names</p><p>the ‘output of a process or an element associated with a process’ (Grimshaw 1990:</p><p>49). The English nominal destruction illustrates the distinction. In (3a), destruction</p><p>refers to the process of destroying the city; the enemy brings about the destruction</p><p>(i.e. it is the agent), while the city undergoes the destruction (i.e. it is the theme).</p><p>In (3b), on the other hand, destruction refers to the outcome, the result of the</p><p>destroying. (Examples from Grimshaw 1990: 52.)</p><p>(3) a. The enemy’s destruction of the city was awful to watch. (process)</p><p>b. The destruction was awful to see. (result)</p><p>Dekydtspotter et al. (1997) investigate dyadic process and result nominals in</p><p>L2 French, that is nominals taking an agent and a theme argument (as in (3a)).</p><p>There are restrictions on the arguments of such nominals, in particular, restrictions</p><p>on the form of the agent. Consider the sentences in (4) (a result nominal) and (5)</p><p>(a process nominal). In both cases, the theme is expressed in a phrase introduced</p><p>by de (‘of’), while the agent can be introduced by par (‘by’). In (4a), la 9e is the</p><p>theme and Karajan is the agent; in (5), Tokyo is the theme, while Godzilla is the</p><p>agent. So far, then, result and process nominals pattern alike.</p><p>(4) Result nominal</p><p>a. la version de la 9e par Karajan</p><p>the version of the 9th by Karajan</p><p>‘Karajan’s version of the Ninth symphony’</p><p>b. la version de la 9e de Karajan</p><p>the version of the 9th of Karajan</p><p>(5) Process nominal</p><p>a. la destruction de Tokyo par Godzilla</p><p>the destruction of Tokyo by Godzilla</p><p>‘Godzilla’s destruction of Tokyo’</p><p>b. *la destruction de Tokyo de Godzilla</p><p>the destruction of Tokyo of Godzilla</p><p>32 Principles of Universal Grammar in L2 acquisition</p><p>The two kinds of nominals differ, however, in that result nominals also permit</p><p>the agent to be introduced by de, as in (4b). In process nominals, on the other</p><p>hand, this is not possible, as shown by the ungrammaticality of (5b). Thus, result</p><p>nominals permitmultiple phrases introduced by de (namely, both theme and agent),</p><p>occurring in either order, whereas process nominals do not. (Nevertheless, native</p><p>speakers have a preference for the agent of a result nominal to be introduced by</p><p>par rather than de.)</p><p>This difference between result and process nominals within French (as well</p><p>as other languages) stems from independent restrictions on argument structure</p><p>and event structure originating in UG. In particular, in process nominals with</p><p>multiple de arguments there are problems in assigning the agent role, violating</p><p>the principle of Full Interpretation, which requires that arguments be interpretable</p><p>at LF. (Unfortunately, Dekydspotter et al. give little detail on precisely how this</p><p>constraint is played out in the case of the different types of nominals that they</p><p>investigate.2) In consequence of an independent parametric difference between</p><p>French and English, the distinction between these nominals is partially obscured</p><p>in English. English disallows multiple of phrases in these contexts, as can be seen</p><p>by looking at the literal glosses for (4b) and (5b). Instead, the agent appears as a</p><p>prenominal genitive, whether the nominal is process or result, as can be seen by the</p><p>translations of (4a), where the agent is expressed as Karajan’s, and (5a) where it</p><p>appears asGodzilla’s. (Alternatively, English can express the agent in a by-phrase,</p><p>equivalent to the par-phrases in French: the destruction of Tokyo by Godzilla; the</p><p>version of the 9th by Karajan – again, there is no distinction between the two types</p><p>of nominal in this respect.)</p><p>Dekydtspotter et al., following</p><p>Carstens (1991), attribute this crosslinguistic</p><p>difference (namely, the possibility of multiple de arguments in French and the im-</p><p>possibility of multiple of arguments in English) to a parametric difference between</p><p>the two languages. As described in chapter 1, section 1.4.1, nouns in French raise</p><p>from N to the head of a functional projection, Num (Bernstein 1993; Carstens</p><p>1991; Valois 1991). When a nominal raises to Num, it governs both its arguments</p><p>and case is assigned under government (via de-insertion) to both theme and agent.</p><p>In languages like English, in contrast, nouns do not raise to Num; in consequence,</p><p>a nominal will never govern its agent and case assignment to the agent via of</p><p>insertion is ruled out, case assignment being achieved, instead, by an alternative</p><p>mechanism (Spec–head agreement).</p><p>Turning to the L2 logical problem for English-speaking learners of French,</p><p>at issue is whether they will acquire the distinction between process and result</p><p>nominals, with respect to the possibility of multiple de-phrases. It is unlikely that</p><p>the L2 input is sufficient to signal this difference, since what has to be acquired</p><p>2.1 UG and the logical problem of L2 acquisition 33</p><p>is knowledge that sentences like (5b) are ungrammatical, whereas sentences like</p><p>(4b) are possible. Since both types of nominal allow the agent to be introduced by</p><p>par (as in (4a) and (5a)), the potential for overgeneralization is considerable. Nor</p><p>is this a topic that is specifically taught in L2 French classrooms (though contrasts</p><p>between de and par in other contexts may be).</p><p>It also seems highly unlikely that properties of the L1Englishwould be sufficient</p><p>to allow learners to arrive at the relevant distinction in the L2. Although there are</p><p>subtle distinctions between process and result nominals in English, these are not</p><p>manifested in the same way. Because English does not have N-raising, it does not</p><p>allow multiple of -phrases. There is no distinction in how the agent is realized in</p><p>process and result nominals; in both cases, the agent is found either as a prenominal</p><p>genitive (Godzilla’s destruction; Karajan’s version) or in a postnominal by-phrase</p><p>(the destruction by Godzilla; the version by Karajan). Thus, the L2 acquisition of</p><p>the distinction between process and result nominals with respect to multiple de-</p><p>phrases constitutes a learnability problem. If English-speaking learners of French</p><p>show knowledge of the distinction, this would suggest that UG constrains the</p><p>interlanguage grammar in this domain.</p><p>Dekydtspotter et al. tested for knowledge of the process/result distinction with</p><p>respect to multiple de-phrases in an experiment involving English-speaking adults</p><p>learning French. (See box 2.3.) The task was an acceptability judgment task,</p><p>with written scenarios providing a context (i.e. somewhat similar to the proce-</p><p>dure adopted by Pérez-Leroux and Glass (1997, 1999), except that Pérez-Leroux</p><p>and Glass asked for translations not judgments). Since the context is crucial to</p><p>establish the interpretation of the nominal as either process or result, the scenarios</p><p>were presented in the L1 (again, like Pérez-Leroux and Glass), in order to ensure</p><p>that the L2 learners would make their judgments on the basis of the appropriate</p><p>interpretation. The test sentences to be judged were presented in French.</p><p>Native speakers of French distinguished sharply and significantly between the</p><p>two types of nominals, accepting multiple de-arguments in result nominals to a</p><p>significantly greater extent than in process nominals, where they were largely</p><p>rejected. (See table 2.3.1.) The relatively low rate of acceptance even in result</p><p>nominals reflects the fact that par is the preferred way of realizing the agent. The</p><p>L2 learners at all levels showed the same distinction between process and result</p><p>nominals, with the advanced learners not being significantly different from the</p><p>controls. The beginners and intermediate subjects in general showed amuch higher</p><p>acceptance rate of multiple de phrases with both types of nominals; nevertheless</p><p>there is a significant difference between the two sentence types, in the expected</p><p>direction, suggesting that the distinction is represented in the grammar; this is true</p><p>both at the group level and at the individual level.</p><p>34 Principles of Universal Grammar in L2 acquisition</p><p>Box 2.3 Process and result nominals (Dekydtspotter et al. 1997)</p><p>Languages: L1 = English, L2 = French.</p><p>Task:Acceptability judgments of sentences containing both a theme and an agent</p><p>introduced by de. Each sentence preceded by a paragraph (in English) to provide</p><p>a context.</p><p>Sample stimuli:</p><p>Result nominals (grammatical) Process nominals (ungrammatical)</p><p>Jean loves nineteenth-century train</p><p>stations. La Gare du Nord, Paris’s northern</p><p>train station, still has some of that feel,</p><p>even though the steam engines are long</p><p>gone. Whenever Jean looks at a catalogue</p><p>of Monet’s works (in particular the work</p><p>entitled La Gare du Nord), he gets a feel</p><p>for what it would have been like to be</p><p>there. Not surprisingly,</p><p>Shocking and disturbing, yes, but</p><p>nonetheless true! The executioner’s wife</p><p>was having an affair, and the only time</p><p>she could meet with her lover was when</p><p>her husband the executioner was on the</p><p>job. She read in the newspaper that a</p><p>couple of traitors had been sentenced to</p><p>death on Friday, so she sent a note to her</p><p>lover, saying:</p><p>Jean adore la peinture de la Gare Viens chez moi vendredi pendant</p><p>du Nord de Monet. l’éxécution de mon mari des traitres</p><p>Feels possible in the context? Feels possible in the context?</p><p>Yes No Cannot decide Yes No Cannot decide</p><p>Results:</p><p>Table 2.3.1 Acceptances of multiple de arguments (in %)</p><p>Result nominals Process nominals</p><p>(# = 10) (# = 10)</p><p>L2 groups Beginners (n = 38) 69.21 53.51</p><p>Intermediate (n = 32) 71.56 48.96</p><p>Advanced (n = 20) 63.5 24.17</p><p>Native speakers French (n = 48) 50.42 15.65</p><p>English (n = 24) 22.08 10.53</p><p>In summary, L2 learners at all levels of proficiency showed an asymmetry in their</p><p>acceptance of multiple phrases introduced by de in result and process nominals.</p><p>The results suggest that the interlanguage grammar is constrained by UG, since</p><p>the distinction between process and result nominals appears to come neither from</p><p>the L1 nor from the L2 input.</p><p>2.1 UG and the logical problem of L2 acquisition 35</p><p>Nevertheless, a caveat is in order. Like Kanno (1997), Dekydtspotter et al.</p><p>included a group of native speakers of the L1 who judged English translations</p><p>of the French test items, all of which are ungrammatical in English, in order</p><p>to see whether learners might simply be judging L2 sentences on the basis of</p><p>their mother tongue. Overall, multiple of -phrases were rejected in both types of</p><p>nominal in contrast to the L2 results, suggesting that the L1 grammar is unlikely</p><p>to be the source of knowledge of the distinction with respect to the L2 sentences.</p><p>Nevertheless, the difference in their judgments on the two sentence types in the L1</p><p>was significant; in other words, even though both sentence types were considered</p><p>fully ungrammatical, the process nominals with two of -phrases were even less</p><p>likely to be accepted than the result nominals. Thus, one cannot totally exclude the</p><p>possibility that the distinction shown by the L2 learners was in some way based</p><p>on the L1 grammar.</p><p>2.1.3 Principles of UG in early interlanguage grammars: the ECP</p><p>So far, the experimental research that we have considered has mainly</p><p>targeted intermediate or advanced learners, a tacit assumption being made that if</p><p>some UG principle can be shown to constrain later interlanguage grammars, it</p><p>would also have constrained the grammars of earlier stages. Indeed, if interlan-</p><p>guage grammars are UG-constrained, this should be so from the earliest stages</p><p>of L2 acquisition.3 In other words, we expect to find evidence of UG principles</p><p>functioning in the interlanguage grammars of beginners or low-proficiency L2</p><p>learners, all other things being equal. However, a methodological problem arises</p><p>in this context. UG principles are proposed as an explanation of</p><p>very subtle and ab-</p><p>stract linguistic phenomena, as we have seen; in many cases, these principles relate</p><p>to properties exemplified in complex sentences. In consequence, some of the sen-</p><p>tence types that are typically investigated by linguists and acquisition researchers</p><p>are, for reasons independent of the UG issue, too difficult for beginner-level L2</p><p>learners to deal with.</p><p>Kanno (1996) has shown that Japanese case drop, a phenomenon that occurs in</p><p>simple sentences, can be used to investigate whether the interlanguage grammars</p><p>of beginners and low-proficiency learners are constrained by UG. Japanese has</p><p>case particles, including nominative /-ga/ and accusative /-o/, as shown in (6a)</p><p>(examples from Kanno). When an object is marked with accusative case, the</p><p>particle may be dropped in informal spoken language, as in (6b); however, it</p><p>is ungrammatical to drop the nominative particle marking the subject, as in (6c).</p><p>Thus, there is an asymmetry here, with respect to the type of particle that can be</p><p>dropped.</p><p>36 Principles of Universal Grammar in L2 acquisition</p><p>(6) a. John ga sono hon o yonda.</p><p>John nom that book acc read-past</p><p>‘John read that book’</p><p>b. John ga sono hon yonda.</p><p>John nom that book read-past</p><p>c. *John sono hon o yonda.</p><p>John that book acc read-past</p><p>Japanese case dropmeets the criteria discussed abovewith respect to learnability</p><p>considerations. Firstly, the relevant generalization (that accusative particles may</p><p>be dropped while nominative may not) is not derivable on the basis of the L1,</p><p>assuming an L1 without case particles, such as English. Secondly, it does not</p><p>appear that this property could be acquired on the basis of L2 input alone, since</p><p>the input underdetermines the case drop property. If the learner were to try and</p><p>make the relevant generalization on the basis of ‘noticing’ dropped particles in the</p><p>input, the potential for overgeneralization is considerable. In addition to permitting</p><p>accusative case drop, Japanese has a topic marker /-wa/ which can be dropped</p><p>(Kuno 1973), as shown in (7a), where the topic marker is present, and (7b), where</p><p>it has been omitted.4</p><p>(7) a. Johni wa Hanako ga proi sono hon o yonda to itta.</p><p>Johni top Hanako nom (hei) that book acc read-past that said</p><p>‘Speaking of John, Hanako said that he (= John) read that book.’</p><p>b. Johni Hanako ga proi sono hon o yonda to itta.</p><p>Johni Hanako nom (hei) that book acc read-past that said</p><p>‘Speaking of John, Hanako said that he (= John) read that book.’</p><p>Since subjects can be topicalized and /-wa/ can be omitted from a topicalized</p><p>subject, it would not be unreasonable to assume that any particle marking a subject</p><p>can be dropped. In addition, the nominative /-ga/ can be dropped when it occurs on</p><p>the complement of a stative verb or when it marks the subject of an unaccusative</p><p>verb.5 Furthermore, case drop is apparently not taught in Japanese L2 classrooms,</p><p>although it is, presumably, exemplified in classroom input.</p><p>Fukuda (1993) attributes the asymmetry in case particle deletion (nomina-</p><p>tive prohibited, accusative permissible) to the Empty Category Principle (ECP)</p><p>(Chomsky 1981), a principle which accounts for a variety of subject–object asym-</p><p>metries across languages. The ECP requires nonpronominal empty categories (i.e.</p><p>traces or variables) to be properly governed. An empty category in object position</p><p>is properly governed by the verb, whereas an empty category in subject position</p><p>is not properly governed. Hence, in a number of different situations, empty cate-</p><p>gories are permitted in object position but not subject position. Assuming that null</p><p>case particles are empty categories of the relevant sort, it is permissible to omit</p><p>an accusative case particle because the empty particle is properly governed by</p><p>2.1 UG and the logical problem of L2 acquisition 37</p><p>Box 2.4 Case-particle deletion (ECP) (Kanno 1996)</p><p>Languages: L1 = English, L2 = Japanese.</p><p>Task: Grammaticality judgments. The naturalness of sentences is assessed on a</p><p>scale of 1 (unnatural) to 3 (natural).</p><p>Sample stimuli:</p><p>Missing accusative particle in sentences</p><p>with 2 overt arguments (grammatical)</p><p>Missing accusative particle in sentences</p><p>with 1 overt argument (grammatical)</p><p>Suzuki-san wa dono biiru nomimasita ka? Dono biiru nomimasita ka?</p><p>(Which beer did Mr(s) Suzuki drink?) (Which beer did (s/he) drink?)</p><p>Missing nominative particle in sentences</p><p>with 2 overt arguments (ungrammatical)</p><p>Missing nominative particle in sentences</p><p>with 1 overt argument (ungrammatical)</p><p>Dono gakusee biiru o nomimasita ka? Dono gakusee nomimasita ka?</p><p>(Which student drank beer?) (Which student drank (it)?)</p><p>Results:</p><p>Table 2.4.1 Case drop: mean naturalness scores (from 1 to 3)</p><p>2 overt arguments (# = 8) 1 overt argument (# = 8)</p><p>acc nom acc nom</p><p>L2 learners (n = 26) 2.4 1.76 2.58 1.64</p><p>Native speakers (n = 20) 2.6 1.36 2.86 1.31</p><p>the verb.6 In contrast, a null nominative particle would not be properly governed,</p><p>hence nominative particles must be overt.</p><p>Kanno (1996) investigated whether beginners learning Japanese are sensitive</p><p>to this asymmetry in case-particle deletion, that is, whether they distinguish be-</p><p>tween grammatical sentences with dropped accusative particles like (6b) and</p><p>ungrammatical sentences with dropped nominative particles like (6c). Subjects</p><p>were tested on a grammaticality-judgment task in which they had to assess</p><p>the naturalness of sentences, like those in (6), with and without case particles.</p><p>(See box 2.4.)</p><p>Results revealed significantly greater acceptance of accusative case drop over</p><p>nominative case drop, and no significant differences betweenL2 learners and native</p><p>speakers. This is true for sentences with one or two arguments expressed overtly.</p><p>(See table 2.4.1.) These results, then, are consistent with the proposal that a UG</p><p>principle (the ECP) functions in the early interlanguage grammar. Kanno was able</p><p>38 Principles of Universal Grammar in L2 acquisition</p><p>to establish this without having to resort to the kinds of complex sentences that</p><p>have previously been used to investigate knowledge of the ECP in L2 English (e.g.</p><p>Bley-Vroman, Felix and Ioup 1988; White 1988) or L2 German (Felix 1988).</p><p>If the interlanguage grammars of beginners show evidence of being constrained</p><p>by the ECP, one expects the same in the case of more advanced learners. In a</p><p>subsequent study, Kanno (1998a) reports results which largely confirm the orig-</p><p>inal findings. The second study involved intermediate-level learners of Japanese.</p><p>Group results again revealed significant differences between acceptance of miss-</p><p>ing nominative and accusative case particles, consistent with the operation of the</p><p>ECP. In this follow-up study, Kanno also looked at individual results and found</p><p>that most individuals exhibited the relevant contrast, although not necessarily as</p><p>strongly as native speakers.</p><p>In contrast, Kellerman, van Ijzendoorn and Takashima (1999), testing Dutch-</p><p>speaking L2 learners, were unable to replicate Kanno’s finding of a sensitivity to</p><p>the subject–object asymmetry in case-particle deletion. This attempted replication,</p><p>however, is somewhat problematic, in that it involved teaching aminiature artificial</p><p>language to students not previously exposed to Japanese, in a single session. This</p><p>artificial language had the following characteristics: it was SOV, subjects and</p><p>objects could be omitted (in other words, it was a prodrop language), and overt</p><p>subjects and objects were always case-marked. Students were explicitly taught</p><p>these properties, by means of translation; they had no exposure to input involving</p><p>missing case particles and theywere not taught that casemarkers could be dropped.</p><p>Immediately after the teaching phase, they were explicitly told that there was one</p><p>rule of this language which they had not been taught and which they were to</p><p>try to discover. Test sentences manipulated dropped nominative or accusative case</p><p>particles, the idea being that, if interlanguagegrammars are constrainedby theECP,</p><p>dropping accusative case would be permissible, while dropping nominative would</p><p>not. Learners</p><p>showed some acceptance of dropped case marking, both nominative</p><p>and accusative. A second studymodified properties of the artificial language (using</p><p>real Japanese lexical items instead of invented ones) and the test instructions, as</p><p>well as involving somewhat older subjects; it was otherwise similar to the first</p><p>study. Again, results showed no subject–object asymmetry in acceptance of case</p><p>particle deletion.</p><p>Kellerman et al. claim that these results cast doubts on Kanno’s conclusion that</p><p>early L2 grammars are constrained by the ECP. However, there are a number of</p><p>problemswith these two studieswhich in turn cast doubts on their own conclusions.</p><p>The ‘L2’ was artificial (the learners knew this – they were told that they were</p><p>dealing with an extra-terrestrial language); learners were taught just three explicit</p><p>rules, via translation. Exposure to the ‘L2’was extremely brief. It seems very likely</p><p>that learners simply treated the whole thing as a problem-solving exercise, rather</p><p>2.2 Alternative accounts 39</p><p>than as a genuine language-learning situation, in which case the ECP is simply</p><p>irrelevant, since it is a constraint on natural language systems, not on systems</p><p>arrived at by other means.7 Hence, the absence of ECP effects in Kellerman et al.’s</p><p>study is hardly surprising, leaving Kanno’s results unchallenged. (Indeed, even if</p><p>Kellerman et al.’s results had supported Kanno’s, one would still have to question</p><p>their validity.)</p><p>Potentially more problematic for Kanno’s claim that L2 learners observe the</p><p>ECP is an additional experiment by Kellerman and Yoshioka (1999). This experi-</p><p>ment involvedgenuineL2acquisition, testingDutch-speaking learners of Japanese.</p><p>The task was a grammaticality-judgment task. Whereas each of Kanno’s test sen-</p><p>tences contained one dropped case particle, either nominative or accusative (aswell</p><p>as a case-marked NP in those sentences which contained two overt arguments),</p><p>Kellerman and Yoshioka included additional sentence types, to provide a greater</p><p>variety of overt and null case-marking combinations. Their reason for doing so</p><p>was the incorrect and unmotivated assumption that the ECP predicts a hierarchy</p><p>of acceptability, as follows (where + means overt case marking and −means no</p><p>case marking):</p><p>(8) +nom+acc ≥ +nom–acc > –nom+acc > –nom–acc</p><p>Kellerman and Yoshioka’s results show no evidence for this hierarchy; hence,</p><p>they conclude that the ECP does not constrain the interlanguage grammar of these</p><p>learners. In fact, as pointed out by Kanno (1999), no such hierarchy is predicted</p><p>by the ECP, which simply says that nominative case particles in subject position</p><p>should not be deleted. The ECP is neutral as to whether accusative case particle</p><p>deletion will take place at all (the ECP certainly does not require this) and whether</p><p>overt accusative marking is preferable to null. As Kanno points out, if one holds</p><p>the accusative case markers constant (comparing +nom+acc to –nom+acc</p><p>or +nom–acc to –nom–acc ), Kellerman and Yoshioka’s subjects do show a</p><p>significant difference in their acceptance of sentences with andwithout nominative</p><p>case particles, in favour of the former, as predicted by the ECP account.</p><p>2.2 The logical problem of L2 revisited: alternative accounts</p><p>So far, we have considered three different linguistic properties assumed</p><p>to stem fromUG: (i) differences between overt and null pronouns (attributed to the</p><p>Overt PronounConstraint); (ii) the distinction between process and result nominals</p><p>(attributed, indirectly, to restrictions on argument structure); (iii) subject–object</p><p>asymmetries in case-particle deletion (a consequence of the ECP). In each case,</p><p>the claim has been that there is a learnability problem, the L2 input alone (whether</p><p>40 Principles of Universal Grammar in L2 acquisition</p><p>naturalistic input or classroom input including instruction) being insufficient to</p><p>allow the learner to arrive at the relevant distinctions. The assumption has also</p><p>been that the relevant properties could not have been arrived at on the basis of</p><p>the L1: in the case of the studies involving the Overt Pronoun Constraint, learners</p><p>were chosen whose L1s do not permit null subjects, with overt pronouns behaving</p><p>quite differently in the two languages; in the case of process and result nominals,</p><p>the L1 and L2 differed in terms of how arguments of these nominals are realized;</p><p>in the case of case particle deletion, the L1 was a language without case particles</p><p>at all. In all the studies described above, L2 learners showed behaviour which</p><p>revealed unconscious knowledge of subtle properties of the L2, consistent with</p><p>the claim that their grammars are UG-constrained.</p><p>In this section, we consider attempts to provide alternative accounts of L2</p><p>learners’ successes in acquiring the kinds of subtle distinctions discussed above.</p><p>Certain researchers deny that there is an underdetermination problem in such</p><p>cases; hence, they claim, there is no need to assume that interlanguage grammars</p><p>are UG-constrained. Two different lines of argument have been advanced. The first</p><p>questions the poverty-of-the-stimulus claim, by trying to show that the L2 input</p><p>is in fact sufficient to allow the relevant contrasts to be induced without recourse</p><p>to principles of UG. The second accepts that the L2 input underdetermines the</p><p>unconscious knowledge that L2 learners attain but maintains that this knowledge</p><p>derives from the L1 grammar rather than UG.</p><p>2.2.1 L2 input</p><p>Kellerman and colleagues take the position that properties of the L2 in-</p><p>put are sufficient to explain L2 learners’ differential treatment of nominative and</p><p>accusative case drop.While they failed to find a subject–object asymmetry in case-</p><p>particle deletion in their three studies, they accept the validity of Kanno’s (1996)</p><p>finding of a robust difference between acceptances of deleted accusative versus</p><p>nominative particles.</p><p>Kellerman and Yoshioka (1999) and Kellerman et al. (1999) account for the</p><p>differences in their results and Kanno’s in terms of differences in the L2 input.</p><p>They suggest that case-particle drop would be relatively easy to formulate as a</p><p>pedagogical rule and would be easy to learn on the basis of instruction. However,</p><p>as neither their subjects nor Kanno’s had received such instruction, this is a moot</p><p>point. They also suggest that naturalistic input may show a statistical bias in favour</p><p>of dropped object particles and that L2 learners in Hawaii (where Kanno’s subjects</p><p>were tested) are likely to have been exposed to such input. Their performance,</p><p>then, could be explained in terms of properties of the L2 input and, in their view,</p><p>in terms of the input alone. But this begs the original question: while input with</p><p>2.2 Alternative accounts 41</p><p>dropped accusative case particles is certainly necessary to motivate case drop in</p><p>the first place, this does not mean that it is sufficient to account for the knowledge</p><p>that L2 learners acquire. The input will exemplify object NPs with and without</p><p>case markers, subject topics with and without topic markers, as well as nominative</p><p>markersmissing on subjects of a subclass of verbs. The potential to overgeneralize,</p><p>thus dropping nominative case markers on subjects in general, remains even if</p><p>there is a preponderance of accusative particle drop in the input. It is important to</p><p>understand that the UG approach does not deny the importance of input. But the</p><p>claim is that input alone is not enough.</p><p>However, suppose, for the sake of the argument, that one could somehowdemon-</p><p>strate that a particular phenomenon (in this case, dropping of accusative particles)</p><p>could successfully be acquired on the basis of statistical frequency in the in-</p><p>put. While this might show that case-particle drop does not constitute a genuine</p><p>learnability problem, it would not dispose of the learnability problem in general.</p><p>The fact that there may not be a logical problem of L2 acquisition with respect</p><p>to one phenomenon does not mean that there is never a logical problem of L2</p><p>acquisition.</p><p>2.2.2 The L1 grammar as the source of knowledge of UG principles</p><p>According to another approach which argues against the need to invoke</p><p>UG, the complexity of the mental representations achieved by L2 learners is ac-</p><p>knowledged, aswell as the fact that input alone cannot account for such complexity.</p><p>Instead, the unconscious knowledge attained by L2 learners is claimed to derive</p><p>from the L1 grammar (Bley-Vroman 1990; Clahsen andMuysken 1989; Schachter</p><p>1989, 1990). For example, there are a number of subject–object asymmetries in</p><p>English which do not involve case particles but which do implicate the ECP.</p><p>The L2 learner of Japanese, then, may have been able to arrive at the relevant</p><p>properties of case drop in Japanese on the basis of very different properties of</p><p>English which are subject to the same constraint. Similarly, in the case of process</p><p>and result nominals, there are a number of subtle differences between them in</p><p>English (e.g. Grimshaw 1990), which might somehow explain the results obtained</p><p>by Dekydtspotter et al. (1997) for L2 French. Thus, it might be argued that knowl-</p><p>edge of the properties described in this chapter (aswell as other principles discussed</p><p>in the literature) stems from the mother tongue and, hence, fails to provide evi-</p><p>dence that interlanguage grammars are UG-constrained independently of the L1</p><p>grammar.</p><p>Indeed, formany principles ofUG, it appears that the L1 can never be completely</p><p>ruled out as a source of the L2 learner’s unconscious knowledge. Since the L1 is a</p><p>natural language and since many UG principles manifest themselves in the L1 in</p><p>42 Principles of Universal Grammar in L2 acquisition</p><p>some form or other, it will often be hard or impossible to disentangle the two (Hale</p><p>1996).8 Nevertheless, the results described so far suggest that L2 learners are able to</p><p>apply UG principles to totally new domains, including data that do not occur in the</p><p>L1. As we have seen, L2 learners observe UG constraints even when they apply to</p><p>very different phenomena in the L2 (restrictions on arguments of process nominals;</p><p>ECP and case-particle drop), ruling out surface transfer as an explanation. Thus,</p><p>it appears that there is considerable flexibility in the system; UG constraints can</p><p>be applied to new data, and to situations that are entirely different from what</p><p>pertains in the L1. It is not clear that this would be the case if UG principles</p><p>were only somehow being ‘reconstructed’ on the basis of how they operate in</p><p>the L1. Furthermore, as will be discussed in chapter 4, where parameters of UG</p><p>are concerned, it is a relatively straightforward matter to distinguish between L1</p><p>parameter settings and other settings, hence to discover whether or not L2 learners</p><p>are restricted to options realized in the L1.</p><p>In any case, an approach which sees the issue as an either/or matter (UG or the</p><p>L1) is misconceived. This is a false dichotomy. It is inappropriate to contrast UG</p><p>with the L1 as the source of UG-like knowledge; rather, both appear to be involved.</p><p>This issue will be considered in greater detail in chapters 3 and 4.</p><p>2.3 Problems for the UG claim: wild interlanguage grammars</p><p>So far, results from the studies discussed in this chapter are consistentwith</p><p>the claim that UG constrains interlanguage grammars. L2 learners demonstrate</p><p>unconscious knowledge of subtle contrasts which are by no means transparent</p><p>in the L2 input and which are not realized in any obvious way in the L1. This</p><p>unconscious knowledge is unexpected if interlanguage grammars are not UG-</p><p>constrained. While the source of this knowledge may, at least in some cases, be</p><p>UG as instantiated in the L1, it is clear that the constraints are applied to completely</p><p>new data and to phenomena that do not exist in the L1. Learner grammars, then,</p><p>conform with UG.</p><p>There is another way of exploring whether or not UG constrains interlanguage</p><p>grammars. UG defines what a grammar is, determining what mental represen-</p><p>tations can and cannot be like. Natural language grammars fall within a range</p><p>sanctioned by UG. L1 acquirers are limited by the hypothesis space provided by</p><p>UG, which reduces the number of logical possibilities that have to be entertained</p><p>in order to arrive at a grammar for the language being acquired. Developing L1</p><p>grammars, in other words, are ‘possible’ grammars (White 1982). At least with</p><p>respect to the properties we have considered so far, the same appears to be true</p><p>of interlanguage grammars, in the sense that they fall within the range sanctioned</p><p>2.3 Problems for the UG claim: wild grammars 43</p><p>by UG. Interlanguage grammars, then, exhibit characteristics typical of natural</p><p>language.</p><p>Grammars that do not conform to principles ofUGhavebeenvariously described</p><p>as impossible (White 1982, 1988), rogue (Thomas 1991a), illicit (Hamilton 1998)</p><p>or wild (Goodluck 1991; Klein 1995a); the latter term will be adopted here. If</p><p>interlanguage grammars are UG-constrained, wild grammars are predicted not</p><p>to occur in L2 acquisition. In other words, interlanguage grammars should be</p><p>restricted to properties found in the L1 and/or the L2, and/or natural languages</p><p>in general. If it can be shown that interlanguage grammars do not conform to</p><p>properties of natural language, this would suggest that the operation of UG is in</p><p>some way impaired.</p><p>Recently, there have been proposals that interlanguage grammars are in fact</p><p>sometimes wild. Two phenomena will be considered in this context: reflexives</p><p>(Christie and Lantolf 1998) and null prepositions (Klein 1993b, 1995a). In both</p><p>cases, it has been suggested that the interlanguage grammar shows a cluster of</p><p>properties that is illicit, hence that the grammar is not sanctioned by UG.</p><p>2.3.1 Reflexive binding</p><p>There has been considerable research which investigates whether reflex-</p><p>ives in interlanguage grammars are constrained by Principle A of the Binding</p><p>Theory. Binding Theory places constraints on coreference between various kinds</p><p>of NPs, Principle A being concernedwith properties of anaphors, such as reflexives</p><p>(himself , herself , etc.) (Chomsky 1981a). According to Principle A, an anaphor</p><p>must be bound in its governing category; effectively, it must take an antecedent</p><p>in a local domain (usually the same clause), as shown in (9). In (9a), coreference</p><p>is possible between the reflexive, herself , and the subject, Mary, because they are</p><p>within the same clause. In (9b), on the other hand, herself cannot refer to the sub-</p><p>ject of the higher clause, Mary, but it can refer to the subject of the lower clause,</p><p>namely Susan.</p><p>(9) a. Maryi blamed herselfi</p><p>b. Maryi thought that Susanj blamed herself*i/j</p><p>As is well known, reflexives differ crosslinguistically as to the domain in which</p><p>theymust be bound.While English reflexives require their antecedents to be within</p><p>the same clause (i.e. local), there are many languages, such as Japanese, which</p><p>permit the antecedent to be in a different clause. The Japanese sentence in (10) is</p><p>ambiguous, with zibun (‘self’) able to have either the local subject, Susan, or the</p><p>main-clause subject,Mary, as its antecedent. When the antecedent is in a different</p><p>clause from the reflexive, this is referred to as non-local or long-distance binding.</p><p>44 Principles of Universal Grammar in L2 acquisition</p><p>(10) Maryi ga Susanj ga zibuni/j o semeta to omotta.</p><p>Mary nom Susan nom self acc blamed that thought</p><p>‘Mary thought that Susan blamed herself.’</p><p>In addition to domain (locality), orientation is also important, that is, whether</p><p>the antecedent is a subject or not. Again, languages differ in this respect. An</p><p>English reflexive can take a subject or a non-subject as its antecedent, whereas</p><p>Japanese reflexives are restricted to subject antecedents. An English sentence like</p><p>(11a) is potentially ambiguous, with either the subject or the object available as the</p><p>antecedent, though context will usually favour one interpretation over the other.</p><p>The Japanese sentence in (11b), on the other hand, only has one interpretation,</p><p>where the subject, kanja-ga, is the antecedent.</p><p>(11) a. The patienti asked the nursej about herselfi/j</p><p>b. Kanjai ga kangofuj ni zibuni/∗j no koto</p><p>nitsuite tazuneta.</p><p>patient nom nurse dat self gen matter about asked</p><p>‘The patient asked the nurse about herself.’</p><p>Much of the earlier L2 research on reflexives was conducted within the frame-</p><p>work of Manzini and Wexler (1987) and Wexler and Manzini (1987). They pro-</p><p>posed two parameters to handle crosslinguistic differences in how Principle A</p><p>operates: the Governing Category Parameter, which dealt with domain, and the</p><p>Proper Antecedent Parameter, dealing with orientation. Explicit discussion of the</p><p>possibility of wild interlanguage grammars first arose in this context (Eckman</p><p>1994; Thomas 1991a). However, most of the L2 research conducted on the Gov-</p><p>erning Category Parameter and the Proper Antecedent Parameter was not directed</p><p>at the issue of wild grammars but rather to the question of whether L2 learners can</p><p>reset parameters and whether or not they observe the Subset Principle (Finer 1991;</p><p>Finer and Broselow 1986; Hirakawa 1990; Thomas 1991b). (SeeWhite (1989) for</p><p>review.)</p><p>According to subsequent linguistic analyses, parameterization of this kind is</p><p>not involved. Instead, two different types of anaphors are assumed, with differing</p><p>properties. Each anaphor type has a cluster of properties associated with it, as</p><p>shown in (12).</p><p>(12) a. Xmax (phrasal) reflexives. These are morphologically complex, require local</p><p>antecedents and allow subjects and (in some languages) non-subjects as an-</p><p>tecedents, e.g. himself , herself in English; taziji (‘himself’) in Chinese; kare-</p><p>zisin, kanojo-zisin (‘himself’, ‘herself’) in Japanese.</p><p>b. X0 (head) reflexives. These are monomorphemic, allow non-local antecedents</p><p>(as well as local), and require the antecedent to be a subject, e.g. ziji in Chinese;</p><p>zibun in Japanese.</p><p>While accounts differ in their details, they share the central insight that both</p><p>phrasal and head anaphors undergo LF movement (Cole, Hermon and Sung 1990;</p><p>2.3 Problems for the UG claim: wild grammars 45</p><p>Katada 1991; Pica 1987; Reinhart and Reuland 1993, amongst others). Differences</p><p>in domain and orientation fall out from the categorial differences between the</p><p>reflexives: X0 reflexives raise by head movement to Infl (itself a head). In complex</p><p>sentences, they can move out of the clause in which they originate, raising from</p><p>one Infl to another. In consequence, they can be interpreted with a long-distance</p><p>antecedent in a higher clause; this antecedent must be a subject because only a</p><p>subject will c-command the reflexive in Infl. Xmax reflexives, on the other hand,</p><p>are maximal projections which can adjoin only to the nearest maximal projection,</p><p>namely the VP in which they originate. There they remain in the binding domain</p><p>of either a local subject or a local object. In general, then, current accounts of</p><p>reflexives agree on the following properties:</p><p>(13) a. Long-distance anaphors must be subject-oriented</p><p>b. Anaphors which allow non-subject antecedents must be local</p><p>It is in this context that there has been extensive consideration ofwhatwould con-</p><p>stitute a wild grammar (Christie and Lantolf 1998; Hamilton 1998; Thomas 1995;</p><p>White, Hirakawa and Kawasaki 1996; Yuan 1998). Christie and Lantolf (1998)</p><p>hypothesize that a UG-constrained grammar will show a correlation between do-</p><p>main and orientation: if the interlanguage grammar allows long-distance reflex-</p><p>ives, then reflexives will be subject-oriented; if the interlanguage grammar has</p><p>local reflexives, then non-subject antecedents will be permitted. If interlanguage</p><p>grammars are not UG-constrained, on the other hand, there will be a breakdown</p><p>in these relationships.</p><p>Christie and Lantolf (1998) investigate the acquisition of reflexives in L2 Chi-</p><p>nese (which has a long-distance reflexive, ziji) and L2 English.9 (See box 2.5.) In</p><p>the case of L2 Chinese, the learners are speakers of English, a language which only</p><p>allows local reflexives. In the case of L2 English, learners are native speakers of</p><p>Chinese or Korean, both of which have long-distance reflexives. (These languages</p><p>also have phrasal reflexives which behave like English himself/herself .)</p><p>In order to understand what might be going on in the case of reflexives, it is</p><p>necessary to establish how L2 learners interpret certain sentences, whether, for</p><p>example, an English sentence such as (9b), repeated here as (14), has only one</p><p>possible interpretation in the interlanguage grammar (with Susan as the antecedent</p><p>of herself ), as is the case in English, or whether it is ambiguous (with either Mary</p><p>or Susan as the antecedent of herself ), as it would be in Japanese or Chinese.</p><p>(14) Mary thought that Susan blamed herself.</p><p>Christie and Lantolf developed a truth-value-judgment task for this purpose. In a</p><p>truth-value-judgment task, a particular sentence is paired with a particular context</p><p>which is provided by a story, a picture, or a scenario acted out in front of the</p><p>subjects (on video, for example). Subjects have to indicate whether the sentence</p><p>46 Principles of Universal Grammar in L2 acquisition</p><p>Box 2.5 Reflexives (Christie and Lantolf 1998)</p><p>Languages: L1 = English, L2 = Chinese and L1 = Chinese/Korean,</p><p>L2 = English.</p><p>Task: Truth-value judgments. Subjects indicate whether a sentence provides a</p><p>true statement about what is going on in an accompanying picture.</p><p>Sample stimuli:</p><p>Local domain (true) Non-local domain (false)</p><p>(Picture showing Grover hitting Grover)</p><p>Bert says that Grover is hitting himself on</p><p>the head.</p><p>(Picture showing Grover hitting Bert)</p><p>Bert says that Grover is hitting himself</p><p>on the head.</p><p>Subject-oriented (true) Object-oriented (true)</p><p>(Picture showing a book about Big Bird)</p><p>Big Bird is giving Oscar a book about</p><p>himself.</p><p>(Picture showing a book about Oscar)</p><p>Big Bird is giving Oscar a book about</p><p>himself.</p><p>is true in the context provided. In Christie and Lantolf’s tasks, pictures were used</p><p>(see box 2.5 for example stimuli). The advantage of such tasks is that subjects</p><p>are not being asked to make explicit grammaticality judgments as to the form of</p><p>the sentences. Indeed, all sentences in a truth-value-judgment task are generally</p><p>grammatical under some interpretation. Rather, subjects are ostensibly being asked</p><p>something about the meaning of the sentence. Nevertheless, their judgments reveal</p><p>something about the form of the grammar, in this case the range of interpretations</p><p>the grammar permits for reflexives.</p><p>As Christie and Lantolf recognize, the crucial issue is the grammar of the indi-</p><p>vidual; in other words, for any particular individual, is the grammar UG-consistent</p><p>or is it wild? (For related discussion, see also Eckman 1994; Thomas 1995; White</p><p>et al. 1996.) Results were analysed using cluster analysis, to see whether there is a</p><p>correlation between domain and orientation in individual grammars. Christie and</p><p>Lantolf found no evidence for such clustering in the case of either L2. Somewhat</p><p>surprisingly (given their assumptions), the control groups (Chinese and English</p><p>native speakers) also failed to show evidence of clustering. It is hard to argue that</p><p>L2 learners do (or do not) have wild grammars if even native speakers behave in</p><p>a way that suggests their grammars are wild, a problem that Christie and Lantolf</p><p>recognize.</p><p>Furthermore, Christie and Lantolf’s assumption about clustering of domain and</p><p>orientation is questionable. As Thomas (1995, 1998) points out, the properties de-</p><p>scribed in (13) only carry a one-way implication. That is, while long-distance</p><p>2.3 Problems for the UG claim: wild grammars 47</p><p>anaphors must be subject-oriented, it is not the case that all subject-oriented</p><p>anaphors must allow long-distance antecedents. There are languages with subject-</p><p>oriented anaphors which are bound only in a local domain, for example, Japanese</p><p>zibun-zisin (Katada 1991) and the French reflexive clitic se (Pica 1987). There</p><p>is nothing in the LF movement account to prohibit this. Similarly, while a non-</p><p>subject antecedent implies a locally bound reflexive, there is no requirement that</p><p>all locally bound reflexives permit object antecedents.</p><p>As Thomas discusses, Christie and Lantolf</p><p>incorrectly presuppose a two-way</p><p>implication between domain and orientation. They assume that a local binding</p><p>domain necessarily implies the possibility of non-subject antecedents, leading</p><p>them to the conclusion that L2 learners had arrived at illicit grammars. However,</p><p>this is not in fact the case. Consider, for example, L2 learners of English who</p><p>permit only local antecedents for reflexives like himself and herself , thus having</p><p>the domain right, but also reject object antecedents, thus having orientation wrong.</p><p>Such a grammar would fail to show the clustering relationship that Christie and</p><p>Lantolf expected. Nevertheless, the grammar would in fact be perfectly legitimate,</p><p>although not the appropriate grammar for English – this is precisely how Japanese</p><p>zibun-zisin behaves, for example.</p><p>In addition, as Thomas points out, numerous investigations of the acquisition</p><p>of English reflexives have found that learners and native speakers have a strong</p><p>preference for subject antecedents even where object antecedents are possible.</p><p>For example, given a sentence like John showed Bill a picture of himself , native</p><p>speakers are more likely to interpret John (the subject) as the antecedent of the</p><p>reflexive. In certain contexts, L2 learners and native speakers may reject interpre-</p><p>tations which their grammars in fact permit. This makes the results of the cluster</p><p>analysis even harder to interpret, since such preferencesmay conceal the full extent</p><p>of grammatical knowledge.10</p><p>Indeed, Thomas (1995) suggests that languageswhich do not have long-distance</p><p>reflexives, such as English, should not be used for investigating whether do-</p><p>main and orientation are correlated. Instead, one should concentrate on L2s with</p><p>long-distance reflexives, such as Chinese or Japanese. But here, too, determin-</p><p>ing whether learners arrive at wild grammars is not as straightforward as Christie</p><p>and Lantolf imply. If learners accept long-distance antecedents for reflexives, thus</p><p>getting the domain right, and also accept objects as antecedents, this is not nec-</p><p>essarily indicative of a wild grammar. It depends crucially on whether the object</p><p>antecedent is found within a local domain or whether it is a long-distance object.</p><p>In other words, there is an important distinction between (15a), where the object</p><p>antecedent is in a local domain, and (15b), a biclausal sentence, where the reflex-</p><p>ive is construed as having a non-subject antecedent outside the clause in which it</p><p>appears.</p><p>48 Principles of Universal Grammar in L2 acquisition</p><p>(15) a. The nurse asked the patienti about herselfi</p><p>b. *Black-san ga Whitei-san ni [Grey-san ga zibuni o</p><p>Black-Mr nom White-Mr dat Grey-Mr nom self acc</p><p>mita to] iimashita</p><p>saw that said</p><p>‘Mr Black said to Mr White that Mr Grey saw himself.’</p><p>Long-distance reflexives in some languages (Icelandic sig or Serbo-Croatian</p><p>sebe) can take a local subject or object as antecedent; non-local antecedents must</p><p>always be subjects, however. In other words, equivalents of (15a) are grammatical,</p><p>while interpretations like (15b) are never possible. Evidence for a wild grammar,</p><p>then, would be provided if non-local non-subjects were considered acceptable</p><p>antecedents for a reflexive, in other words if sentences equivalent to (15b) were</p><p>permitted. In this case, there would be a violation of the requirement that the</p><p>antecedent of the reflexive must c-command the reflexive at LF.</p><p>In their cluster analysis, Christie and Lantolf compare the possibility of local</p><p>object antecedents (such as Chinese equivalents of (15a)) with long-distance</p><p>subject antecedents (such as Chinese equivalents of (10)). Finding that both are</p><p>accepted, they conclude, incorrectly, that the grammar is wild. (Eckman (1994)</p><p>does the same, and draws the same conclusion.) In fact, Christie and Lantolf only</p><p>had one test item relevant to the issue of long-distance objects, i.e. only one item</p><p>similar to (15b). Some learners accepted the long-distance non-subject antecedent</p><p>in this case, which makes it appear that their interlanguage grammars might in-</p><p>deed not be UG-constrained. However, given the fact that only one sentence was</p><p>involved, there is considerable doubt as to the validity and generalizability of this</p><p>result.11</p><p>Thomas (1995) overcomes these shortcomings by investigating the L2 acqui-</p><p>sition of the Japanese reflexive zibun, by learners at low and high levels of profi-</p><p>ciency. Again, the task was a truth-value-judgment task (see box 2.6). Crucially,</p><p>stimuli included four biclausal sentences with long-distance objects as potential</p><p>antecedents, similar to (15b).</p><p>Table 2.6.1 presents group results, where it can be seen that long-distance object</p><p>antecedents are indeed accepted by low-proficiency learners, to about the same</p><p>extent that local object antecedents are accepted. The results on local objects,</p><p>though inappropriate for Japanese, do not provide evidence of a wild grammar</p><p>but the acceptance of long-distance object antecedents is problematic. Thomas</p><p>provides a further analysis in terms of individual performance, since it is properties</p><p>of the individual grammar that are at issue. Using a criterion of three or four</p><p>responses (out of four) to determine individual consistency,12 Thomas concentrates</p><p>on the twenty-three learners who consistently accept non-local antecedents for</p><p>zibun, excluding from consideration those who treat zibun only as a locally bound</p><p>2.3 Problems for the UG claim: wild grammars 49</p><p>Box 2.6 Reflexives (Thomas 1995)</p><p>Languages: L1 = English, L2 = Japanese.</p><p>Task: Truth-value judgments. Contexts provided by written scenarios (3–5 sen-</p><p>tences), illustrated by pictures. Each scenario is followed by a statement. Subjects</p><p>indicate whether the statement is true.</p><p>Sample stimuli:</p><p>Biclausal: LD subject (true) Biclausal: LD object (false)</p><p>Scenario: B hits A Scenario: A likes B’s book</p><p>A wa B ga zibun o butta to iimasita</p><p>(A said that B hit self)</p><p>C wa B ni A ga zibun no hon ga suki da</p><p>to iimasita</p><p>(C told B that A likes self’s book)</p><p>Monoclausal: local subject (true) Monoclausal: local object (false)</p><p>Scenario: A describes A’s problems to B Scenario: A asks B questions about B</p><p>A wa B ni zibun no mondai ni tuite</p><p>hanasimasita</p><p>A wa B ni zibun no koto ni tuite iroiro na</p><p>situmon o simasita</p><p>(A spoke with B about self’s problems) (A asked B various questions concerning</p><p>self )</p><p>Results:</p><p>Table 2.6.1 Responses of true (in %)</p><p>LD subject *LD object Local subject *Local object</p><p>(# = 4) (# = 4) (# = 4) (# = 4)</p><p>L2 groups</p><p>Low (n = 34) 55 54 85 47</p><p>Advanced (n = 24) 57 14 96 17</p><p>Native speakers (n = 34) 89 18 93 13</p><p>reflexive. The question, then, is whether these learners have unconscious knowl-</p><p>edge of the requirement for non-local antecedents to be subjects.While themajority</p><p>(70%) showed appropriate orientation, 30% of the learners (six in the low-</p><p>proficiency group and one advanced) consistently allowed long-distance object an-</p><p>tecedents (accepting interpretations like (15b)), a property not permitted in natural</p><p>language.</p><p>Other researchers have also reported acceptance of binding to long-distance</p><p>objects by L2 learners and controls, though to a much lesser extent. White et al.</p><p>(1996) report for L2 Japanese that one learner (out of thirteen) consistently</p><p>50 Principles of Universal Grammar in L2 acquisition</p><p>accepted long-distance object antecedents, as did one native speaker (out of ten).</p><p>Thomas (1995) also found that 10%of her Japanese native speaker controls showed</p><p>the same ‘wild’ behaviour. For L2 Chinese, Yuan (1998) found that three learn-</p><p>ers out of fifty consistently accepted long-distance objects, while no controls</p><p>did so.</p><p>It is not unreasonable to assume a certain amount of ‘noise’ in the data; perfor-</p><p>mance at 100% accuracy is unusual in any experimental attempts to get at linguistic</p><p>competence. Investigation of binding principles is no exception. However, while</p><p>performance factors may provide a reasonable explanation of the occasional native</p><p>speaker or L2 learner who deviates from expected behaviour, it is unlikely that</p><p>noisy data can provide an adequate</p><p>account of the performance of Thomas’s sub-</p><p>jects, where 30% of those who have a long-distance reflexive in L2 Japanese also</p><p>allow its antecedent to be a non-local object. Thomas herself assumes that there is a</p><p>competence issue here; she suggests that these subjects have misanalysed zibun as</p><p>a pronoun rather than a reflexive.13 As such, it can take any non-local antecedent.</p><p>In other words, these L2 learners adopt an alternative analysis, which happens to</p><p>be inappropriate for the L2, such that their grammars nevertheless fall within the</p><p>bounds of UG.</p><p>This solution in turn raises the issue of falsifiability. By changing the analysis</p><p>of zibun from anaphor to pronoun, the prospect of a wild interlanguage grammar</p><p>has been avoided. Does this mean that one can always change the analysis in</p><p>order to avoid such problems? It is important to understand that the alternative</p><p>proposal itself makes testable predictions. In order to explore further the pos-</p><p>sibility that zibun is a pronoun in the interlanguage grammar of at least some</p><p>learners, one would have to include additional sentence types which investigate</p><p>how learners treat pronouns in L2 Japanese. Hyams and Sigurjonsdottir (1990)</p><p>raised a similar possibility for L1 acquisition of Icelandic, namely that children</p><p>treat the reflexive sig as a pronoun; they rejected this possibility precisely be-</p><p>cause there turned out to be differences in the way children treated other pronouns</p><p>and sig.</p><p>Another possibility is suggested by Hamilton (1998), who argues that English-</p><p>speaking learners of Japanese may be treating reflexives as logophoric, i.e.</p><p>anaphors which are exempt from binding principles and which can be bound</p><p>non-locally within the discourse (e.g. A picture of myself would be nice on that</p><p>wall). (See Reinhart and Reuland (1993) for a recent treatment of logophoricity,</p><p>including constraints on when an anaphor can be logophoric.) Again, this would</p><p>mean that these interlanguage grammars are in fact licit. And, again, this would</p><p>have to be demonstrated independently with relevant stimuli, as Hamilton (1998)</p><p>tries to do for L2 English.</p><p>2.3 Problems for the UG claim: wild grammars 51</p><p>To sum up, as far as reflexive binding is concerned, it has been suggested by</p><p>Christie and Lantolf (1998) that L2 learners arrive at grammars where domain and</p><p>orientation fail to cluster, with a result that their grammars may be wild rather than</p><p>UG-constrained. Thomas (1995) notes, however, that this claim rests on a mis-</p><p>conception of what the permissible and impermissible grammars are. Controlling</p><p>for this by examining only the issue of whether a long-distance reflexive permits</p><p>a non-local non-subject antecedent (which is illicit), Thomas finds that the major-</p><p>ity of learners have UG-constrained grammars of anaphora. However, a minority</p><p>appears to have a wild grammar. Thomas (1995) and Hamilton (1998) offer alter-</p><p>native analyses of this behaviour which suggest that the interlanguage is, after all,</p><p>UG-constrained, though not the appropriate grammar for Japanese.</p><p>2.3.2 Null prep</p><p>Another example of a purported wild interlanguage grammar is provided</p><p>by Klein (1993b, 1995a). Klein investigates a phenomenon which she calls null</p><p>prep, whereby L2 learners of English omit prepositions in obligatory contexts.</p><p>English has a number of prepositional verbs (verbs taking PP complements), where</p><p>the preposition cannot be deleted, as shown in (16a, b). The preposition also appears</p><p>in questions, as in (16c, d), as well as relatives (16e, f), whether these involve</p><p>preposition stranding (16c, e) or the more formal pied-piping (16d, f). It is not</p><p>possible to omit the preposition in such contexts, as shown in (17). It is omission</p><p>of the kind illustrated in (17) which Klein terms null prep.</p><p>(16) a. The student is worrying about the exam.</p><p>b. *The student is worrying the exam.</p><p>c. Which exam is the student worrying about?</p><p>d. About which exam is the student worrying?</p><p>e. Here’s the exam that the student is worrying about.</p><p>f. Here’s the exam about which the student is worrying.</p><p>(17) a. *Which exam is the student worrying?</p><p>b. *Here’s the exam that the student is worrying.</p><p>Although null prep is not possible in English, it is found in certain languages,</p><p>such as Brazilian Portuguese, several dialects of Spanish and French, as well as</p><p>Haitian Creole. However, it is relatively rare and, according to Klein’s (1993b)</p><p>review of a number of languages that exhibit null prep, it is found only in relative</p><p>clauses (equivalents of (17b)), being prohibited in wh-questions (equivalents of</p><p>(17a)). The examples in (18), from Haitian Creole (Klein 1995a) illustrate this</p><p>point:</p><p>52 Principles of Universal Grammar in L2 acquisition</p><p>(18) a. Twa zanmi-yo ap pale de sinema sa a.</p><p>three friend-p are talking about movie this top</p><p>‘The three friends are talking about this movie.’</p><p>b. Men sinema que twa zanmi-yo ap pale a.</p><p>here movie that three friend-p are talking top</p><p>‘Here is the movie that the three friends are talking (about)’</p><p>c. *Ki sinema twa zanmi-yo ap pale a?</p><p>what movie three friend-p are talking top?</p><p>What movie are the three friends talking (about)?</p><p>Klein observes that relative clauses in languages allowing null prep show char-</p><p>acteristics which suggest that they are not derived by syntactic movement: (i) in</p><p>lieu of a null prep, relative clauses can contain an overt resumptive PP, consisting</p><p>of a pronoun with a preposition cliticized to it, as shown in (19a) (from a dialect</p><p>of Greek) – resumptives in general are characteristic of lack of movement (Sells</p><p>1984); (ii) relative clauses are introduced by complementizers rather than relative</p><p>pronouns, as shown in (19b, c), from Brazilian Portuguese – relative pronouns are</p><p>characteristic of movement.</p><p>(19) a. to rafio pou douleve (s’afto) ine mikro.</p><p>the office that he-works (in-it) it-is small</p><p>‘The office that he works (in) is small.’</p><p>b. a mô que eu falei</p><p>the girl that I spoke</p><p>‘The girl that I spoke (to)’</p><p>c. *a mô quem eu falei</p><p>the girl who I spoke</p><p>‘The girl who I spoke (to)’</p><p>Klein analyses null-prep relatives as containing a null resumptive PP, which</p><p>alternateswith an overt resumptive PP.14 Null prep is not permitted inwh-questions</p><p>or relative clauses derived bymovement (as in English), on the other hand, because</p><p>this would constitute an ECP violation: the null preposition would be unable to</p><p>properly govern the empty category resulting from wh-movement.</p><p>Given such restrictions, if null prep were to be found in interlanguage grammars</p><p>either in relative clauses derived by movement or in wh-questions, this would</p><p>constitute evidence of a wild grammar, violating the ECP. Previous research has</p><p>reported sporadic use of null prep in L2 (Bardovi-Harlig 1987; Mazurkewich</p><p>1984a). Klein devised a series of experiments to investigate whether null prep is</p><p>in fact a robust phenomenon in interlanguage grammars, and whether it occurs</p><p>illicitly, i.e. in movement contexts.</p><p>Klein (1993b, 1995a) tested adult learners of English from a variety of language</p><p>backgrounds and at different levels of proficiency. Learners were asked to judge</p><p>the grammaticality of sentences lacking prepositions and to correct them if they</p><p>2.3 Problems for the UG claim: wild grammars 53</p><p>Box 2.7 Null prep (Klein 1995a)</p><p>Languages: L1s = various, L2 = English.</p><p>Task: Grammaticality judgments and corrections.</p><p>Sample stimuli (ungrammatical):</p><p>Declarative: The delivery boy applied a new job last week</p><p>Relative: This is the job which/that/Ø the delivery boy applied last week</p><p>Question: Which job did the delivery boy apply last week?</p><p>Results:</p><p>Table 2.7.1 Acceptances of null prep in questions and relatives (in %)</p><p>Questions (# = 9) Relatives (# = 9)</p><p>L2 groups Beginners (n = 55) 69 78</p><p>Intermediate (n = 66) 52 57</p><p>Advanced (n = 75) 30 35</p><p>Native speakers (n = 40) 1 2</p><p>considered them to be wrong. (See box 2.7.) In other words, they were expected</p><p>to insert the missing preposition. As Klein points out, it is crucial to establish</p><p>learners’ knowledge of the subcategorization properties of the verbs in question.</p><p>For example,</p><p>for L2 parameter resetting: more on Verb Movement 162</p><p>5.6 A role for negative evidence in triggering? 164</p><p>5.7 Triggering in L2: manipulating the input 166</p><p>5.7.1 Manipulating the input: assessment 169</p><p>5.8 Beyond explicit teaching 171</p><p>5.9 Conclusion 175</p><p>Topics for discussion 175</p><p>Suggestions for additional reading 176</p><p>6. Morphological variability and the morphology/syntax</p><p>interface 178</p><p>6.1 Morphological variability: identifying the problem 178</p><p>6.2 Surface versus abstract morphology 180</p><p>6.3 Accounts of morphological variability in L1 acquisition 181</p><p>6.3.1 Morphology-before-syntax 182</p><p>6.3.2 Syntax-before-morphology 182</p><p>6.4 Perspectives on the morphology/syntax interface in L2 184</p><p>6.4.1 Morphology-before-syntax: incompleteness and deficits 184</p><p>6.4.2 Syntax-before-morphology: the data 187</p><p>6.4.3 Missing surface inflection: explanations 193</p><p>6.5 Methodological considerations 199</p><p>6.6 The morphology/syntax interface: conclusion 201</p><p>Topics for discussion 201</p><p>Suggestions for additional reading 202</p><p>7. Argument structure 203</p><p>7.1 Argument structure 203</p><p>7.2 Lexical entries 203</p><p>7.3 Mapping from lexicon to syntax: the logical problem of</p><p>argument-structure acquisition 205</p><p>7.4 Semantic constraints on argument-structure alternations 206</p><p>7.5 Crosslinguistic differences in conflation patterns 212</p><p>7.5.1 Conflation patterns in L2 motion verbs 213</p><p>7.5.2 Lexical parameters and conflation 218</p><p>7.6 Thematic properties of arguments and their syntactic</p><p>consequences 223</p><p>7.6.1 Thematic hierarchies, UTAH and psych verbs 224</p><p>7.6.2 The Unaccusative Hypothesis 228</p><p>x Contents</p><p>7.7 Transitivity alternations and effects of argument-changing</p><p>morphology 234</p><p>7.8 Methodological considerations 238</p><p>7.9 Argument structure: conclusion 238</p><p>Topics for discussion 239</p><p>Suggestions for additional reading 240</p><p>8. Ultimate attainment: the nature of</p><p>the steady state 241</p><p>8.1 Introduction 241</p><p>8.2 Convergence versus divergence 242</p><p>8.3 How to identify an endstate grammar 244</p><p>8.4 Age effects on ultimate attainment 245</p><p>8.4.1 Violations of Subjacency 245</p><p>8.4.2 Subjacency violations: a reanalysis 248</p><p>8.5 Age effects in near-native speakers 249</p><p>8.6 Convergence or not: more on near-native speakers 252</p><p>8.7 Non-UG structures revisited 254</p><p>8.8 Divergence: L1 influence 258</p><p>8.9 Non-native ultimate attainment: optionality revisited 263</p><p>8.10 Summary: endstate competence 266</p><p>8.11 Conclusion: initial to steady state 266</p><p>Topics for discussion 271</p><p>Suggestions for additional reading 271</p><p>Glossary 273</p><p>Notes 283</p><p>References 288</p><p>Index 312</p><p>Preface</p><p>This book examines the extent to which the underlying linguistic compe-</p><p>tence of learners or speakers of a second language (L2) is constrained by the same</p><p>universal principles that govern natural language in general. It is presupposed that</p><p>there is an innately given Universal Grammar (UG), which constrains first lan-</p><p>guage (L1) grammars, placing limits on the kinds of hypotheses that L1 acquirers</p><p>entertain as to the nature of the language that they are acquiring. Assuming the cor-</p><p>rectness of this general approach, the question arises as to whether UG constrains</p><p>grammars in non-primary language acquisition as well. This book will present and</p><p>discuss research which investigates whether or not interlanguage grammars can be</p><p>characterized in terms of principles and parameters of UG, and which explores the</p><p>nature of interlanguage competence during the course of L2 acquisition, from the</p><p>initial state onwards. It is hoped that the book will provide sufficient background</p><p>for the reader to understand current research conducted within the framework of</p><p>UG and L2 acquisition.</p><p>The generative perspective on L2 acquisition is sometimes dismissed because</p><p>it has a rather circumscribed goal, namely to describe and explain the nature of</p><p>interlanguage competence, defined in a technical and limited sense. Researchers</p><p>whose work is discussed in this book do not seek to provide an all encompassing</p><p>theory of L2 acquisition, or to account the role of performance factors, psycholog-</p><p>ical processes and mechanisms, sociolinguistic variables, etc. In fact, it is doubtful</p><p>whether there is any one theory that can achieve all this; certainly, no theory has</p><p>succeeded so far.</p><p>It will be presupposed that the reader has some familiarity with the concepts and</p><p>mechanisms assumed in current generative grammar, including the Government</p><p>and Binding framework andMinimalism. The book will not be concerned with the</p><p>precise technical details as to how UG principles and parameters are formulated,</p><p>nor with the intricacies of current linguistic theory. Indeed, the intention is to</p><p>consider the L2 issues without being tied down to a particular version of generative</p><p>theory. The linguistic principles and parameters that will be discussed are those</p><p>that have attracted attention in the L2 field. Out of context, these principles may</p><p>sometimes seem ad hoc. It is important to understand that they are part of a system</p><p>xi</p><p>xii Preface</p><p>of knowledge, accounting for far more than whatever we happen to touch on in this</p><p>book. A list of abbreviations and a glossary are provided which give definitions of</p><p>the main linguistic and acquisition terminology used throughout the book.</p><p>This book is not intended to be a revised version of my earlier work (White</p><p>1989), which examined the first decade of research (conducted during the 1980s)</p><p>on UG and L2 acquisition, looking at claims for the availability of principles and</p><p>parameters of UG. There has been an enormous increase in research conducted</p><p>within this general framework since that time and it is not possible to do justice</p><p>to all of it. The current work takes a somewhat different perspective, a perspective</p><p>which is more representative of research conducted during the 1990s. The book</p><p>is organized as follows: chapter 1 provides a general introduction to UG and the</p><p>logical problem of language acquisition; chapter 2 considers the logical problem</p><p>of L2 language acquisition and the issue of whether principles of UG constrain in-</p><p>terlanguage grammars; chapter 3 examines hypotheses as to the nature of the initial</p><p>state (the L2 learner’s earliest assumptions about the L2), including the influence of</p><p>the L1 grammar; chapter 4 looks at the issue of developing grammars in the con-</p><p>text of parameters and parameter resetting; chapter 5 considers what properties</p><p>of the L2 input might stimulate grammar change; chapter 6 investigates disso-</p><p>ciations between morphology and syntax in interlanguage grammars; chapter 7</p><p>explores the nature of argument structure and the influence of the L1 on argument</p><p>structure representations; finally, in chapter 8 the nature of the ultimate attainment</p><p>of L2 learners is discussed. Each chapter ends with some suggestions for general</p><p>discussion, often on broader issues than those raised in the chapter in question, as</p><p>well as further reading.</p><p>Throughout the book, where experiments are described, the main details of</p><p>the experiment (including the languages involved, example stimuli, results, etc.)</p><p>are summarized in boxes, offset from the main text. In many cases, it has been</p><p>necessary to be selective in deciding which aspects of a particular experiment</p><p>to focus on, in order to fit with the general themes of the book. If this has led</p><p>to misrepresentation, I apologize! Readers are strongly encouraged to go to the</p><p>original sources for further details, especially if they are themselves intending to</p><p>pursue experimental research.</p><p>The terms L2 learner and L2 speaker are adopted as convenient cover terms</p><p>for non-native acquisition or the learning of any number of languages (L2, L3,</p><p>L4, Ln). No distinction will be made between second language acquisition and</p><p>foreign language learning. In principle at least, any kind of non-native acquisition</p><p>or learning should be subject to the same constraints, although lack of suitable input</p><p>may be a major inhibiting factor in certain foreign language learning contexts.</p><p>Many people have provided helpful input on the manuscript, at various stages.</p><p>For their thoughtful and detailed comments and</p><p>if a learner does not know that the verb worry requires a PP comple-</p><p>ment introduced by about, absence of the preposition does not constitute a case</p><p>of the null prep phenomenon but simply indicates a lack of knowledge of the sub-</p><p>categorization requirements of this particular verb. For this reason, the first step</p><p>in Klein’s analysis of the results was to concentrate on those declaratives which</p><p>learners identified as ungrammatical and corrected by the insertion of a preposition.</p><p>Such corrections indicate that learners are aware that the verbs in question require</p><p>a PP complement. Where the declarative was correctly identified as requiring a</p><p>preposition, Klein then looked at whether learners know that the preposition is also</p><p>obligatory in the corresponding questions and relatives. If questions and/or rela-</p><p>tives with missing prepositions are accepted while the corresponding declaratives</p><p>are rejected, this is taken to be evidence of illicit null prep.</p><p>Results show extensive acceptance of null prep forms in both wh-questions</p><p>and relative clauses, even at the highest proficiency level (see table 2.7.1); all</p><p>groups are significantly different from each other and from the controls. There is</p><p>a decline in acceptance of null prep with increasing level of proficiency. Klein’s</p><p>other studies (Klein 1993a, 1995b) suggest that null prep is systematic, occurring</p><p>across different age groups and L1s (including L1s with and without null prep), in</p><p>both relatives and wh-questions.</p><p>54 Principles of Universal Grammar in L2 acquisition</p><p>While null prep is permissible in relative clauses in certain languages, Klein</p><p>argues that it is always illicit in wh-questions. Since the learners she studied ac-</p><p>cepted null prep inwh-questions, and since her methodology rules out the possibil-</p><p>ity that this is just a general subcategorization error, the results appear to suggest</p><p>a wild interlanguage grammar, violating the ECP, even in the case of those of</p><p>advanced proficiency. In addition, if relative clauses are derived by movement,</p><p>their grammars are wild in this respect as well, since null prep is not permitted in</p><p>such cases (compare (18b and c)). However, Dekydtspotter, Sprouse and Ander-</p><p>sen (1998) have challenged Klein’s assumption that null prep is prohibited in the</p><p>case of structures involving wh-movement. They show that null prep is found in</p><p>wh-questions in Yoruba, as well as in popular French. Dekydtspotter et al. pro-</p><p>pose an alternative account of the null prep phenomenon in terms of preposition</p><p>incorporation, arguing that it obeys general constraints on incorporation, hence</p><p>suggesting that a null prep interlanguage grammar is not wild after all. (See Klein</p><p>(2001) for counter-arguments.)</p><p>2.4 Methodological issues</p><p>The experiments described in this chapter have used a variety of different</p><p>methodologies to try to determine the nature of the interlanguage grammar. Tasks</p><p>have included elicited production (translation), grammaticality and acceptabil-</p><p>ity judgments, coreference judgments, and truth-value judgments. The judgment</p><p>tasks fall into two distinct types, some concentrating on the form of sentences</p><p>(for example, Kanno’s (1996) study of case drop – see section 2.1.3) and some</p><p>concentrating on interpretation (for instance, the studies by Christie and Lantolf</p><p>(1998) and Thomas (1995) on reflexives – section 2.3.1). In the case of judgments</p><p>designed to probe learners’ interpretations of particular sentence types, we have</p><p>seen that it is crucial to provide some kind of context for the interpretation in ques-</p><p>tion. Even within the judgments directed at interpretation, some tasks are more</p><p>metalinguistic than others. For example, Kanno (1997, 1998b) requires subjects to</p><p>choose explicitly between potential antecedents for pronouns; Dekydtspotter et al.</p><p>(1997) ask learners whether a sentence ‘sounds possible’ in a given context. In</p><p>contrast, truth-value-judgment tasks merely require the learner to indicate whether</p><p>or not a particular sentence is true in a particular context. (However, see White,</p><p>Bruhn-Garavito, Kawasaki, Pater and Prévost (1997) for discussion of problems</p><p>associated with such tasks: even with this methodology, the effects of preferences</p><p>for certain interpretations over others cannot be fully eliminated.)</p><p>Another welcome trend is the move towards analysing data in terms of the per-</p><p>formance of individual subjects. Although group results can be quite informative,</p><p>2.4 Methodological issues 55</p><p>they can also be misleading, concealing properties of individual grammars. For</p><p>example, if a given group of L2 learners shows a 50% acceptance rate for some</p><p>structure, this could be due to half the subjects accepting all sentences and half</p><p>rejecting all, or it could be the result of all subjects accepting half of the sentences.</p><p>Since the claim that the interlanguage grammar is (or is not) UG-constrained is</p><p>a claim about individual linguistic competence, it is crucial to determine what is</p><p>going on at the individual level.</p><p>A further methodological advance concerns the use of control groups. It has</p><p>been fairly standard for many years to include native speakers of the L2 as controls.</p><p>Research described in this chapter has also included a more recent development,</p><p>namely the use of native speakers of the L1 (e.g. Dekydtspotter et al. 1997 – see</p><p>section 2.1.2; Kanno 1998b – see section 2.1.1). This allows the experimenter to</p><p>determine whether or not interlanguage performance could be accounted for in</p><p>terms of properties of the L1 grammar.</p><p>However, the use of native-speaker control groups, whether they are speakers of</p><p>the L1 or L2, raises the issue of the so-called comparative fallacy (Bley-Vroman</p><p>1983). Bley-Vroman remarked that ‘the learner’s system is worthy of study in its</p><p>own right, not just as a degenerate formof the target system’ (1983: 4). A number of</p><p>researchers have emphasized the need to consider interlanguage grammars in their</p><p>own right with respect to principles and parameters of UG (e.g. du Plessis, Solin,</p><p>Travis and White 1987; Finer and Broselow 1986; Martohardjono and Gair 1993;</p><p>Schwartz and Sprouse 1994; White 1992c; Lakshmanan and Selinker 2001). As</p><p>Birdsong points out with respect to grammaticality-judgment data: ‘the relevant</p><p>data are learners’ judgments – not their similarity to or deviance from natives’</p><p>judgments’ (1989: 119). This focus on the interlanguage grammar remains the</p><p>current perspective: much of the research described in this chapter (as well as later</p><p>in the book) is committed to this position. That is, the crucial question is whether</p><p>or not interlanguage grammars are UG-constrained, rather than whether or not</p><p>they are native-like.</p><p>Nevertheless, avoiding the comparative fallacy does not require the experimenter</p><p>to exclude native-speaker controls altogether. First of all, control groups are nec-</p><p>essary simply to ensure: (i) that the tasks devised by the experimenter in fact are</p><p>successful in testingwhat they are supposed to test; and (ii) that the facts in question</p><p>are indeed as the experimenter supposes them to be (or as claimed in the theoretical</p><p>linguistics literature). For example, in Christie and Lantolf’s study (section 2.3.1),</p><p>the performance of the native speakers was unexpected, given the researchers’</p><p>assumptions about what properties of reflexives should cluster. Secondly, there</p><p>are legitimate reasons for asking whether the learner in fact shows unconscious</p><p>knowledge of principles and parameters relevant to the L2. What is problematic is</p><p>when certain conclusions are drawn on the basis of failure to perform like native</p><p>56 Principles of Universal Grammar in L2 acquisition</p><p>speakers. Failure to acquire L2 propertiesmay nevertheless involve acquiring prop-</p><p>erties different from the L1, properties of other natural languages, properties that</p><p>are underdetermined by the L2 input. Such failure does not necessarily imply lack</p><p>of UG, as we have seen.</p><p>A related problem is how to interpret significant differences between native</p><p>speaker and L2 learner performance on some task. Consider, for</p><p>example, the re-</p><p>sults of the study by Dekydtspotter et al. (1997) on process and result nominals</p><p>(section 2.1.2.). Their beginner and intermediate subjects showed a significantly</p><p>higher acceptance rate of multiple de-phrases with process nominals (ungrammat-</p><p>ical in French). Thus, it might be claimed that their grammars are not constrained</p><p>by the relevant principles governing argument structure. It has been suggested</p><p>in this chapter that a more appropriate comparison is to look at whether certain</p><p>sentence types are treated significantly differently from other sentence types by</p><p>the same group of learners (c.f. Grimshaw and Rosen 1990). If such distinctions</p><p>are found, this suggests that the interlanguage grammar represents the relevant</p><p>distinction (whatever it may be), even if the degree to which they observe it differs</p><p>from native speakers. This was in fact the case in Dekydtspotter et al.’s study: all</p><p>groups distinguished between process and result nominals in terms of acceptability</p><p>of multiple de-phrases.</p><p>2.5 Conclusion</p><p>In conclusion, this chapter has considered evidence that interlanguage</p><p>grammars are constrained by principles of UG. While earlier research on UG</p><p>and L2 acquisition concentrated largely (though certainly not exclusively) on L2</p><p>English, many other L2s are now being investigated, including, in the experiments</p><p>described here, L2 Spanish, Japanese, French and Chinese. Results from several</p><p>experiments suggest that learners of a variety of L2s demonstrate unconscious</p><p>knowledge of subtle distinctions that are unlikely to have come from the L2 input</p><p>(including instruction) or from the L1, consistent with the claim that principles</p><p>of UG constrain interlanguage grammars. The claim that interlanguage grammars</p><p>are sometimes wild (hence, not falling within the bounds laid down by UG) has</p><p>also been considered. We have seen that analyses adopted by L2 learners may in</p><p>fact be true of natural language, even if they happen not to be appropriate for the</p><p>L1 or L2 of the learners in question. Of course, the reason why L2 learners should</p><p>arrive at such alternative analyses still requires explanation.</p><p>In the next chapter, a different issue is discussed, namely the nature of the initial</p><p>state in L2 acquisition. We will consider the extent to which the L1 grammar</p><p>determines properties of the early interlanguage grammar, and whether or not UG</p><p>constitutes the initial state, as in L1 acquisition (see chapter 1).</p><p>Suggestions for additional reading 57</p><p>Topics for discussion</p><p>� Is it ever possible to eliminate the L1 as a source of knowledge of princi-</p><p>ples ofUGor is this a non-issue?According toDekydtspotter et al. (1997),</p><p>if one adopts the Minimalist Program whereby computational principles</p><p>are universally the same and invariant, it no longer makes sense to distin-</p><p>guish between direct access to UG or indirect access via the L1.</p><p>� According to Schwartz and Sprouse (2000a, b), too much current L2</p><p>research uses linguistic theory to provide relatively sophisticated and</p><p>detailed analyses of interlanguage data, without considering the logical</p><p>problem of L2 acquisition. They suggest that such research does not help</p><p>us to understand the nature of L2 acquisition or interlanguage compe-</p><p>tence. In contrast, Hawkins (2001b) argues that many researchers are</p><p>overly preoccupied with the logical problem of L2 acquisition. Instead,</p><p>he suggests, a better way to reach an understanding of L2 acquisition is</p><p>to focus on differences between native speaker and L2 learner grammars</p><p>(described in terms of current linguistic constructs). Are these positions</p><p>incompatible?</p><p>� In this chapterwe have seen several exampleswhere problematic interlan-</p><p>guage data are reanalysed in terms of some other theory. To what extent</p><p>does this render the claim that UG constrains interlanguage grammars</p><p>unfalsifiable?</p><p>� Linguistic theory is constantly changing and undergoing development;</p><p>proposals as to the precise nature of UG have changed considerably over</p><p>the years. What are the implications for L2 acquisition research, and</p><p>particularly for theories that assume a role for UG? (See Schwartz and</p><p>Sprouse (2000b) and White (1995a) for discussion.)</p><p>� Davies and Kaplan (1998) and Lantolf (1990) advocate group grammat-</p><p>icality judgments, where learners (in pairs or groups) discuss test items</p><p>together, in order to arrive at decisions about their grammaticality. Why</p><p>is this approach problematic?</p><p>Suggestions for additional reading</p><p>� For more detailed discussion of the logical problem of L2 acquisition, see</p><p>Bley-Vroman (1990), White (1989: chapter 2) and Schwartz and Sprouse</p><p>(2000a, b).</p><p>3</p><p>The initial state</p><p>3.1 What is the initial state?</p><p>In chapter 2, it was argued that there is a logical problemof L2 acquisition.</p><p>Experimental evidence was reviewed which suggests that interlanguage grammars</p><p>allow the representation of subtle and abstract distinctions whose source could not</p><p>be theL1grammaror theL2 input, hencemust beUG. Inotherwords, interlanguage</p><p>representations are constrained by UG, conforming to principles such as the Overt</p><p>Pronoun Constraint and the ECP. While some researchers have proposed that</p><p>interlanguage grammars are ‘wild’, hence not fully UG-constrained, there are</p><p>alternative analyses of the phenomena in question which can accommodate the</p><p>potentially problematic data.</p><p>In this chapter, we turn to a different (though related) issue, namely the nature</p><p>of the initial state in L2 acquisition. The term initial state is variously used to</p><p>mean the kind of unconscious linguistic knowledge that the L2 learner starts out</p><p>with in advance of the L2 input and/or to refer to characteristics of the earliest</p><p>grammar. As Schwartz and Eubank (1996) point out, the interlanguage initial state</p><p>was a neglected topic until the mid 1990s. In earlier work on UG in L2 acquisition,</p><p>assumptions about the initial state were usually implicit. Even where they were</p><p>explicit, the initial state was not the main focus of research. For example, White</p><p>(1985b) proposed that L2 learners start out with L1 parameter settings. Although</p><p>not presented as such at the time, this is clearly an initial state claim, since it</p><p>presupposes that at least part of the L1 grammar (namely, L1 parameter settings)</p><p>determines how the learner initially approaches the L2 data. Rather than focus-</p><p>ing on the initial state, early research explored the question of whether different</p><p>stages of interlanguage development could be characterized as exemplifying dif-</p><p>ferent parameter settings and whether the L2 learner could achieve settings which</p><p>differ from the L1 settings. In other words, research addressed the question of</p><p>whether parameters can be reset and under what conditions; initial grammars as</p><p>such were rarely considered. More recently, a number of explicit hypotheses have</p><p>been advanced as to the nature of the initial state in L2 acquisition, which also</p><p>58</p><p>3.1 What is the initial state? 59</p><p>L2 PLD = L1 Ss</p><p>S0 ILG1 ILGn IL Ss</p><p>Figure 3.1 L2 acquisition without UG</p><p>make claims about the kind of development (or lack thereof) that can be expected</p><p>subsequently.</p><p>In L1 acquisition, UG is the initial state (Chomsky 1981b), determining, in</p><p>advance, the form and the functioning of language-particular grammars (see</p><p>chapter 1, section 1.2). While UG is the initial state (or S0), it is not entirely</p><p>clear what happens subsequently, that is, whether UG somehow ‘turns into’ a</p><p>particular steady-state grammar (SS) in the course of language acquisition or</p><p>whether it remains distinct from specific instantiations. Possibly because this</p><p>matter is of little consequence for researchers interested in L1 acquisition or in</p><p>native speaker competence, the issue has been relatively little discussed; where</p><p>it is discussed, the former assumption is often adopted. As DeGraff (1999: 15)</p><p>puts it: ‘L1A is the process by which exposure to PLD transforms the innately</p><p>specified experience-independent faculté de langage into a language-particular</p><p>grammar by assigning fixed values to parameter arrays specified by UG.’</p><p>In the context</p><p>of L2 acquisition, the question ofwhetherUGbecomes a particular</p><p>grammar or remains distinct from particular grammars is central. If UG is trans-</p><p>formed into a grammar which may subsequently be modified during the course</p><p>of acquisition (S0.... S1.... SS) then only the particular steady-state instantiation of</p><p>UG would remain available in non-primary language acquisition. Perhaps the first</p><p>person to raise this issue in the L2 context was Bley-Vroman (1990: 18–19), who</p><p>suggested the following computer analogy:</p><p>It is as if an application program camewith an installation-configuration program,</p><p>with which you set parameters to customize the application to your computer and</p><p>your tastes. You use this installation program just once, it sets up the application to</p><p>operate properly, often stripping it down, removing options your machine cannot</p><p>implement.Younever use the installation programagain. The application program</p><p>is now a particular program for your machine.</p><p>In otherwords,UG survives only as the language-specificmother-tongue grammar.</p><p>Bley-Vroman’s Fundamental Difference Hypothesis rests on the assumption that</p><p>UG as a distinct ‘entity’ does not survive L1 acquisition.1 On this view, the initial</p><p>state of L2 acquisition is, necessarily, the L1 grammar (L1 SS), as schematized in</p><p>figure 3.1. Subsequently, there may be development away from the L1 grammar,</p><p>until a steady state interlanguage grammar is attained (IL SS).</p><p>60 3 The initial state</p><p>Arguing against the position that UG becomes a language-specific grammar and</p><p>in favour of the position that UG remains constant and distinct from any particular</p><p>grammar, Flynn and Martohardjono (1994) and Epstein et al. (1996) point out that</p><p>bilingual first language acquisition would be hard to account for on the former</p><p>view, given that the two languages that a bilingual child is acquiring will often</p><p>require contradictory parameter settings. Since bilingual children are known to</p><p>acquire two distinct grammars (Meisel 1989; Müller and Hulk 2000; Paradis and</p><p>Genesee 1996), this suggests that UGmust be distinct from both grammars and that</p><p>it constrains both grammars. (See Schwartz (1987) and Cook (1991) for related</p><p>observations.)</p><p>All the initial-state proposals to be considered in this chapter presuppose the</p><p>following: UG is constant (that is, unchanged as a result of L1 acquisition); UG is</p><p>distinct from the learner’s L1 grammar; UG constrains the L2 learner’s interlan-</p><p>guage grammars. In spite of this common ground, there is considerable disagree-</p><p>ment over the nature of the interlanguage initial state.</p><p>Two logical possibilities will be considered here: the grammar of the mother</p><p>tongue (the L1) is the initial state or UG is the initial state. (It is of course con-</p><p>ceivable that neither UG nor the L1 constitutes the interlanguage initial state, an</p><p>alternative which will not be discussed.) It may be useful to think of the issue</p><p>by asking what unconscious preconceptions the learner has about the nature of</p><p>the L2. In advance of input, does the learner start out with a language-specific</p><p>grammar, namely, the L1 grammar? Alternatively, does the learner start with no</p><p>preconceptions other than the ‘blueprint’ provided by UG?</p><p>We first consider proposals that the initial state is indeed a specific grammar. In</p><p>particular, the L2 learner is assumed to start out with grammatical representations</p><p>derived from the L1 grammar, in whole or in part. Falling into this category are</p><p>the Full Transfer Full Access Hypothesis of Schwartz and Sprouse (1994, 1996),</p><p>the Minimal Trees Hypothesis of Vainikka and Young-Scholten (1994, 1996a, b)</p><p>and the Valueless Features Hypothesis of Eubank (1993/1994, 1994, 1996). These</p><p>proposals contrast with others where the interlanguage initial state is argued not</p><p>to be a particular grammar but rather UG itself, similar to the situation in L1</p><p>acquisition. Falling into this latter category are the Initial Hypothesis of Syntax</p><p>(Platzack 1996), where this claim is explicit, and the Full Access Hypothesis of</p><p>Epstein et al. (1996, 1998), where it is implicit. It is important to understand</p><p>that all the hypotheses to be considered here presuppose that UG constrains in-</p><p>terlanguage grammars, although some accounts imply an impairment to certain</p><p>UG-related domains, as we shall see. In other words, the fact that the L2 learner</p><p>may start off by adopting a particular grammatical representation (based on the</p><p>L1) does not preclude UG-constrained changes in response to properties of the</p><p>L2 input.</p><p>3.2 A grammar as the initial state 61</p><p>3.2 A grammar as the initial state</p><p>3.2.1 The Full Transfer Full Access Hypothesis</p><p>Webeginwith an examination of the Full Transfer FullAccessHypothesis</p><p>of Schwartz and Sprouse (1994, 1996), according to which the initial state in L2</p><p>acquisition is a particular grammar. Faced with accounting for L2 input, learners</p><p>adopt the grammar that they already have, the steady-state grammar of the mother</p><p>tongue. In contrast to other researchers who argue for less than total involvement of</p><p>the L1 (see sections 3.2.2 and 3.2.3), Schwartz and Sprouse propose full transfer:</p><p>the entire L1 grammar (in the sense of all abstract properties but excluding specific</p><p>lexical items) constitutes the initial state. Furthermore, it is hypothesized that</p><p>changes to the initial grammar can take place; in other words, the learner is not</p><p>‘stuck’ with representations based on the L1 steady state. When the L1 grammar is</p><p>unable to accommodate properties of the L2 input, the learner has recourse to UG</p><p>options not instantiated in the L1, including new parameter settings, functional</p><p>categories and feature values, in order to arrive at an analysis more appropriate</p><p>to the L2 input, although this may turn out not to be the same analysis as that</p><p>found in the native-speaker grammar. The resulting interlanguage grammars are</p><p>UG-constrained, hence, the term full access. Full transfer, then, is Schwartz and</p><p>Sprouse’s claim about the initial state; full access is their claim about subsequent</p><p>grammar restructuring during the course of development. Full Transfer Full Access</p><p>is schematized in figure 3.2 (adapted from White (2000)).</p><p>3.2.1.1 Full Transfer Full Access: evidence</p><p>Two kinds of evidence serve to support the claims of Full Transfer Full</p><p>Access: (i) evidence of L1 properties in the interlanguage grammar; (ii) evidence</p><p>of restructuring away from the L1 grammar. A case study by Haznedar (1997) sup-</p><p>ports Full Transfer Full Access, providing evidence of an L1-based initial state, as</p><p>well as subsequent changes to the interlanguage grammar. Haznedar examines</p><p>spontaneous production data gathered from a Turkish-speaking child, named</p><p>UG</p><p>L2 PLD = L1 Ss</p><p>S0 IL SsILG1 ILGn</p><p>Figure 3.2 Full Transfer Full Access</p><p>62 3 The initial state</p><p>Erdem, who was learning English. Erdem was initially interviewed at the age of 4,</p><p>after three months in England. For the first two months he had been almost</p><p>entirely in a Turkish-speaking environment at home; thereafter, he spent a month</p><p>in an English nursery school. Thus, the data that Haznedar reports are relevant to</p><p>the initial state.</p><p>Turkish and English differ as to word order, particularly headedness of both</p><p>lexical (in this case, VP) and functional (NegP) projections: Turkish is verb final</p><p>while English is verb initial and Turkish is Neg final while English is Neg initial.</p><p>Haznedar reports that, for the first three months of recording, Erdem consistently</p><p>(almost 100%) produced head-final word order, suggesting transfer of Turkish</p><p>headedness. For example, he would produce utterances like (1a) (head-final VP)</p><p>and (1b) (head-final NegP). In the fourth month, Erdem switched headedness of</p><p>both VP and NegP to their English values, now consistently producing head-initial</p><p>utterances like (1c) and (1d).</p><p>(1) a. I something eating.</p><p>b. Finish no.</p><p>c. You eating apple.</p><p>d. I not eat cornflake.</p><p>Although the data are somewhat limited, due to the fact that there were rel-</p><p>atively few relevant utterances in the first three months of recording, they are</p><p>nevertheless</p><p>highly suggestive, supporting both components of Full Transfer Full</p><p>Access: Erdem’s initial grammar clearly exhibits Turkish word order but he also</p><p>successfully switches to English order.</p><p>Haznedar’s (1997) study is a case study, involving spontaneous production data.2</p><p>Case studies have the advantage of following an individual or individuals over</p><p>time, so that changes in the interlanguage grammar can be observed. However,</p><p>there are disadvantages to relying on case studies and there are disadvantages to</p><p>relying on production data alone. Firstly, one cannot be sure that case studies are</p><p>representative of anything more than the behaviour of the individuals in question.</p><p>Secondly, it is not clear to what extent spontaneous production data accurately</p><p>reflect properties of the underlying grammar. In particular, if certain forms are</p><p>absent in spontaneous production, this does not necessarily reflect absence of</p><p>some corresponding abstract grammatical category, a point that will be considered</p><p>in more detail in section 3.2.2.2, as well as in chapter 6. However, in these data</p><p>from Erdem, this problem does not arise since he did initially produce word orders</p><p>clearly based on his L1 rather than on the L2; in other words, nothing is being</p><p>inferred about the grammar based on absence of some phenomenon.</p><p>Experimental data provide another way of exploring the claims of the Full</p><p>Transfer Full Access Hypothesis. One possibility is to conduct experiments with</p><p>learners of a particular L1, looking for evidence of properties of that L1 in the</p><p>3.2 A grammar as the initial state 63</p><p>interlanguage grammar. But more compelling evidence can be provided by con-</p><p>sidering learners of different L1s acquiring the same L2 (Schwartz and Sprouse</p><p>1996, 2000b). Such learners are predicted to behave differently, reflecting different</p><p>initial states. Indeed, some of the earliest experimental research on L2 parameter</p><p>setting is relevant in this context. White (1985b, 1986), for example, showed</p><p>that French-speaking and Spanish-speaking learners of English behaved differ-</p><p>ently with respect to null subjects in English. In grammaticality-judgment tasks,</p><p>Spanish speakers were significantly more likely to accept null subjects in English</p><p>than French speakers were. This differential behaviour based on properties of the</p><p>L1 (Spanish but not French being a null subject language) supports Full Transfer</p><p>Full Access. That is, the difference in behaviour with respect to null subjects</p><p>reflects different L2 initial states (and, as yet, no subsequent restructuring).</p><p>More recent examples of experimental research supporting Full Transfer Full</p><p>Access are provided by Yuan (1998) and Slabakova (2000). In both these studies,</p><p>learners of different L1s show distinctly different behaviour with respect to the</p><p>linguistic properties under investigation, suggesting different representations, by</p><p>inference due to different initial states.</p><p>Yuan (1998) investigated acquisition of the long-distance reflexive ziji in L2</p><p>Chinese, considering both domain (long-distance versus local antecedents) and</p><p>orientation (subject versus object antecedents) (see chapter 2, section 2.3.1). Here</p><p>we focus on domain only. (See box 3.1.)</p><p>Box 3.1 Full Transfer Full Access – Reflexives (Yuan 1998)</p><p>Languages: L1s = English/Japanese, L2 = Chinese.</p><p>Task: Coreference judgments. Following each test item, subjects select from a</p><p>list of potential antecedents for the reflexive ziji.</p><p>Sample stimulus (favouring long-distance binding):</p><p>Wang Ming bu gaoxing de shuo Li Dong jingchang bu xiangxin ziji.</p><p>(Wang Ming said unhappily that Li Dong often does not trust self .)</p><p>Results:</p><p>Table 3.1.1 Acceptances of long-distance antecedents from embedded</p><p>finite clauses (in %)</p><p>L2 groups L1 Japanese (n = 24) 92</p><p>L1 English – intermediate (n = 32) 53</p><p>L1 English – advanced (n = 25) 71</p><p>Native speakers (n = 24) 94</p><p>64 3 The initial state</p><p>Yuan found that English-speaking and Japanese-speaking learners of Chinese</p><p>at the same level of proficiency (intermediate, as determined by a proficiency test)</p><p>behaved quite differently with respect to their treatment of ziji. Japanese, like</p><p>Chinese, has a long-distance reflexive, whereas English does not. The Japanese</p><p>speakers recognized the long-distance nature of ziji and their performance did not</p><p>differ significantly from that of the native speakers of Chinese. The English speak-</p><p>ers, on the other hand, were much less likely to accept long-distance antecedents</p><p>for the reflexive (especially in finite contexts), even when a long-distance inter-</p><p>pretation was pragmatically favoured; their judgments differed significantly from</p><p>those of the Japanese-speaking group and from the native speakers of Chinese.</p><p>These results, then, suggest that the two intermediate-level L2 groups are treating</p><p>long-distance reflexives very differently, reflecting properties of these anaphors</p><p>in their respective L1s, thus supporting Full Transfer. The advanced English-</p><p>speaking group showed evidence of acquiring the long-distance properties of ziji,</p><p>suggesting that L2 learners are not confined to L1 properties, in support of Full</p><p>Access.</p><p>Onemight questionwhy results from learners of intermediate proficiency should</p><p>be relevant to the initial state. While Full Transfer Full Access predicts evidence of</p><p>L1-based properties in early interlanguage, it crucially does not make any predic-</p><p>tions about how long L1 influence should last. It is not the case that restructuring</p><p>of the initial-state grammar necessarily takes place early on, nor is it the case that</p><p>the whole grammar must be changed at once. In the case of the initially adopted</p><p>Turkish word order (discussed in section 3.2.1.1), very basic properties of the</p><p>English input reveal that head final is an inappropriate analysis. This, presumably,</p><p>is why the L1-based initial state lasted only a short while. There was an early</p><p>reanalysis (within the first three months) to the head-initial categories appropriate</p><p>for the L2. But in other situations, the L2 input to motivate change may be more</p><p>obscure or even lacking altogether. In the case of reflexives, Yuan’s study suggests</p><p>that intermediate-level learnerswith English as amother tongue still have problems</p><p>in recognizing the long-distance nature of Chinese ziji. The situation regarding re-</p><p>flexives is quite different from the situation regarding word order. Since Chinese</p><p>also has a local reflexive (taziji) and since the long-distance reflexive ziji can take</p><p>both local or long-distance antecedents, the L2 input is more ambiguous and it</p><p>may be insufficient to lead the learner to postulate a long-distance reflexive for</p><p>the L2. In other words, an L2 learner who always produces reflexives restricted</p><p>to local antecedents will not be wrong; any confusion, then, is likely to arise in</p><p>interpreting long-distance reflexives used by other people. Even if misinterpreta-</p><p>tion results, the alternative local interpretation will not be ungrammatical though</p><p>it may be inappropriate at times.</p><p>Another experiment whose results support Full Transfer Full Access was con-</p><p>ducted by Slabakova (2000). Slabakova investigates a crosslinguistic aspectual</p><p>3.2 A grammar as the initial state 65</p><p>contrast in telicity. A clause is telic if the situation it describes includes a natural</p><p>endpoint; it is atelic if there is no such endpoint. For example, in (2a), the activity</p><p>is understood as having been completed, whereas (2b) does not necessarily imply</p><p>completion.</p><p>(2) a. Angela made a cake.</p><p>b. Angela made cakes.</p><p>Telicity can be realized in different ways. In languages like English and Spanish,</p><p>there is no special verbalmorphology to indicatewhether an event is telic or atelic.3</p><p>Rather, in clauses involving transitive verbs, like those in (2), telicity depends on</p><p>properties of the direct object: an event is telic if the object has specified cardinality,</p><p>as in (2a), that is, if it can be exhaustively counted or measured, as is the case for</p><p>DPs with determiners (an apple, three apples, the apple(s)); an event is atelic if</p><p>the object is of unspecified</p><p>cardinality, i.e. if it lacks a determiner (apples, cake),</p><p>as in (2b). In Slavic languages such as Bulgarian, on the other hand, telicity is</p><p>generally indicated bymeans of verbalmorphology, telic events beingmarkedwith</p><p>a preverb and atelic unmarked, while the cardinality of the object is irrelevant to</p><p>the aspectual interpretation. Slabakova (2001) argues that the difference between</p><p>Slavic languages and languages like English is a consequence of a parametric</p><p>difference relating to a functional category, Aspect. (See also Smith (1991) and</p><p>Snyder (1995a).)</p><p>Turning to L2 acquisition, Slabakova investigates the acquisition of English by</p><p>native speakers of Bulgarian, a language whose setting of the aspectual parameter</p><p>differs from English, and by native speakers of Spanish, a language with the</p><p>same setting as English. If the L1 grammar forms the interlanguage initial state,</p><p>differences are expected, with respect to aspectual interpretation, between these</p><p>two groups of learners.</p><p>Determining aspectual interpretations is not a simple matter, since both telic</p><p>and atelic sentences are grammatical. However, there are contexts where one or</p><p>other interpretation is more natural. Slabakova took advantage of such contextual</p><p>differences to devise an aspectual interpretation task. In the test sentences, the</p><p>context (provided by the first clause) is held constant, as well as the form of</p><p>the verb in the second clause. (See examples in box 3.2.) The only thing that</p><p>varies is the cardinality of the direct object in the second clause; in other words,</p><p>there are no other aspectual cues. To native speakers of English and Spanish,</p><p>the presence or absence of a determiner in the object DP in these sentences is</p><p>sufficient to determine the telicity of the second clause and hence the naturalness</p><p>of the sentence as a whole. Given a first clause which sets up the expectation of a</p><p>habitual, non-completed event, an atelic second clause sounds more natural than</p><p>a telic one. In Bulgarian, however, presence or absence of the determiner has no</p><p>effect on aspectual interpretation.</p><p>66 3 The initial state</p><p>Box 3.2 Full Transfer Full Access – Aspect (Slabakova 2000)</p><p>Languages: L1s = Bulgarian/Spanish, L2 = English.</p><p>Task: Aspectual interpretation task. Test sentences contain two clauses. Subjects</p><p>judge (on a scale from −3 to +3) how well the two clauses go together.</p><p>Sample stimuli:</p><p>1st clause habitual, 2nd clause telic</p><p>Antonia worked in a bakery and made a cake.</p><p>1st clause habitual, 2nd clause atelic</p><p>Antonia worked in a bakery and made cakes.</p><p>Results:</p><p>Table 3.2.1 Aspectual interpretation: mean ratings (from −3 to +3)</p><p>Telic (# = 6) Atelic (# = 6)</p><p>L2 groups L1 Bulgarian (n = 22) 1.44 1.95</p><p>L1 Spanish (n = 21) 0.55 2.04</p><p>Native speakers American English (n = 16) 0.19 2.09</p><p>British English (n = 16) 0.81 2.41</p><p>If Bulgarian-speaking learners of English initially represent telicity as it is</p><p>represented in the L1, they should have particular difficulties, since English verbal</p><p>morphology provides no indication of aspect. Slabakova found that native speakers</p><p>ofEnglish andSpanish-speaking learners ofEnglish distinguished sharply between</p><p>the two sentence types, finding the ones with an atelic second clause significantly</p><p>more natural than those with a telic second clause. (Both possibilities are gram-</p><p>matical, so the issue is naturalness rather than grammaticality.) The Bulgarian</p><p>speakers, on the other hand, showed a non-significant difference between the two</p><p>sentence types. (See table 3.2.1.) Looking only at the telic sentences, there were</p><p>highly significant differences between the groups, attributable solely to the perfor-</p><p>mance of the Bulgarian speakers. There were no significant differences between</p><p>the groups on the atelic sentences. Slabakova suggests that the Bulgarian speakers</p><p>are relatively accurate on the atelic sentences because these, in Bulgarian, would</p><p>not carry overt aspectual morphology. Hence, these sentences are interpreted as</p><p>atelic and are considered natural in the L2 English; in consequence the Bulgarian</p><p>speakers’ judgments pattern with those of the Spanish speakers and the native</p><p>speakers of English. In the case of the telic sentences, on the other hand, the</p><p>Bulgarian speakers are misled by the lack of aspectual morphology in the L2 and</p><p>treat these as sometimes telic, sometimes atelic, being unaware of the significance</p><p>of the cardinality of the object.</p><p>3.2 A grammar as the initial state 67</p><p>3.2.1.2 Full Transfer Full Access: conceptual issues</p><p>So far we have considered evidence that suggests that the L1 grammar is</p><p>implicated in the interlanguage initial state. It is important to remember that, ac-</p><p>cording to the Full Transfer Full Access Hypothesis, the L1 grammar in its entirety</p><p>is involved. The studies that we have considered have not, however, looked at the</p><p>initial state as a whole; indeed, it is unrealistic to expect anyone to do so. Rather,</p><p>particular properties have been investigated and, in the case of the ones we have</p><p>considered so far, there is indeed evidence of L1 properties in the interlanguage</p><p>grammar; furthermore, learners of different L1s behave differently with respect</p><p>to the same L2, consistent with the assumption that the L1 is the initial state.</p><p>It is also necessary to consider what might constitute counter-evidence to the</p><p>claim that there is full transfer in the initial state. That is, what would demon-</p><p>strate that the L1 grammar in its entirety is not the initial state? One kind of</p><p>potential counter-evidence can be dismissed immediately. If at some point L2</p><p>learners fail to show evidence of L1 effects, or if L2 learners of different L1s</p><p>behave in the same way with respect to some particular phenomenon in the L2,</p><p>this does not automatically disconfirm the hypothesis. Since Full Transfer Full</p><p>Access crucially assumes that the interlanguage grammar will be restructured in</p><p>response to properties of the L2 input interacting with UG, it is conceivable that</p><p>the grammars of L2 learners of different L1s will at some point converge on the</p><p>relevant properties of the L2 (or that they will converge on some non-L1, non-L2</p><p>properties).</p><p>Such a possibility raises the general issue of falsifiability: in the absence of</p><p>L1 effects, proponents of Full Transfer Full Access can maintain that learners are</p><p>already beyond the full transfer stage. How, then, could one ever show that this</p><p>hypothesis might be wrong? In fact, there are situations which are predicted by</p><p>Full Transfer Full Access not to occur, thus rendering the hypothesis falsifiable.</p><p>If learners of different L1s learning the same L2 can be shown to have the same</p><p>initial state and the same early stages of development, despite differences in how the</p><p>two L1s treat the linguistic phenomenon being investigated, this would constitute</p><p>counter-evidence. Experimental evidence of precisely this kind is provided by</p><p>Yuan (2001), who shows that French-speaking and English-speaking learners of</p><p>Chinese treat verb placement in exactly the same way from the earliest stage of</p><p>L2 acquisition, even though French and English differ in the relevant respects. We</p><p>will postpone discussion of this evidence until section 3.2.3.1.</p><p>3.2.1.3 Full Transfer Full Access: summary</p><p>To summarize, the Full Transfer Full Access Hypothesis makes claims</p><p>about the initial state, about grammars during development and about the steady</p><p>state:</p><p>68 3 The initial state</p><p>a. The initial state in L2 acquisition is the L1 steady state grammar in its</p><p>entirety. One needs to think of this as in some sense a copy (or clone)</p><p>of the L1 grammar, a copy which can be modified without affecting</p><p>the original. Although Full Transfer Full Access presupposes that the</p><p>L2 learner restructures the interlanguage grammar, the mother tongue</p><p>grammar does not (usually) get altered in response to L2 input (but see</p><p>Sorace (2000) for cases where the L2 may indeed have effects on the L1</p><p>grammar.)</p><p>b. The L2 learner is not limited to L1-based representations. If the L1-based</p><p>analysis fails for some reason, restructuring of the grammar</p><p>will occur;</p><p>in other words, L2 input will trigger grammar change. L2 development</p><p>is UG-constrained, with interlanguage grammars falling within the range</p><p>sanctioned by UG. (See chapter 4.)</p><p>c. Final outcome – convergence on a grammar identical to that of a native</p><p>speaker is not guaranteed, because properties of the L1 grammar or</p><p>subsequent interlanguage grammars may lead to analyses of the input</p><p>that differ from those of native speakers. (See chapter 8.)</p><p>3.2.2 The Minimal Trees Hypothesis</p><p>In this section, we examine another perspective on the interlanguage</p><p>initial state, namely theMinimalTreesHypothesis ofVainikka andYoung-Scholten</p><p>(1994, 1996a, b), which also proposes that the initial state is a grammar, with early</p><p>representations based on the L1. However, in contrast to Full Transfer Full Access,</p><p>only part of the L1 grammar is seen as constituting the initial state. Under this</p><p>approach, the initial grammar is claimed to lack functional categories altogether,</p><p>hence, L1 functional categories will not be present, nor will functional categories</p><p>from any other source (such as UG).</p><p>Vainikka and Young-Scholten claim that grammars in the earliest stage of</p><p>development are different from later grammars, lacking certain properties which</p><p>subsequently emerge. This claim is made in the context of the Weak Continuity</p><p>Hypothesis for L1 acquisition (Clahsen, Eisenbeiss and Penke 1996; Clahsen,</p><p>Eisenbeiss and Vainikka 1994; Clahsen, Penke and Parodi 1993/1994; Vainikka</p><p>1993/1994). According to this hypothesis, while functional categories are avail-</p><p>able in the UG inventory, initial grammars lack the full complement of functional</p><p>categories, containing lexical categories and their projections (NP, VP, PP, AP),</p><p>and possibly one underspecified functional projection, FP (Clahsen 1990/1991).</p><p>Det, Infl and Comp and associated projections (IP, CP and DP) emerge gradually,</p><p>triggered by input.</p><p>It is this conception of early grammars that Vainikka and Young-Scholten</p><p>develop in the context of L2 acquisition. According to their proposal, the initial</p><p>3.2 A grammar as the initial state 69</p><p>state in L2 acquisition consists of a grammar partly based on the L1: the lexical</p><p>categories of the mother tongue are found in the initial interlanguage grammar,</p><p>together with associated L1 properties, in particular, headedness. Functional cate-</p><p>gories, however, are lacking. Although functional categories are not realized in the</p><p>initial grammar, the full UG inventory of functional categories remains available.</p><p>L2 learners gradually add functional categories to the interlanguage grammar, on</p><p>the basis of L2 input, and are eventually able to project the associated projections</p><p>(IP, CP, DP, etc.). The claim is that functional categories are added ‘bottom up’,</p><p>in discrete stages, so that there is an IP stage before CP. In other words, presence</p><p>of CP in the grammar implicates IP: one can have IP without CP but not CP with-</p><p>out IP. Thus, although the emergence of functional categories is claimed to be</p><p>triggered by input, there must presumably be some kind of built-in sequence that</p><p>dictates this order. After all, there seems to be no reason in principle why a learner</p><p>should not ‘notice’ properties in the L2 input which would motivate a CP before</p><p>properties which would motivate IP.</p><p>On the Minimal Trees account, the initial states of learners of different L1s will</p><p>differ, depending on the headedness characteristics of lexical categories in the L1s</p><p>in question. Vainikka andYoung-Scholten (1996a) claim that headedness of lexical</p><p>categories will be reset to the value appropriate for the L2 before the appearance of</p><p>any functional categories. Emergence of functional categories, on the other hand,</p><p>in no way depends on properties of the L1 grammar; in other words, there is pre-</p><p>dicted to be no transfer in this domain, no stage or grammar in which properties</p><p>of the mother-tongue functional categories are found, an assumption which dif-</p><p>fers from Full Transfer Full Access. Rather, the L2 learner acquires L2 functional</p><p>categories, with L2 properties. Thus, L1 and L2 acquisition of any particular lan-</p><p>guage are generally assumed to be identical with respect to functional categories</p><p>and projections. (Vainikka and Young-Scholten (1998) do propose one difference,</p><p>relating to what properties of the input trigger the emergence of functional cate-</p><p>gories, namely boundmorphology in L1 acquisition versus freemorphemes in L2.)</p><p>3.2.2.1 Minimal Trees: evidence</p><p>In a series of papers, Vainikka andYoung-Scholten (1994, 1996a, b, 1998)</p><p>examine spontaneous and elicited production data from adult learners of German,</p><p>immigrants to Germany who had had no formal instruction in the L2. A variety</p><p>of L1s are represented in their studies, including Turkish and Korean, which, like</p><p>German, have head-final VPs, as well as Spanish and Italian, which are head initial.</p><p>Some of the data are longitudinal (following the same learners over time), some</p><p>cross-sectional (drawn from different learners who are hypothesized, post hoc, to</p><p>be at different stages of development).</p><p>As described above, the early interlanguage grammar, according to Vainikka</p><p>and Young-Scholten, has: (a) lexical categories with headedness characteristics</p><p>70 3 The initial state</p><p>from the L1; (b) no functional categories. In other words, in the initial state of</p><p>a Korean-speaking or Turkish-speaking learner of German, sentences would be</p><p>represented as in (3a), whereas in the case of a Spanish speaker or Italian speaker</p><p>the representation would look like (3b). Sentences are represented as VPs, because</p><p>there are as yet no higher functional projections like IP or CP; VPs accord with</p><p>the headedness of the VP in the L1.</p><p>(3) Stage 1 – the lexical stage</p><p>a.</p><p>Spec</p><p>NP V</p><p>b.</p><p>V NP</p><p>V' Spec V'</p><p>VP VP</p><p>Evidence for L1-based headedness of VPs in the early grammar is quite robust.</p><p>Vainikka and Young-Scholten (1994) found that over 95% of VPs were head final</p><p>in the case of the three least advanced Turkish and Korean speakers. (Unfortu-</p><p>nately, no independent measure of proficiency is provided. Stages of development</p><p>are determined in terms of performance on the syntactic and morphological prop-</p><p>erties being investigated, which is somewhat circular.) These L2 learners produced</p><p>utterances like those in (4) (from Vainikka and Young-Scholten 1994). Such sen-</p><p>tences are ungrammatical in German, where main clauses require a finite verb in</p><p>second position.</p><p>(4) a. Oya Zigarette trinken.</p><p>Oya cigarette drink-inf</p><p>‘Oya smokes cigarettes.’</p><p>a. Eine Katze Fisch alle essen.</p><p>a cat fish all eat-inf</p><p>‘A cat ate the entire fish.’</p><p>Data from speakers of head-initial languages show something quite different. Four</p><p>Romance speakers at a similar stage of development show predominantly head-</p><p>initial VPs. Typical productions are shown in (5) (from Vainikka and Young-</p><p>Scholten (1996a)):</p><p>(5) a. Trinke de orange oder?</p><p>drink the orange or?</p><p>‘(She’s) drinking the orange (juice), right?</p><p>a. De esse de fis.</p><p>she eat the fish</p><p>‘She’s eating the fish’</p><p>Although the VP in German is head final, finite verbs must move to second</p><p>position in main clauses; this is known as verb second (V2) (see chapter 1,</p><p>3.2 A grammar as the initial state 71</p><p>section 1.4.1). The data from the Turkish and Korean speakers considered alone</p><p>might be taken simply as evidence of acquisition of the verb-final nature of the L2,</p><p>rather than as evidence of L1 headedness. The data from the Romance speakers</p><p>considered alone might be taken as evidence of acquisition of V2 in the L2, rather</p><p>than as evidence of L1 headedness. But taken together, the data are highly sug-</p><p>gestive: the least advanced learners of different L1s adopt different word orders</p><p>in the early interlanguage grammar, consistent with the claim that the initial state</p><p>includes L1 lexical categories and their headedness.</p><p>But the crucial question is whether the early grammar is restricted to lexi-</p><p>cal categories, which is, after all, the central proposal of the Minimal Trees</p><p>Hypothesis. Let us</p><p>consider what kind of evidence is advanced to support the</p><p>claim that functional categories are initially lacking. Vainikka andYoung-Scholten</p><p>assume that spontaneous production data provide a relatively direct and reliable</p><p>window onto the underlying grammar: if some form is absent in production, the</p><p>underlying category associated with it is absent from the grammar. In the sit-</p><p>uation considered here, absence of particular lexical items (function words and</p><p>inflectional morphology) is taken to imply absence of corresponding functional</p><p>categories. Thus, Vainikka and Young-Scholten look at a number of morpho-</p><p>logical and lexical properties to determine whether or not functional categories</p><p>are present in the interlanguage. At the morphological level, they look for pres-</p><p>ence or absence of an agreement paradigm, productive person and/or number</p><p>morphology implicating at least IP. At the lexical level, they look for presence/</p><p>absence of auxiliary and modal verbs, since these are assumed to be generated</p><p>in Infl.</p><p>Vainikka and Young-Scholten argue that the language of the least advanced</p><p>Turkish/Korean speakers and the least advanced Romance speakers has the fol-</p><p>lowing characteristics: (a) incidence of correct subject – verb agreement is low;</p><p>instead, where a finite verb with agreement should be found, infinitives or bare</p><p>stems or default suffixes predominate; and (b) modals and auxiliaries are almost</p><p>non-existent. Another kind of evidence that Vainikka and Young-Scholten adduce</p><p>is syntactic: according to them there is no evidence that the verb raises out of theVP.</p><p>These characteristics suggest the lack of IP. In addition, there are no wh-questions</p><p>or subordinate clauses introduced by complementizers, suggesting lack of CP.</p><p>On this account, the lexical (VP) stage constitutes the initial state. The next stage,</p><p>according to Vainikka and Young-Scholten, is characterized by the emergence of a</p><p>functional categorywhichdoes not exist in adultGerman (or in anyother language).</p><p>They call this projection FP (finite phrase), following Clahsen (1990/1991). At this</p><p>stage, a German sentence will be represented as in (6), regardless of L1. In other</p><p>words, where necessary, headedness of lexical categories has been restructured to</p><p>accord with L2 headedness (Vainikka and Young-Scholten 1996b).</p><p>72 3 The initial state</p><p>(6) FP stage</p><p>FP</p><p>Spec F'</p><p>F VP</p><p>Spec V'</p><p>NP V</p><p>Their reason for postulating an FP as opposed to an IP is that some learners</p><p>showed evidence of properties that are characteristic of the presence of IP; at the</p><p>same time, these properties were not consistently present. This group of learn-</p><p>ers produced an increasing number of auxiliaries and modals; they also produced</p><p>sentences where the main verb had raised out of the VP, as well as verb-final utter-</p><p>ances, giving the impression that verb raising is optional.At the same time, subject –</p><p>verb agreement was still largely absent, nor were there any complementizers.</p><p>Examples of utterances at this stage (from a speaker of Turkish) are given in</p><p>(7) (from Vainikka and Young-Scholten 1994). In (7a), the verb appears to the left</p><p>of the object, suggesting that it has moved out of the VP. At the same time, it is</p><p>not inflected. In (7b), the uninflected verb is still within the VP, in final position.</p><p>(7) a. Ich sehen Schleier.</p><p>I see-inf veil</p><p>‘I see the veil’</p><p>b. Immer jeden Tag fünfhundert Stück machen.</p><p>always every day 500 unit make-inf</p><p>‘(I) always make 500 units every day’</p><p>Verb raising implicates a functional projection higher than the VP for the verb</p><p>to move to; lack of overt agreement morphology suggests to Vainikka and Young-</p><p>Scholten that this projection is somehowdifferent from IP. They resolve the conflict</p><p>occasioned by presence of verb raising but lack of accurate inflectionalmorphology</p><p>by proposing a category F, to which the verb moves. F is underspecified (lacking</p><p>agreement features); FP is head initial.</p><p>In the next stage (represented by yet another group of L2 learners), raising of</p><p>finite verbs becomes obligatory and the correct agreement paradigm is present, as</p><p>shown in the examples in (8), suggesting that agreement features are now avail-</p><p>able. Vainikka and Young-Scholten propose that IP has been added (replacing FP)</p><p>and that it is head initial (see (9)). According to them, CP is not yet moti-</p><p>vated because embedded clauses introduced by complementizers are still lacking</p><p>(see (8)).</p><p>3.2 A grammar as the initial state 73</p><p>(8) a. Er hat gesagt, nimmst du Lokomotive?</p><p>he has-3sg said take-2sg you train</p><p>‘He said, will you take the train?’</p><p>b. Ich kaufe dich Eis.</p><p>I buy-1sg you ice-cream</p><p>‘I will buy you some ice-cream’</p><p>(9) IP stage</p><p>IP</p><p>I'</p><p>I VP</p><p>V'</p><p>NP V</p><p>Spec</p><p>Spec</p><p>The final stage would involve the acquisition of CP. However, Vainikka and</p><p>Young-Scholten report that none of their subjects show evidence of reaching this</p><p>stage.</p><p>There are data from a number of different studies which are inconsistent with the</p><p>Minimal Trees Hypothesis. According to Minimal Trees, there are no functional</p><p>categories in the initial state. To argue against this view, it is sufficient to show</p><p>that at least some functional categories are in fact present from the beginning.</p><p>As Haznedar (1997) points out, the data from Erdem’s early productions (see</p><p>section 3.2.1.1) are clearly inconsistent with the Minimal Trees Hypothesis: at</p><p>least one functional category (NegP) is present initially. Indeed, Vainikka and</p><p>Young-Scholten themselves provide evidence of a functional category in the early</p><p>stage data: the examples in (4) and (5) include definite and indefinite articles,</p><p>suggesting that the functional categoryDetmust be present. Other evidence against</p><p>Minimal Trees is provided by Grondin and White (1996) who show that two</p><p>English-speaking children learning French have determiners firmly in place from</p><p>the earliest recordings. In addition, they report a number of reflexes of IP; for</p><p>example, finite verbs appear to the left of negative pas, suggesting that they have</p><p>raised out of the VP. Lakshmanan (1993/1994) shows that a 4-year-old Spanish-</p><p>speaking child learning English has IP early on, as evidenced by extensive use of</p><p>the copula be in spontaneous production.</p><p>As for CP, Lakshmanan and Selinker (1994) report that the same Spanish-</p><p>speaking child, as well as a French-speaking child of the same age, produced</p><p>tensed embedded clauses early on (in the third interviews), with null complemen-</p><p>tizers (which are, of course, permitted and preferred in English), while infinitival</p><p>complements are found from the second interview. Examples are given in (10).</p><p>74 3 The initial state</p><p>(10) a. I think it’s for me.</p><p>b. I don’t want to play with you.</p><p>Gavruseva and Lardiere (1996) find evidence for CP in the first transcripts of</p><p>spontaneous production data from Dasha, an 8-year-old Russian-speaking child</p><p>learning English, two months after her initial exposure to the L2. Dasha produces</p><p>subject auxiliary inversion in yes/no and wh-questions, as shown in (11a) and</p><p>(11b), consistent with movement of auxiliaries to C, with the wh-phrase in (11b)</p><p>in Spec CP; embedded clauses are found from the third recording session onwards</p><p>(after less than three months of exposure to the L2), as shown in (11c).</p><p>(11) a. Can I see please?</p><p>b. What are we going to do?</p><p>c. Mama know that we go outside.</p><p>Furthermore, asGavruseva andLardiere point out, while there is robust evidence</p><p>for CP in the early interlanguage grammar, IP must be considered to be absent if</p><p>Vainikka and Young-Scholten’s criterion for determining presence of a category</p><p>is adopted. As will be discussed in section 3.2.2.2, Vainikka and Young-Scholten</p><p>define a category as being present in the grammar only if lexical items/inflectional</p><p>morphology associated with it occur in 60% or more of obligatory contexts.</p><p>Gavruseva and Lardiere found that incidence of CP-related phenomena, such as</p><p>subject – auxiliary inversion in questions, was 100%. At the same time, suppli-</p><p>ance of subject – verb agreement in obligatory contexts was less than 40%,</p><p>while</p><p>production of modals and auxiliaries generally fell below 52%. As the example</p><p>in (11c) shows, although CP is implicated because of the presence of the com-</p><p>plementizer that, verbal inflection is lacking (know rather than knows). Since the</p><p>Minimal Trees Hypothesis claims that IP emerges before CP and that IP is not</p><p>present until its reflexes are found in 60% of obligatory contexts, these findings</p><p>are contradictory and cannot be accounted for. On the other hand, if it is recog-</p><p>nized that a functional category can be present in the abstract, even though not</p><p>consistently realized lexically, there is no such contradiction; we return to this</p><p>point below (and in chapter 6).</p><p>Data that suggest the influence of the L1 grammar in the functional domain can</p><p>also be used to argue against the Minimal Trees Hypothesis. Recall that Minimal</p><p>Trees expects transfer of lexical categories only. When functional categories</p><p>emerge, they will exhibit properties relevant to the L2 (triggered by L2 input), not</p><p>properties derived from the L1. But, as Schwartz and Sprouse (1996) point out,</p><p>there is evidence for L1 effects in the functional domain. For example, as described</p><p>above, Erdem’s NegP has the headedness of NegP in the L1 Turkish. Further-</p><p>more, in a series of studies conducted byWhite and colleagues (Trahey andWhite</p><p>1993; White 1990/1991, 1991a, 1992a), French-speaking children (aged 10–12)</p><p>learning English allow lexical verbs to appear to the right of adverbs, as shown in</p><p>3.2 A grammar as the initial state 75</p><p>(12), suggesting that the verb has raised to I and that I has strong features, a char-</p><p>acteristic of French but not English (see chapter 1, section 1.4.1). (These studies</p><p>will be discussed in more detail in chapters 4 and 5.)</p><p>(12) a. Susan plays often the piano.</p><p>b. Susan playsi [often [VP ti the piano]]</p><p>One possible response might be that the data implicating functional categories,</p><p>as discussed so far, are drawn from child L2 acquisition. Could the Minimal Trees</p><p>Hypothesis be recast as a claimonly about adult L2 acquisition?AlthoughVainikka</p><p>and Young-Scholten do advance the hypothesis in the adult context, the logic of</p><p>their argument implies that it applies equally to child L2 acquisition. The Minimal</p><p>Trees Hypothesis is a hypothesis about initial-state grammars in general; Vainikka</p><p>and Young-Scholten motivate it on the basis of similar proposals for child L1</p><p>acquisition. Thus, child L2 acquisition can hardly be excluded. Any data, whether</p><p>from child or adult L2 learners, that suggests the presence of functional material</p><p>in the initial state, or emergence of CP before IP, or L1 effects on functional</p><p>categories, is problematic for the Minimal Trees Hypothesis.</p><p>Vainikka and Young-Scholten (1996b) counter some of the problematic data</p><p>which seem to suggest the initial presence of functional categories by questioning</p><p>whether such data are genuinely relevant to the initial state. For example, they point</p><p>out, correctly, that the data discussed in Grondin andWhite (1996) may not in fact</p><p>be relevant, since the children had been exposed to French for some months before</p><p>they were first recorded (even though they did not speak any French during that</p><p>time). They also question the data discussed by Lakshmanan and Selinker (1994),</p><p>since evidence for CP is scanty until the fifth and sixth recordings, at which point</p><p>the children could be deemed to be beyond the initial state. Recognizing that the</p><p>majority of adults in their own studies do not appear to have grammars totally</p><p>lacking functional categories, they suggest that they may already have passed the</p><p>no functional category stage (Vainikka and Young-Scholten 1994).</p><p>This again raises the issue of falsifiability – how early is early enough? If</p><p>researchers can resort to the claim that an earlier stage lacking functional categories</p><p>would have been found had data elicitation started early enough or that a ‘silent pe-</p><p>riod’ preceding L2 production would have had no functional categories, it renders</p><p>the Minimal Trees Hypothesis unfalsifiable. Indeed, to investigate the possibility</p><p>that there might be a stage prior to the emergence of L2 speech in which functional</p><p>categories are lacking, we need methodologies that do not rely on production data.</p><p>Comprehension tasks where functional properties are manipulated are not easy to</p><p>construct.</p><p>It should be noted, however, that the falsifiability problem is by nomeans unique</p><p>to the Minimal Trees Hypothesis. As discussed in section 3.2.1.2, in the absence</p><p>of evidence of transfer, the proponents of the Full Transfer Full Access Hypothesis</p><p>76 3 The initial state</p><p>could resort to the same appeal: there would have been transfer if an earlier stage</p><p>had been examined. We return to this general issue in section 3.4.</p><p>3.2.2.2 Minimal Trees: conceptual issues</p><p>In addition to empirical evidence suggesting that the Minimal Trees</p><p>Hypothesis is misconceived, there have been a number of objections on conceptual</p><p>grounds. Regardless of whether one accepts the Weak Continuity Hypothesis for</p><p>L1 acquisition, there are problemswith the claim that the initial state of L2 learners</p><p>is limited to lexical categories. Given a steady-state grammar (the L1) with func-</p><p>tional categories, as well as UG with an inventory of functional categories, why</p><p>should these be totally absent in the interlanguage initial state, indeed how could</p><p>they be? As Schwartz and Sprouse (1996: 66) point out, ‘It is difficult to imagine</p><p>what sort of cognitive mechanismwould be involved in extracting a proper subpart</p><p>of the L1 grammar and using that proper subsystem as the basis for a new cognitive</p><p>state.’</p><p>We have seen above that Vainikka and Young-Scholten postulate two stages</p><p>involving a head-initial functional category, first FP, then IP. In both cases, this is</p><p>different from German. On most analyses of German, VP and IP are head final,</p><p>while CP is head initial (see chapter 1, section 1.4.1); this is the analysis that</p><p>Vainikka and Young-Scholten adopt for the grammars of native speakers. Finite</p><p>verbs in main clauses raise from V to I and then to C. Some other phrase (the</p><p>subject, the object or an adjunct) moves to the Spec of CP, yielding the V2 effect.</p><p>In other words, the position of the finite verb in main clauses depends on there</p><p>being a functional head (C) to the left of the VP, as shown in (13).</p><p>(13) CP</p><p>C'</p><p>C IP</p><p>Spec I'</p><p>VP I</p><p>V'</p><p>NP V</p><p>Spec</p><p>Spec</p><p>The L2 learners who are described by Vainikka and Young-Scholten as being</p><p>beyond the initial state show clear (though not necessarily consistent) evidence of</p><p>verb movement out of the VP. And as the verb is found following the subject, a</p><p>fairly obvious analysis would be that these learners in fact have the representation</p><p>3.2 A grammar as the initial state 77</p><p>in (13), namely a full CP (Epstein et al. 1996; Schwartz 1998a; Schwartz and</p><p>Sprouse 1994), with the verb in C and the subject in Spec of CP. Since Vainikka</p><p>and Young-Scholten assume that there is no CP at this stage, they are forced to</p><p>assume that FP and IP are head initial in order to account for the observed word</p><p>order; if FP and IP were head final, finite verbs should remain at the end of the</p><p>sentence, contrary to fact.</p><p>What, then, drives Vainikka and Young-Scholten’s hypothesis that CP is absent,</p><p>even given the presence of data consistent with a CP projection (namely finite verbs</p><p>occurring outside the VP near the front of the clause)? Recall that Vainikka and</p><p>Young-Scholten crucially assume that absence of some form in production data</p><p>means absence of the corresponding abstract category. In the case of CP, they fail</p><p>to find evidence of overt complementizers and conclude that there is no C, hence</p><p>no CP. Similarly, they conclude that, in the absence of an overt verbal agreement</p><p>paradigm, there is no IP.</p><p>Inmaking these assumptions, Vainikka andYoung-Scholten are adopting a form</p><p>of ‘morphological bootstrapping’ (Clahsen et al. 1996), the idea being that the</p><p>learner’s acquisition of functional categories is a consequence of having already</p><p>acquired regular inflectional</p><p>paradigms, with overt morphology acting as a trigger</p><p>for projecting functional structure. In other words, in the absence of overt morphol-</p><p>ogy or other lexical items associated with functional categories (such as modals</p><p>and complementizers), the associated syntactic positions are assumed not to be in</p><p>place.</p><p>A number of researchers have questioned the assumption that, if a form is</p><p>absent in production data (or used inaccurately or with variability), the associated</p><p>functional category must necessarily be lacking. There is a difference between</p><p>‘knowing’ abstract functional properties and knowing how these happen to be</p><p>lexically realized in a particular language; problems with the latter do not neces-</p><p>sarily indicate problems with the former (Epstein et al. 1996; Grondin and White</p><p>1996; Haznedar and Schwartz 1997; Lardiere 1998a, b, 2000; Prévost and White</p><p>2000a, b; Schwartz 1991; Schwartz and Sprouse 1996). If learners show evidence</p><p>of syntactic properties associated with functional categories, this suggests that</p><p>the categories are present, even in the absence of particular lexical items or mor-</p><p>phology. This issue will be discussed in more detail in chapter 6.</p><p>Another problem with Vainikka and Young-Scholten’s account is that presence</p><p>and absence are defined in an arbitrary, and apparently unmotivated, way. A cat-</p><p>egory is defined as present in the grammar if lexical items associated with it are</p><p>produced in 60% or more of obligatory contexts; otherwise, it is assumed to be</p><p>absent. In other words, the criterion for acquisition is an accuracy level of 60%.</p><p>In earlier L1 acquisition research, as well as in early L2 research that investigated</p><p>acquisition orders relating to functionwords andmorphology (Bailey,Madden and</p><p>78 3 The initial state</p><p>Krashen 1974; Dulay and Burt 1974), a criterion of 90% accuracy was adopted</p><p>(Brown 1973). This is, of course, equally arbitrary and is the result of equating</p><p>performance (almost totally accurate use of inflectional morphology) with acqui-</p><p>sition. Vainikka and Young-Scholten recognize that the 90% criterion is too high</p><p>but it is not clear what a criterion of 60% achieves. Indeed, it would be more appro-</p><p>priate to take evidence of emergence of some property as evidence of acquisition</p><p>(Meisel, Clahsen and Pienemann 1981; Stromswold 1996).</p><p>3.2.2.3 Minimal Trees: summary</p><p>In summary, according to the Minimal Trees Hypothesis:</p><p>a. The interlanguage initial state is a grammar containing lexical categories</p><p>(drawn from the L1 grammar) but no functional categories.</p><p>b. Developmental stages involve the addition of functional categories (avail-</p><p>able fromUG),which emerge gradually, triggered byL2 input. Functional</p><p>categories are added to the representation from the bottom up (i.e. CP</p><p>could not be acquired before IP).</p><p>c. Final outcome – L2 learners should, presumably, converge on the L2</p><p>grammar, at least as far as functional projections and their conse-</p><p>quences are concerned, since L2 data triggering the relevant properties</p><p>are available.</p><p>3.2.3 The Valueless Features Hypothesis</p><p>We turn now to a third hypothesis concerning the interlanguage initial</p><p>state, the Valueless Features Hypothesis of Eubank (1993/1994, 1994, 1996). Like</p><p>Full Transfer Full Access and Minimal Trees, the Valueless Features Hypothesis</p><p>claims that the initial state is a grammar. Eubank argues for ‘weak’ transfer, main-</p><p>taining that the L1 grammar largely – but not entirely – determines the interlan-</p><p>guage initial state. Like Full Transfer Full Access and unlike Minimal Trees, the</p><p>Valueless Features Hypothesis claims that L1 lexical and functional categories are</p><p>present in the earliest interlanguage grammar. However, although L1 functional</p><p>categories are available, their feature values are claimed not to be. That is, feature</p><p>strength does not transfer. Instead of being either strong or weak, features are</p><p>valueless or ‘inert’ in the initial state.</p><p>As described in chapter 1, section 1.4.1, feature strength has consequences for</p><p>word order. In a language like English, where I has weak V-features, finite verbs</p><p>remain within the VP. In languages like French, where I is strong, the verb raises</p><p>to I to check its features. In other words, finite lexical verbs either must raise</p><p>(as in French) or may not raise (as in English).</p><p>3.2 A grammar as the initial state 79</p><p>On theValuelessFeaturesHypothesis, feature values are neitherweaknor strong.</p><p>According to Eubank, this has the following consequence: when features are not</p><p>specified for strength, finite verbs can alternate between raising or not raising. In</p><p>other words, if the L2 learner is acquiring a language with strong features like</p><p>French, both the word orders in (14) are expected in early stages, the grammatical</p><p>(14a), where the finite verb has raised out of the VP, and the ungrammatical (14b),</p><p>where the verb has not raised. The predictions are identical for an L2 with weak</p><p>features, like English. That is, both of the word orders in (15) are predicted to</p><p>occur because, in the absence of a specification of feature strength, the verb can</p><p>‘choose’ to raise or not, as the case may be.</p><p>(14) a. Marie regardei [souvent [VP ti la télévision]]</p><p>b. Marie [souvent [VP regarde la télévision]]</p><p>(15) a. Mary [often [VP watches television]]</p><p>b. Mary watchesi [often [VP ti television]]</p><p>The assumption that valueless features implies optional verb raising is a stipu-</p><p>lation whose justification is unclear; we return to this point in section 3.2.3.2.</p><p>3.2.3.1 Valueless Features: evidence</p><p>In support of the hypothesis, Eubank examines data from a variety of</p><p>sources. In earlier work on adverb placement, White (1990/1991, 1991a) showed</p><p>that French-speaking learners of English produce and accept both the word orders</p><p>in (15). She suggested that (15b) is the result of transfer of the strong feature</p><p>value from French. Eubank argues that White’s data in fact support the Valueless</p><p>Features Hypothesis, given the fact that the order in (15a), without verb raising,</p><p>was also found. On a strong transfer account, this order should be impossible,</p><p>since the strong feature should force verb raising. (See chapters 4 and 5 for more</p><p>detailed presentation of these data and their implications.)</p><p>Aproblem for theValueless Features account is that verbs should raise optionally</p><p>over negatives as well. That is, a French-speaking learner of English would be</p><p>expected to produce both preverbal and postverbal negation, as in (16).</p><p>(16) a. The children likei [not [VP ti spinach]]</p><p>b. The children (do) [not [VP like spinach]]</p><p>However, such variability has not been reported. In other words, there is less</p><p>variability here than expected under the inert features proposal. White (1992a)</p><p>found that French-speaking learners of English consistently reject sentences like</p><p>(16a) where not follows the lexical verb. (See chapter 4, section 4.5.2.1.) In addi-</p><p>tion, Eubank (1993/1994) examines spontaneous production data (from Gerbault</p><p>(1978) and Tiphine (undated)) and finds the same thing: French-speaking children</p><p>80 3 The initial state</p><p>learning English produce negatives like (16b) but not like (16a). Eubank accounts</p><p>for the absence of optional verb raising over negation in terms of the interaction</p><p>of inert features in Tense and weak features in Agr. But the lack of sentences like</p><p>(16a) is in fact consistent with these learners having acquired the relevant feature</p><p>value of English, namely weak agreement.4</p><p>Experimental evidence that goes against the Valueless Features Hypothesis is</p><p>provided byYuan (2001). Yuan examines the L2 acquisition of Chinese, a language</p><p>withweak features, hence lacking verbmovement. In Chinese, sentences like (17a)</p><p>are grammatical, whereas (17b) is not.</p><p>(17) a. Zhangsan changchang kan dianshi.</p><p>Zhangsan often watch television</p><p>b. *Zhangsan kan changchang dianshi.</p><p>Zhangsan watch often television</p><p>Subjects were adult native speakers of French and English, learning Chinese; they</p><p>were at various levels of proficiency, including beginners. (See box 3.3.) The</p><p>L1 of one group (English) shares the</p><p>property of weak feature strength with the</p><p>L2 Chinese, while the L1 of the other group (French) has the opposite strength.</p><p>According to the Valueless Features Hypothesis, both groups should initially</p><p>behave in the same way, regardless of L1 feature strength, permitting optional</p><p>verb placement in Chinese. However, subjects at the lowest level of proficiency</p><p>(who had studied Chinese for less than sixmonths) showed no evidence of optional</p><p>verb placement. This was true of both the French speakers and the English speak-</p><p>ers. In two different tasks, production and acceptance of the grammatical order, as</p><p>in (17a), were very high, while production and acceptance of the ungrammatical</p><p>order, where the verb has raised, as in (17b), were correspondingly low, as shown</p><p>in table 3.3.1.</p><p>Not only are these results inconsistent with the Valueless Features Hypothesis,</p><p>since they provide no evidence for optional verb raising, they also appear to be in-</p><p>consistent with the Full Transfer Full AccessHypothesis, since there is no evidence</p><p>that I is initially strong in the grammars of the French speakers. (However, pro-</p><p>ponents of both the Valueless Features Hypothesis and Full Transfer Full Access</p><p>could argue that the two first-year groups had already had sufficient exposure to</p><p>the L2 to acquire the weak feature strength appropriate for Chinese, once again</p><p>raising the question of falsifiability.)</p><p>What happens beyond the initial state? According to the Valueless Features</p><p>Hypothesis as originally propounded by Eubank, inertness is a temporary phe-</p><p>nomenon, characteristic only of the early interlanguage grammar. Subsequent</p><p>acquisition of feature strength (strong or weak) is claimed to depend on the</p><p>emergence of inflectional morphology. Eubank follows Rohrbacher (1994) in</p><p>assuming that feature strength is determined by the nature of morphological</p><p>3.2 A grammar as the initial state 81</p><p>Box 3.3 Parameter setting – features (Yuan 2001)</p><p>Languages: L1s = English/French, L2 = Chinese.</p><p>Tasks:</p><p>i. Oral production.</p><p>ii. Grammaticality judgments. Pairs of sentences differing only as to verb</p><p>position. Subjects indicate whether or not both sentences are acceptable.</p><p>Sample stimuli:</p><p>a. Wo gege he pingchang Deguo jiu.</p><p>b. Wo gege pingchang he Deguo jiu.</p><p>(My brother drinks usually German wine.</p><p>My brother usually drinks German wine.)</p><p>Results:</p><p>Table 3.3.1 Production and judgments of grammatical and ungrammatical</p><p>word orders</p><p>Production Grammaticality judgments</p><p>SAVO *SVAO Other SAVO *SVAO Both</p><p>L2 groups Level 1 223 9 8 136 3 4</p><p>(L1 English) (n = 24) (93%) (95%)</p><p>Level 2 140 0 0 88 0 2</p><p>(n = 15) (100%) (98%)</p><p>Level 3 160 0 0 91 0 4</p><p>(n = 16) (100%) (96%)</p><p>Level 4 120 0 0 63 1 8</p><p>(n = 12) (100%) (88%)</p><p>L2 groups Level 1 148 0 2 70 5 2</p><p>(L1 French) (n = 15) (99%) (91%)</p><p>Level 2 141 0 19 93 2 0</p><p>(n = 16) (88%) (97%)</p><p>Level 3 167 0 3 98 0 1</p><p>(n = 17) (98%) (99%)</p><p>Native speakers (n = 10) 92 0 8 60 0 0</p><p>(92%) (100%)</p><p>paradigms: Infl is strong if and only if inflectional morphology is rich, a term that</p><p>is variously defined but which essentially means that verbs show distinct person</p><p>and/or number morphology. In other words, strong I will be triggered by rich</p><p>morphology, weak I otherwise. (Arguments against this claim will be discussed in</p><p>chapter 6.)</p><p>82 3 The initial state</p><p>On the assumption that morphology and feature strength are correlated in this</p><p>way, the following predictions can be made for L2 acquisition: learners who have</p><p>not yet acquired overt agreement should show variability in verb placement (char-</p><p>acteristic of inertness in the initial state), whereas learners who have acquired</p><p>agreement should consistently raise or fail to raise the verb, depending on the</p><p>L2 in question. Eubank and Grace (1998) and Eubank, Bischof, Huffstutler, Leek</p><p>and West (1997) conducted experiments testing these predictions. In both stud-</p><p>ies, the L1 (Chinese) and L2 (English) share the same feature strength, namely</p><p>weak. As we have seen, the Valueless Features Hypothesis predicts optional verb</p><p>raising even in such cases. Thus, Chinese-speaking learners of English who have</p><p>not yet acquired English third-person-singular agreement should sometimes raise</p><p>the main verb over an adverb, a possibility not permitted in either language,</p><p>whereas learners who have acquired third-person-singular agreement should not</p><p>do so.</p><p>Both studies included an oral translation task to determine whether or not third-</p><p>person-singular agreement morphology had been acquired, as well as another task</p><p>to determine whether or not learners permit verb raising. Eubank and Grace (1998)</p><p>use a sentence-matching task for the latter purpose, while Eubank et al. (1997) use</p><p>a truth-value-judgment task.</p><p>The sentence-matchingmethodology (Freedman andForster 1985) involves pre-</p><p>senting pairs of sentences on a computer screen. Subjects have to press a response</p><p>key indicating whether the two sentences are the same (matched) or different</p><p>(unmatched). It has been established that native speakers respond significantly</p><p>faster tomatched grammatical pairs thanmatched ungrammatical pairs, for a range</p><p>of constructions. Thus, response times can be used as a diagnostic of grammat-</p><p>icality, even though subjects are not making explicit grammaticality judgments.</p><p>The sentence-matching methodology has recently received considerable attention</p><p>in L2 acquisition research, the rationale being that it may also provide a diag-</p><p>nostic of grammaticality in interlanguage grammars (Beck 1998a; Bley-Vroman</p><p>and Masterson 1989; Clahsen and Hong 1995; Duffield andWhite 1999; Duffield,</p><p>White, Bruhn de Garavito, Montrul and Prévost 2002; Eubank 1993). (For a recent</p><p>critique of the use of sentence-matching tasks in L2 acquisition research, see Gass</p><p>(2001).)</p><p>To test verb raising, Eubank and Grace include pairs of sentences like (15a) and</p><p>other pairs like (15b), repeated here as (18). (See box 3.4.)</p><p>(18) a. Mary often watches television.</p><p>b. Mary watches often television.</p><p>If feature values are inert, verb raisingwill be permitted but not required and both</p><p>sentence typeswill be grammatical. In otherwords, there should be no difference in</p><p>3.2 A grammar as the initial state 83</p><p>Box 3.4 Valueless Features Hypothesis (Eubank and Grace 1998)</p><p>Languages: L1 = Chinese, L2 = English.</p><p>Task: Sentence-matching. Pairs of sentences presented on computer screen.</p><p>Subjects decide whether the two sentences are the same or different.</p><p>Sample stimuli:</p><p>Grammatical (No V-raising) Ungrammatical (V-raising)</p><p>The woman often loses her books The boy takes often the flowers</p><p>The woman often loses her books The boy takes often the flowers</p><p>Results:</p><p>Table 3.4.1 Mean response times in ms.</p><p>V-raising (ungram) No raising (gram)</p><p>L2 groups No agreement (n = 14) 3038 2841 sig</p><p>Agreement (n = 18) 2594 2618 ns</p><p>Native speakers (n = 36) 1546 1491 sig</p><p>response times to pairs like (18a) compared to pairs like (18b). On the other hand,</p><p>if the L2 learners have a grammar where the weak value has been established, the</p><p>pairs involving verb raising, like (18b), will be ungrammatical. Hence, it should</p><p>take significantly longer to respond to this type than to grammatical pairs like</p><p>(18a).</p><p>Native speakers of English behaved as predicted, responding significantly more</p><p>slowly to the ungrammatical pairs with the raised verbs. (See table 3.4.1.) The L2</p><p>learners were divided into two groups on the basis of the translation task: (i) a</p><p>no agreement group whose suppliance of agreement was inconsistent – this group</p><p>is assumed still to be in the initial state; and (ii) an agreement group, who con-</p><p>sistently produced third-person-singular morphology – by hypothesis, this group</p><p>is beyond the initial state. In the case of the no agreement group, no significant</p><p>difference is predicted between the two types of sentence pairs, since they are</p><p>both grammatical in a grammar that tolerates but does not require verb raising</p><p>because of inert features. The agreement group, in contrast, is expected to show</p><p>the same contrast as native speakers, that is, to take significantly longer to respond</p><p>to the pairs</p><p>suggestions, I would particularly</p><p>Preface xiii</p><p>like to thank:KevinGregg,DonnaLardiere, DawnMacLaughlin, Bonnie Schwartz</p><p>and Antonella Sorace, as well as the anonymous reviewers for Cambridge Univer-</p><p>sity Press. Thematerial in this book has formed the core of my graduate seminar on</p><p>L2 acquisition for several years and I would like to acknowledge the contribution</p><p>of many former and current graduate students of the Linguistics Department at</p><p>McGill University: their stimulating discussion and questioning of many of the</p><p>issues presented here has been invaluable, as well as their ability in catching typos.</p><p>Abbreviations</p><p>Adj adjective</p><p>AdjP adjective phrase</p><p>A(dv) adverb</p><p>acc accusative case</p><p>Agr the functional category Agreement</p><p>AgrP Agreement Phrase</p><p>asp aspect marker</p><p>Asp the functional category Aspect</p><p>AspP Aspect Phrase</p><p>caus causative</p><p>cl classifier</p><p>cli clitic</p><p>CNPC Complex Noun Phrase Constraint</p><p>comp complementizer</p><p>C(omp) the functional category Complementizer</p><p>CP Complementizer Phrase</p><p>dat dative</p><p>dec declarative marker</p><p>D(et) the functional category Determiner</p><p>dp Determiner Phrase</p><p>fem feminine</p><p>F finite</p><p>FP finite phrase</p><p>gen genitive</p><p>ger gerund</p><p>imp imperfect</p><p>inf infinitive</p><p>I(nfl) the functional category Inflection</p><p>IP Inflection Phrase</p><p>masc masculine</p><p>n number of subjects</p><p>xiv</p><p>Abbreviations xv</p><p># number of stimuli</p><p>Neg the functional category Negation</p><p>NegP Negation Phrase</p><p>nom nominative case</p><p>N noun</p><p>NP noun phrase</p><p>ns not significant</p><p>NS native speaker</p><p>Num the functional category Number</p><p>NumP Number Phrase</p><p>O object</p><p>pl plural</p><p>P preposition</p><p>PP prepositional phrase</p><p>pass passive</p><p>perf perfective</p><p>pol politeness marker</p><p>pres present</p><p>pret preterite</p><p>prt particle</p><p>ps person</p><p>Q question marker</p><p>S subject</p><p>sg singular</p><p>sig significant</p><p>Spec specifier</p><p>subj subjunctive</p><p>T the functional category Tense</p><p>TP Tense Phrase</p><p>top topic marker</p><p>V verb</p><p>VP verb phrase</p><p>V2 verb second</p><p>V3 verb third</p><p>1 1st person</p><p>2 2nd person</p><p>3 3rd person</p><p>1</p><p>Universal Grammar and language</p><p>acquisition</p><p>1.1 Introduction</p><p>This book will be concerned with characterizing and explaining the lin-</p><p>guistic systems that second language (L2) learners develop, considering in par-</p><p>ticular the extent to which the underlying linguistic competence of L2 speakers</p><p>is constrained by the same universal principles that govern natural language in</p><p>general. Following Chomsky (1959, 1965, 1975, 1980, 1981a, b, 1986b, 1999),</p><p>a particular perspective on linguistic universals will be adopted and certain as-</p><p>sumptions about the nature of linguistic competence will be taken for granted. In</p><p>particular, it will be presupposed that the linguistic competence of native speakers</p><p>of a language can be accounted for in terms of an abstract and unconscious linguis-</p><p>tic system, in other words, a grammar, which underlies use of language, including</p><p>comprehension and production.Native-speaker grammars are constrained by built-</p><p>in universal linguistic principles, known as Universal Grammar (UG).</p><p>Throughout this book, non-native grammars will be referred to as interlanguage</p><p>grammars. The concept of interlanguage was proposed independently in the late</p><p>1960s and early 1970s by researchers such as Adjémian (1976), Corder (1967),</p><p>Nemser (1971) and Selinker (1972). These researchers pointed out that L2 learner</p><p>language is systematic and that the errors produced by learners do not consist of</p><p>random mistakes but, rather, suggest rule-governed behaviour. Such observations</p><p>led to the proposal that L2 learners, like native speakers, represent the language</p><p>that they are acquiring by means of a complex linguistic system.</p><p>The current generative linguistic focus on the nature of interlanguage has its</p><p>origins in the original interlanguage hypothesis. Explicit claims are made about</p><p>the underlying grammars of L2 learners and L2 speakers, the issues including a</p><p>consideration of the role of UG and the extent to which interlanguage grammars</p><p>exhibit properties of natural language. Such questions will be explored in detail in</p><p>this book. It will be suggested that the linguistic behaviour of non-native speakers</p><p>can be accounted for in terms of interlanguage grammars which are constrained</p><p>by principles and parameters of UG. At the same time, it will be recognized</p><p>1</p><p>2 1 Universal Grammar and language acquisition</p><p>that interlanguage grammars differ in various ways from the grammars of native</p><p>speakers, and some of these differences will be explored.</p><p>1.2 Universal Grammar in L1 acquisition</p><p>Amajor task for the first language (L1) acquirer is to arrive at a linguistic</p><p>system which accounts for the input, allowing the child to build linguistic repre-</p><p>sentations and to understand and produce language. UG is proposed as part of an</p><p>innate biologically endowed language faculty (e.g. Chomsky 1965, 1981b; Pinker</p><p>1984, 1994), which permits the L1 acquirer to arrive at a grammar on the basis of</p><p>linguistic experience (exposure to input). UG provides a genetic blueprint, deter-</p><p>mining in advance what grammars can (and cannot) be like. In the first place, UG</p><p>places requirements on the form of grammars, providing an inventory of possible</p><p>grammatical categories and features in the broadest sense, i.e. syntactic, mor-</p><p>phological, phonological and semantic. In addition, it constrains the functioning</p><p>of grammars, by determining the nature of the computational system, including</p><p>the kinds of operation that can take place, as well as principles that grammars</p><p>are subject to. UG includes invariant principles, that is, principles that are gener-</p><p>ally true across languages, as well as parameters which allow for variation from</p><p>language to language.</p><p>Throughout this book it will be presupposed that UG constrains L1 acquisition,</p><p>as well as adult native-speaker knowledge of language. That is, grammars of chil-</p><p>dren and adults conform to the principles and parameters of UG. The child acquires</p><p>linguistic competence in the L1. Properties of the language are mentally repre-</p><p>sented by means of an unconscious, internalized linguistic system (a grammar).</p><p>As Chomsky (1980: 48) puts it, there is : ‘a certain mental structure consisting of</p><p>a system of rules and principles that generate and relate mental representations of</p><p>various types’.1</p><p>UG constitutes the child’s initial state (S0), the knowledge that the child is</p><p>equipped with in advance of input. The primary linguistic data (PLD) are critical</p><p>in helping the child to determine the precise form that the grammar must take. As</p><p>the child takes account of the input, a language-specific lexicon is built up, and</p><p>parameters of UG are set to values appropriate for the language in question. The</p><p>grammar (G) may be restructured over the course of time, as the child becomes</p><p>responsive to different properties of the input. In due course, the child arrives at</p><p>a steady state grammar for the mother tongue (SS). This model of acquisition is</p><p>schematized in figure 1.1.</p><p>As linguistic theories such as Government–Binding (Chomsky 1981a), Mini-</p><p>malism (Chomsky 1995) orOptimality Theory (Archangeli andLangendoen 1997)</p><p>1.3 Why UG? 3</p><p>(UG)PLD</p><p>G1 G2 Gn</p><p>So</p><p>Ss</p><p>Figure 1.1 Model of L1 acquisition</p><p>have developed, there have been changes in how universal principles and param-</p><p>eters have been formalized, in other words, changes in what UG is assumed to</p><p>consist of. For example, the numerous and very specific principles of the early</p><p>days of generative theory, such as many of the original Island Constraints (Ross</p><p>1967), have been replaced with more general, invariant economy principles (e.g.</p><p>Chomsky 1991), as well as computational operations, such as Move and Merge</p><p>(see Marantz 1995). Parameters have gradually become more constrained, now</p><p>being largely associated with the lexicon: properties of items that enter into a com-</p><p>putation, for example, may vary in feature composition and feature strength, with</p><p>associated syntactic consequences.</p><p>Such ongoing changes in the definition ofUGare a reflection of development and</p><p>growth within linguistic theory. Nevertheless, regardless of howUG is formalized,</p><p>there remains a consensus (within the</p><p>involving verb raising, which are ungrammatical in a grammar with</p><p>weak agreement. However, the results showed the reverse: the no agreement group</p><p>responded significantly more slowly to the ungrammatical pairs (like the native</p><p>speakers), whereas the agreement group did not.</p><p>84 3 The initial state</p><p>Box 3.5 Valueless Features Hypothesis (Eubank et al. 1997)</p><p>Languages: L1 = Chinese, L2 = English.</p><p>Task:Truth-value judgments. Short narratives, each followed by a sentence. Sub-</p><p>jects indicate whether the sentence is true or false in the context of the narrative.</p><p>Sample stimulus:</p><p>Tom loves to draw pictures of monkeys in the zoo. Tom likes his pictures to be</p><p>perfect, so he always draws them very slowly and carefully. All the monkeys</p><p>always jump up and down really fast.</p><p>Tom draws slowly jumping monkeys.</p><p>True False</p><p>Results:</p><p>Table 3.5.1 Responses to V-raising items (in %)</p><p>False True</p><p>L2 groups No agreement (n = 14) 69.5 30.4</p><p>Agreement (n = 18) 81.7 18.3</p><p>Native speakers (n = 28) 91 9</p><p>Eubank et al. (1997) examined the same issues, using a modified truth-value-</p><p>judgment task, to determine whether the unexpected results from the sentence-</p><p>matching task are attributable to the methodology rather than being a reflection</p><p>of grammatical knowledge. The rationale, assumptions and predictions were the</p><p>same as for Eubank and Grace but different subjects were tested. (See box 3.5.)</p><p>Their truth-value-judgment task centres on test sentences which are ambiguous in</p><p>the case of a grammar that allows optional verb raising (thus permitting a response</p><p>of true or false) but unambiguous in a grammar that does not (allowing only false).</p><p>In the example in box 3.5, the sentence Tom draws slowly jumping monkeys is false</p><p>if slowly is construed as modifying jumping, that is, if it is analysed as in (19a).</p><p>In contrast, the sentence is true if it is interpreted with raising of the verb over the</p><p>adverb, with the structure in (19b), where slowly is understood as modifying the</p><p>verb draws.</p><p>(19) a. Tom draws [NP [slowly jumping] monkeys]</p><p>b. Tom drawsi [slowly [VP ti jumping monkeys]]</p><p>In a grammar which prohibits verb raising, only the former interpretation, in-</p><p>volving modification within the NP, is possible (but see below). The prediction,</p><p>then, is that native speakers of English and learners with agreement (as defined</p><p>3.2 A grammar as the initial state 85</p><p>above) will respond in the same manner, treating all such items as false. The no</p><p>agreement group, on the other hand, is expected to alternate between responses,</p><p>reflecting the ambiguity of these sentences in a grammarwith optional verb raising.</p><p>Contrary to predictions, Eubank et al. found that all three groups differ sig-</p><p>nificantly from each other in their responses; that is, the agreement group did</p><p>not perform exactly like the native speakers. (See table 3.5.1.) Nevertheless, it is</p><p>clear from the results that the response patterns of the three groups are in fact quite</p><p>similar: all groups, including the native speakers, give some responses of true, and,</p><p>for all groups, including the no agreement group, responses of false predominate.</p><p>If verb raising truly were optional in interlanguage grammars, one might expect</p><p>a much higher proportion of true responses. In several of the crucial scenarios,</p><p>the test sentence is extremely odd if the verb has not raised, involving NPs like</p><p>slowly jumping monkeys and quietly toasted bread. If the grammar really sanctions</p><p>verb raising, one might expect the more natural verb-raised interpretation to be</p><p>adopted (leading to a higher number of responses of true). An alternative possibil-</p><p>ity is that subjects resorted to interpretations involving verb raising even though</p><p>their grammars prohibit it, precisely because the alternatives were pragmatically so</p><p>odd.5 Also, one cannot exclude the possibility that subjects read the sentences with</p><p>the adverb ‘in parentheses’ (e.g. Tom draws, slowly, jumping monkeys), in which</p><p>case a response of true would be given and yet the verb has not raised over the</p><p>adverb.</p><p>There are a number of respects in which Eubank et al.’s task departed from</p><p>more standard truth-value-judgment tasks. In the first place, the test items should</p><p>all be grammatical; appropriateness is then determined on the basis of the story</p><p>(or picture) which supplies the context. (See examples testing knowledge of re-</p><p>flexives, chapter 2, boxes 2.5 and 2.6.) However, in Eubank et al.’s task, English</p><p>sentences with raised verbs are ungrammatical for native speakers and for learners</p><p>who have acquired the weak English feature value. Furthermore, they are ungram-</p><p>matical but appropriate, given the contexts. It really is not clear that there is any</p><p>prediction as to what subjects (native speakers or otherwise) should do in such</p><p>circumstances. Eubank et al. appear to assume that the interpretation where the</p><p>adverb modifies the verb simply will not come to mind (because the sentences are</p><p>ungrammatical) but this has not been demonstrated, and clearly it does come to</p><p>mind some of the time.</p><p>A related problem is that each context in Eubank et al.’s task sets one interpre-</p><p>tation off against the other, for example, there is something slow going on in the</p><p>story as well as something fast (see box 3.5). Again, this diverges from the usual</p><p>practice (see chapter 2, boxes 2.5 and 2.6), where the context describes just one</p><p>situation and subjects have to judge whether or not the test sentence is true of that</p><p>situation. Because there is no difference in interpretation depending solely on the</p><p>86 3 The initial state</p><p>position of the verb (slowly draws versus draws slowly), it is in fact impossible</p><p>to match contexts with interpretations without introducing such additional com-</p><p>plications. The truth-value-judgment methodology is simply unsuitable for testing</p><p>issues relating to feature strength.</p><p>To sum up, neither the sentence-matching experiment nor the truth-value-</p><p>judgment experiment bears out the predictions of the Valueless Features</p><p>Hypothesis. In both studies, the results from the group who had passed the crite-</p><p>rion for acquiring inflection are particularly problematic, since these learners have</p><p>overt morphological inflection but are still, ostensibly, permitting verb raising. In</p><p>consequence, Eubank et al. (1997) and Eubank and Grace (1998), following Beck</p><p>(1998a), interpret their results as providing evidence of an even stronger version of</p><p>the Valueless Features Hypothesis, whereby inert features are a permanent prop-</p><p>erty of interlanguage grammars, rather than just being found in the initial state.</p><p>This position, the Local Impairment Hypothesis, will be discussed in chapter 4.</p><p>In addition, setting aside the methodological problems discussed above, there</p><p>are problematic inconsistencies in the results from the two tasks as regards the</p><p>no agreement groups. If both groups have grammars with inert features, why</p><p>are they behaving differently? In the sentence-matching task, the no agreement</p><p>group performed like native speakers, taking significantly longer to respond to</p><p>ungrammatical sentences, suggesting absence of verb raising. In the truth-value-</p><p>judgment task, the no agreement group gave more responses of true than other</p><p>groups (although responses of false predominated), interpreted by Eubank et al.</p><p>as indicating that verb raising is permitted. To accommodate the former finding,</p><p>Eubank and Grace suggest that the Minimal Trees Hypothesis must be the cor-</p><p>rect account of the initial state (which is then followed by a grammar with inert</p><p>features): in the sentence-matching task, the results from the no agreement group</p><p>are consistent with an initial grammar with only a VP projection, in which case</p><p>the sentences with raised verbs would be ungrammatical, because there would be</p><p>no functional category for the verb to raise to. Unfortunately for this proposal,</p><p>the results from the truth-value-judgment task do not support such an account:</p><p>if the no agreement subjects had only a VP, verb raising would be impossible and</p><p>their responses should have been exclusively false, contrary</p><p>to fact.</p><p>3.2.3.2 Valueless Features: conceptual issues</p><p>Such difficulties reflect deeper problems with the Valueless Features</p><p>Hypothesis, at the conceptual level. It really is not clear what it means for feature</p><p>strength to be inert, or what motivates this proposal. Nor is it clear whether in-</p><p>ertness is confined to features of Infl or why this should be so. Eubank suggests</p><p>that inertness is somewhat similar to underspecification of functional categories</p><p>proposed for the grammars of L1 acquirers (e.g.Wexler 1994). However, proposals</p><p>for underspecification in L1 acquisition are, in fact, quite different: underspecified</p><p>3.3 UG as the initial state 87</p><p>Tense, for example, results in variability (in verb-raising languages) between finite</p><p>and non-finite verbs. Finite verbs always raise, whereas non-finite verbs (optional</p><p>infinitives) do not. On Eubank’s account, on the other hand, it is finite verbs that</p><p>show variability. Again, it is not clear what this claim follows from. Indeed, one</p><p>might just as well predict that variability should not occur, as noted by Schwartz</p><p>(1998b) and by Robertson and Sorace (1999): in the absence of a strong feature</p><p>forcing raising, the verb should not move at all, since it is the strong feature value</p><p>that motivates movement. Since inert implies not strong, all verbs should remain</p><p>within the VP. Even if inertness could somehow be rendered less stipulative, it</p><p>is, in any case, not obvious why feature strength should be inert in interlanguage</p><p>grammars. As Schwartz and Sprouse (1996) point out, why should all properties of</p><p>the L1 grammar be found in the initial state with the exception of feature strength?</p><p>3.2.3.3 Valueless Features: summary</p><p>To summarize, the claims of the Valueless Features Hypothesis are as</p><p>follows:</p><p>a. The interlanguage initial state is a grammar containing lexical and func-</p><p>tional categories, as well as features, drawn from the L1 grammar. Feature</p><p>strength is inert.</p><p>b. L2 feature strength will be acquired during the course of development,</p><p>when morphological paradigms are acquired.</p><p>c. Final outcome – ultimately, L2 learners should converge on the L2</p><p>grammar.</p><p>3.3 UG as the initial state</p><p>The three hypotheses considered so far (Full Transfer Full Access,</p><p>Minimal Trees and Valueless Features) agree that the L1 grammar forms the inter-</p><p>language initial state, although they disagree on whether the whole L1 grammar</p><p>is implicated. We turn now to two hypotheses which reject the possibility that</p><p>any properties of the L1 grammar are involved in the interlanguage initial state.</p><p>Instead, something quite different is proposed, namely that the L2 learner starts</p><p>out with UG rather than with any particular grammar.</p><p>3.3.1 The Initial Hypothesis of Syntax</p><p>According to Platzack’s (1996) Initial Hypothesis of Syntax, the initial</p><p>states of L1 and L2 acquisition are identical. The initial state is UG; it includes</p><p>functional categories with all features set at default or unmarked strength, namely</p><p>88 3 The initial state</p><p>weak.Weak is claimed to be the default value, on the grounds that overt movement</p><p>(motivated by strong features) is costly (Chomsky 1993, 1995). All learners (L1 or</p><p>L2), then, will initially assume weak features. In the case of L2 acquisition, this is</p><p>claimed to be so even if the L1 grammar has strong feature values. Subsequently,</p><p>the learner has to work out which features should in fact be set to strong, on</p><p>the basis of L2 input (such as input showing evidence of overt movement). The</p><p>Initial Hypothesis of Syntax represents an updated version of earlier markedness</p><p>proposals, whereby L2 learners were argued to resort to unmarked options made</p><p>available by UG regardless of the situation in the L1 (Liceras 1986; Mazurkewich</p><p>1984a).</p><p>The Initial Hypothesis of Syntax has not, as yet, been pursuedwith any degree of</p><p>detail in the L2 context. There is, however, experimental evidence that casts doubts</p><p>on such claims. As mentioned above, results from White (1990/1991, 1991a)</p><p>suggest that L2 learners do not start off with all features set at weak values:</p><p>French-speaking learners of English transfer strong features from the L1, hence</p><p>allowing verb movement over adverbs in the L2. On the other hand, the results of</p><p>Yuan (2001) could be seen as offering support, since French-speaking learners of</p><p>Chinese, whose L1 has strong I, do not at any stage permit verb raising in the L2.</p><p>Sprouse (1997) and Schwartz (1997) question the claim that L2 learners start</p><p>off with weak features. The Initial Hypothesis of Syntax presupposes Kayne’s</p><p>(1994) antisymmetry hypothesis, whereby all languages are underlyingly SVO.</p><p>SOV word order is the result of a strong object feature which must be checked in</p><p>AgrO, causing the object to raise over the verb. The Initial Hypothesis of Syntax,</p><p>then, predicts that all L2 learners will start off with SVO order because object</p><p>features are initially weak, so that no object movement is possible. This prediction</p><p>holds regardless of what the L2 word order is and regardless of feature strength</p><p>and word order in the L1: thus, learners whose mother tongue is SOV are predicted</p><p>to have an initial SVO stage, even if the L2 is also SOV. As we have already seen,</p><p>this prediction is false. Turkish- and Korean-speaking learners of German initially</p><p>assume that German is SOV (Schwartz and Sprouse 1994; Vainikka and Young-</p><p>Scholten 1994), as do Turkish-speaking learners of English (Haznedar 1997).</p><p>3.3.2 Full Access (without Transfer)</p><p>The final proposal to be considered in this chapter is the Full Access</p><p>Hypothesis of Flynn and Martohardjono (1994), Flynn (1996), and Epstein et al.</p><p>(1996, 1998). According to Epstein et al. (1996: 750), the Full Access Hypothesis</p><p>is not, strictly speaking a hypothesis about the initial state. Nevertheless, although</p><p>not proposed as an explicit initial-state hypothesis, in fact it has clear implications</p><p>for the nature of the initial state, as we shall see. Furthermore, this hypothesis</p><p>3.3 UG as the initial state 89</p><p>implies, like the Initial Hypothesis of Syntax, that UG must constitute the initial</p><p>state in L2 acquisition.</p><p>What is meant by the Full Access Hypothesis? Epstein et al. argue that the</p><p>interlanguage grammar is UG-constrained at all stages; grammars conform to the</p><p>principles of UG and learners are limited to the hypothesis space allowed by UG.</p><p>In other words, UG remains accessible in non-primary acquisition. So far, the</p><p>assumptions are identical to those advanced by proponents of Full Transfer Full</p><p>Access; in other words, the full access of Full Transfer Full Access is the full</p><p>access of Epstein et al. Furthermore, this assumption is shared by Vainikka and</p><p>Young-Scholten, who argue that all of UG is available in L2 acquisition, although</p><p>someproperties (functional categories) emerge after others. TheValueless Features</p><p>Hypothesis was also originally intended as a full access theory, interlanguage</p><p>grammars being constrained by UG, with inertness of features only a temporary</p><p>property.</p><p>What, then, makes Epstein et al.’s Full Access Hypothesis different from the</p><p>positions considered so far? In contrast to the first three initial-state proposals,</p><p>Epstein et al. (1996: 751) specifically reject the possibility that the L1 grammar</p><p>forms the initial state. For this reason, I will sometimes refer to their position</p><p>as Full Access without Transfer. In spite of this rejection, they do recognize the</p><p>presence of L1 effects in interlanguage grammars. But if these effects are not due</p><p>to L1-based initial representations, then it is difficult to understand just what they</p><p>have in mind.6</p><p>If the initial state is not the L1 grammar, what is it? The (implicit) logic of</p><p>Epstein et al.’s argumentation necessitates that it is not a grammar at all but rather</p><p>UG. In other words, the initial state in L2 acquisition is the same as the initial state</p><p>in L1. In fact, however, Epstein et al. (1996: 751) reject this possibility as well,</p><p>stating that the initial state in L2 is not S0, so presumably not UG itself. It is hard</p><p>to conceive what the initial state</p><p>could possibly be, if it is neither at least partially</p><p>the L1 grammar nor UG. I will continue to interpret their hypothesis as implying</p><p>that UG must be the initial state, although they fail to recognize that this is the</p><p>logical outcome of their position.</p><p>To understand how it is that Epstein et al.’s position implicates UG as the initial</p><p>state, consider that they specifically argue in favour of the Strong Continuity or</p><p>Full Competence Hypothesis as the correct account of functional categories in</p><p>L2 grammars and against the Minimal Trees Hypothesis of Vainikka and Young-</p><p>Scholten. According to the Strong Continuity Hypothesis, all functional categories</p><p>are present in L1 grammars from the beginning (Borer and Rohrbacher 1997;</p><p>Hyams 1992; Lust 1994; Wexler 1998). In contrast, as discussed in section 3.2.2,</p><p>the Minimal Trees Hypothesis claims that initially no functional categories are</p><p>present and that they emerge gradually. Since the Minimal Trees Hypothesis is a</p><p>90 3 The initial state</p><p>= UGL2 PLD</p><p>L1</p><p>S0</p><p>L1 Ss L2 SsILG1 ILGn</p><p>Figure 3.3 Full Access (without transfer)</p><p>claim about the nature of the interlanguage initial state, Epstein et al.’s refutation</p><p>necessarily involves an alternative initial-state claim, in particular the claim that</p><p>the earliest interlanguage grammar will contain a full complement of functional</p><p>categories. And, since they reject the possibility that the L1 grammar forms the</p><p>initial state, the source of functional categories in the early interlanguage grammar</p><p>can only be UG itself.</p><p>Hence, whether they recognize it or not, the clear implication of Epstein et al.’s</p><p>Full Access Hypothesis is that UG is the initial state in L2 as well as in L1. The Full</p><p>Access Hypothesis is illustrated in figure 3.3 (adapted from White 2000), which</p><p>shows the L1 grammar dissociated from the interlanguage grammars, in other</p><p>words, Full Access without Transfer. (Figure 3.3 applies equally to the Initial</p><p>Hypothesis of Syntax.)</p><p>The logical outcome of this position is that interlanguage grammars of learners</p><p>of different L1s will be the same, because of the influence of UG, with no effects</p><p>attributable to the L1, since the L1 does not form the initial state. In fact, Epstein</p><p>et al. do not exclude the possibility of L1-effects and differences between learners</p><p>of different L1s, although, as already mentioned, it is not at all clear how the L1</p><p>fits into their scheme of things. As to the final state achievable in L2 acquisition,</p><p>this should in principle be a representation identical to that of native speakers of</p><p>the L2 (Flynn 1996: 150).</p><p>3.3.2.1 Full Access: evidence</p><p>Epstein et al. (1996) conducted an experiment with child and adult</p><p>Japanese-speaking learners ofEnglishwhich, they argue, provides evidence against</p><p>theMinimal Trees Hypothesis and in favour of Strong Continuity and Full Access.</p><p>In addition, they claim that the results demonstrate lack of L1 influence in the early</p><p>grammar. By implication, then, this experiment should be relevant to the initial</p><p>state, since the Minimal Trees Hypothesis involves claims about the initial state,</p><p>as does Full Transfer Full Access.</p><p>The experiment involved an elicited imitation task. Stimuli were designed</p><p>with the intention of testing knowledge of various morphological and syntactic</p><p>3.3 UG as the initial state 91</p><p>Box 3.6 Full Access (Epstein et al. 1996)</p><p>Languages: L1 = Japanese, L2 = English.</p><p>Task: Elicited imitation.</p><p>Sample stimuli:</p><p>Items testing for IP</p><p>The nervous doctor wanted a new lawyer in the office. (past tense)</p><p>The happy janitor does not want the new television. (neg/do support)</p><p>The little girl can see a tiny flower in the picture. (modal)</p><p>Items testing for CP</p><p>Which secret message does the young girl find in the basket? (wh-question)</p><p>Breakfast, the wealthy business man prepares in the kitchen. (topicalization)</p><p>The lawyer slices the vegetables which the father eats. (relative clause)</p><p>Results:</p><p>Table 3.6.1 Accurate imitations (in %)</p><p>IP (# = 12) CP (# = 12)</p><p>L2 groups Children (n = 33) 69 50</p><p>Adults (n = 18) 68 45</p><p>properties associated with the functional projections IP and CP. (See box 3.6.)</p><p>The rationale behind this choice of methodology is the assumption that learners</p><p>(L1 or L2) can only successfully imitate sentences that are analysable by the</p><p>current grammar (Lust, Flynn and Foley 1996). If so, L2 learners should not be</p><p>able to imitate sentences exemplifying grammatical properties normally analysed</p><p>in terms of functional categories unless those categories are present in the grammar.</p><p>For example, if learners can imitate sentences containing verbs marked for third-</p><p>person-singular agreement, then they are assumed to have the functional category</p><p>that hosts agreement, namely IP.</p><p>Results showed no significant differences between child and adult learners of</p><p>English in their imitation abilities. (See table 3.6.1.) Both groups were more ac-</p><p>curate at imitating sentences testing for presence of IP than those testing for CP.</p><p>(Epstein et al. suggest that the lower performance on CPs is not due to absence of</p><p>a CP projection but rather to processing problems associated with long-distance</p><p>movement, compounded by the lack of syntactic wh-movement in the L1 gram-</p><p>mar, leading to problems withwh-movement sentences in the L2. This explanation</p><p>provides a good example of their somewhat ambivalent attitude to the role of the</p><p>92 3 The initial state</p><p>L1: here the L1 is used to explain away problemswith the data; yet at the same time</p><p>they deny a role to the L1 as far as representation of such sentences is concerned.)</p><p>Let us consider (a) whether their experiment in fact tells us anything about</p><p>the initial state; and (b) what it shows more generally about Full Access. These</p><p>results do not provide evidence against Minimal Trees or against Full Transfer Full</p><p>Access, despite claims to the contrary by Epstein et al., because they tell us nothing</p><p>about the presence or absence of functional categories in the initial state. Epstein</p><p>et al.’s adult subjects were of low-intermediate proficiency, with an average of one</p><p>year of exposure to English in an English-speaking country, as well as seven years</p><p>of instruction in the L2. The children, whose proficiency level is not reported, had</p><p>lived in the USA for an average of three years, with an average of three years of</p><p>formal instruction in English. Thus, these learners must be well beyond an initial-</p><p>state grammar. Strong Continuity (argued for by Epstein et al.) and Minimal Trees</p><p>(advocated by Vainikka and Young-Scholten) differ only in their claims about</p><p>the presence of functional categories at the outset of L2 acquisition. None of</p><p>the theories we have considered claims that intermediate level learners will have</p><p>grammars totally lacking functional categories. Epstein et al.’s results are irrelevant</p><p>as far as the Minimal Trees Hypothesis is concerned; these intermediate learners</p><p>could simply be past the lexical stage. Similarly, the results are irrelevant as an</p><p>argument against Full Transfer Full Access. Epstein et al. claim that the fact that</p><p>their subjects show evidence of L2 functional categories argues against an early</p><p>L1-based grammar, on the assumption that Japanese lacks functional categories</p><p>(Fukui and Speas 1986). But Full Transfer Full Access is not just a full transfer</p><p>theory; it is also a full access theory. So the fact that intermediate-level learners</p><p>show evidence of L2 categories is not an argument against an L1-based initial</p><p>state. The data are neutral on this point.</p><p>Clearly, Epstein et al.’s results are irrelevant to the initial state. In what sense</p><p>do they otherwise support Full Access? All the initial-state hypotheses included in</p><p>this chapter presuppose that UG constrains interlanguage grammars. All assume</p><p>that L2 functional categories will be attainable. In other words, they are all full</p><p>access theories. So, in so far as Epstein et al.’s data could provide evidence of full</p><p>access, they are neutral between a number of different full access theories.</p><p>However, evidence for full access on the basis of this</p><p>study is weak, at best.</p><p>Epstein et al. fail to demonstrate that there is a poverty of the stimulus with respect</p><p>to learning English inflectional morphology and function words (Schwartz and</p><p>Sprouse 2000a); thus, it is not clear in what sense UG is implicated at all. Even if</p><p>it is, there is a problem with their use of elicited imitation. Setting aside general</p><p>questions as to the suitability of elicited imitation for investigating the nature of</p><p>linguistic competence (see Bley-Vroman and Chaudron (1994) and the commen-</p><p>taries on Epstein et al. (1996)), their task included no ungrammatical sentences.</p><p>3.3 UG as the initial state 93</p><p>If learners can only imitate sentences that fall within their current grammatical</p><p>competence, then, given ungrammatical sentences to imitate (for example, lacking</p><p>inflection or with incorrect inflection), they should not imitate these but rather</p><p>correct them in accordance with their current grammar. In the absence of ungram-</p><p>matical stimuli, the data are uninterpretable: successful imitation could indicate</p><p>a reflection of grammatical properties or simply an ability to imitate whatever</p><p>stimuli are presented.</p><p>3.3.2.2 Full Access: conceptual issues</p><p>The Initial Hypothesis of Syntax is underdeveloped in the L2 context.</p><p>This leaves the Full Access Hypothesis of Epstein et al. as the only version of Full</p><p>Access without Transfer that has been explored in any detail. In principle, Full</p><p>Access without Transfer is a perfectly coherent position. Unfortunately, Epstein</p><p>et al.’s proposal suffers from several problems, including equivocation over the</p><p>role of the L1, which makes it almost impossible to establish precisely what their</p><p>predictions are for the initial-state or subsequent grammars.</p><p>With these caveats in mind, let us consider whether Full Transfer Full Access</p><p>and Full Access without Transfer are irreconcilable. Epstein et al. recognize the</p><p>existence of L1 effects but are reluctant to attribute these to an initial represen-</p><p>tation based on the L1, partly, perhaps, because they feel that there is less L1</p><p>influence than might be expected on a Full Transfer account, as well as consid-</p><p>erable commonalities between learners of different L1s. But suppose that initial</p><p>representations are in fact based on the L1 but that for some properties L1-based</p><p>representations are fleeting, with almost immediate triggering of L2 properties,</p><p>based on L2 input and drawing on UG. Indeed, we have seen that some L1 word-</p><p>order effects are quite short-lived. Thus, it might appear that there are L1 effects in</p><p>some areas of the grammar but not others, whereas in fact there are short-lived and</p><p>long-lived effects. In that case, one needs an account of why some L1 character-</p><p>istics are easily overridden, while others have lasting effects in the interlanguage</p><p>representation.A theory of triggering can perhaps contribute to such an explanation</p><p>(see chapter 5).</p><p>3.3.2.3 Full Access: summary</p><p>In summary, the implications of the Initial Hypothesis of Syntax and the</p><p>Full Access Hypothesis are as follows:</p><p>a. The L2 initial state is not the L1 grammar. UG is the initial state.</p><p>b. L2 development is UG-constrained, with interlanguage grammars falling</p><p>within the range sanctioned by UG.</p><p>Ta</p><p>bl</p><p>e</p><p>3.</p><p>1</p><p>In</p><p>it</p><p>ia</p><p>ls</p><p>ta</p><p>te</p><p>an</p><p>d</p><p>be</p><p>yo</p><p>nd</p><p>:</p><p>hy</p><p>po</p><p>th</p><p>es</p><p>es</p><p>co</p><p>m</p><p>pa</p><p>re</p><p>d</p><p>Fu</p><p>ll</p><p>T</p><p>ra</p><p>ns</p><p>fe</p><p>r</p><p>Fu</p><p>ll</p><p>A</p><p>cc</p><p>es</p><p>s</p><p>M</p><p>in</p><p>im</p><p>al</p><p>T</p><p>re</p><p>es</p><p>V</p><p>al</p><p>ue</p><p>le</p><p>ss</p><p>Fe</p><p>at</p><p>ur</p><p>es</p><p>Fu</p><p>ll</p><p>A</p><p>cc</p><p>es</p><p>s</p><p>(w</p><p>ith</p><p>ou</p><p>tT</p><p>ra</p><p>ns</p><p>fe</p><p>r)</p><p>In</p><p>iti</p><p>al</p><p>st</p><p>at</p><p>e</p><p>L</p><p>1</p><p>le</p><p>xi</p><p>ca</p><p>la</p><p>nd</p><p>fu</p><p>nc</p><p>tio</p><p>na</p><p>l</p><p>N</p><p>o</p><p>fu</p><p>nc</p><p>tio</p><p>na</p><p>lc</p><p>at</p><p>eg</p><p>or</p><p>ie</p><p>s</p><p>L</p><p>1</p><p>le</p><p>xi</p><p>ca</p><p>la</p><p>nd</p><p>fu</p><p>nc</p><p>tio</p><p>na</p><p>l</p><p>Fu</p><p>ll</p><p>co</p><p>m</p><p>pl</p><p>em</p><p>en</p><p>to</p><p>f</p><p>le</p><p>xi</p><p>ca</p><p>l</p><p>ca</p><p>te</p><p>go</p><p>ri</p><p>es</p><p>,f</p><p>ea</p><p>tu</p><p>re</p><p>s</p><p>an</p><p>d</p><p>L</p><p>1</p><p>le</p><p>xi</p><p>ca</p><p>lc</p><p>at</p><p>eg</p><p>or</p><p>ie</p><p>s</p><p>ca</p><p>te</p><p>go</p><p>ri</p><p>es</p><p>;(</p><p>so</p><p>m</p><p>e)</p><p>in</p><p>er</p><p>t</p><p>an</p><p>d</p><p>fu</p><p>nc</p><p>tio</p><p>na</p><p>lc</p><p>at</p><p>eg</p><p>or</p><p>ie</p><p>s,</p><p>fe</p><p>at</p><p>ur</p><p>e</p><p>st</p><p>re</p><p>ng</p><p>th</p><p>fe</p><p>at</p><p>ur</p><p>es</p><p>.</p><p>fe</p><p>at</p><p>ur</p><p>es</p><p>an</p><p>d</p><p>fe</p><p>at</p><p>ur</p><p>e</p><p>st</p><p>re</p><p>ng</p><p>th</p><p>D</p><p>ev</p><p>el</p><p>op</p><p>m</p><p>en</p><p>t</p><p>D</p><p>if</p><p>fe</p><p>re</p><p>nt</p><p>pa</p><p>th</p><p>fo</p><p>r</p><p>le</p><p>ar</p><p>ne</p><p>rs</p><p>E</p><p>m</p><p>er</p><p>ge</p><p>nc</p><p>e</p><p>of</p><p>fu</p><p>nc</p><p>tio</p><p>na</p><p>l</p><p>In</p><p>er</p><p>tf</p><p>ea</p><p>tu</p><p>re</p><p>s</p><p>re</p><p>pl</p><p>ac</p><p>ed</p><p>by</p><p>N</p><p>o</p><p>de</p><p>ve</p><p>lo</p><p>pm</p><p>en</p><p>tr</p><p>eq</p><p>ui</p><p>re</p><p>d</p><p>of</p><p>di</p><p>ff</p><p>er</p><p>en</p><p>tL</p><p>1s</p><p>,a</p><p>tl</p><p>ea</p><p>st</p><p>in</p><p>iti</p><p>al</p><p>ly</p><p>.</p><p>ca</p><p>te</p><p>go</p><p>ri</p><p>es</p><p>in</p><p>st</p><p>ag</p><p>es</p><p>,i</p><p>n</p><p>L</p><p>2</p><p>fe</p><p>at</p><p>ur</p><p>e</p><p>st</p><p>re</p><p>ng</p><p>th</p><p>in</p><p>ab</p><p>st</p><p>ra</p><p>ct</p><p>pr</p><p>op</p><p>er</p><p>tie</p><p>s</p><p>of</p><p>R</p><p>es</p><p>tr</p><p>uc</p><p>tu</p><p>ri</p><p>ng</p><p>of</p><p>fu</p><p>nc</p><p>tio</p><p>na</p><p>l</p><p>re</p><p>sp</p><p>on</p><p>se</p><p>to</p><p>L</p><p>2</p><p>in</p><p>pu</p><p>t</p><p>fu</p><p>nc</p><p>tio</p><p>na</p><p>lc</p><p>at</p><p>eg</p><p>or</p><p>ie</p><p>s</p><p>pr</p><p>op</p><p>er</p><p>tie</p><p>s</p><p>in</p><p>re</p><p>sp</p><p>on</p><p>se</p><p>to</p><p>L</p><p>2</p><p>in</p><p>pu</p><p>t</p><p>St</p><p>ea</p><p>dy</p><p>st</p><p>at</p><p>e</p><p>L</p><p>n</p><p>(L</p><p>2-</p><p>lik</p><p>e</p><p>gr</p><p>am</p><p>m</p><p>ar</p><p>po</p><p>ss</p><p>ib</p><p>le</p><p>L</p><p>2-</p><p>lik</p><p>e</p><p>gr</p><p>am</p><p>m</p><p>ar</p><p>L</p><p>2-</p><p>lik</p><p>e</p><p>gr</p><p>am</p><p>m</p><p>ar</p><p>L</p><p>2-</p><p>lik</p><p>e</p><p>gr</p><p>am</p><p>m</p><p>ar</p><p>bu</p><p>tn</p><p>ot</p><p>in</p><p>ev</p><p>ita</p><p>bl</p><p>e)</p><p>3.4 Assessing initial-state hypotheses 95</p><p>c. Final outcome – the linguistic competence of L2 learners will be</p><p>effectively identical to that of native speakers (Flynn 1996), any apparent</p><p>differences being attributable to performance factors.</p><p>3.4 Assessing initial-state hypotheses: similarities</p><p>and differences</p><p>It can be seen that there is considerable overlap in the various initial-state</p><p>proposals that we have considered. In consequence, there is also an overlap in</p><p>their predictions, sometimes making it hard to find suitable evidence to distinguish</p><p>between them. Full Transfer FullAccess,Minimal Trees and theValueless Features</p><p>Hypothesis share the assumption that L1 properties are implicated. Full Transfer</p><p>Full Access, the Valueless Features Hypothesis and Full Access without Transfer</p><p>coincide in assuming a full complement of functional categories in the initial</p><p>state. All of the theories assume that L2 functional properties will be present</p><p>in later grammars. All hypotheses assume that interlanguage grammars will be</p><p>UG-constrained in the course of development. The implication of all hypotheses</p><p>except Full Transfer Full Access is that the steady-state grammar of an L2 speaker</p><p>will, in principle at least, converge on L2 functional properties (setting aside the</p><p>possibility that L2 learners may not get adequate input). In the case of Full Transfer</p><p>Full Access, convergence is possible but not guaranteed, depending on the L1s and</p><p>L2s in question. (See chapter 8 for further discussion.) In table 3.1, the four major</p><p>proposals are summarized in terms of what they have to say about the initial state,</p><p>development and the final state.</p><p>What kind of evidence is needed to distinguish between the various claims, as</p><p>well as to resolve the falsifiability problem? As we have seen, where data do not</p><p>support a particular hypothesis, there has been a tendency to resort to the claim</p><p>that the data do not, in fact, come from initial-state learners and that earlier data,</p><p>if found, would support the hypothesis in question. This would appear to render</p><p>several of the hypotheses untestable but in fact it is possible to get round this</p><p>problem, at least in certain cases.</p><p>a. Absence of functional categories. The Minimal Trees Hypothesis is the</p><p>only proposal to argue for absence of functional categories in the initial</p><p>state. Thus, a demonstration of the presence of functional categories in</p><p>the earliest stages is in principle sufficient to show that Minimal Trees</p><p>does not obtain. Here, however, we run into the problem of falsifiability:</p><p>if evidence for functional categories is found in some group of learners,</p><p>it is always possible to claim that their grammars were not, after all, in</p><p>the initial state.</p><p>96 3 The initial state</p><p>b. Feature strength. Evidence of strong or weak feature values in the initial</p><p>state is evidence against Valueless Features. (Evidence for strong feature</p><p>values in earliest stages is also evidence against the Initial Hypothesis of</p><p>Syntax.) Again, in the face of data suggesting that features are not inert,</p><p>it is always possible to claim that learners were not in the initial stage.</p><p>c. L1 effects. The presence of L1 functional categories, features or feature</p><p>strength (or other L1-based properties) at any point (in the initial state,</p><p>during the course of development, or at the endstate) would provide evi-</p><p>dence against Full Access without Transfer, as well as the Minimal Trees</p><p>Hypothesis, since both of these proposals claim that L1-based properties</p><p>will never be found in the functional domain.</p><p>On the other hand, absence</p><p>of L1 properties at some point does not necessarily argue against Full</p><p>Transfer Full Access, since this hypothesis assumes the possibility of re-</p><p>structuring away from the L1-based initial state, after which L1 effects</p><p>should disappear. Again, this raises the question of falsifiability: in the</p><p>absence of L1 effects, how can one tell what an earlier grammar would</p><p>have looked like?</p><p>d. Developmental sequences. The Full Transfer Full Access Hypothesis pre-</p><p>dicts differences between learners of different L1s learning the same L2,</p><p>in particular, different initial states. It is also the only hypothesis to pre-</p><p>dict different development paths for learners of different L1s (at least until</p><p>the learners converge on the relevant L2 properties). Thus, if learners of</p><p>different L1s learning the same L2 can be shown to have the same initial</p><p>state and the same early stages of development, this would be counter-</p><p>evidence to Full Transfer Full Access and would constitute evidence in</p><p>favour of Full Access without Transfer. (Even here, unfortunately, one</p><p>could still resort to the claim that some supposed initial state was not in</p><p>fact the initial state.)</p><p>3.5 Interlanguage representation: defective or not?</p><p>Hypotheses about the interlanguage initial state can be broadly classified</p><p>into two types, as described above. On the one hand are proposals that the initial</p><p>state is a grammar, the L1 grammar, in whole or in part. This contrasts with the</p><p>position that the initial state is not yet a language-specific grammar but, rather,</p><p>UG itself. All the hypotheses that we have considered in this chapter agree that L2</p><p>learners can acquire functional categories and feature values not instantiated in the</p><p>L1, though not necessarily immediately. Despite their differences over the nature</p><p>of the initial state, all these positions agree that interlanguage representations are</p><p>3.5 Interlanguage representation: defective or not? 97</p><p>constrained by UG. Indeed, all these hypotheses are full access theories in some</p><p>sense.</p><p>Nevertheless, althoughUGavailability is assumed, the various hypotheses differ</p><p>quite radically in terms of their assumptions about the nature of an interlanguage</p><p>grammar. There is disagreement over: (a)whether or not a full complement of func-</p><p>tional categories is initially available; (b) whether or not L1 functional categories</p><p>and feature values are found in the initial state; (c) whether or not default feature</p><p>values or inert feature values occur. There is another area of disagreement, which</p><p>is the logical outcome of these different views on the nature of the grammar, which</p><p>relates to the issue of ‘completeness’. On the one hand are theories that presuppose</p><p>that any particular interlanguage grammar will be complete, in the sense that it will</p><p>manifest a full complement of lexical and functional categories, features and fea-</p><p>ture strength. In other words, the interlanguage grammar is a grammar with all the</p><p>properties of natural languages (though not necessarily the L2) and interlanguage</p><p>grammars are not wild (in the sense discussed in chapter 2). The Full Transfer Full</p><p>Access Hypothesis falls into this category, since the initial state is the L1 represen-</p><p>tation, a natural-language grammar, manifesting L1 functional categories, features</p><p>and feature values. When the grammar is subsequently restructured, it remains a</p><p>natural-language grammar. Similarly, Full Access without Transfer is of this type,</p><p>as is the Initial Hypothesis of Syntax; the interlanguage representation is always a</p><p>fully fledged, UG-constrained grammar, including functional categories, features</p><p>and feature values (though these are not seen as stemming from the L1).</p><p>In contrast, theMinimalTreesHypothesis and theValueless FeaturesHypothesis</p><p>imply that an interlanguage grammar in early stages is in some sense temporarily</p><p>defective or impaired, in that it lacks properties that are assumed to be given by UG</p><p>and that are found in the grammars of adult native speakers. As we have seen, the</p><p>Minimal Trees Hypothesis assumes that functional categories are initially lacking;</p><p>in other words, not all UG properties are available at once. The Valueless Features</p><p>Hypothesis assumes that feature strength is initially inert. In both cases, the idea</p><p>is that these properties will eventually be acquired.</p><p>This view of early grammars as temporarily lacking certain characteristics con-</p><p>trasts with yet another position, whereby the interlanguage grammar is said to</p><p>suffer from permanent impairment. Beck (1998a) argues for the Local Impairment</p><p>Hypothesis,whereby inert features are not just a property of early grammars; rather,</p><p>feature strength is never acquirable, on this view. As a result, the interlanguage</p><p>grammar is different in nature from the L1 grammar, from the L2 grammar, and</p><p>from natural-language grammars in general. In other words, the interlanguage rep-</p><p>resentation is never fullyUG-constrained.This impairment is considered tobequite</p><p>local, confined to feature strength (possibly only strength of Infl-related features).</p><p>Others argue that interlanguage grammars suffer from more global impairment,</p><p>98 3 The initial state</p><p>being quite different from natural languages, and not being UG-constrained at all</p><p>(Meisel 1997). Whether the impairment is local or global, interlanguage gram-</p><p>mars are effectively wild on these views. We will consider these views and their</p><p>implications in more detail in chapter 4.</p><p>3.6 Conclusion</p><p>In this chapter, we have considered claims about the initial state, com-</p><p>paring several hypotheses that have focused on the nature and role of functional</p><p>categories and their features in early interlanguage grammars. Full Access without</p><p>Transfer and Full Transfer Full Access would seem to represent the most logical</p><p>possibilities, at least in principle: either UG is the initial state or the L1 gram-</p><p>mar is the initial state. The Minimal Trees Hypothesis and the Valueless Features</p><p>Hypothesis fall somewhere in between: neither UG in its entirety nor the L1 in its</p><p>entirety constitute the initial state. These two proposals are influenced by corre-</p><p>sponding hypotheses for L1 acquisition, Minimal Trees by the Weak Continuity</p><p>Hypothesis and Valueless Features by the hypothesis that features can be under-</p><p>specified in L1 acquisition (Wexler 1994). Even if such hypotheses are correct for</p><p>L1 acquisition (and this is much debated), the motivation for assuming that they</p><p>apply to L2 grammars is not strong.</p><p>Whilemany of the hypotheses considered in this chapter are directed specifically</p><p>at the initial state, they also make predictions for later development and ultimate</p><p>attainment, topics that will be considered in later chapters.</p><p>Topics for discussion</p><p>� The initial-state proposals discussed in this chapter crucially depend on</p><p>the assumption that UG is not transformed into a particular grammar in</p><p>the course of L1 acquisition. If UG does turn into a language-specific</p><p>grammar, what are the implications for theories of L2 acquisition?</p><p>� Several L1 acquisition theories assume that early grammars are in some</p><p>sense defective (even though this is not always recognized). For example,</p><p>underspecified Tense (Wexler 1994) or underspecified number (Hoekstra</p><p>and Hyams 1998) appear to be characteristic only of grammars in the</p><p>course of acquisition. Does this mean that grammars in L1 acquisition</p><p>can, after all, be wild? What are the implications?</p><p>� According to Lakshmanan and Selinker (2001), theories that argue for</p><p>transfer (whether full or partial) are in danger of incurring the comparative</p><p>Suggestions for additional reading 99</p><p>fallacy. That is, if interlanguage data are to be assessed in their own right,</p><p>then it is as much of a problem to try and explain them in terms of the</p><p>L1 grammar as it is to compare the interlanguage grammar with the L2</p><p>grammar. Do you agree?</p><p>� Design a series of ‘thought experiments’ to show how one might falsify</p><p>each of the hypotheses discussed in this chapter.</p><p>Suggestions for additional reading</p><p>� A special issue of Second</p><p>Language Research (vol. 12.1, 1996), edited</p><p>by Schwartz and Eubank, is devoted to the initial state, containing some</p><p>of the papers discussed in this chapter.</p><p>� Hawkins (2001a) proposes a combination of Minimal Trees and Full</p><p>Transfer. He concurs with Vainikka and Young-Scholten that the initial</p><p>state has only lexical categories. However, when functional categories</p><p>emerge, his assumption is that they show L1 characteristics. Hawkins</p><p>argues against Full Access in the functional domain (see chapter 4).</p><p>4</p><p>Grammars beyond the initial state:</p><p>parameters and functional categories</p><p>4.1 Introduction</p><p>In the previous chapter, a variety of hypotheses were considered as to</p><p>the nature of the grammatical representations adopted by learners in the earliest</p><p>stages of L2 acquisition. In this chapter, we examine developing interlanguage</p><p>grammars, exploring the issue of whether grammars change over time and, if</p><p>so, in what respects. We will consider whether interlanguage grammars can be</p><p>characterized in terms of parameters of UG, concentrating particularly on the</p><p>situation that obtains when the L1 and L2 differ in parameter values.</p><p>As discussed in chapter 2, there is considerable evidence to suggest that inter-</p><p>language grammars are constrained by invariant principles of UG, since learners</p><p>are sensitive to subtle properties of the L2 that are underdetermined by the input.</p><p>L2 learners successfully acquire highly abstract unconscious knowledge, despite</p><p>a poverty of the L2 stimulus, suggesting that this knowledge must originate from</p><p>UG. Nevertheless, in some cases one cannot totally eliminate the L1 as the source</p><p>of such abstract knowledge: even where languages differ considerably at the sur-</p><p>face level, the same universal principles may apply at a more abstract level. For</p><p>this reason, the issue of parameters and parameter resetting is of crucial impor-</p><p>tance in assessing the role of UG in L2 acquisition. If the L1 and L2 differ in their</p><p>parameter settings and if the learner’s linguistic behaviour is consistent with pa-</p><p>rameter values appropriate for the L2, this strongly supports the position that UG</p><p>constrains interlanguage grammars. Conversely, failure to achieve L2 parameter</p><p>settings is often taken as clear evidence against a role for UG. However, it will be</p><p>suggested in this chapter (and subsequently) that failure to acquire L2 parameter</p><p>settings does not necessarily indicate failure of UG.</p><p>4.2 Parameters in interlanguage grammars</p><p>In this chapter, we will consider two general positions, as well as two</p><p>subcategories within each, on the status of parameters in interlanguage grammars.</p><p>100</p><p>4.2 Parameters in interlanguage grammars 101</p><p>The first position argues for a breakdown in parametric systems, either global or</p><p>local. The implication of global breakdown is that there are no parameters at all</p><p>in interlanguage grammars; claims for more local breakdown, on the other hand,</p><p>assume that parameters are found in interlanguage grammars but that some of them</p><p>are defective.</p><p>Proponents of global breakdown in the parameter system include Clahsen and</p><p>Hong (1995) and Neeleman and Weerman (1997), who argue that interlanguage</p><p>grammars are construction specific, hence very different from UG-constrained</p><p>grammars. Their claim is that, in L2 acquisition, each construction theoretically</p><p>associated with a given parameter has to be learned separately, on a construction-</p><p>by-construction basis. According to some researchers, this is achieved by means</p><p>of what is often called pattern matching: the learner concentrates on surface prop-</p><p>erties, unconsciously taking account of similarities and differences across various</p><p>linguistic forms (Bley-Vroman 1997). Such proposals are characteristic of the</p><p>Fundamental Difference Hypothesis (Bley-Vroman 1990), according to which</p><p>UG does not constrain interlanguage grammars (or does so only weakly, by means</p><p>of properties that can be ‘reconstructed’ via the L1).</p><p>Claims for a more local breakdown in interlanguage parameters are made in</p><p>the context of feature strength. Extending the Valueless Features Hypothesis (see</p><p>chapter 3, section 3.2.3), Beck (1998a) argues that ‘inert’ feature values are a per-</p><p>manent property of interlanguage grammars. In consequence, some parameters are</p><p>never set, neither the L1 value nor the L2 value being realized in the interlanguage</p><p>grammar.</p><p>Breakdown (whether global or local) implies that there will be impair-</p><p>ment to grammatical representations: interlanguage grammars are not fully UG-</p><p>constrained andmay demonstrate properties which are not otherwise characteristic</p><p>of natural language. In other words, they are in some sense defective or ‘wild’ (see</p><p>chapter 2, section 2.3).</p><p>The alternative perspective maintains that interlanguage grammars are unim-</p><p>paired: they can be characterized in terms of UG parameters and in general exhibit</p><p>properties of natural language. Again, there are two types of account which fall</p><p>into this category. The first is the No Parameter Resetting Hypothesis, according</p><p>to which only L1 parameter settings are exemplified in interlanguage grammars</p><p>(Hawkins 1998;Hawkins andChan 1997). Since the L1 is a natural language, inter-</p><p>language grammars can indeed be described in terms of UG parameters. However,</p><p>the range of parametric options is totally restricted, ‘new’ parameter settings being</p><p>unavailable.</p><p>The No Parameter Resetting Hypothesis contrasts with the hypothesis that pa-</p><p>rameters can be reset. That is, interlanguage grammars can realize parameter values</p><p>distinct from those found in the L1; these values may either be appropriate for the</p><p>102 4 Grammars beyond the initial state</p><p>L2 or they may be settings found in other languages. Thus, full access to UG is</p><p>assumed, with new parameter values being, in principle, achievable. Most of the</p><p>initial-state hypotheses considered in chapter 3 adopt this view. According to the</p><p>Full Transfer Full Access Hypothesis, parameters in interlanguage grammars will</p><p>initially be set at the values that obtain in the L1 (i.e. full transfer). In response to</p><p>L2 input, parameters can be reset to values more appropriate to the L2 (i.e. full</p><p>access). On this kind of account, developing grammars will be characterized by</p><p>parameter resetting, from the L1 value to some other value. According to Full</p><p>Access without Transfer, on the other hand, L1 settings are never adopted; rather,</p><p>appropriate L2 settings are, in principle, effective immediately.1 On this view, then,</p><p>there is parameter setting (as in L1 acquisition) but there is no need for resetting.</p><p>To sum up, in the rest of this chapter four perspectives on parameters in interlan-</p><p>guage grammarswill be considered: (i) global impairment, implying no parameters</p><p>at all; (ii) local impairment, or breakdown in the case of some parameters; (iii) no</p><p>parameter resetting, according towhich only L1 settings are available; (iv) parame-</p><p>ter resetting, which assumes the possibility of acquiring parameter settings distinct</p><p>from those found in the L1. Under the first two views, interlanguage grammars</p><p>fail to conform to properties of natural language. Under the two latter perspec-</p><p>tives, interlanguage grammars are natural-language systems in which parameters</p><p>are instantiated.</p><p>4.3 Global impairment</p><p>If interlanguage grammars are constrained by UG, syntactic and morpho-</p><p>logical properties related by a single parameter should cluster together. In other</p><p>words, characteristics typical of either the L1 parameter setting or the L2 setting</p><p>should be exemplified. Given a range of phenomena which are associated together</p><p>under a particular parameter setting, these same phenomena should be character-</p><p>istic of the interlanguage grammar and should, ideally, be acquired at more or less</p><p>the same time. Consequently, researchers who argue against parameters in L2 ac-</p><p>quisition seek to support their position by demonstrating the absence of clustering</p><p>effects in interlanguage grammars.</p><p>4.3.1 Breakdown of the Null Subject Parameter</p><p>One such case is advanced by Clahsen and Hong (1995: 59), who put</p><p>forward</p><p>what they call the ‘weak UG’ view. On this view, while the grammars of</p><p>adult learners are constrained by UG principles (via the L1), access to parameters</p><p>has been lost. In otherwords, there is a total breakdown in the domain of parameters</p><p>4.3 Global impairment 103</p><p>only; principles remain intact (or intact in so far as they are found in the L1</p><p>grammar). Clahsen and Hong’s hypothesis is that properties that would ‘co-vary’</p><p>(i.e. cluster) under some parameter setting in L1 acquisition no longer do so in L2</p><p>acquisition.</p><p>In order to investigate this hypothesis, Clahsen and Hong look at adult L2 ac-</p><p>quisition of German by speakers of Korean, in the context of the Null Subject</p><p>Parameter. (See chapter 1, section 1.4.) Korean is a [+null subject] language,</p><p>permitting empty subjects in a variety of contexts. German, in contrast, is a</p><p>[–null subject] language; subjects must be overt, with a few limited exceptions.</p><p>Examples in (1) and (2), from Korean and German respectively, illustrate the rel-</p><p>evant properties (Clahsen and Hong’s (4)). As can be seen by comparing (2) and</p><p>(3), subject pronouns must be overt in German.</p><p>(1) Peter-ka Inge-lul sarangha-n-tako malha-n-ta.</p><p>Peter-nom Inge-acc love-pres-that say-pres-dec</p><p>‘Peter says that (he) loves Inge.’</p><p>(2) *Peter sagt, dass Inge liebt.</p><p>Peter says that Inge loves</p><p>‘Peter says that (he) loves Inge.’</p><p>(3) Peter sagt, dass er Inge liebt.</p><p>Peter says that he Inge loves</p><p>‘Peter says that he loves Inge.’</p><p>Clahsen and Hong adopt an account that attributes the distribution of null sub-</p><p>jects to two independent but interacting parameters (Jaeggli and Safir 1989; Rizzi</p><p>1986). On the one hand, null subjects must be licensed. That is, there must be</p><p>some property that permits null subjects in principle. Licensing is a necessary but</p><p>not a sufficient condition. In addition to being licensed, a null subject must be</p><p>identified. In order to interpret a missing argument, one must be able to work out</p><p>what it refers to; there must be some way of recovering the content of the null</p><p>subject from other properties of the sentence. Various types of identification have</p><p>been proposed in the literature. Rich verbal agreement (formalized in terms of</p><p>so-called phi-features in Agr) allows null subjects to be identified in Romance lan-</p><p>guages such as Italian and Spanish (see Jaeggli and Safir 1989 for an overview). In</p><p>languages lacking agreement, such as Chinese, Japanese and Korean, a preceding</p><p>topic in the discourse provides the means to identify a null element (Huang 1984).</p><p>Null subjects are in fact licensed in German as well as Korean. (Hence, German</p><p>permits null expletives, which do not need to be identified, since they are non-</p><p>referential.) Thus, the licensing parameter value is the same in both languages.</p><p>The two languages differ, however, as to identification. Null subjects are identified</p><p>in Korean by a preceding topic. German, on the other hand, is a language with</p><p>104 4 Grammars beyond the initial state</p><p>Table 4.1 Licensing and identification of null subjects</p><p>Korean German Italian</p><p>Licensed Yes Yes Yes</p><p>Identified Yes, via topics No Yes</p><p>Agr = [−pronominal] Agr = [+pronominal]</p><p>relatively rich verbal agreement. Even so, null subjects cannot be identified. This is</p><p>because Agr lacks the feature [+pronominal], required for identification purposes.</p><p>In Romance null subject languages, on the other hand, Agr is [+pronominal] and</p><p>null subjects can be identified. These differences are summarized in table 4.1</p><p>(adapted from Clahsen and Hong’s (5)).</p><p>In the case of German L2 acquisition, Korean speakers have to reset the pa-</p><p>rameter that determines how null subjects are identified. They have to establish</p><p>two things: (i) German is a language with agreement; (ii) Agr is non-pronominal.</p><p>According to Clahsen and Hong, in the L1 acquisition of German, these two</p><p>properties co-vary. That is, children allow null subjects in L1 German until they</p><p>acquire the agreement paradigm (Clahsen 1990/1991; Clahsen and Penke 1992).</p><p>Once subject-verb agreement is acquired, subjects are used systematically. In other</p><p>words, when agreement emerges in L1 German, the child recognizes that Agr is</p><p>[–pronominal] and null subjects are eliminated from the grammar.2</p><p>According to Clahsen and Hong (1995), if properties have been shown to co-</p><p>vary in L1 acquisition, in this case, presence of agreement (non-pronominal Agr)</p><p>with absence of null subjects, then one expects the same to obtain in L2 acquisition</p><p>if parameters are unimpaired. If such clustering fails to occur, this demonstrates</p><p>that interlanguage grammars cannot be characterized in terms of UG parameters.</p><p>Clahsen and Hong test for evidence of clustering, using the sentence-matching</p><p>procedure, which requires subjects to identify pairs of sentences as the same or dif-</p><p>ferent (see chapter 3, section 3.2.3.1). Recall that it takes native speakers longer to</p><p>recognize ungrammatical sentence pairs as being the same (Freedman and Forster</p><p>1985). Clahsen and Hong hypothesize that, if L2 learners have successfully ac-</p><p>quired the German value of the identification parameter, it should take longer to</p><p>recognize ungrammatical sentence pairs involving incorrect agreement than gram-</p><p>matical pairs with correct agreement; at the same time, ungrammatical pairs with</p><p>null subjects should take longer than grammatical pairs with lexical subjects. (See</p><p>box 4.1 for examples of the stimuli.)</p><p>As a group, the native speakers showed significant differences in response times</p><p>to grammatical and ungrammatical sentence pairs, being slower on the ungram-</p><p>matical sentences, as expected. This was true of sentences testing for ± agreement,</p><p>4.3 Global impairment 105</p><p>Box 4.1 Global impairment – null subjects (Clahsen and Hong 1995)</p><p>Languages: L1 = Korean, L2 = German.</p><p>Task: Sentence matching.</p><p>Sample stimuli:</p><p>Grammatical (+agreement) Grammatical (−null subjects)</p><p>Peter und Inge wohnen in Düsseldorf Maria sagt, dass sie die Zeitung liest</p><p>Peter und Inge wohnen in Düsseldorf Maria sagt, dass sie die Zeitung liest</p><p>(Peter and Inge live-3pl in Düsseldorf ) (Maria says that she the newspaper</p><p>reads)</p><p>Ungrammatical (−agreement) Ungrammatical (+null subjects)</p><p>Peter und Inge wohnt in Düsseldorf Maria sagt, dass oft die Zeitung liest</p><p>Peter und Inge wohnt in Düsseldorf Maria sagt, dass oft die Zeitung liest</p><p>(Peter and Inge lives-3sg in Düsseldorf) (Maria says that often the</p><p>newspaper reads)</p><p>Results:</p><p>Table 4.1.1 Numbers of subjects whose grammars show clustering</p><p>L2 learners (n = 33) −agreement +agreement</p><p>+null subject 2 5</p><p>−null subject 13 13</p><p>Native speakers (n = 20)</p><p>+null subject 0 1</p><p>−null subject 1 18</p><p>as well as sentences testing for ± null subjects. In order to determine whether</p><p>these properties do or do not co-vary in the grammars of individuals, Clahsen</p><p>and Hong present results in terms of how individuals behave with respect to the</p><p>two properties in question. (See table 4.1.1.) Native speakers are predicted to re-</p><p>spond faster to grammatical pairs with overt subjects, showing that they know</p><p>that German is not a null-subject language [–null subject], as well as to grammat-</p><p>ical agreement pairs, showing that they know that German is a language with</p><p>overt agreement [+agreement]; 18 out of 20 native speakers behaved as pre-</p><p>dicted. In other words, almost all native speakers recognized the necessity of</p><p>106 4 Grammars beyond the initial state</p><p>agreement in German at the same time as recognizing the impossibility of null</p><p>subjects.</p><p>TheL2 learners presented amore varied picture. (See table 4.1.1.) Thirteen out of</p><p>33 of theKorean speakers showed the same response pattern as the controls, namely</p><p>slower responses to the ungrammatical sentences of both types, suggesting that</p><p>they had successfully reset the parameter, recognizing that German requires overt</p><p>subjects and agreement. Clahsen and Hong suggest that these learners may simply</p><p>have acquired the two properties independently but this, of course, is also true of</p><p>the native speakers. When looking at properties of a grammar at a particular point</p><p>in</p><p>time, one cannot tell what the grammar might have been like at an earlier stage.</p><p>This requires comparing subjects at different stages of development or following</p><p>the same subjects over an extended period of time.</p><p>Two subjects failed to distinguish in their response times between grammati-</p><p>cal and ungrammatical sentences in either condition. This behaviour, in fact, is</p><p>consistent with operating under the L1 value of the parameter: they have not yet</p><p>acquired the fact that German requires agreement and overt subjects, treating it,</p><p>rather, as a null subject language without agreement, like Korean. Of more inter-</p><p>est are the subjects whose response latencies show a significant grammaticality</p><p>effect for only one of the two properties (±agreement or ±null subjects). Five</p><p>subjects failed to distinguish in their response times between overt and null subject</p><p>sentences, suggesting that the interlanguage is [+null subject]. At the same time,</p><p>they did distinguish in their response times between sentences with correct and</p><p>incorrect agreement, suggesting that the interlanguage is [+agreement]. Thirteen</p><p>subjects showed the reverse pattern; in their interlanguage grammars, German is</p><p>not a null subject language but it also does not have overt agreement. In these</p><p>cases, the two properties fail to co-vary, which is taken by Clahsen and Hong as</p><p>evidence of a failure of parameters; the presence of agreement is presumed to be</p><p>the trigger for loss of null subjects, so one should not be able to have one without</p><p>the other.</p><p>However, an alternative account is possible. Clahsen and Hong’s assumption</p><p>about the co-varying of agreement and the requirement for overt subjects is clearly</p><p>incorrect for languages like Italian,where rich agreement andnull subjects coincide</p><p>because Agr is [+pronominal] (see table 4.1). Indeed, the results of the fiveKorean</p><p>speakers who distinguished in response latencies between the ungrammatical and</p><p>grammatical agreement pairs but not between pairs with or without null subjects</p><p>are consistent with the Italian value of the identification parameter. While it may</p><p>be true that these Korean speakers may not have acquired the German parameter</p><p>value, they appear to have acquired a value other than the Korean one, treating</p><p>German like Italian, a null subject language with rich agreement. If learners of</p><p>German acquire rich agreement while also treating Agr as [+pronominal], this</p><p>4.3 Global impairment 107</p><p>is precisely the result one would expect. On such an analysis, null subjects are</p><p>grammatical and agreement is grammatical, hence learners show response latency</p><p>differences in the case of the ± agreement sentences but not in the case of the</p><p>± null subject sentences.</p><p>Assuming this reanalysis of Clahsen and Hong’s results, 20 of their 33 subjects</p><p>demonstrated behaviour consistent with some value of the identification parameter</p><p>(Korean value, German value or Italian value). On the other hand, the results</p><p>from the remaining 13 subjects who took longer on the sentence pairs involving</p><p>null subjects but did not distinguish between grammatical and ungrammatical</p><p>agreement pairs are, ostensibly, more supportive of Clahsen and Hong’s claim that</p><p>there are no parameters in interlanguage grammars.</p><p>Even here it is not necessarily the case that the results are as problematic as</p><p>Clahsen and Hong suggest. Like Vainikka and Young-Scholten (1994, 1996a, b)</p><p>and Eubank (1993/1994, 1994, 1996), Clahsen and Hong equate acquisition of</p><p>abstract agreement with accuracy in surface morphology. However, as will be</p><p>discussed in greater detail in chapter 6, there is a difference between having ab-</p><p>stract features or categories (in this case, agreement) realized in the grammar</p><p>and knowing the particular surface morphology associated with them. L2 learn-</p><p>ers might know that German has agreement at an abstract level, and that this</p><p>is non-pronominal. At the same time, they might not yet have fully acquired</p><p>the morphology by which agreement is realized on the surface, thus failing to</p><p>recognize deviant morphology in some cases. If so, the failure to show differ-</p><p>ences in response times between grammatical and ungrammatical sentence pairs</p><p>testing agreement is relatively uninformative. We simply cannot tell what the</p><p>status of abstract agreement is in these grammars, hence we cannot conclude any-</p><p>thing about this group of subjects. (In later chapters, we will address in more</p><p>detail the question of how one can show that some property is present at an ab-</p><p>stract level in the absence of appropriate morphology.) In contrast, in the case</p><p>of learners who do distinguish between sentence pairs involving grammatical</p><p>and ungrammatical agreement morphology, it seems reasonable to conclude that</p><p>they have both the relevant abstract representation and the appropriate surface</p><p>realization.</p><p>A methodological issue with the sentence-matching procedure is worth men-</p><p>tioning here, relating to the sentence pairs contrasting overt and null subjects.</p><p>At issue is whether learners distinguish between grammatical pairs of sentences</p><p>like (3) and ungrammatical pairs like (2). The sentence-matching methodology</p><p>requires that grammatical and ungrammatical pairs be of equivalent length, so</p><p>that grammaticality is the only difference between them. Since a sentence with</p><p>an overt subject is longer than one without an explicit subject, Clahsen and Hong</p><p>insert adverbs in the null subject sentences (see examples in box 4.1). This means</p><p>108 4 Grammars beyond the initial state</p><p>that the grammatical and ungrammatical pairs differ in two ways, presence versus</p><p>absence of an overt subject and presence versus absence of an adverb. Thus, it is</p><p>not clear that the sentence-matching procedure is in fact successful in isolating the</p><p>relevant property, making the results harder to interpret.</p><p>4.3.2 Breakdown of a word-order parameter</p><p>Using the Null Subject Parameter as a particular example, Clahsen and</p><p>Hong (1995) argue that there are no parameters in L2 acquisition, based on pre-</p><p>sumed absence of clustering. In this section, we will continue to pursue this issue,</p><p>considering another parameter, the OV/VO parameter (Neeleman and Weerman</p><p>1997). If interlanguage representations do not conform to UG parameter settings,</p><p>the issue arises as to the nature of the interlanguage grammatical system in the</p><p>absence of parameters. Learners must nevertheless come up with analyses of the</p><p>L2 input; they must be able to interpret and produce L2 sentences. A number of</p><p>researchers have proposed that interlanguage grammars differ from native-speaker</p><p>grammars in being ‘construction specific’ (Bley-Vroman 1996, 1997).</p><p>One explicit proposal for a construction-specific interlanguage grammar is ad-</p><p>vanced by Neeleman and Weerman (1997) who maintain that the grammars of L1</p><p>acquirers (and adult native speakers) are constrained by a word-order parameter</p><p>which is lacking in the interlanguage grammars of L2 learners. The parameter that</p><p>they propose is theOV/VOparameter, which accounts for head-final or head-initial</p><p>word order crosslinguistically.3 Dutch and English, the languages under investi-</p><p>gation, differ as to the values of this parameter that they instantiate. The settings</p><p>of the parameter, as Neeleman and Weerman conceive it, have a range of asso-</p><p>ciated consequences, namely: (i) whether or not scrambling is permitted (that is,</p><p>whether the direct object can appear distant from the verb which assigns case to it);</p><p>(ii) the distribution of particles; (iii) the possibility of extraction of objects from</p><p>particle constructions; and (iv) exceptional casemarking (ECM) (that is, the ability</p><p>of some verbs to case mark an embedded subject). The properties in question are</p><p>listed in table 4.2.</p><p>Examples are given in (4) to (8) (from Neeleman and Weerman 1997). The</p><p>sentences in (4) show the basic word-order difference between Dutch, where VPs</p><p>are head final, and English, where VPs are head initial. In (5), we see that the</p><p>direct object in Dutch can be scrambled; that is, it does not have to occur adjacent</p><p>to the verb, in contrast</p><p>to English. The sentences in (6) show that particles like</p><p>up in English can appear before or after the direct object, whereas in Dutch the</p><p>particle must be adjacent to the verb. A further property of particle constructions is</p><p>illustrated in (7): in both languages, extraction of a phrase fromwithin the object of</p><p>4.3 Global impairment 109</p><p>Table 4.2 The OV/VO parameter</p><p>Word order OV VO</p><p>Scrambling + −</p><p>O adverb V * V adverb O</p><p>Particles Prt V V (X) Prt</p><p>Always adjacent May be separated</p><p>Extraction from object of particle verb + +/−</p><p>ECM − +</p><p>a verb-particle construction is possible if the verb and particle are adjacent. Where</p><p>the verb and particle are separated (only possible in English), such extraction is</p><p>not permitted. Finally, (8) illustrates contrasts in exceptional case marking, where</p><p>the subject of a non-finite embedded clause receives accusative case from the verb</p><p>in a higher clause. This is not possible in Dutch (except to a very limited extent)</p><p>but is possible in English.</p><p>(4) OV vs. VO</p><p>a. Ik heb de poes gezien.</p><p>I have the cat seen</p><p>b. I have seen the cat.</p><p>(5) Scrambling vs. case adjacency (O adv V/*V adv O)</p><p>a. . . . dat Jan langzaam het boek las.</p><p>. . . that John slowly the book read</p><p>b. . . . dat Jan het boek longzaam las.</p><p>. . . that John the book slowly read</p><p>c. John read the book slowly.</p><p>d. *John read slowly the book.</p><p>(6) Particles</p><p>a. . . . dat Jan Marie geregeld uit lacht.</p><p>. . . that John Mary regularly out laughs</p><p>b. . . . *dat Jan Marie uit geregeld lacht.</p><p>. . . that John Mary out regularly laughs</p><p>c. John looks up the information.</p><p>d. John looks the information up.</p><p>(7) Extraction from object of a verb particle construction</p><p>a. Waar heeft Jan informatie over op gezocht?</p><p>what has John information about up looked</p><p>b. What did John look up information about?</p><p>c. *What did John look information about up?</p><p>110 4 Grammars beyond the initial state</p><p>(8) ECM</p><p>a. *dat Jan Marie verwacht Shakespeare te lezen.</p><p>that John Mary expects Shakespeare to read</p><p>b. John expects Mary to read Shakespeare.</p><p>The question arises as to how to determine whether a set of syntactic properties</p><p>is indeed unified under some parameter setting in L1 acquisition. If structures</p><p>which have been independently identified on linguistic grounds as likely to fall</p><p>under some parameter emerge at about the same time, this is usually considered</p><p>to provide strong support for the claim that they are indeed associated under some</p><p>parameter setting (Hyams 1986; Snyder 1995a; Snyder and Stromswold 1997).</p><p>Because other factors, such as sentence complexity, may come into play, Neeleman</p><p>and Weerman make the questionable assumption that the crucial issue is not that</p><p>the properties in question should be acquired together but that they be acquired in</p><p>an error-free manner. However, this proposal is problematic: there are many things</p><p>that children acquire in an error-free manner which should not be included in this</p><p>particular parameter. Furthermore, there are properties associated with parameters</p><p>in L1 acquisition that are not acquired in an error-free manner. (For example,</p><p>according to the analysis of null subjects offered by Clahsen and Hong (section</p><p>4.3.1), German-speaking children initially mistakenly assume that German is a</p><p>null subject language, with topic identification of null subjects.)</p><p>Using spontaneous production data available on the CHILDES database</p><p>(MacWhinney 1995), as well as from other sources, Neeleman and Weerman</p><p>show that basic word-order acquisition is error free for both Dutch and English</p><p>acquired as first languages. As for scrambling, violations of case adjacency by</p><p>children learning English are non-existent. Rather, verbs and direct objects are</p><p>found together and are never interrupted by some other constituent such as an</p><p>adverb. In L1 Dutch, on the other hand, both scrambled and non-scrambled word</p><p>orders are found. (In fact, Neeleman and Weerman are incorrect in stating that</p><p>acquisition of scrambling is error free in L1 Dutch. Schaeffer (2000) shows that</p><p>Dutch 2-year-olds fail to scramble in contexts where scrambling is obligatory</p><p>for adults.) ECM constructions are common in L1 English and infrequent in L1</p><p>Dutch, where they are limited to the subset of contexts in which they can occur</p><p>in that language. In the case of verb-particle constructions, in the Dutch L1 data,</p><p>the particle is always adjacent to the verb, whereas in English it appears either</p><p>adjacent or separated. Neeleman and Weerman offer no data on extraction of</p><p>objects from verb-particle constructions on the grounds that the relevant construc-</p><p>tions do not occur in early L1 data from either language. To summarize, the L1</p><p>data suggest early acquisition of word order (OV/VO) and somewhat later ac-</p><p>quisition of the associated properties, while for one of them there is no relevant</p><p>evidence.</p><p>4.3 Global impairment 111</p><p>Box 4.2 Construction-specific grammars (Neeleman and Weerman</p><p>1997)</p><p>Languages: L1 = English, L2 = Dutch and L1 = Dutch, L2 = English.</p><p>Task: Grammaticality judgments.</p><p>Sample stimuli (English version):</p><p>Sentence types Grammatical Ungrammatical</p><p>OV/VO David wants to finish his</p><p>homework.</p><p>I think I have that cat seen</p><p>before.</p><p>Scrambling My father will wash his car</p><p>tomorrow.</p><p>Andrea is reading slowly</p><p>the newspaper.</p><p>Particles (and related</p><p>constructions)</p><p>Steve will paint the door</p><p>black.</p><p>The barman kicked right</p><p>out the man.</p><p>Extraction from object of</p><p>particle verb</p><p>What did Cindy cut open a</p><p>box of?</p><p>What did you send a</p><p>message about back?</p><p>ECM The teacher expects us to</p><p>do our homework.</p><p>Brenda wants very much</p><p>Kelly to be her best friend</p><p>again.</p><p>Results:</p><p>Table 4.2.1 Accuracy on five constructions (in %)</p><p>OV/VO Scrambling Particles Extraction ECM</p><p>English NS (n = 14) 98 93 97 89 96</p><p>Dutch NS (n = 15) 100 96 100 90 91</p><p>L2 English (n = 15) 88 77 67 41 73</p><p>L2 Dutch (n = 14) 86 88 72 60 46</p><p>The situation for L2 acquisition was investigated by means of an experiment</p><p>using a grammaticality-judgment and correction task. (See box 4.2.) This was</p><p>a bidirectional study which looked at the acquisition of Dutch and English by</p><p>adolescent speakers of English and Dutch, respectively. Unfortunately, however,</p><p>the two groups had not had equivalent amounts of exposure to the L2. A criterion</p><p>of 75% accuracy (i.e. in accepting grammatical and rejecting ungrammatical test</p><p>sentences) is taken as indicating knowledge of a particular construction. Overall</p><p>accuracy rates are given in table 4.2.1, where it can be seen that the native-speaker</p><p>controls for both languages responded as expected, showing accuracy on all five</p><p>constructions. The L2 learners are relatively accurate only on word order and</p><p>scrambling; again, this applies to both L2s. Furthermore, when individual data are</p><p>112 4 Grammars beyond the initial state</p><p>considered, most learners did not demonstrate accuracy on all five constructions</p><p>in the L2. Nor did they demonstrate behaviour consistent with the L1 value of the</p><p>parameter. Thus, the data suggest that the various properties are independent of</p><p>each other and have to be learned on a construction-by-construction basis, rather</p><p>than being acquired together, linked by a parameter.</p><p>Neeleman andWeerman claim to have shown that there are considerable differ-</p><p>ences between L1 and L2 acquisition of the five constructions. While comparison</p><p>between L1 and L2 acquisition can be very useful when investigating claims about</p><p>UG, such comparisons are not without problems. In this particular case, different</p><p>methodologies and different kinds of data are compared and a different criterion</p><p>for success is adopted for each situation. Spontaneous production data are used</p><p>in the case of the L1 acquirers; we are provided with a few examples, drawn</p><p>from different children, and there is no attempt at quantifying the data. The re-</p><p>searchers conclude, incorrectly, that acquisition is error free, largely because of</p><p>absence of errors (but see Schaeffer 2000). However, one cannot simply conclude</p><p>that the non-occurrence of some structure in production data is due to the fact that</p><p>it is ungrammatical</p><p>for the child; non-occurrence might simply reflect an acci-</p><p>dental gap in the data. The L2 learners were asked to accept or reject sentences;</p><p>we have no idea what the L1 acquirers would have done given a grammaticality-</p><p>judgment task, in particular whether they would have rejected the ungrammatical</p><p>sentences.</p><p>A serious flaw in the L1 data is that they are months, sometimes years, apart:</p><p>the earliest word-order data come from children aged 1;9; scrambling data are</p><p>included from a child of 3;3; the oldest child for whom data is cited is a 4;10-year-</p><p>old who produces a verb-particle construction. At no point are we provided with</p><p>any data showing that the same child or children have acquired all (or at least some)</p><p>of the constructions. This considerably weakens the argument for this particular</p><p>parameter in L1 acquisition.</p><p>The L2 data are drawn from a grammaticality-judgment task which includes a</p><p>wider range of constructions than those reported on for L1 acquisition. For ex-</p><p>ample, in the case of particle constructions, the L1 data are simply reported in</p><p>terms of whether or not the particle can be separated from the verb (no in Dutch;</p><p>yes in English). The grammaticality-judgment task, in contrast, includes resul-</p><p>tatives (e.g. Steve will paint the door black) within this category (see box 4.2).</p><p>While there may indeed be valid theoretical reasons to link resultatives to par-</p><p>ticles, since they both implicate a telic interpretation (Slabakova 2001; Snyder</p><p>1995b) (see chapter 3, section 3.2.1.1), it is unreasonable to include them in the</p><p>L2 study without having relevant L1 data. Furthermore, no detail is given as to</p><p>what is going onwithin any one construction when it includes several subtypes; for</p><p>example, we are not told to what extent problems with resultatives as opposed to</p><p>4.3 Global impairment 113</p><p>particles might be contributing to the results. In the case of the L1 children, there is</p><p>no evidence whatsoever concerning extraction from objects of verb-particle com-</p><p>binations (since there were no relevant examples in the production data) and yet</p><p>this is still included as one of the structures on which the L2 learners are tested.</p><p>In fact, the L1 data at best only weakly support the argument that a parameter</p><p>is implicated: the five constructions do not emerge at anything like the same time</p><p>in the course of L1 acquisition, making the claim for a cluster of properties less</p><p>than compelling. Neeleman and Weerman dismiss this objection on the grounds</p><p>that other factors (e.g. limited processing capacity) lead to delays in the emergence</p><p>of some of the more complex constructions involved. This may well be true but</p><p>it applies equally to L2 acquisition: other factors may explain why L2 learners</p><p>perform less successfully on some of the constructions than on others. For example,</p><p>one of the constructions (ECM) involves embedded clauses, and some of the test</p><p>items included embedded clauses even when this was not necessary for the issue</p><p>being tested, but there is no independent attempt to show that these L2 learners</p><p>have mastered embedding. As many of them have been learning the L2 for only a</p><p>year (two hours a week), it is conceivable that some of the sentences in the task</p><p>were simply too difficult.</p><p>In conclusion, it is true that theL2 data suggest that learners have notmastered all</p><p>the properties in question. However, given the absence of compelling evidence for</p><p>a parameter in L1 acquisition, it would be premature to conclude that interlanguage</p><p>grammars are construction specific in this case whereas L1 grammars are not. If</p><p>anything, data from both L1 and L2 acquisition fail to implicate a parameter here.</p><p>4.3.3 Global impairment: assessment</p><p>Clahsen and Hong (1995) and Neeleman and Weerman (1997) start from</p><p>a reasonable premise, namely that properties that are the consequence of a par-</p><p>ticular parameter setting should cluster together in L2 acquisition, if parameters</p><p>of UG are unimpaired. In both cases, comparisons are made with L1 acquisition</p><p>(emergence of overt subjects and morphological agreement in L1 German; error-</p><p>free acquisition of properties associated with OV or VO in L1 Dutch and English).</p><p>Data are then presented which are interpreted as showing that the properties in</p><p>question are dissociated in L2 acquisition, hence parameters are not available.</p><p>In both studies, there are methodological and theoretical grounds for question-</p><p>ing this interpretation of the results. In the case of Clahsen and Hong’s study,</p><p>sentence pairs which tested overt versus null subjects in the sentence-matching</p><p>task differed in more than one respect, so that one cannot fully determine what</p><p>the learners are responding to, while failure to distinguish between grammati-</p><p>cal and ungrammatical agreement is open to two quite different interpretations,</p><p>114 4 Grammars beyond the initial state</p><p>depending on how one perceives the relationship between overt morphology and</p><p>abstract agreement. (See chapter 6 for further discussion.) As for Neeleman and</p><p>Weerman’s study, they adopt very different criteria for determining whether or not</p><p>a set of constructions is the consequence of one parameter. For L1 acquisition, they</p><p>argue that it is sufficient to show that the presumed cluster of properties emerges in</p><p>an error-free fashion; for L2 acquisition, on the other hand, they require not only</p><p>that all the constructions be acquired in an error-free manner but also that they</p><p>be acquired at the same time. In conclusion, the issue of whether interlanguage</p><p>grammars conform to parameters is still open. We consider other approaches to</p><p>this issue in the following sections.</p><p>4.4 Local impairment</p><p>According to the proposals discussed in the previous sections, parameters</p><p>of UG break down in a global way in L2 acquisition, with the consequence that</p><p>interlanguage grammars are construction specific. The effects of breakdown are</p><p>claimed to be pervasive. Such proposals have largely been phrased in terms of the</p><p>‘classic’ parameters of the Principles and Parameters framework (null subjects,</p><p>word order, etc.). Another approach to the question of parametric breakdown fo-</p><p>cuses on parameters associated with functional categories. Beck (1998a) suggests</p><p>that breakdown is much more local than previously proposed. She claims that in-</p><p>terlanguage grammars suffer from some kind of permanent grammatical deficit as</p><p>far as feature strength is concerned. However, Beck assumes that in other respects</p><p>the interlanguage grammar will be UG-constrained. Beck terms this the Local</p><p>Impairment Hypothesis.</p><p>Beck’s proposal is an extension of Eubank’s Valueless Features Hypothesis</p><p>(1993/1994, 1994, 1996). As discussed in chapter 3 (section 3.2.3), Eubank pro-</p><p>poses that the features of Infl are inert in the initial state, rather than being either</p><p>strong or weak. Strong features are responsible for verb raising and weak features</p><p>for lack of raising; the idea is that when features are inert, raising is optional.</p><p>Eubank related this inertness to absence of overt inflection in the early stages,</p><p>claiming that, as learners acquire morphology, they also acquire appropriate fea-</p><p>ture strength.</p><p>Like Eubank, Beck assumes that the consequence of impaired features is op-</p><p>tional verb raising. However, her proposal differs from Eubank’s in two respects.</p><p>One difference relates to development in the interlanguage grammar. On the</p><p>Valueless Features Hypothesis, there is an early stage with inert features, followed</p><p>by later stages where the grammar is restructured to include the feature strength</p><p>appropriate for the L2. On the Local Impairment Hypothesis, in contrast, feature</p><p>4.4 Local impairment 115</p><p>strength is considered to be permanently impaired. Thus, there is predicted to be no</p><p>development in this domain; even the grammars of advanced interlanguage speak-</p><p>ers are assumed to suffer from this impairment. The second difference is that</p><p>Beck does not assume a causal relationship between overt morphology and feature</p><p>strength. Even if learners show evidence of development in the domain of mor-</p><p>phology,</p><p>generative linguistic perspective) that certain</p><p>properties of language are too abstract, subtle and complex to be acquired without</p><p>assuming some innate and specifically linguistic constraints on grammars and</p><p>grammar acquisition. Furthermore, there is fairly widespread agreement as to what</p><p>these problematic phenomena are. This issue will be considered in more detail in</p><p>the next section.</p><p>1.3 Why UG? The logical problem of language acquisition</p><p>The arguments for some sort of biological basis to L1 acquisition arewell-</p><p>known (e.g. Aitchison 1976; Chomsky 1959, 1965, 1981b, 1986b; O’Grady 1997;</p><p>Pinker 1994): the language capacity is species specific; ability to acquire language</p><p>is independent of intelligence; the pattern of acquisition is relatively uniform across</p><p>different children, different languages and different cultures; language is acquired</p><p>with relative ease and rapidity and without the benefit of instruction; children</p><p>show creativity which goes beyond the input that they are exposed to. All of these</p><p>observations point to an innate component to language acquisition. However, it</p><p>4 1 Universal Grammar and language acquisition</p><p>is conceivable that an innate capacity for language acquisition could be general</p><p>rather than domain specific and that cognitive principles not unique to language</p><p>might be implicated (for relevant proposals, see O’Grady 1987, 1996, 1997, 2003).</p><p>Thus, it is important to understand the arguments in favour of an innate component</p><p>that is specifically linguistic in character.</p><p>UG is motivated by learnability arguments: the primary linguistic data underde-</p><p>termine unconscious knowledge of language in ways which implicate specifically</p><p>linguistic principles. In other words, there is a mismatch between the input (the</p><p>utterances that the child is exposed to), and the output (the unconscious gram-</p><p>matical knowledge that the child acquires). This mismatch gives rise to what is</p><p>known as the problem of the poverty of the stimulus or the logical problem of</p><p>language acquisition. Given such underdetermination, the claim is that it would</p><p>be impossible to account for the L1 acquirer’s achievement without postulating</p><p>a built-in system of universal linguistic principles and grammatical properties</p><p>(Baker andMcCarthy 1981; Hornstein and Lightfoot 1981). UG, then, is proposed</p><p>as an explanation of how it is that language acquirers come to know, uncon-</p><p>sciously, properties of grammar that go far beyond the input in various respects.</p><p>The idea is that such properties do not have to be learned; they are part of the</p><p>‘advance knowledge’ that the child brings to bear on the task of acquiring a</p><p>language.</p><p>The child’s linguistic experience includes what is known as positive evidence;</p><p>that is, the primary linguistic data include utterances that in some sense reveal</p><p>properties of the underlying grammar (but see chapter 5). Negative evidence, or</p><p>information about ungrammaticality, is not (reliably) available. Nevertheless, chil-</p><p>dren come to know that certain sentence types are disallowed; furthermore, they</p><p>acquire knowledge that certain interpretations are permitted only in certain con-</p><p>texts (see section 1.3.1). This kind of knowledge is acquired even though children</p><p>are not taught about ungrammaticality, explicitly or implicitly.</p><p>1.3.1 An example: the Overt Pronoun Constraint</p><p>As an example of abstract knowledge which children successfully ac-</p><p>quire despite an underdetermination problem, we consider here subtle interpretive</p><p>phenomena relating to subject pronouns. It will be suggested that these properties</p><p>could not be acquired solely on the basis of input; rather, a universal linguistic</p><p>principle is implicated.</p><p>Languages differ as to whether or not subject pronouns must be phonetically</p><p>realized, that is whether pronouns are overt or null (Chomsky 1981a; Jaeggli</p><p>1982; Rizzi 1982). In languages like English, known as [−null subject] languages,</p><p>pronouns must be overtly expressed, as can be seen by comparing (1a) and (1b).</p><p>1.3 Why UG? 5</p><p>However, in null subject or prodrop languages (in other words, [+null subject]</p><p>languages), pronouns may be null, taking the form of an empty category, pro.</p><p>Typical examples are Romance languages like Spanish and Italian, as well as East</p><p>Asian languages such as Chinese, Japanese and Korean. The Spanish example in</p><p>(1c) and the Japanese example in (1d) illustrate this point. (Spanish examples in</p><p>this section are drawn from Montalbetti (1984); Japanese examples come from</p><p>Kanno (1997).)</p><p>(1) a. John believes that he is intelligent.</p><p>b. *John believes that is intelligent.</p><p>c. Juan cree que es inteligente.</p><p>John believes that is intelligent.</p><p>‘John believes that (he) is intelligent.’</p><p>d. Tanaka-san wa kaisya de itiban da to itte-iru.</p><p>Tanaka-Mr TOP company in best is that saying-is</p><p>‘Mr Tanaka says that (he) is the best in the company.’</p><p>It is not the case that null subject languages require all pronouns to be unexpressed:</p><p>both overt and null subject pronouns are possible. However, as described below,</p><p>overt and null pronouns do not occur in identical contexts and there are subtle</p><p>restrictions on their distribution.</p><p>The particular restriction at issue here relates to pronominal subjects of embed-</p><p>ded clauses, as in (1). There are interesting differences between [± null subject]</p><p>languages in terms of what can serve as a potential antecedent for the pronoun,</p><p>in other words, limitations on what the pronoun may refer to. In particular, there</p><p>are restrictions on when it is possible for a pronoun to have a quantified expres-</p><p>sion (such as everyone, someone, no one) or a wh-phrase (e.g. who, which) as its</p><p>antecedent.</p><p>In the following examples, the lower, or embedded, clause has a pronoun subject,</p><p>with the main clause subject serving as a potential antecedent of that pronoun. In</p><p>English, an overt pronoun in an embedded clause can be interpreted as coreferen-</p><p>tial with a referential NP in the main clause. As shown in (2), the subject of the</p><p>embedded clause, she, refers to the matrix clause subject, Mary. (Where expres-</p><p>sions are coindexed with the same subscripts, coreference is intended; different</p><p>subscripts indicate disjoint reference.)</p><p>(2) [Maryi thinks [that shei will win]]</p><p>It is also possible for the pronoun subject of the lower clause to have a quantified</p><p>phrase in the main clause as its antecedent, as in (3a), or a wh-phrase, as in (3b).</p><p>(3) a. [Everyonei thinks [that shei will win]]</p><p>b. [Whoi thinks [that shei will win?]]</p><p>6 1 Universal Grammar and language acquisition</p><p>To get the relevant interpretations, imagine a room full of women about to take</p><p>part in a race. In (3a), every person in the room thinks herself a likely winner: she,</p><p>then, does not refer to a particular individual. The same thing applies in (3b): there</p><p>can be many people, each of whom thinks herself a likely winner. In such cases,</p><p>the pronoun is said to receive a bound variable interpretation.</p><p>In the examples so far, the pronoun in the embedded clause is interpreted in</p><p>terms of some other NP within the same sentence, either a referential NP, as in</p><p>(2), or a quantified expression or wh-phrase, as in (3). In addition, a pronoun can</p><p>refer to some other person in the discourse altogether. This is true whether the</p><p>matrix subject is a referring expression or a quantified expression, as shown in</p><p>(4), where the pronoun subject of the lower clause refers to another individual,</p><p>Jane.</p><p>(4) a. Janej is a great athlete. [Maryi thinks [that shej will win]]</p><p>b. Janej is a great athlete. [Everyonei thinks [that shej will win]]</p><p>c. Janej is a great athlete. [Whoi thinks [that shej will win?]]</p><p>Note that, in principle, a sentence like Everyone thinks that she will win is</p><p>ambiguous, with she being interpretable either as a variable bound to the quantifier</p><p>everyone (as in (3a)) or as referring to a particular person, such as Jane, as in (4b).</p><p>Similarly, Mary thought that she would win is ambiguous, with she referring to</p><p>Mary or to some other individual. Usually, the context will favour one of the</p><p>potential interpretations.</p><p>To summarize</p><p>this is not expected to have effects on verb raising: interlanguage feature</p><p>strength will remain impaired even if inflectional morphology is totally accurate.</p><p>4.4.1 Local impairment: evidence</p><p>Beck tests the Local Impairment Hypothesis in the context of the adult L2</p><p>acquisition of German by speakers of English. As discussed in chapter 1 (section</p><p>1.4.1), VP and IP in German are underlyingly head final. In main clauses, the finite</p><p>verb moves to C (via I), in order to check strong features. Some other constituent</p><p>raises to the Spec of CP, resulting in verb second (V2). In the case of a grammatical</p><p>sentence like (9a), the verb has raised to C, while the subject, Maria, has moved</p><p>to Spec CP. In (9b), it is the adverb which has moved to Spec CP; consequently</p><p>the subject remains in Spec IP. The sentence in (9c) is ungrammatical because the</p><p>verb has failed to raise from final position.</p><p>(9) a. [CP Mariaj trinkti [IP tj [oft [VP Kaffee ti] ti]]</p><p>Maria drinks often coffee</p><p>b. [CP Jetztj trinkti [IP Maria [tj [VP Kaffee ti] ti]]</p><p>Now drinks Maria coffee</p><p>c. *[Maria [oft [VP Kaffee trinkt]]]</p><p>Maria often coffee drinks</p><p>According to theLocal ImpairmentHypothesis, feature strength in interlanguage</p><p>grammars is impaired. In the absence of strong or weak features, placement of the</p><p>finite verb in simple clauses is expected to be optional, such that it sometimes</p><p>remains in the VP (unraised) and sometimes occurs in the V2 position (raised</p><p>to C).4</p><p>Beck used a translation task (English to German) to establish whether learners</p><p>ever produce raised verbs; correct translations of three of the sentences would</p><p>require the verb to precede the subject, as in (9b), demonstrating unambiguously</p><p>that it must havemoved out of the VP into C. On the basis of this task, Beck divided</p><p>subjects into two groups, an ‘inversion’ group who demonstrated evidence of verb</p><p>raising to C (although three test items are hardly sufficient to establish this point),</p><p>and a ‘no inversion’ group who failed to do so. The implication appears to be that</p><p>the former are more proficient than the latter.</p><p>116 4 Grammars beyond the initial state</p><p>Box 4.3 Local Impairment Hypothesis (Beck 1998a)</p><p>Languages: L1 = English, L2 = German.</p><p>Task: Sentence matching.</p><p>Sample stimuli:</p><p>Grammatical (V-raising – SVAO) Ungrammatical (No raising – SAVO)</p><p>Der Nachbar kauft bald das Bier. Der Lehrer erst kauft einen Saft.</p><p>Der Nachbar kauft bald das Bier. Der Lehrer erst kauft einen Saft.</p><p>(The neighbour buys soon the beer.) (The teacher first buys a juice.)</p><p>Results:</p><p>Table 4.3.1 Mean response times (z-scores)</p><p>V-raising No raising</p><p>(# = 15) (# = 15)</p><p>L2 groups No inversion (n = 21) −0.0878 0.0878 sig</p><p>Inversion (n = 26) −0.0316 0.0175 ns</p><p>Native speakers (n = 27) −0.1385 0.1403 sig</p><p>The main task was a sentence-matching task. Test sentences consisted of gram-</p><p>matical pairs like (9a) as well as ungrammatical pairs like (10). (See box 4.3.)</p><p>(10) *Maria oft trinkt Kaffee.</p><p>Mary often drinks coffee</p><p>One might question why Beck’s ungrammatical sentences do not take the form of</p><p>(9c), namely, SOV order without verb raising. This is because she assumes, along</p><p>with Full Transfer Full Access, Minimal Trees and Valueless Features, that the</p><p>initial-state grammar of English-speaking learners of German will have SVO as</p><p>the basic word order, this being the order found in the L1. Thus, in her experiment,</p><p>L2 learners are being assessed on whether or not the English order in (10) is</p><p>grammatical for them as well as the German order in (9a).</p><p>According to the Local Impairment Hypothesis, verb raising is predicted to be</p><p>optional regardless of acquisition stage, because feature strength is impaired in</p><p>initial and subsequent grammars. This predicts the following for both groups of</p><p>L2 learners: regardless of how they performed on the translation task, there will be</p><p>no differences in response latencies in the sentence-matching task to sentence pairs</p><p>with raised or unraised verbs, since they are both presumed to be grammatical in</p><p>grammars with impaired feature strength. For native speakers of German, on the</p><p>4.4 Local impairment 117</p><p>other hand, response latencies to sentence pairs like (10) should be significantly</p><p>slower than to pairs where the verb has raised, as in (9a).</p><p>Controls responded as predicted, that is, their response times to the gram-</p><p>matical sentence pairs were significantly faster than to the ungrammatical pairs.</p><p>(See table 4.3.1.) L2 learners, on the other hand, did not behave as expected. The</p><p>‘no inversion’ group, like the native speakers, showed a significant difference be-</p><p>tween grammatical and ungrammatical pairs, taking longer to respond to the latter.</p><p>This is unexpected on the Local Impairment Hypothesis; this group should have</p><p>treated both sentence types alike if their grammars are characterized by optionality.</p><p>The ‘inversion’ group’s results are consistent with the hypothesis: their response</p><p>latencies to the two sentence types were not significantly different, apparently</p><p>supportive of the claim for optional verb raising.</p><p>However, these results are problematic for the Local Impairment Hypothesis as</p><p>well. Beck does not in fact establish what the underlying interlanguage word order</p><p>is for any of her subjects. Even if she is correct in assuming that some subjects</p><p>still treat German as underlyingly SVO, one would expect more proficient learn-</p><p>ers (as the ‘inversion’ group presumably are) no longer to adopt the L1 English</p><p>SVO order but, rather, to have restructured to SOV. (See du Plessis et al. (1987)</p><p>and Vainikka and Young-Scholten (1994) for evidence of restructuring from SVO</p><p>to SOV in L2 German.) Recall that Beck’s ungrammatical test pairs do not in-</p><p>clude sentences with unraised verbs in final position (such as (9c)). Yet, for a</p><p>grammar with appropriate SOV order, it is precisely such sentences that should</p><p>have been compared to the grammatical sentences like (9a). In the absence of</p><p>the relevant data, it is hard to accept that this group has optional verb raising in</p><p>German.</p><p>In summary, the experimental evidence for impaired features as a permanent</p><p>property of interlanguage grammars is not compelling. Beck’s results are prob-</p><p>lematic for the Local Impairment Hypothesis. The ‘no inversion’ group behaved</p><p>like native speakers on the sentence-matching task, distinguishing between gram-</p><p>matical and ungrammatical sentences, contrary to expectations. Although the ‘in-</p><p>version’ group failed to make this distinction, it is not clear that the sentence types</p><p>being tested were in fact appropriate.</p><p>4.4.2 Local Impairment: assessment</p><p>The Local Impairment Hypothesis suffers from the same conceptual</p><p>problems as the Valueless Features Hypothesis. The claim that features are in-</p><p>ert or impaired, and that the consequence of this impairment is variability in verb</p><p>placement, is stipulative. It does not follow from any theory of the effects of</p><p>118 4 Grammars beyond the initial state</p><p>feature strength on grammars. Nor is it clear why impairment should be con-</p><p>fined to strength of features in I and C. As noted in chapter 3 (section 3.2.3.2),</p><p>in the absence of strong features, there should be no raising at all, rather than</p><p>optional verb raising (Robertson and Sorace 1999; Schwartz 1998b; Schwartz and</p><p>Sprouse 2000b).</p><p>There is also a question of plausibility: what are the grounds (theoretical or</p><p>empirical) for claiming that impairment is permanent? The Local Impairment</p><p>Hypothesis would seem to predict that interlanguage grammars never recover</p><p>from variability in verb placement. Beck (1997) recognizes that this is not in fact</p><p>the case. In other words, advanced L2 speakers do not demonstrate the predicted</p><p>variability. In consequence, Beck (1997) suggests that where L2 learners appear to</p><p>perform in a manner appropriate for the L2, this is achieved not on the basis of the</p><p>grammar itself but is due to the operation of additional, agrammatical, explicitly</p><p>learned mechanisms that lead to superficially correct L2 performance. Such a</p><p>proposal is only of interest if it can be demonstrated that there are in fact</p><p>empirical</p><p>differences between a grammar that has the relevant L2 properties and a grammar</p><p>that lacks those properties but that can resort to additional ‘patch up’ mechanisms.</p><p>Otherwise, it renders the Local Impairment Hypothesis unfalsifiable: in the event</p><p>of ‘success’ in the acquisition of L2 word order (i.e. lack of variability), it can</p><p>always be claimed that learners achieve the same surface performance as native</p><p>speakers by radically different means.</p><p>4.5 UG-constrained grammars and parameter setting</p><p>So far, we have considered proposals that interlanguage grammars can-</p><p>not be characterized in terms of parameter settings, instead being impaired either</p><p>globally or locally, such that no parameters are instantiated at all or at least some</p><p>parameters have broken down. The implication of impairment is that interlan-</p><p>guage grammars are not fully UG-constrained; rather, they demonstrate properties</p><p>which are not characteristic of natural language in general. We now turn to the</p><p>alternative, namely that interlanguage grammars can be characterized in terms</p><p>of parameter settings. As we shall see, some researchers maintain that interlan-</p><p>guage grammars exemplify only L1 settings, others that only L2 settings are to</p><p>be found, yet others argue for development and change in the form of parameter</p><p>resetting, from the L1 value to the L2 value or to settings found in other languages.</p><p>Regardless of these differences, these proposals share the intuition that interlan-</p><p>guage representations are unimpaired, in that they conform to possible parameter</p><p>settings.</p><p>4.6 No parameter resetting 119</p><p>4.6 No parameter resetting</p><p>The first hypothesis to be considered in this context is the No Parameter</p><p>Resetting Hypothesis. Although this position was not discussed in the context of</p><p>initial-state proposals in chapter 3, it nevertheless carries an implicit initial-state</p><p>claim, as well as an explicit claim about the kind of grammar development that can</p><p>or cannot be expected. The interlanguage grammar is assumed to have recourse</p><p>only to those parameter settings realized in the L1. Thus, this hypothesis differs</p><p>from Full Transfer Full Access in claiming that there is no subsequent parameter</p><p>resetting in response to L2 input – new parameter values cannot be acquired.</p><p>Hence, representations like those of native speakers of the L2 will necessarily be</p><p>unattainable whenever the L1 and L2 differ in parameter values. On this account,</p><p>then, there is full transfer but not full access.</p><p>According to the No Parameter Resetting Hypothesis, the L1 grammar consti-</p><p>tutes the learner’s representation of the L2 initially and subsequently. Unlike the</p><p>proposals discussed earlier in this chapter, parameters as such are not assumed to</p><p>break down. Rather, the problem lies in resetting, which is considered to be impos-</p><p>sible. However, some proponents of this hypothesis argue that the interlanguage</p><p>grammar (with L1 settings) is able to accommodate L2 data that differ considerably</p><p>from L1 data, imposing an analysis which is UG-constrained and L1-based, even</p><p>though it has no exact parallel in the L1 (e.g. Tsimpli and Roussou 1991). Others</p><p>argue that the effect of being restricted to L1 parameter settings is that learners</p><p>have to resort to ad hoc local fixes to their grammars (Liceras 1997).</p><p>Smith and Tsimpli (1995: 24), advancing a version of the No-Parameter Reset-</p><p>ting Hypothesis that emphasizes parameterization of functional categories, make</p><p>the following claim:</p><p>We will maintain that the set of functional categories constitutes a sub-module</p><p>of UG, namely the UG lexicon. Each functional category is associated with an</p><p>entry specified for relevant functional features . . . Parameterization is then de-</p><p>fined in terms of a finite set of alternative values that a functional category</p><p>can be associated with. Cross-linguistic variation is thus restricted to differ-</p><p>ences in the parametric values of functional categories . . .Moreover, if we as-</p><p>sume that the critical period hypothesis is correct, maturational constraints on</p><p>the functional module can be interpreted as entailing its complete inaccessibility</p><p>after the end of this period . . . UG may still be available but parameter-setting</p><p>cannot be.</p><p>It is not entirely clear from the above quotation whether ‘values’ refers to features</p><p>of functional categories or to their strength. Here it will be assumed that both are</p><p>intended. The implication of the claim for a functional module which becomes</p><p>120 4 Grammars beyond the initial state</p><p>inaccessible is that no new functional categories, features or feature strength can</p><p>be acquired by adult L2 learners.</p><p>4.6.1 No parameter setting: evidence</p><p>Following this line of reasoning, Hawkins and Chan (1997) propose the</p><p>Failed Functional Features Hypothesis, according to which adult L2 learners are</p><p>unable to acquire features differing from the those found in the L1. In spite of</p><p>being restricted to L1 parameter values, the interlanguage grammar is able to</p><p>generate representations that account for the L2 data and that fall within the general</p><p>constraints of UG.</p><p>As described in chapter 1 (section 1.4.1), verbal features within C motivate verb</p><p>raising inGerman, fromV to I toC.Here,we consider another feature characteristic</p><p>of the CP domain, namely a feature relevant to wh-movement. Hawkins and Chan</p><p>investigate the acquisition of English restrictive relative clauses by native speakers</p><p>of Cantonese. They assume a parametric difference between Chinese and English</p><p>relating to presence or absence of a [±wh] feature in C.</p><p>In English, relative clauses can be introduced by a wh-phrase, by the com-</p><p>plementizer that, or without any explicit indicator, as shown in (11a, b, c). The</p><p>complementizer that cannot co-occur with an overt wh-phrase, as in (11d). Re-</p><p>sumptive pronouns are not possible, as shown in (11e). (Examples in this section</p><p>come from Hawkins and Chan (1997).)</p><p>(11) a. The girl who you like</p><p>b. The girl that you like</p><p>c. The girl you like</p><p>d. *The girl who that you like</p><p>e. *The girl who you like her</p><p>English relative clauses are derived by operator movement. An overt operator,</p><p>such as a wh-phrase, or a null operator moves to the Spec of CP, leaving a variable</p><p>(or trace) t. In English, C has a [±wh] feature (Rizzi 1990). It is this feature which</p><p>motivates operator movement, in relative clauses as well as in wh-questions. The</p><p>[wh] feature in English is strong. As a result, in relative clauses a wh-phrase has</p><p>to move overtly into the Spec of CP to check the [+wh] feature in the head C (by</p><p>Spec–head agreement); a null operator has to move if the feature is [–wh], again</p><p>for feature checking purposes. The complementizer that is the lexical realization</p><p>of [–wh] in C, hence it cannot co-occur with a wh-phrase in Spec CP because this</p><p>would result in a feature clash ([–wh] in C but [+wh] in Spec). The structure of the</p><p>relative clause is shown in (12); wh-movement is shown in (12a), corresponding</p><p>to (11a), and null operator movement in (12b), corresponding to (11b) and (11c).</p><p>4.6 No parameter resetting 121</p><p>(12) a. CP</p><p>Spec</p><p>whoi</p><p>C'</p><p>C</p><p>[+wh]</p><p>IP</p><p>Spec</p><p>you</p><p>I'</p><p>I VP</p><p>V</p><p>like</p><p>DP</p><p>ti</p><p>b. CP</p><p>Spec</p><p>OPi</p><p>C'</p><p>C</p><p>[–wh]</p><p>that/∅</p><p>IP</p><p>Spec</p><p>you</p><p>I'</p><p>I VP</p><p>V</p><p>like</p><p>DP</p><p>ti</p><p>One other property of relative clauses is relevant. Wh-movement is constrained</p><p>by a principle of UG known as Subjacency (Chomsky 1981a, b, 1986a), which</p><p>prevents awh-phrase frommoving ‘too far’ from its original position. The specific</p><p>formulation of Subjacency has changed over the years. InGovernment andBinding</p><p>theory it was couched in terms of bounding nodes (Chomsky 1981a); subsequently,</p><p>the constraint was expressed in terms of barriers (Chomsky 1986a). In either case,</p><p>the idea is that a phrase which moves, such as a fronted wh-phrase occurring</p><p>in questions or relative clauses, may not cross more than one bounding node</p><p>or barrier at a time, where DP and IP are bounding nodes in English. In the</p><p>relative clause in (13), who has crossed a contiguous DP and IP and the sentence is</p><p>ungrammatical:</p><p>(13) *This is the boy [CP whoi</p><p>[IP Mary described [DP the way [CP ti that [IP Bill attacked</p><p>ti]]]]]</p><p>Relative clauses in Chinese are somewhat different from English; Mandarin and</p><p>Cantonese are alike in the relevant respects. DPs containing relative clauses are</p><p>head final. Relative clauses are introduced by complementizers (ge in Cantonese;</p><p>de in Mandarin) rather than bywh-phrases. Resumptive pronouns are found where</p><p>there would be an empty category in English. As can be seen in the Mandarin</p><p>examples below, the resumptive pronoun can be overt, as in (14a), or it can be</p><p>null, as in (14b). Where it is null, the assumption is that the empty category is pro,</p><p>122 4 Grammars beyond the initial state</p><p>rather than a variable bound to amoved operator (Huang 1984;Xu andLangendoen</p><p>1985).</p><p>(14) a. Wo xihuan ta de neige nuhai.</p><p>I like her comp that girl</p><p>‘the girl that I like’</p><p>b. Wo xihuan de neige nuhai.</p><p>I like comp that girl</p><p>‘the girl that I like’</p><p>Chinese C, in contrast to English, is assumed to lack the [±wh] feature. As</p><p>a result, there is no motivation for operator movement, since there is no strong</p><p>feature in C requiring a wh-phrase or operator to move to Spec CP for feature</p><p>checking. In consequence, there can be no variable within the relative clause in</p><p>Chinese; instead, the relative clause contains either an overt resumptive pronoun</p><p>or a null pronoun, pro, bound to a null topic, base-generated in the Spec of CP.</p><p>This has consequences for Subjacency, as we shall see below. The structure of the</p><p>Chinese relative clause is shown in (15), corresponding to (14):</p><p>(15) CP</p><p>Spec</p><p>null topici</p><p>C'</p><p>C</p><p>de</p><p>IP</p><p>Spec</p><p>wo</p><p>I'</p><p>I VP</p><p>V</p><p>xihuan</p><p>DP</p><p>tai/proi</p><p>As far as Subjacency is concerned, Chinese sentences which superficially would</p><p>appear to violate Subjacency are in fact grammatical, as the example in (16) shows</p><p>(from Xu and Langendoen 1985).</p><p>(16) zheben shui [NP [IP du guo proi de] ren] bu duo</p><p>this booki read asp proi comp man not many</p><p>‘This book, the people who read (it) aren’t many.’</p><p>Here, the topicalized DP (this book) is associated with an empty category within</p><p>the relative clause, with a contiguous IP and NP intervening. Such structures are</p><p>licit because the relationship between pro (a null resumptive pronoun) and the</p><p>fronted phrase is not one of movement; rather, it is the same as the relationship</p><p>between any pronoun and its antecedent, a relationship which is not subject to</p><p>4.6 No parameter resetting 123</p><p>Subjacency. Thus, the topic and pro can have two or more bounding nodes inter-</p><p>vening between them.</p><p>Hawkins and Chan (1997) maintain that L2 learners have access only to those</p><p>functional features instantiated in the L1. Since C in Chinese lacks a [wh] feature,</p><p>the same will be true of C in the interlanguage grammar. Learners will be able to</p><p>acquire the complementizer that but this will not be a lexical realization of [–wh];</p><p>rather, like ge inCantonese or de inMandarin, it will have no [wh] feature at all. The</p><p>Failed Functional Features Hypothesis predicts that Cantonese speakers will not</p><p>be able to derive English relative clauses by means of operator movement, hence</p><p>that they will have particular difficulties in recognizing that resumptive pronouns</p><p>and Subjacency violations are ungrammatical in English. In contrast, learners of</p><p>English whose L1 also has a [wh] feature, such as French, should experience no</p><p>such difficulties.</p><p>To test the predictions of the Failed Functional Features Hypothesis, an experi-</p><p>ment was conducted. (See box 4.4.) Subjects were post-puberty learners of English</p><p>(adolescents and young adults), with Cantonese or French as their mother tongue,</p><p>at three different levels of L2 proficiency. If L2 learners have acquired the [wh]</p><p>feature of English, they should acquire the associated properties of relative clauses,</p><p>including knowledge of the grammaticality of sentences with fronted wh-phrases</p><p>and a gap within the relative clause, as well as the prohibition against doubly filled</p><p>comps (*who that), resumptive pronouns and Subjacency violations. It is predicted</p><p>that Chinese speakers will fail to acquire this knowledge, because of the lack of a</p><p>[wh] feature in the L1, in contrast to French speakers.</p><p>These predictions were tested by means of a grammaticality-judgment task,</p><p>which included a variety of grammatical and ungrammatical relative clauses.</p><p>Results showed significant differences due to L1, with the French speakers</p><p>at all levels of proficiency outperforming the Chinese on all aspects tested.</p><p>(See table 4.4.1.) The advanced French speakers did not differ from the native-</p><p>speaker controls and neither did the intermediate and elementary French speakers</p><p>on some of the sentence types. There was significant improvement with increasing</p><p>proficiency for both L1 groups, which would seem to argue against the Failed</p><p>Functional Features Hypothesis. Although not attaining the level of performance</p><p>of the French speakers, the Chinese speakers do show increasing accuracy on most</p><p>aspects of English relative clauses. The advancedChinese group appears to be quite</p><p>successful in acquiring most properties of English relatives, such as fronting of</p><p>wh-pronouns and the impossibility of resumptive pronouns.</p><p>According to Hawkins and Chan, these appearances are deceptive. They main-</p><p>tain that, in the interlanguage grammars of the Chinese speakers, relative clauses</p><p>are not derived by operator movement. Instead, an L1-based analysis is adopted:</p><p>124 4 Grammars beyond the initial state</p><p>Box 4.4 No parameter resetting (Hawkins and Chan 1997)</p><p>Languages: L1 = Cantonese/French, L2 = English.</p><p>Task: Grammaticality judgments.</p><p>Sample stimuli:</p><p>Grammatical:</p><p>The boy who I hit broke the window.</p><p>The lady that I met yesterday was my former teacher.</p><p>The girl John likes is studying at the university.</p><p>Doubly filled CP: *The dog which that hurt a child ran away.</p><p>Resumptive pronoun: *The patient that I visited him was very sick.</p><p>Subjacency (CNPC): *This is the boy who Mary described the way that Bill</p><p>attacked.</p><p>Results:</p><p>Table 4.4.1 Accurate judgments (in %)</p><p>Ungrammatical sentences</p><p>Double CP Resumptives Subjacency</p><p>Grammatical</p><p>sentences</p><p>L2 groups Elementary 56 50 38 71</p><p>(n = 47)</p><p>(L1 Chinese) Intermediate 67 68 55 61</p><p>(n = 46)</p><p>Advanced 79 83 90 38</p><p>(n = 54)</p><p>L2 groups Elementary 81 91 81 72</p><p>(n = 33)</p><p>(L1 French) Intermediate 88 95 90 79</p><p>(n = 40)</p><p>Advanced 92 98 96 90</p><p>(n = 40)</p><p>Native speakers (n = 32) 96 99 98 85</p><p>wh-phrases are base-generated as topics in Spec CP, binding a null resumptive</p><p>pro, an analysis which falls within the bounds of UG (for similar proposals, see</p><p>Martohardjono and Gair 1993; White 1992c). Evidence in favour of this analysis</p><p>is as follows. If one looks at the results on relative clauses containing resumptive</p><p>pronouns, the lower-proficiency Chinese speakers accept these to a considerable</p><p>extent, in contrast to the advanced group who reject them with a high degree of</p><p>accuracy. (See table 4.4.1.) At the same time, subjects at lower proficiency levels</p><p>4.6 No parameter resetting 125</p><p>are significantly more likely to reject Subjacency violations than the advanced</p><p>group. Hawkins and Chan suggest that what is going on here is a change from a</p><p>requirement for overt resumptive pronouns (based on properties of the L1) to the</p><p>recognition that English does not permit overt resumptives (based, presumably,</p><p>on observations of the L2 data). The lower-proficiency learners accept overt re-</p><p>sumptives in general; they reject Subjacency violations not because of Subjacency</p><p>but because they are expecting an overt pronoun within the relative clause. The</p><p>advanced group recognizes that English does not permit overt resumptives and</p><p>instead assumes that the gap in a relative clause is a null resumptive pro. Hence,</p><p>they accept apparent Subjacency violations, which are not in fact violations in their</p><p>grammars, since no movement has taken place.</p><p>Hawkins andChan’s data showclear differences based onL1. French shareswith</p><p>English the property of having a [wh] feature, hence wh-movement in relatives.</p><p>French speakers outperform Chinese speakers at every level</p><p>of proficiency. While</p><p>these results are taken as evidence of failed features in the interlanguage grammar,</p><p>Hawkins and Chan nevertheless maintain that the Chinese speakers have shown</p><p>evidence of accommodatingL2 input in away that is not simply the result of surface</p><p>transfer, and in a way that is UG-consistent. Fronted wh-phrases, which would not</p><p>be encountered in Chinese relative clauses, are interpreted as topics (an option</p><p>available and much used in Chinese). However, there does seem to be something</p><p>very odd about a grammar which permits null resumptives and disallows lexical</p><p>ones; it is not clear that this is in fact a possibility realized elsewhere.</p><p>4.6.2 No parameter resetting: assessment</p><p>The No Parameter Resetting Hypothesis, including the Failed Functional</p><p>Features Hypothesis, which is a particular version of the more general hypothesis,</p><p>predicts that interlanguage grammars will be confined to L1 feature values, even</p><p>if there is ample positive evidence to motivate resetting. Triggers for parameter</p><p>setting in the input (see chapter 5) will be ignored, or can no longer function as</p><p>triggers. Even though there appears to be ample positive evidence that English is</p><p>a language with wh-movement, it is claimed that Chinese speakers are never able</p><p>to acquire the [wh] feature because this is unrealized in the L1.</p><p>However, there are other results that suggest, contra Hawkins and Chan, that</p><p>Chinese speakers are not confined to an interlanguage lacking a [wh] feature. If</p><p>so, this would suggest that interlanguage grammars are not restricted to L1 fea-</p><p>ture values, even if these are initially adopted. While Hawkins and Chan are by</p><p>no means alone in finding that speakers of East Asian languages have problems</p><p>in recognizing Subjacency violations (e.g. Johnson and Newport 1991; Schachter</p><p>1989, 1990), there is evidence to show that high-proficiency Chinese speakers do</p><p>126 4 Grammars beyond the initial state</p><p>acquire the relevant properties of English, at least with respect to wh-movement in</p><p>questions. White and Juffs (1998) show that a group of Chinese-speaking learners</p><p>of English who were immersed in English as adults in China performed extremely</p><p>accurately on a timed grammaticality-judgment task involving Subjacency viola-</p><p>tions. These subjects did not differ significantly from the native-speaker controls</p><p>in their rejections of ungrammatical sentences. These results are quite different</p><p>from those reported by Hawkins and Chan (1997) and suggest that the [wh]feature</p><p>had been acquired. (Most studies of Subjacency other than Hawkins and Chan</p><p>have concentrated on wh-questions. Wh-questions in Chinese do involve a [+wh]</p><p>feature but it is weak (i.e. there is no overt movement). Hawkins and Chan do not</p><p>appear to take this into account.)</p><p>Another point is relevant, which again suggests that Chinese speakers cannot be</p><p>limited to representations lacking a [±wh] feature.Adistinctionhas beenmadebe-</p><p>tween strong and weak violations of Subjacency (Chomsky 1986a; Cinque 1990).</p><p>Some violations are relatively worse than others, in consequence of the number</p><p>and types of barriers or bounding nodes that are crossed. For example,many people</p><p>consider (17a), involving wh-movement from within the lower relative clause, to</p><p>be worse than (17b), where extraction is from a complex NP.</p><p>(17) a. *This is the book whichi John met a friend who had read ti</p><p>b. *?This is the book whichi John heard a rumour that you had read ti</p><p>If relative clauses (and other wh-structures) are not derived by movement but</p><p>by base-generated topics associated with null resumptive pronouns, L2 learners</p><p>should not treat strong and weak violations differently, since all ‘violations’ would</p><p>be grammatical. However, several studies have shown that L2 learners of English</p><p>whose L1s lack wh-movement nevertheless judge certain kinds of Subjacency</p><p>violations as being worse than others (Epstein et al. 1996; Martohardjono 1993;</p><p>Pérez-Leroux and Li 1998). Given an L1 without wh-movement, hence with no</p><p>basis to make a distinction between strong and weak violations, such sensitivity</p><p>goes beyond what could be established via the L1 grammar alone.</p><p>Finally, although Hawkins and Chan do not discuss this, the Failed Functional</p><p>FeaturesHypothesis appears to predict that English-speaking learners of languages</p><p>without syntactic wh-movement, such as Chinese, Japanese or Korean, will mis-</p><p>takenly assume wh-movement to be possible, in relative clauses and elsewhere.</p><p>That is, they should be unable to lose the [±wh] feature in C, leading to the</p><p>converse of the behaviour reported above. Even advanced learners of Chinese</p><p>should presumably reject resumptive pronouns in Chinese relative clauses; they</p><p>should front wh-phrases (and reject wh-in-situ questions); they should reject sen-</p><p>tences which are ungrammatical Subjacency violations in English but grammati-</p><p>cal in Chinese. While I am not aware of relevant research on this point, it seems</p><p>4.7 Parameter setting and resetting 127</p><p>unlikely that English speakers would continue to impose a movement analysis in</p><p>such circumstances.</p><p>According to the No Parameter Resetting Hypothesis, the interlanguage gram-</p><p>mar is UG-constrained because it has the same properties as the L1 grammar. But</p><p>while the grammar itself is not impaired, no development is possible, in the sense of</p><p>grammar restructuring. Full Transfer Full Access appears to fare better at account-</p><p>ing for the combined results of various studies on wh-movement in interlanguage</p><p>grammars, since it assumes a major role for transfer, hence, predicting L1-based</p><p>analyses in interlanguage grammars, while also allowing for the possibility of</p><p>eventual parameter resetting.</p><p>Hawkins and Chan (1997) discuss No Parameter Resetting only in terms of</p><p>features, but logically it would seem that the hypothesis should apply to feature</p><p>strength as well, that is, L2 learners should not be able to reset strength from strong</p><p>to weak or vice versa (see Hawkins 2001a: 254). In other words, there does not</p><p>appear to be any principled reason for the hypothesis to distinguish between fea-</p><p>tures and feature strength. However, as will be discussed in the following sections,</p><p>learners are able to acquire L2 word orders that differ from L1 order and that de-</p><p>pend on feature strength, providing further counter-evidence to the No Parameter</p><p>Resetting Hypothesis.</p><p>4.7 Parameter setting and resetting</p><p>According to the claims discussed so far in this chapter, interlanguage</p><p>grammars suffer from a breakdown of parameters (globally or locally) or are</p><p>restricted to L1 parameter values. In the remaining sections, we consider proposals</p><p>that interlanguage grammars are not limited to the parameter settings realized in</p><p>the L1 grammar. Rather, functional categories, features and feature values absent</p><p>from the L1 grammar are instantiated in the interlanguage representation. This is</p><p>the position adopted by the various full access hypotheses discussed in chapter 3</p><p>(e.g. Epstein et al. 1996; Schwartz and Sprouse 1996). The interlanguage grammar</p><p>is fully UG-constrained in the functional domain, such that the grammar can draw</p><p>on the full range of functional categories, features and feature strength; hence,</p><p>parameter resetting is possible.</p><p>Chapter 3 surveyed a variety of hypotheses about the interlanguage initial state,</p><p>including Full Transfer Full Access, which maintains that the L1 grammar con-</p><p>stitutes the initial state but that there will be subsequent restructuring in response</p><p>to properties of the L2 input (Schwartz and Sprouse 1994, 1996). In the case of</p><p>parameters, then, these will be initially set at their L1 values but will subsequently</p><p>be reset. That is to say, parameter resetting is in principle possible. The alternative</p><p>128 4 Grammars beyond the initial state</p><p>full access approach, Full Access without Transfer, maintains that the L1 is not</p><p>implicated in the interlanguage representation, initially or subsequently (Epstein</p><p>et al. 1996; Flynn 1996). Rather, parameters are set to L2 values, on the basis of</p><p>UG interacting with L2</p><p>input, without a prior stage of L1 settings. Thus, these</p><p>approaches agree that L2 parameter values are attainable; they differ over whether</p><p>L1 settings are ever found. At issue is whether or not there are changes in inter-</p><p>language parameter settings during the course of development: on the first view</p><p>there are, while on the second view there are not.</p><p>As discussed in chapter 1 (section 1.4), there are three potential sources of</p><p>crosslinguistic variation as far as functional projections are concerned: the cate-</p><p>gories themselves, their features and the strength of those features. In subsequent</p><p>sections, we will review studies which suggest that L2 learners can indeed acquire</p><p>functional categories, features and feature strength which differ from those found</p><p>in the L1. We will also consider whether or not successful attainment of L2 func-</p><p>tional projections is preceded by a stage implicating L1 categories and features,</p><p>and whether this is always the case. Finally, we will consider whether L2 learn-</p><p>ers sometimes arrive at interlanguage grammars exhibiting parameter settings (or</p><p>combinations of parameter settings) which are found in neither the L1 nor the L2</p><p>but which are present in some other language. We will examine these issues by</p><p>considering research on clausal (IP) and nominal (DP) projections.</p><p>4.7.1 The Verb Movement Parameter: acquiring new feature strength</p><p>A major source of parametric variation is provided by differences in</p><p>feature strength (Chomsky 1995). Hence, if L2 learners can be shown to acquire</p><p>L2 feature strength where this differs from L1 feature strength, this constitutes</p><p>evidence in favour of parameter resetting and against global or local impairment</p><p>or the No Parameter Resetting Hypothesis. In this section, we consider whether or</p><p>not L2 learners can reset a parameter which depends on feature strength, namely,</p><p>the Verb Movement or Verb Raising Parameter, which accounts for differences in</p><p>the placement of lexical verbs in a variety of languages. In this case, the languages</p><p>under consideration are English and French.</p><p>As discussed in chapter 1 (section 1.4.1), whether or not a finite verb raises</p><p>overtly is determined by strength of features in higher functional categories.</p><p>Pollock (1989) (building on earlier work by Emonds (1978)) attributes a variety of</p><p>word-order differences between French and English to a parametric difference in</p><p>feature strength, French having strong I and English weak. While simple declara-</p><p>tive word order looks identical in the two languages, the difference between them</p><p>is revealed in a range of constructions, including, but not limited to, negatives,</p><p>4.7 Parameter setting and resetting 129</p><p>Table 4.3 Verb placement differences between French and English</p><p>Strong I (V in I) (French) Weak I (V in VP) (English)</p><p>Declaratives Les chats attrapent les souris. Cats catch mice.</p><p>Negation Les chats (n’)attrapent (V) pas Cats do not (neg) catch (V)</p><p>(V neg vs. neg V) (neg) les chiens. dogs.</p><p>*Les chats pas attrapent les *Cats catch not dogs.</p><p>chiens.</p><p>Adverbs Les chats attrapent (V) souvent Cats often (A) catch (V) mice.</p><p>(SVAO vs. SAVO) (A) les souris.</p><p>*Les chats souvent attrapent *Cats catch often mice.</p><p>les souris.</p><p>Questions Attrapent (V) -ils (S) les souris? Do they (S) catch (V) mice?</p><p>(VS vs. SV) *Catch they mice?</p><p>adverb placement, and questions, as summarized in table 4.3. In French, finite</p><p>lexical verbs move to I; hence, they appear to the left of negation and adverbs and</p><p>they may appear to the left of the subject in questions. (This is because, once the</p><p>verb has left the VP, it can move from I to C.) In English, on the other hand, the</p><p>lexical verb does not move; hence, it appears to the right of negation, adverbs and</p><p>subjects. In other words, there is a cluster of seemingly disparate constructions</p><p>where the position of the verb can be attributed to just one parametric difference</p><p>between the two languages.</p><p>In chapter 3, a study by Yuan (2001) was presented (see chapter 3, box 3.3)</p><p>which demonstrates that L2 learners can set features of I to the value appropriate</p><p>for the L2. In Yuan’s study, the L2 was Chinese, which has weak I. Hence, verb</p><p>raising is not possible. The L1s in question were French (with strong I) and En-</p><p>glish (with weak I). The French speakers and the English speakers, regardless of</p><p>proficiency level, recognized the impossibility of verb raising in Chinese. That is,</p><p>they neither accepted nor produced ungrammatical sentences like (18b), whereas</p><p>they did accept and produce grammatical sentences like (18a).</p><p>(18) a. Zhangsan changchang kan dianshi. (SAVO)</p><p>Zhangsan often watch television</p><p>b. *Zhangsan kan changchang dianshi. (SVAO)</p><p>Zhangsan watch often television</p><p>Both groups of learners behaved in the same way: there were no effects of the</p><p>strong feature value of French in the interlanguage grammar, even in the group</p><p>that had only been learning Chinese for a few months. In other words, the results</p><p>are consistent with Full Access without Transfer.</p><p>130 4 Grammars beyond the initial state</p><p>Box 4.5 Verb Movement Parameter (White 1992)</p><p>Languages: L1 = French, L2 = English.</p><p>Tasks: (i) Elicited production; (ii) Preference task. Subjects read pairs of sen-</p><p>tences and decide whether or not one of the sentences is better than the other.</p><p>Sample stimuli (preference task):</p><p>Questions Negatives Adverbs</p><p>Like you pepperoni</p><p>pizza?</p><p>The boys like not the</p><p>girls.</p><p>Linda takes always the metro.</p><p>Do you like pepperoni</p><p>pizza?</p><p>The boys do not like the</p><p>girls.</p><p>Linda always takes the metro.</p><p>Results:</p><p>Table 4.5.1 Production of questions by</p><p>L2 learners</p><p>Total questions produced 1171</p><p>Total with main verb raising 2 (0.17%)</p><p>Table 4.5.2 Preference task. Rejections of main verb raising (in %)</p><p>Questions Negatives Adverbs</p><p>L2 learners (n = 72) 86 85 23</p><p>Native speakers (n = 29) 97 98 95</p><p>Yuan’s results contrast with results from White (1992a), who investigated the</p><p>L2 acquisition of English by native speakers of French.Whereas Yuan (2001) only</p><p>investigated whether verbs in Chinese interlanguage raise over VP initial adverbs,</p><p>White investigated all three constructions listed in table 4.3, namely question for-</p><p>mation, negative placement and adverb placement. According to Full Access Full</p><p>Transfer, French-speaking learners of English will start out with the L1 parameter</p><p>setting (strong I); this value must be reset to weak, if the relevant L2 properties are</p><p>to be acquired. Before resetting, it is predicted that verb raising will be possible</p><p>in all three constructions; once the parameter is reset, there should be verb raising</p><p>in none of them. According to Full Access without Transfer, on the other hand,</p><p>there should be no stage implicating verb raising. The results, ostensibly at least,</p><p>are consistent with neither hypothesis, as we shall see.</p><p>4.7 Parameter setting and resetting 131</p><p>Subjects were children in intensive English as a Second Language programmes,</p><p>beginners at the time of testing. (See box 4.5.)5 Data on one of the three properties,</p><p>namely question formation, were gathered by means of an elicited oral production</p><p>task. In addition, there was a preference-judgment task involving all three sentence</p><p>types.</p><p>Results from the oral production task show that lexical verbs were never raised</p><p>in questions (see table 4.5.1); that is, questions of the form in (19a) were not</p><p>produced.</p><p>(19) a. Like you pepperoni pizza?</p><p>b. The boys like not the girls.</p><p>c. Linda takes always the metro.</p><p>Results from the preference task showed that questions and negatives patterned</p><p>together: sentences with verbs raised in questions, as in (19a), or in negatives, as</p><p>in (19b), were rejected with a high degree of accuracy. (See table 4.5.2.) At the</p><p>same time, judgments on adverb placement were quite different. Sentences like</p><p>(19c) were accepted to a considerable extent, being judged to be as good as or</p><p>better than the sentences they were paired with in the preference task. Thus, in the</p><p>case of negation and questions, the learners seem to have discovered that English</p><p>has weak I whereas in the case of</p><p>adverbs they have not.</p><p>White’s solution to this apparent contradiction is to account for the results</p><p>in terms of the split-Infl hypothesis of Pollock (1989). That is, rather than one</p><p>category, Infl, there are in fact two, namely, Tense and Agr, as shown in (20).6</p><p>(20) CP</p><p>TP</p><p>Spec T'</p><p>T NegP</p><p>not AgrP</p><p>Agr</p><p>takesi</p><p>adv</p><p>often</p><p>V</p><p>ti</p><p>NP</p><p>the metro</p><p>VP</p><p>VP</p><p>TheFrench-speaking learners of English, then, have reset the strength of features</p><p>in T to weak. Hence, the lexical verb cannot raise as far as T and so verbs never</p><p>132 4 Grammars beyond the initial state</p><p>Table 4.4 DP differences between French and English</p><p>Romance (e.g. French) English</p><p>Adjective placement le livre (N) anglais (Adj) the English (Adj) book (N)</p><p>(N Adj vs. Adj N)</p><p>±Gender le livre (masc ) the book</p><p>la bière (fem) the beer</p><p>Agreement le (mascsg ) livre (mascsg ) the English book</p><p>anglais (mascsg )</p><p>la (femsg) the English beer</p><p>bière (femsg) anglaise (femsg)</p><p>appear before negation and they do not invert with the subject in questions. These</p><p>same learners, on the other hand, have not reset the strength of Agr. Because it</p><p>is still strong, the lexical verb can raise out of the VP, but only as far as Agr,</p><p>accounting for the Verb Adverb order found in the interlanguage. It is not clear,</p><p>however, why feature strength of different categories should be reset at different</p><p>times. Possibly, the presence of do-support in questions and negatives provides a</p><p>trigger for resetting the strength of T (see chapter 5). Another problem is that verbs</p><p>do not raise consistently over adverbs in the grammars of these learners; there is</p><p>variability here, which is unexpected if Agr is strong.7</p><p>In conclusion, these data suggest successful setting of the strength of T to the</p><p>L2 value. (Indeed, they provide no evidence as to whether the L1 value of T was</p><p>ever adopted.) At the same time, the data are consistent with the adoption of L1</p><p>strength of Agr and failure to reset, at this point in development. In chapter 5, we</p><p>will consider what kind of evidence might lead to resetting of the Verb Movement</p><p>Parameter.</p><p>4.7.2 Nominal projections: feature strength, features and categories</p><p>Characteristics of DPs in interlanguage grammars provide additional evi-</p><p>dence that parameters can be reset. As described in chapter 1 (section 1.4.1), many</p><p>current analyses of the DP include a functional category Num, whose N-features</p><p>are strong in Romance languages like Spanish and French (Bernstein 1993; Ritter</p><p>1992; Valois 1991), in contrast to Germanic languages like English and German</p><p>where they are weak. This difference in feature strength accounts for differences</p><p>in word order within the DP (parallel to the account for word-order differences</p><p>within IP): nouns raise overtly to Num in Romance, yielding the order N Adj, as</p><p>shown in (21a), whereas they do not do so in English, yielding the order Adj N,</p><p>as shown in (21b).8</p><p>4.7 Parameter setting and resetting 133</p><p>(21) a. DP</p><p>D</p><p>le</p><p>NumP</p><p>Num</p><p>livrei</p><p>NP</p><p>AdjP</p><p>anglais</p><p>NP</p><p>|</p><p>N</p><p>ti</p><p>b. DP</p><p>D</p><p>the</p><p>NumP</p><p>Num NP</p><p>AdjP</p><p>English</p><p>NP</p><p>|</p><p>N</p><p>book</p><p>In addition, nouns in Romance languages are classified in terms of a gender</p><p>feature, while determiners and adjectives show gender and number agreement</p><p>with the head noun. English, on the other hand, lacks gender features and gen-</p><p>der agreement. These differences between Romance and English are summa-</p><p>rized in table 4.4, using examples from French. In the following sections, we</p><p>consider to what extent learners of Romance languages acquire these proper-</p><p>ties and to what extent there is evidence of L1 influence. We will also consider</p><p>whether or not the properties in question are attributable to the settings of a single</p><p>parameter.</p><p>4.7.2.1 New feature strength</p><p>In the interlanguage grammar of an English-speaking learner of French or</p><p>Spanish, Num featuresmust be reset fromweak to strong, if word order appropriate</p><p>to the L2 is to be achieved. The various approaches to parameters in interlanguage</p><p>grammars that we have been consideringmake different predictions with respect to</p><p>acquisition of L2 feature strength in such circumstances.As far as theNoParameter</p><p>Resetting Hypothesis is concerned, L2 learners should be confined to L1 feature</p><p>strength. In other words, English-speaking learners of French should be unable</p><p>to acquire N Adj order, whereas French-speaking learners of English should be</p><p>unable to lose it. According to the Local Impairment Hypothesis, variable Adj N</p><p>and N Adj orders are predicted to occur, even in advanced L2 speakers, parallel to</p><p>claims made for the verbal domain. (Beck (1998a), Eubank and Grace (1998) and</p><p>134 4 Grammars beyond the initial state</p><p>Table 4.5 Noun/adjective order in interlanguage DPs</p><p>Study Languages Proficiency level *Adj N *N Adj</p><p>Hawkins 1998 L1 English, Advanced 1 (n = 10) 5%</p><p>L2 French Advanced 2 (n = 10) 0</p><p>Gess and Herschensohn L1 English, Beginners 1 (n = 6) 66%</p><p>2001 L2 French Beginners 2 (n = 8) 9%</p><p>Intermediate (n = 29) 7%</p><p>Advanced 1 (n = 14) 1%</p><p>Advanced 2 (n = 20) 0</p><p>Advanced 3 (n = 7) 0</p><p>Bruhn de Garavito L1 French, Low (n = 30) 2%</p><p>and White 2002 L2 Spanish Intermediate (n = 12) 9%</p><p>Parodi et al. 1997 L1s Korean and Low/intermediate 0.5%</p><p>Turkish, (n = 11)</p><p>L2 German</p><p>L1s Romance, Low/intermediate 20%</p><p>L2 German (n = 4)</p><p>Eubank et al. (1997) do not, in fact, discuss nominal feature strength. However,</p><p>there seems no reason not to extend their hypothesis to this area. That is, on an</p><p>approach that argues for inertness, there appears to be no principled reason not</p><p>to claim that all features are inert in interlanguage grammars.) According to Full</p><p>Transfer Full Access, following a stage of L1 feature strength, L2 strength should</p><p>be attainable; this predicts a stage of Adj N order prior to attaining the appropriate</p><p>N Adj order. Full Access without Transfer predicts successful acquisition of NAdj</p><p>order with no initial stage of Adj N order.</p><p>There are several recent studies which suggest that L2 learners of Romance</p><p>languages successfully acquireNAdj order. These studies examineDPs containing</p><p>adjectives in spontaneous production or by means of elicited production of such</p><p>DPs. Incidence of incorrect adjective–noun (*Adj N) word order is quite low,</p><p>often non-existent. Table 4.5 provides a summary of the relevant studies. As far</p><p>as English-speaking learners of French are concerned, Hawkins (1998) (reported</p><p>in Hawkins 2001a) found that two groups of advanced L2 proficiency showed few</p><p>errors. In other words, nouns were almost always raised over adjectives in relevant</p><p>contexts. Gess and Herschensohn (2001) report similar results from advanced</p><p>learners, on a written sentence-completion task. However, it should be noted that</p><p>Gess and Herschensohn’s task is very like a classroom activity, possibly drawing</p><p>on explicit learned knowledge, rather than unconscious knowledge of the L2. Even</p><p>so, they find English adjective–noun order produced by beginners, suggesting an</p><p>4.7 Parameter setting and resetting 135</p><p>initial stage where Num features are weak. The results generally fail to support</p><p>the No Parameter Resetting Hypothesis.</p><p>Is there any evidence of impaired or inert features? Recall that, according to the</p><p>Local Impairment Hypothesis, feature values are inert regardless of the situation in</p><p>the L1 and the L2. Even if both the L1 and the L2 have strong features, variability</p><p>is predicted. Bruhn de Garavito and White (2000, 2002) show that, in the case of</p><p>French-speaking learners of Spanish, there is no period of variability, contrary to</p><p>the predictions of the Local Impairment Hypothesis: correct N Adj order is present</p><p>and consistently used by low-proficiency learners.</p><p>Research by Parodi, Schwartz and Clahsen (1997) also suggests that there is</p><p>no variability when the L1 and L2 have the same feature strength. Parodi et al.</p><p>examined data from untutored L2 learners of German, whose mother tongues were</p><p>Korean, Turkish, Italian and Spanish. German, like English, has weak Num, hence</p><p>Adj N order is required and N Adj order is ungrammatical (*N Adj). The same is</p><p>true of Korean and Turkish. Results</p><p>show that raising of nouns over adjectives was</p><p>totally non-existent in the German interlanguage of the Korean and Turkish speak-</p><p>ers. In contrast to these results were those from the Romance speakers. These</p><p>subjects produced *N Adj word order to varying degrees in their interlanguage</p><p>German (from 0% to 37.5% of all DPs including adjectives, depending on</p><p>proficiency level).</p><p>The results from the studies described above, taken together, suggest that feature</p><p>strength is resettable. Furthermore, as the subjects discussed by Parodi et al. were</p><p>untutored learners, one can exclude the possibility, in their case at least, that correct</p><p>word order was simply an effect of instruction. At the same time, the data suggest</p><p>there is a prior stage in which L1 feature strength is adopted. Thus, results appear</p><p>to favour the Full Transfer Full Access Hypothesis.</p><p>Nevertheless, the data suggest some variability when L1 and L2 differ in feature</p><p>strength, which is somewhat problematic for Full Transfer Full Access. It is not the</p><p>case that there is a stage where only the L1 feature strength word order is found.</p><p>The beginner group studied by Gess and Herschensohn (2001) produced ap-</p><p>propriate French order at the same time as L1 order (34% N Adj versus 66%</p><p>*Adj N). The Romance speakers studied by Parodi et al. produced German order</p><p>at the same time as L1 order (80%Adj N versus 20% *NAdj). Furthermore, Parodi</p><p>et al. show that such variability occurs at the individual level. In other words, while</p><p>there is L1 influence, this is not consistent. L1 effects alternate with L2 properties.</p><p>4.7.2.2 New features</p><p>In the cases of L2 feature strength described so far, the L1 and L2 share</p><p>features (V features in Infl, N features in Num), the only difference being in their</p><p>136 4 Grammars beyond the initial state</p><p>Table 4.6 Gender agreement accuracy in L2 DPs</p><p>Accurate gender</p><p>Study Languages Proficiency level agreement</p><p>Hawkins 1998 L1 English, Advanced 1 (n = 10) 82.5%</p><p>L2 French Advanced 2 (n = 10) 89.5%</p><p>Gess and L1 English, Beginners 1 (n = 6) 0%</p><p>Herschensohn 2001 L2 French Beginners 2 (n = 8) 43%</p><p>Intermediate (n = 29) 67%</p><p>Advanced 1 (n = 14) 91%</p><p>Advanced 2 (n = 20) 86%</p><p>Advanced 3 (n = 7) 98%</p><p>Bruhn de Garavito L1 French, Low (n = 30) 81.5%</p><p>and White 2002 L2 Spanish Intermediate (n = 12) 89%</p><p>strength.We turn now to the question of whether new features can be acquired, that</p><p>is, features required in L2 representations but altogether absent from the L1, taking</p><p>gender in Romance as a case in point. At issue is whether learners of languages like</p><p>French and Spanish can acquire the gender feature, given an L1 such as English,</p><p>where nouns are not classified according to gender and where there is no gender</p><p>agreement.</p><p>In Romance languages, nouns fall into two gender classes, masculine and fem-</p><p>inine. Gender is an inherent feature of nouns (Corbett 1991). Gender of this type,</p><p>often referred to as grammatical gender, is arbitrary. This contrasts with natural</p><p>or biological gender. In addition to a gender feature on nouns, gender agreement</p><p>(or concord) is found on adjectives and determiners, which agree in gender with</p><p>the head noun. On current accounts, this is achieved by means of feature checking</p><p>(Carstens 2000).</p><p>Table 4.6 presents results from several recent studies (the same as those reported</p><p>in table 4.5) which examine spontaneous and elicited production data for accuracy</p><p>of gender agreement between determiners and nouns or determiners, adjectives and</p><p>nouns. Hawkins (1998) (reported inHawkins 2001a) examines data from advanced</p><p>learners of French, considering gender agreement between determiners and nouns</p><p>(adjectives not being included). Subjects exhibited persistent problems: (i) showing</p><p>greater accuracy with gender agreement on definite determiners than on indefinite</p><p>and (ii) adopting a ‘default’ gender on determiners (leading to overuse of one or</p><p>other gender). Hawkins attributes these problems to the lack of a gender feature</p><p>in the L1 English, hence supporting the Failed Functional Features Hypothesis.</p><p>However, it is noteworthy that his subjects were relatively accurate on gender</p><p>overall (see table 4.6). Furthermore, Bruhn de Garavito and White (2000, 2002)</p><p>show that similar phenomena (that is, greater problems with agreement in the case</p><p>4.7 Parameter setting and resetting 137</p><p>of indefinite determiners, as well as the use of default gender) also occur in the</p><p>acquisition of Spanish by French speakers. Since both languages have gender, the</p><p>absence of gender in the L1 cannot be the crucial factor. (In chapter 6, the issue of</p><p>incorrect or default forms will be considered in more detail.)</p><p>Gess and Herschensohn (2001) report that advanced proficiency learners of</p><p>French achieved a high degree of accuracy on gender and number agreement</p><p>between determiners, adjectives and nouns, again suggesting that L2 learners can</p><p>acquire features which are absent in the L1. On the other hand, learners at lower</p><p>proficiency levels were quite inaccurate on agreement. As Gess and Herschensohn</p><p>do not separate gender and number agreement in their results, it is impossible</p><p>to determine whether the problem at lower levels of proficiency is restricted to</p><p>gender.</p><p>It is conceivable that failure to achieve greater accuracy in gender agreement is a</p><p>problem specific to production and that it does not reflect underlying competence.</p><p>In other words, L2 learners, even at low levels of proficiency, might acquire an</p><p>abstract gender feature, together with the requirement for agreement, but fail to</p><p>implement it all the time. (See chapter 6 for further discussion.)</p><p>This issue is investigated by White, Valenzuela, Macgregor, Leung and Ben-</p><p>Ayed (2001), who devised an interpretation task in order to investigate the acquisi-</p><p>tion of gender agreement without relying on production data. The L2 was Spanish.</p><p>Subjects were native speakers of French (which has gender) and English (which</p><p>does not), at various levels of L2 proficiency. If L2 learners can acquire features not</p><p>present in the L1, then both English and French speakers should exhibit knowledge</p><p>of Spanish gender. If there are L1 effects initially, then low-proficiency learners</p><p>whose mother tongue is English should perform worse than those whose mother</p><p>tongue is French. Furthermore, English speakers should perform better on number</p><p>features than gender, since number is present in the L1.</p><p>In order to test for unconscious knowledge of gender and number agreement,</p><p>White et al. make use of a phenomenon known as N-drop, whereby Romance DPs</p><p>can contain ‘null nominals’, in other words, nouns which are not overtly realized</p><p>(Bernstein 1993; Snyder, Senghas and Inman, 2002). This phenomenon is very</p><p>productive in Spanish. The null nominal is licensed and identified by gender and</p><p>number agreement on the remaining adjectives and/or determiners (e.g. Snyder</p><p>1995a); that is, its content is recoverable from these agreement features (similar to</p><p>the situationwith respect to null subjects, as described in section 4.3.1).An example</p><p>is given in (22). In (22a), the N, libro, is overt. In (22b), libro has been dropped;</p><p>the determiner uno identifies the missing noun as masculine and singular. (In this</p><p>particular example, the adjective grande is invariant in form.) The corresponding</p><p>English sentence with N-drop, (11c), is ungrammatical. Instead, the pronoun one</p><p>is required, as in (11d).</p><p>138 4 Grammars beyond the initial state</p><p>(22) a. Uno libro grande está encima de la mesa.</p><p>a-mascsg book-mascsg big is on-top of the table</p><p>‘There is a big book on the table.’</p><p>b. Uno grande está encima de la mesa.</p><p>a-mascsg big is on-top of the table</p><p>c. *There is a big on the table.</p><p>d. There is a big one on the table.</p><p>White et al. developed a picture-identification task, involving a story which con-</p><p>sisted of a number of sentences, each followed by three pictures, equally plausible</p><p>in the context. (See box 4.6.) Subjects had to indicate which picture was the ap-</p><p>propriate one for any given sentence. Each test sentence contained a null nominal,</p><p>as in (23).</p><p>(23) ¿Me compro este negro?</p><p>cli buy this-mascsg black-mascsg</p><p>‘Shall I buy this black one?’</p><p>The phrase este negro contains a null nominal, which is masculine and singu-</p><p>lar, as shown by the form of the determiner este (‘this’) and the adjective negro</p><p>(‘black’). This sentence is followed by pictures of a black sweater (el suéter, mas-</p><p>culine), a black shirt (la camiseta, feminine) and a black tie (la corbata, feminine).</p><p>Crucially, the vocabulary is not supplied. If learners have gender agreement per-</p><p>colating through the DP, they should pick the picture of the sweater, this being</p><p>the only noun whose gender is masculine, like that of the adjective in the null</p><p>nominal. Otherwise, they are expected to pick randomly. (An independent vocab-</p><p>ulary test established whether subjects knew the lexical items and their gender in</p><p>isolation. Subjects who did not pass the vocabulary test were eliminated.) Since</p><p>the gender/number of the missing noun could only be established on the basis</p><p>of the gender/number of the adjective and/or determiner in the test sentence, this</p><p>task provides a means of determining, via comprehension rather than production,</p><p>whether abstract features are present in learner grammars.</p><p>Table 4.6.1 presents the results, comparing performance on number and gen-</p><p>der. Advanced and intermediate groups showed considerable accuracy on both</p><p>features, regardless of L1, with no significant differences between them and the</p><p>native speakers. Only the low-proficiency English speakers showed a significant</p><p>difference in accuracy between number and gender, suggesting L1 influence at this</p><p>level. These results suggest that new features can be acquired: the English speakers</p><p>proved to be very accurate on gender, a feature not present in the L1.</p><p>To sum up, taking gender as a case in point, L2 learners are able to acquire</p><p>gender and gender agreement, suggesting that features not present in the L1 are</p><p>attainable in the interlanguage grammar At the same time, there are L1 effects in</p><p>the case of learners at lowest levels of proficiency, suggesting initial transfer of</p><p>4.7 Parameter setting and resetting 139</p><p>Box 4.6 Acquiring new features – gender (White et al. 2001)</p><p>Languages: L1s = English/French, L2 = Spanish.</p><p>Task: Picture identification. Sentences containing null nominals, each accompa-</p><p>nied by 3 pictures. Subjects indicate which picture is appropriate for any given</p><p>sentence.</p><p>Sample stimulus:</p><p>Paco se prueba algunas cosas también y le pregunta a Marı́a: ‘¿Me compro este</p><p>negro?’</p><p>(Paco tries on some things too and asks Maria: ‘Shall I buy this black (one) ?’)</p><p>Results:</p><p>Table 4.6.1 Mean accuracy on number and gender</p><p>Number (# = 14) Gender (# = 14)</p><p>L2 groups (L1 English) Low (n = 10) 11.5 7.4</p><p>Intermediate (n = 7) 12.57 11.57</p><p>Advanced (n = 7) 13.57 12.57</p><p>L2 groups (L1 French) Low (n = 5) 11.6 10.2</p><p>Intermediate (n = 8) 12.62 10.86</p><p>Advanced (n = 16) 12.63 12.44</p><p>Native speakers (n = 20) 13.2 13.05</p><p>L1 features (or lack of them). Advanced learners show some problems in produc-</p><p>tion but when other tasks are used to assess their interlanguage competence, the</p><p>problems disappear, an issue we will return to in chapter 6.</p><p>4.7.2.3 New functional categories</p><p>The results discussed so far suggest that L2 learners can acquire feature</p><p>strength which differs from the L1, as well as features which are not instantiated in</p><p>140 4 Grammars beyond the initial state</p><p>Table 4.7 Accurate production of articles in obligatory</p><p>contexts (in %)</p><p>L2 English L2 English L3 French</p><p>Contexts Robertson 2000 Leung 2001 Leung 2001</p><p>Definite 83.2% 85% Beginners: 60%</p><p>Intermediate: 81.3%</p><p>Indefinite 77.9% 99.5% Beginners: 99.6%</p><p>Intermediate: 99.2%</p><p>the L1. We will now consider whether new functional categories can be acquired.</p><p>Before pursuing this question, the issue of universality must briefly be addressed,</p><p>that is, whether or not all functional categories are realized in all languages. If</p><p>all categories are present in all languages, there is no such thing as a functional</p><p>category not exemplified in the L1, hence the issue of grammar development in this</p><p>domain does not arise. A very large set of functional categories has been argued</p><p>for in the recent literature (Cinque 1999; Pollock 1997; Rizzi 1997); it seems</p><p>somewhat unparsimonious to assume that these are necessarily instantiated in every</p><p>grammar (Iatridou 1990). I will therefore assume, for the sake of the argument, that</p><p>languages can differ in terms of the functional categories that they instantiate. (See</p><p>also Bobaljik and Thráinsson (1998), Thráinsson (1996) and Webelhuth (1995).)</p><p>Even so, it is not an easymatter to determinewhether or not particular categories</p><p>are exemplified in particular languages. There has been a longstanding debate over</p><p>whether languages likeChinese and Japanese lack functional categories (Fukui and</p><p>Speas 1986); on the whole, the current consensus is that they do have functional</p><p>categories, although these are not necessarily identical to the categories found in</p><p>languages like English. For example, it has been suggested that Chinese lacks the</p><p>functional category D. Instead, it has a functional category CL (classifier), with</p><p>properties rather different fromD (Cheng and Sybesma 1999). If so, this allows one</p><p>to investigate whether a Chinese-speaking learner of English can acquire D (and</p><p>associated features, such as ± definite) or whether an English-speaking learner</p><p>of Chinese can acquire CL. The No Parameter Resetting Hypothesis presumably</p><p>predicts that new categories cannot be acquired. Full Access (with or without</p><p>Transfer), on the other hand, predicts that they are acquirable.</p><p>Recently, there has been someattention to the determiner system in the grammars</p><p>of L2 speakers whose L1 is Chinese (Leung 2001; Robertson 2000). Table 4.7</p><p>presents production of determiners in obligatory contexts, as reported in these</p><p>two studies. Robertson (2000) examined use of definite and indefinite articles in</p><p>the L2 English of advanced learners whose L1 was Taiwanese and/or Mandarin.</p><p>Production data were gathered by means of an elicited production task, involving</p><p>4.8 Settings of neither L1 nor L2 141</p><p>collaborative problem-solving. Robertson found that suppliance of determiners</p><p>was quite high. Inaccurate responses consisted of omission of articles, rather than</p><p>misuse of definite for indefinite or vice versa. (Individual accuracy ranged from</p><p>67.5% to 97%.) There were pragmatic effects: echo contexts (where the speaker</p><p>repeated what the previous person had said) resulted in a much higher proportion</p><p>of determiner omission. Robertson shows that there are syntactic constraints on</p><p>determiner omission: by and large, determiners are only dropped where the NP in</p><p>question forms a chain with a preceding NP which has an overt determiner.</p><p>Leung (2001) explores the acquisition of French as a third language (L3) by</p><p>Cantonese–English bilinguals, who were advanced speakers of English, having</p><p>learned it in childhood. They were learning the L3, French, as adults. Results from</p><p>an elicited production task (picture description) showed amuch lower incidence of</p><p>determiner omission than Robertson found: omission (in the case of singular count</p><p>nouns) was around 6% in both L2 English and L3 French. However, article usage</p><p>was not error free (see table 4.7). While use of indefinite articles in indefinite</p><p>contexts was extremely accurate (almost 100%) in L2 English and L3 French</p><p>regardless of proficiency level, indefinite articles were also used in contexts where</p><p>definite articleswere expected. The errorwasmost extensive in the case of beginner</p><p>learners of French.</p><p>Both these studies suggest that Chinese-speaking learners of English and French</p><p>acquire articles, possibly implicating the functional category D, together with the</p><p>associated feature ± definite. (However, suppliance is not always fully accurate, a</p><p>point wewill return to in chapter 6.) Nevertheless, a note of caution is in order here:</p><p>one needs to eliminate the possibility that Chinese-speaking learners of English or</p><p>French are categorizing L2 articles</p><p>as classifiers, the functional category found in</p><p>theL1. In otherwords, it is necessary to establish, on independent grounds,whether</p><p>these forms behave like English determiners or like Chinese classifiers, something</p><p>which it is not possible to do on the basis of the data reported by Robertson or by</p><p>Leung.</p><p>4.8 Settings of neither L1 nor L2</p><p>So far, we have examined a variety of recent studies reporting successful</p><p>acquisition of feature strength, features and functional categories not found in the</p><p>L1. Since many of these studies involve adult learners, the results argue against</p><p>Smith and Tsimpli’s (1995) claim that the functional sub-module of UG becomes</p><p>unavailable in post-puberty acquisition. Nevertheless, while resetting appears to</p><p>be possible, it is by no means inevitable that L2 parameter values are achieved.</p><p>Some learners do indeed seem to persist with L1 parameter settings. Furthermore,</p><p>142 4 Grammars beyond the initial state</p><p>D</p><p>E</p><p>F</p><p>G</p><p>Parameter Y</p><p>Parameter Z</p><p>Parameter X</p><p>A</p><p>B</p><p>C</p><p>Figure 4.1 Parameter settings of neither L1 nor L2</p><p>there are other logical possibilities besides the ones we have considered so far.</p><p>Earlier in this chapter (section 4.3.1), it was suggested that some Korean-speaking</p><p>learners of German arrive at the Italian setting of the Null Subject Parameter, that</p><p>is, a setting of neither the L1 nor the L2. In this section, we consider other examples</p><p>of this nature.</p><p>There are two ways in which it is in principle possible for the interlanguage</p><p>grammar to show parameter settings which differ from both the L1 and the L2.</p><p>Consider the situation shown in figure 4.1. The left-hand side illustrates some</p><p>parameter, X, which has three settings, A, B and C. It is logically possible, then,</p><p>that the L1 might instantiate setting A, the L2 setting B and the interlanguage</p><p>grammar setting C.</p><p>The possibility that L2 learners arrive at parameter values exemplified in nei-</p><p>ther the L1 nor the L2 was discussed in earlier research on UG and L2 acquisi-</p><p>tion, one proposal of this type being advanced by Finer and Broselow (1986) and</p><p>Finer (1991). They investigated the Governing Category Parameter, a parameter</p><p>accounting for crosslinguistic differences in the distribution of reflexives, which</p><p>was argued to have five settings (Wexler and Manzini 1987). Finer and Broselow</p><p>(1986) reported that learners of English whose mother tongue was Korean treated</p><p>English reflexives neither likeKorean reflexives (which freely permit long-distance</p><p>antecedents) nor like English ones (which require local antecedents); instead, re-</p><p>flexives in the interlanguage grammar behaved as they do in Russian, in that</p><p>long-distance binding was permitted but only out of non-finite clauses.</p><p>If parameters are restricted to binary values, this possibility disappears. Given</p><p>binary parameters, it is no longer possible to account for parametric variation</p><p>in the distribution of reflexives in terms of a multivalued Governing Category</p><p>Parameter. However, parameters do not work in isolation: grammars include sets</p><p>of interrelated parameters. So another way in which an interlanguage grammar</p><p>might differ from either the L1 or the L2 is in having combinations of parameter</p><p>settings which reflect neither the L1 nor the L2. Consider figure 4.1 again. On</p><p>4.8 Settings of neither L1 nor L2 143</p><p>the right are two parameters, Y and Z, each with two settings. Imagine a situation</p><p>where the L1 has setting D of parameter Y and setting F of parameter Z while</p><p>the L2 has settings E and G. It is logically possible for an interlanguage grammar</p><p>to instantiate combinations which together represent neither the L1 nor the L2</p><p>(D and G; E and F). If one considers more than two parameters, the range of</p><p>available combinations extends accordingly.</p><p>4.8.1 Settings of neither L1 nor L2: reflexives</p><p>An example of L2 acquisition involving this situation is provided</p><p>by MacLaughlin (1996, 1998), again involving the distribution of reflexives.</p><p>MacLaughlin suggests that L2 learners reset parameters relating to reflexives but</p><p>that they will not necessarily acquire the parameter settings appropriate for the L2.</p><p>She adopts Progovac’s (1992, 1993) Relativized Subject framework, which shares</p><p>the following assumptions with the LF movement approach discussed in chapter 2</p><p>(section 2.3.1): crosslinguistically, reflexives can be morphologically simplex</p><p>(a head, X◦) or complex (phrasal, XP); long-distance binding is associated only</p><p>with simplex reflexives and long-distance binding must be subject oriented.</p><p>MacLaughlin investigates the L2 acquisition of English reflexives by native</p><p>speakers of Chinese and Japanese. English reflexives like himself, herself are</p><p>polymorphemic and require local antecedents, which are not necessarily sub-</p><p>jects. Chinese and Japanese both have simplex reflexives (ziji in Chinese, zibun</p><p>in Japanese) which permit long-distance subject antecedents, as well as phrasal</p><p>reflexives which require local antecedents. MacLaughlin (1998) explains the be-</p><p>haviour of reflexives crosslinguistically in terms of two parameters, a reflexive</p><p>parameter and an Agr parameter (following Bennett 1994; Progovac 1992,1993),</p><p>as given in (24):</p><p>(24) Reflexive parameter: a reflexive is monomorphemic or polymorphemic.</p><p>Agr parameter: Agr is null9 or morphological.</p><p>These two parameters account for the crosslinguistic binding facts as follows.</p><p>A reflexive must be bound to its antecedent in a domain containing a subject .10</p><p>On Progovac’s account, a head reflexive must be bound to its antecedent in the</p><p>domain of a subject which is also a head (namely, Agr), whereas a phrasal</p><p>reflexive must be bound within the domain of a subject which is also phrasal</p><p>(namely the subject of a clause or complex NP). XP anaphors will always require</p><p>local binding because the nearest XP subject will be within the same clause (or</p><p>NP) as the reflexive. Head reflexives, on the other hand, must be in the domain of</p><p>some Agr, not necessarily within the same clause. If Agr is morphologically overt,</p><p>the reflexive must be bound within the same clause (i.e. locally). However, if Agr</p><p>144 4 Grammars beyond the initial state</p><p>Table 4.8 Grammar types sanctioned by</p><p>two parameters</p><p>Type 1 Type 2 Type 3</p><p>Reflexive poly mono mono</p><p>Agr overt/null overt null</p><p>is never realized morphologically (as in Chinese or Japanese) or is not realized in</p><p>non-finite clauses (as in Russian), the reflexive can be bound in the domain of a</p><p>higher Agr, thus permitting long-distance antecedents.</p><p>The reflexive parameter distinguishes between different types of reflexives. The</p><p>Agr parameter, on the other hand, distinguishes between null or overt agreement</p><p>morphology. It should be noted that if a reflexive is polymorphemic, the status of</p><p>Agr is irrelevant, since the reflexive will always be bound within the domain of</p><p>a phrasal subject , not Agr. The setting of the Agr parameter becomes crucial</p><p>in the case of monomorphemic reflexives. These two parameters yield the three</p><p>possibilities shown in table 4.8.</p><p>Type 1 represents the situation in English, where all reflexives are polymor-</p><p>phemic and only local binding of reflexives is possible, as well as the situation</p><p>for phrasal reflexives in languages like Chinese and Japanese. Type 3 is found</p><p>in languages like Chinese and Japanese, where long-distance binding is always</p><p>possible with monomorphemic reflexives. Type 2 is characteristic of languages</p><p>like Russian, where long-distance antecedents for reflexives are possible but only</p><p>out of non-finite clauses. These types represent ‘possible’ grammars.</p><p>InMacLaughlin’s experiment (see box 4.7) subjects were given various types of</p><p>sentences to read, all containing reflexives, and asked whether or not the reflexive</p><p>could refer to various NPs mentioned in the sentence. As well as reporting group</p><p>data, MacLaughlin analyses individual performance, looking at consistency of</p><p>response patterns (80% in a particular direction being considered to demonstrate</p><p>consistency).</p><p>The responses of six of the subjects were restricted to local antecedents, suggest-</p><p>ing</p><p>a type 1 grammar (i.e. the parameter combination appropriate for the L2). Two</p><p>subjects showed type 3 behaviour, as in the L1. There were seven subjects whose</p><p>responses suggest a type 2 combination, allowing long-distance binding only out</p><p>of non-finite clauses. These results confirm findings of research conducted in ear-</p><p>lier frameworks, which showed significantly more long-distance binding out of</p><p>non-finite clauses in L2 English (e.g. Finer and Broselow 1986; Hirakawa 1990).</p><p>In such cases, the interlanguage grammar is like neither the L1 nor the L2 but</p><p>it does represent a combination of parameter settings found in other languages.</p><p>4.8 Settings of neither L1 nor L2 145</p><p>Box 4.7 Other parameter settings (MacLaughlin 1998)</p><p>Languages: L1 = Chinese/Japanese, L2 = English.</p><p>Task: Coreference-judgment task.</p><p>Sample stimuli:</p><p>Biclausal finite</p><p>Barbara thinks that Lisa is proud of herself.</p><p>herself can be Lisa Agree— Disagree—</p><p>herself can be Barbara Agree— Disagree—</p><p>Biclausal non-finite</p><p>Michael forced Peter to help himself.</p><p>himself can be Peter Agree— Disagree—</p><p>himself can be Michael Agree— Disagree—</p><p>Results:</p><p>Table 4.7.1 Number of subjects by response patterns</p><p>Type 1 Type 2 Type 3</p><p>L2 learners (n = 15) 6 7 2</p><p>Native speakers (n = 18) 18 0 0</p><p>MacLaughlin suggests that in the case of these seven subjects, the parameter</p><p>relating to reflexive type has not been reset whereas the Agr parameter has been.</p><p>As a result, long-distance binding is possible in principle (because the reflexive</p><p>is monomorphemic) but this can only happen in the case of clauses which are</p><p>non-finite (i.e. where Agr is not overtly realized).</p><p>One might legitimately ask why it is that L2 learners make this kind of mis-</p><p>analysis. What causes them to successfully reset the Agr parameter but to fail to</p><p>reset the parameter relating to reflexive type? Acquisition involves an interaction</p><p>of UG, the learner’s current grammar and the L2 input. Progovac and Connell</p><p>(1991) propose that, for speakers of languages like Japanese or Chinese, L2 in-</p><p>put in the form of third-person-singular agreement provides evidence that Agr</p><p>is overt in English, accounting for the resetting of the Agr parameter. However,</p><p>what is not so clear is why L2 learners fail to reset the reflexive parameter: it</p><p>would seem that the L2 English input provides fairly transparent evidence that</p><p>reflexives are morphologically complex. (Although it might seem odd that the</p><p>input fails to trigger the relevant reflexive type, it is noteworthy that this is not a</p><p>problem unique to L2 acquisition. According to McDaniel, Cairns and Hsu (1990)</p><p>and Thomas (1994), young children acquiring English as an L1 also initially fail</p><p>146 4 Grammars beyond the initial state</p><p>to recognize that English reflexives are polymorphemic and restricted to local</p><p>antecedents.)</p><p>Unfortunately, there is a circularity problem with much of the research that</p><p>depends on the distinction between monomorphemic and polymorphemic reflex-</p><p>ives, in that no independent evidence is provided to show that English reflexives</p><p>are indeed monomorphemic in these interlanguage grammars (e.g. Bennett 1994;</p><p>Bennett and Progovac 1998;MacLaughlin 1998). In addition, such studies also fail</p><p>to provide an independent test of whether or not subjects have acquired morpho-</p><p>logical Agr in English. Most monomorphemic reflexives are deficient in person,</p><p>gender and number agreement, so one might expect L2 learners to fail to use forms</p><p>like himself, herself, themselves, etc., appropriately (i.e. to fail to make them agree</p><p>with their antecedents) if such reflexives are truly monomorphemic. In fact, White</p><p>(1995b) found that Japanese-speaking learners of English were quite accurate on</p><p>person, gender and number agreement between reflexives and their antecedents,</p><p>suggesting that English reflexives are recognized as being polymorphemic; at the</p><p>same time, learners still allowed them to take long-distance antecedents, as if they</p><p>were monomorphemic.</p><p>The most compelling evidence for the L1 parameter setting (type 3) would</p><p>be if one could demonstrate that L2 learners of English not only fail to observe</p><p>agreement between English reflexives and their antecedents but also fail to show</p><p>evidence of morphological agreement on verbs, as well as allowing long-distance</p><p>antecedents. The most compelling evidence for some other setting (type 2) would</p><p>be provided if learners of English fail to observe agreement between English</p><p>reflexives and their antecedents, show morphological agreement on finite verbs</p><p>and treat reflexives as allowing long-distance antecedents. As most studies look</p><p>only at domain and orientation and not at the relevant agreement phenomena, the</p><p>relevant evidence is still lacking.</p><p>4.8.2 Settings of neither L1 nor L2: case checking</p><p>A somewhat different example of an interlanguage grammar with prop-</p><p>erties of neither the L1 nor the L2 is provided by Schwartz and Sprouse (1994) in</p><p>their case study of an adult Turkish-speaking learner of German, known as Cevdet.</p><p>Looking at production data, Schwartz and Sprouse examine Cevdet’s word-order</p><p>development and argue that the interlanguage grammar is restructured in a number</p><p>of different respects during the course of acquisition. Initially, they hypothesize,</p><p>Cevdet produced SOV order. As Turkish and German are both SOV languages,</p><p>presence of SOV order in the initial stage is consistent with L1 headedness or with</p><p>successful acquisition of L2 headedness. However, Turkish differs from German</p><p>in not being a V2 language (see chapter 1, section 1.4.1). V2 begins to emerge in</p><p>4.8 Settings of neither L1 nor L2 147</p><p>what Schwartz and Sprouse characterize as stage 2. Here, some constituent other</p><p>than the subject is fronted, as shown in (25a), and the subject remains in Spec IP,</p><p>giving the appearance of inversion between subject and verb. At the same time,</p><p>Cevdet is still producing sentences without inversion, as shown in (25b) and (25c),</p><p>which are ungrammatical in German.</p><p>(25) a. dann trinken wir bis neun Uhr.</p><p>then drink we until nine o’clock</p><p>‘Then we will drink until nine o’clock.’</p><p>b. In der Türkei der Lehrer kann den Schüler schlagen.</p><p>in the Turkey the teacher can the pupils beat</p><p>‘In Turkey, teachers can hit pupils.’</p><p>c. Ankara ich kenne.</p><p>Ankara I know</p><p>What is particularly interesting about this stage is that the subject of Cevdet’s</p><p>inverted V2 structures (where the verb precedes the subject) is almost always a</p><p>personal pronoun, as in (25a). When the subject is a full DP, as in (25b), inversion</p><p>is hardly ever found: 69 out of 70 examples with inversion involved a pronoun</p><p>subject. In stage 3, inversion is occasionally found with non-pronominal subjects,</p><p>as in (26). However, this is still not extensive, only 10.5% of post verbal subjects</p><p>being full NPs.</p><p>(26) Das hat eine andere Frau gesehen.</p><p>this has an other woman seen</p><p>‘Another woman saw that.’</p><p>While this distinction between DPs and pronouns with respect to inversion is</p><p>not characteristic of German, such a distinction is observed in French. In the case</p><p>of questions in French, pronoun subjects can invert with the verb but full DPs</p><p>cannot.</p><p>Schwartz and Sprouse offer the following analysis of Cevdet’s stage 2 inter-</p><p>language where pronominal subjects are treated differently from non-pronominal</p><p>subjects. They propose that the issue relates to checking of case features. Every</p><p>overt DP has a case feature which must be checked. UGmakes available a number</p><p>of different (and parameterized) mechanisms for case checking (Rizzi and Roberts</p><p>1989). In German, case features are checked by Spec–head agreement (where the</p><p>subject DP raises to the Spec of AgrP, in order to check its case feature against a</p><p>corresponding feature in Agr) or under government (where the verb in C governs</p><p>the subject or its trace in Spec AgrP). Schwartz and Sprouse assume that only the</p><p>Spec–head agreement option is available in Turkish. A third possibility is found in</p><p>French, where case checking is achieved via incorporation. Subject clitics (whose</p><p>case features have to be checked) are incorporated onto</p><p>so far, embedded subject pronouns in [−null subject] languages</p><p>like English can have referential or quantified NPs within the same sentence as</p><p>antecedents, as well as being interpretable with discourse antecedents. In [+null</p><p>subject] languages, on the other hand, it is not the case that any embedded pronom-</p><p>inal subject can take a quantified antecedent: overt and null pronouns behave</p><p>differently in this respect, as described below.</p><p>Embedded null subjects in [+null subject] languages behave very similarly to</p><p>English overt subject pronouns. That is, the null subject of an embedded clause</p><p>can take either a referential or a quantified expression in the main clause as its</p><p>antecedent; in other words, a null pronoun can be interpreted as a bound vari-</p><p>able.2 This is illustrated in (5) for Spanish and in (6) for Japanese; the (a) exam-</p><p>ples show referential antecedents and the (b) examples show quantified/wh-phrase</p><p>antecedents.</p><p>(5) a. [Juani cree [que proi es inteligente]]</p><p>Johni believes that (hei) is intelligent</p><p>b. [Nadiei cree [que proi es inteligente]]</p><p>Nobodyi believes that (hei) is intelligent</p><p>1.3 Why UG? 7</p><p>(6) a. [Tanaka-sani wa [proi kaisya de itiban da to] itte-iru]</p><p>Tanaka-Mri top (hei) company in best is that saying-is</p><p>‘Mr Tanaka says that (he) is the best in the company.’</p><p>b. [Darei ga [proi kuruma o katta to] itta no?]</p><p>Whoi nom (hei) car acc bought that said Q</p><p>‘Who said that (he) bought a car?’</p><p>Overt pronouns in [+null subject] languages, on the other hand, are more</p><p>restricted than null pronouns; furthermore, they are more restricted than overt</p><p>pronouns in [−null subject] languages. In particular, while an overt pronoun sub-</p><p>ject of an embedded clause in Spanish or Japanese can take a sentence-internal</p><p>referential antecedent, it cannot have a quantified expression or wh-phrase as its</p><p>antecedent. In other words, an overt pronoun cannot receive a bound variable</p><p>interpretation. This contrast is shown in (7) for Spanish and in (8) for Japanese.</p><p>(7) a. Juani cree [que éli es inteligente]</p><p>Johni believes that hei is intelligent</p><p>b. *Nadiei cree [que éli es inteligente]</p><p>Nobodyi believes that hei is intelligent</p><p>(8) a. Tanaka-sani wa [karei ga kaisya de itiban da to] itte-iru</p><p>Tanaka-Mri top hei nom company in best is that saying-is</p><p>‘Mr Tanaka is saying that he is the best in the company.’</p><p>b. *Darei ga [karei ga kuruma o katta to] itta no?</p><p>Whoi nom hei nom car acc bought that said Q</p><p>‘Who said that he bought a car?’</p><p>In both Spanish and Japanese, overt and null pronouns can refer to someone</p><p>else in the discourse, just like overt pronouns in English.3 Thus, a sentence with</p><p>a quantified expression as the main-clause subject and with a null subject in the</p><p>embedded clause is potentially ambiguous; the null subject may either be bound</p><p>to the quantifier, as in (5b) or (6b), or may refer to some other individual in the</p><p>discourse. In contrast, a sentencewith a quantifiedphrase as themain-clause subject</p><p>and an embedded overt-pronoun subject is not ambiguous, since the bound variable</p><p>interpretation is not available (see (7b) and (8b)); only an antecedent elsewhere in</p><p>the discourse is possible.</p><p>The relevant differences between languages like Spanish and Japanese and lan-</p><p>guages like English are summarized in table . Crucially, overt subject pronouns in</p><p>[+null subject] languages cannot take quantified antecedents, whereas null sub-</p><p>jects can, as can overt pronouns in [−null subject] languages. In other respects,</p><p>overt and null pronouns behave alike, permitting referential and discourse an-</p><p>tecedents. Adult native speakers of [+null subject] languages unconsciously know</p><p>8 1 Universal Grammar and language acquisition</p><p>Table 1.1 Antecedents for embedded subject pronouns</p><p>[+Null subject] languages [−Null subject] languages</p><p>Null pronouns Overt pronouns Overt pronouns</p><p>Referential antecedents yes yes yes</p><p>Quantified antecedents yes no yes</p><p>Discourse antecedents yes yes yes</p><p>this restriction on antecedents for overt pronouns, that is, they know that overt pro-</p><p>nouns cannot serve as bound variables.</p><p>The question then arises as to how such knowledge is acquired by native speak-</p><p>ers of null-subject languages. This situation constitutes a learnability problem, in</p><p>that there is a mismatch between the adult knowledge and the kind of data that</p><p>the child is exposed to. The phenomenon in question is very subtle. The input</p><p>is surely insufficient to alert the child to the relevant distinction. For one thing,</p><p>utterances involving quantified antecedents are likely to be relatively infrequent.</p><p>Furthermore, in many cases, overt and null pronouns permit the same kinds of</p><p>antecedents (see table 1.1), so it is unlikely that the absence of overt pronouns</p><p>with quantified antecedents under the relevant interpretation would be detected. A</p><p>further complication is that there is nothing ungrammatical about these particular</p><p>surface forms; sentences like (7b) and (8b) are grammatical on the interpreta-</p><p>tion where there is disjoint reference between the embedded pronoun subject and</p><p>the main clause subject. What the child has to discover is that sentences like</p><p>(7b) or (8b) are ungrammatical on the other interpretation. Negative evidence is</p><p>unlikely to be available; it is implausible that L1 acquirers would produce utter-</p><p>ances incorrectly using overt pronouns with quantified antecedents, with intended</p><p>coreference, and then be provided with implicit or explicit feedback as to their</p><p>ungrammaticality.</p><p>It is on grounds such as these that linguists have argued that certain properties</p><p>of grammar must be innately specified. In the present case, knowledge of the</p><p>distinction between overt and null pronouns is argued to be built in as a universal</p><p>constraint, a principle of UG. Montalbetti (1984) proposed the Overt Pronoun</p><p>Constraint in part to account for the differences described above. This constraint</p><p>holds true of null-argument languages in general, including languages unrelated to</p><p>each other, such as Spanish and Japanese. The Overt Pronoun Constraint is given</p><p>in (9) (based on Montalbetti 1984):</p><p>(9) Overt Pronoun Constraint: overt pronouns cannot receive a bound variable inter-</p><p>pretation (i.e. cannot have quantified or wh-antecedents), in situations where a</p><p>null pronoun could occur.4</p><p>1.4 Parameters of Universal Grammar 9</p><p>To summarize, the distinction in the behaviour of overt and null pronouns with</p><p>respect to the kinds of antecedents that they permit provides an example of a</p><p>poverty of the stimulus situation: the unconscious knowledge that adult native</p><p>speakers have of these properties is extremely subtle. It is implausible that the</p><p>child could induce such restrictions from the input alone. In consequence, it is</p><p>argued that this knowledge must stem from a principle of UG, the Overt Pronoun</p><p>Constraint.</p><p>This is just one example of the kind of abstract knowledge that is attributed to</p><p>UG. The linguistic literature is full ofmany other cases, for example, constraints on</p><p>the distribution of reflexives (Binding Principle A) (Chomsky 1981a), constraints</p><p>on the distribution of empty categories (the Empty Category Principle) (Chomsky</p><p>1981a), and constraints onwh-movement (Subjacency) (Chomsky 1977). As men-</p><p>tioned in section 1.2, linguistic theory has developed over time and the formulation</p><p>of many of the proposed principles of UG has changed. In this book, we will not</p><p>be concerned with the precise technical details as to how UG principles have been</p><p>formulated and reformulated. Rather, the crucial question here is the identification</p><p>of linguistic knowledge that could not arise from the input alone and that requires</p><p>the postulation of innate principles.</p><p>As we shall see in chapter 2, the same general issue arises in the context of L2</p><p>acquisition. That is, it appears that L2 learners are also faced with a poverty of the</p><p>stimulus, namely the L2 stimulus (Schwartz and Sprouse 2000a, b; White 1985a,</p><p>1989), and that their interlanguage competence goes beyond the input that they</p><p>are exposed to. Hence, the question arises as to whether interlanguage grammars</p><p>the verb that bears the</p><p>148 4 Grammars beyond the initial state</p><p>associated case features. When acquiring V2, then, Cevdet appears to have hit</p><p>upon the incorporation option, an option utilized in neither the L1 nor the L2, such</p><p>that only pronominal subjects can occur in inversion contexts. If the government</p><p>option were available, the verb in C should be able to govern the subject regardless</p><p>of whether or not it is pronominal. In stage 3, Cevdet acquires the government</p><p>option.</p><p>Once again, there is something puzzling here. This learner has arrived at an</p><p>analysis that is fully compatible with UG. However, it is not at all clear what</p><p>properties of the L2 German input would lead Cevdet to arrive at an analysis in</p><p>terms of incorporation.</p><p>4.9 Parameter setting and resetting: assessment</p><p>To summarize, results discussed in the preceding sections suggest that</p><p>developing interlanguage grammars: (i) show changes in feature strength, from</p><p>strong to weak or vice versa; (ii) contain features not present in the L1 grammar;</p><p>(iii) contain functional categories not present in the L1 grammar. In other words,</p><p>learners can acquire L2 functional categories, features and feature values, together</p><p>with their associated consequences, thus, implicating parameter resetting. These</p><p>findings hold true of child and adult learners, across a variety of languages, tested by</p><p>means of various methodologies. L2 learners may also arrive at parameter settings</p><p>of neither the L1 nor the L2, though it is not always clear what properties of</p><p>the L2 input motivate such misanalyses. Development away from the L1 grammar</p><p>provides additional evidence in favour of interlanguage grammars being fully UG-</p><p>constrained, in so far as these grammars exhibit properties which are not derivable</p><p>from the L1 or from surface properties of the L2 input and which are consistent</p><p>with natural language grammars.</p><p>In some cases, we have seen evidence for a stage where L1 values of parameters</p><p>are instantiated before L2 values come into place; in other cases, the relevant L2</p><p>properties emerge very early, without any obvious stage of L1 influence. On the</p><p>whole, the results are consistent with Full Transfer Full Access: learners start out</p><p>with L1 functional categories, features and feature strength and are able to acquire</p><p>L2 categories, features and feature strength. Nevertheless, some aspects of the</p><p>results are puzzling for this view, in that effects of the L1 reveal themselves more</p><p>often than not in the form of variability, with L1 and L2 properties co-occurring,</p><p>rather than there being initial effects of the L1 setting alone.</p><p>More generally, results are not totally unproblematic as far as claims for param-</p><p>eters in interlanguage grammars are concerned. In the case of the Verb Movement</p><p>Parameter, three properties assumed to be the consequence of feature strength</p><p>Topics for discussion 149</p><p>of I did not cluster in the interlanguage grammar: White (1992a) found that</p><p>French-speaking learners of English consistently prohibited verb raising in ques-</p><p>tions and negatives, as if I were weak, as it is in the L2, while at the same time</p><p>permitting verb raising over adverbs, as if I were strong, as it is in the L1. If this</p><p>is one parameter, then constructions related under it are not being treated in the</p><p>same way. In the case of nominals, we have seen considerable success in three</p><p>different areas (strength, features, categories), which might seem supportive of the</p><p>claim that clusters of properties are implicated. Unfortunately, it is unclear to what</p><p>extent one parameter is involved. Gess and Herschensohn (2001) attribute differ-</p><p>ences between French andEnglish in the nominal domain to aDP parameter, which</p><p>accounts for many of the differences discussed above. However, this purported pa-</p><p>rameter does not appear to be generalizable beyond Romance versus English: it</p><p>seems to suggest that languages with gender and gender agreement will also have</p><p>strong Num, hence N-raising over adjectives, contrary to fact. German, for ex-</p><p>ample, has gender and gender agreement but no N-raising. Ultimately, theories</p><p>of parameter setting in interlanguage grammars depend on convincing proposals</p><p>from theoretical linguists as to the nature of parameters.</p><p>4.10 Conclusion</p><p>In this chapter, we have considered proposals for total or local breakdown</p><p>in parameters, for restriction to L1 parameter values, and for successful setting of</p><p>parameters to values distinct from the L1. Despite conflicting evidence and con-</p><p>flicting theories, results from several studies suggest that interlanguage grammars</p><p>conform to parameters of UG. Nevertheless, L2 performance is by no means per-</p><p>fect. In particular, there is often divergence between syntactic and morphological</p><p>performance, with greater accuracy on syntactic properties than on morphological</p><p>ones. Variability is characteristic of the morphological domain. We address this</p><p>issue in more detail in chapter 6.</p><p>Topics for discussion</p><p>� Several studies in this chapter explore word order in the DP in Romance,</p><p>particularly the relative positions of nouns and adjectives, which are ac-</p><p>counted for in terms of feature strength. Yet it might reasonably be ar-</p><p>gued that this property is something that is taught in the L2 classroom</p><p>and learned prescriptively, hence that it does not constitute a genuine</p><p>poverty of the L2 stimulus situation. How can one distinguish between a</p><p>150 4 Grammars beyond the initial state</p><p>theory that attributes some phenomenon to ‘deep’ properties of an abstract</p><p>grammar and a theory that says it is learned as an isolated construction?</p><p>� Can functional categories be lost in L2 acquisition? For example, in the</p><p>case of Chinese versus English, the L2 learner of Chinese must acquire</p><p>cl and lose Det, while the L2 learner of Chinese must do the oppo-</p><p>site. One possibility is that the new category is acquired without the old</p><p>one being lost. How could one determine this?</p><p>� Impairment. It has been assumed here that a grammar restricted to L1</p><p>settings is not impaired, in the sense that it conforms to principles and</p><p>parameters of UG. On the other hand, it is impaired in the sense that true</p><p>restructuring in response to L2 input is not possible.</p><p>� Clustering.Most of the studies described here assume that clusteringmust</p><p>take place during the course of acquisition. What would it mean if one</p><p>could show that certain properties cluster in the endstate and yet had not</p><p>been acquired together? In other words, can there be different routes to</p><p>the same endstate?</p><p>� Does parameter ‘resetting’ imply that the L1 grammar is also changed?</p><p>� An implication of global impairment and of the Local Impairment</p><p>Hypothesis and the No Parameter Resetting Hypothesis is that it is im-</p><p>possible to be fully ‘successful’ in adult L2 acquisition. Such theories</p><p>concentrate on difficulties that L2 learners have, as well as on differences</p><p>between interlanguage grammars and native-speaker grammars. How can</p><p>such theories account for successful L2 acquisition?</p><p>Suggestions for additional reading</p><p>� Several earlier studies examine the effects of word-order parameters in</p><p>interlanguage, in the context of the debate over whether or not parameters</p><p>can be reset in L2 acquisition (e.g. Clahsen andMuysken 1986; du Plessis</p><p>et al. 1987; Flynn 1987; Hulk 1991; Schwartz and Tomaselli 1990).</p><p>� For additional evidence that new functional categories and feature</p><p>strength can be acquired, see research on the L2 acquisition of clitics</p><p>(Duffield and White 1999; Duffield et al. 2002; White 1996a).</p><p>� For arguments that gender is not represented in the same way in L2 as it</p><p>is in L1, see Carroll (1999).</p><p>� For arguments that L2 grammars are UG-constrained but that the course</p><p>of acquisition differs in L1 and L2, see Herschensohn (2000).</p><p>5</p><p>The transition problem, triggering</p><p>and input</p><p>5.1 Introduction</p><p>As a number of researchers have pointed out, theories of language acquisi-</p><p>tion must explain both properties of linguistic representations (the form and nature</p><p>of the grammar) and transition or development (how</p><p>are constrained by UG, an issue which will be a major focus of this book.</p><p>1.4 Parameters of Universal Grammar</p><p>In addition to universal principles, UG includes principles with a limited</p><p>number of built-in options (settings or values), which allow for crosslinguistic vari-</p><p>ation. Such principles are known as parameters. Most parameters are assumed to</p><p>be binary, that is, they have only two settings, the choices being predetermined by</p><p>UG. L1 acquisition consists, in part, of setting parameters, the appropriate setting</p><p>being triggered by the input that the child is exposed to. A central claim of param-</p><p>eter theory, as originally instantiated in the Principles and Parameters framework,</p><p>is that a single parameter setting brings together a cluster of apparently disparate</p><p>syntactic properties (Chomsky 1981a). This, for example, was part of the rationale</p><p>for the Null Subject Parameter, which related the possibility of null subjects</p><p>to other syntactic and morphological properties found in null subject languages</p><p>(Chomsky 1981a; Jaeggli 1982; Rizzi 1982, amongst others). The insight behind</p><p>10 1 Universal Grammar and language acquisition</p><p>the proposal for parameters is that they should severely reduce the acquisition task.</p><p>Rather than learning a number of seemingly unrelated properties individually, the</p><p>child has only to discover the appropriate setting of a parameter and a range of as-</p><p>sociated syntactic properties follows automatically. Some L1 acquisition research</p><p>has provided evidence in favour of clustering, showing that properties which are</p><p>argued to be consequences of a particular parameter setting emerge at about the</p><p>same time (e.g. Hyams 1986; Snyder and Stromswold 1997).</p><p>Under current proposals, parametric differences between grammars are associ-</p><p>ated with properties of lexical items, particularly so-called functional categories</p><p>(Borer 1984; Chomsky 1995; Ouhalla 1991; Pollock 1989). Linguistic theory</p><p>distinguishes between lexical categories – verb (V), noun (N), adjective (Adj), ad-</p><p>verb (Adv), preposition (P) – and functional categories, including complementizer</p><p>(Comp or C), inflection (Infl or I) (often split into agreement (Agr) and tense (T)),</p><p>negation (Neg), determiner (Det), number (Num), as well as others. Functional</p><p>categories have certain formal features associated with them (such as tense, num-</p><p>ber, person, gender and case). Functional categories and features form part of the</p><p>UG inventory.</p><p>There are three potential sources of crosslinguistic variation relating to func-</p><p>tional categories:</p><p>i. Languages can differ as to which functional categories are realized in the</p><p>grammar. On some accounts, for example, Japanese lacks the category</p><p>Det (Fukui and Speas 1986).</p><p>ii. The features of a particular functional category can vary from language to</p><p>language. For instance, French has a gender feature, while English does</p><p>not.</p><p>iii. Features are said to vary in strength: a feature can be strong in one lan-</p><p>guage and weak in another, with a range of syntactic consequences. For</p><p>example, Infl features are strong in French and weak in English (see</p><p>below), resulting in certain word-order alternations between the two</p><p>languages.</p><p>The lexicons of different languages, then, vary as to which functional categories</p><p>and features are instantiated and what the strength of various features may be.</p><p>Such variation has a variety of syntactic effects.</p><p>1.4.1 An example: feature strength and movement</p><p>In this section, we review the role of feature strength in current accounts</p><p>of syntax, and consider some examples of parametric variation which depend</p><p>1.4 Parameters of Universal Grammar 11</p><p>on feature strength. In later chapters, such variation will become relevant as we</p><p>examine the nature of interlanguage grammars, and the kinds of changes that take</p><p>place in the grammar during the course of L2 development.</p><p>Feature strength is an abstract property which is argued to have syntactic con-</p><p>sequences, particularly for word order. The first example to be considered here</p><p>concerns the strength of features associated with the functional category Infl.</p><p>Finite verbs have tense and agreement features which have to be checked, at some</p><p>point, against corresponding V(erb)-features in Infl (Chomsky 1995). Simplifying</p><p>somewhat, if the V-features in Infl are strong (henceforth, strong I), there is overt</p><p>movement of the finite verb, which raises from the VP to I for feature checking.</p><p>If V-features are weak (henceforth, weak I), overt movement does not take place.</p><p>Instead, features are checked at Logical Form (LF); this movement is not ‘visible’</p><p>in the syntax and is said to be covert.</p><p>This distinction between strong and weak features accounts for a number of</p><p>well-known word-order differences between languages like French and English</p><p>(Emonds 1978; Pollock 1989). In French, finite lexical verbs must appear to the</p><p>left of the negative pas and to the left of VP-adjoined adverbs, as illustrated in</p><p>(10). In English, on the other hand, the lexical verb remains to the right of not and</p><p>to the right of adverbs, as shown in (11).</p><p>(10) a. Marie ne regarde pas la télévision.</p><p>Mary (ne) watches not the television</p><p>‘Mary does not watch television.’</p><p>b. *Marie pas regarde la télévision.</p><p>Mary not watches the television</p><p>c. Marie regarde souvent la télévision.</p><p>Mary watches often the television</p><p>d. *Marie souvent regarde la télévision.</p><p>Mary often watches the television</p><p>(11) a. Mary does not watch television.</p><p>b. *Mary watches not television.</p><p>c. Mary often watches television.</p><p>d. *Mary watches often television.</p><p>These verb placement differences are accounted for in terms of differences in</p><p>feature strength, French having strong I and English weak. At an underlying level,</p><p>the two languages have the same structure (compare (12) and (13)). However,</p><p>because of the difference in feature strength, finite verbs in French must raise to I</p><p>for feature-checking purposes, whereas finite verbs in English remain within the</p><p>VP. This is illustrated in (12) and (13).</p><p>12 1 Universal Grammar and language acquisition</p><p>(12) CP</p><p>Spec C'</p><p>C</p><p>Spec</p><p>Marie</p><p>I</p><p>regardei</p><p>souvent VP</p><p>V</p><p>ti</p><p>NP</p><p>la télévision</p><p>IP</p><p>I'</p><p>VP</p><p>(13) CP</p><p>Spec C'</p><p>C</p><p>Spec</p><p>Mary</p><p>I VP</p><p>often VP</p><p>V</p><p>watches</p><p>NP</p><p>television</p><p>IP</p><p>I'</p><p>Germanic languages provide another example of crosslinguistic differences in</p><p>word order which are partially explained in terms of feature strength. Languages</p><p>like English and German contrast in two respects, namely the underlying position</p><p>of the verb (VP initial in English, final in German), and the verb second (V2)</p><p>phenomenon (characteristic of German but not English). Main clauses in German</p><p>and English both show subject-verb-object (SVO) order when no auxiliaries or</p><p>modals are present, as shown in (14a, b). In such cases, sentences with finite</p><p>main verbs in final position are ungrammatical, as shown in (14c, d). However,</p><p>in German main clauses containing auxiliary or modal verbs, the lexical verb</p><p>appears finally (see (14e)); all verbs appear finally in embedded clauses, as in</p><p>(14f). Furthermore, in German main clauses, any constituent can be fronted; when</p><p>this happens, the verb must appear in the second position (V2) in the clause, as</p><p>shown in (14g–j). That is, the finite verb in main clauses can only be preceded by</p><p>one other constituent, which does not have to be a subject.</p><p>1.4 Parameters of Universal Grammar 13</p><p>(14) a. Maria trinkt Kaffee.</p><p>b. Mary drinks coffee.</p><p>c. *Maria Kaffee trinkt.</p><p>d. *Mary coffee drinks.</p><p>e. Maria möchte Kaffee trinken.</p><p>Mary wants coffee drink-inf</p><p>f. Maria sagt, dass sie Kaffee trinken will.</p><p>Mary says that she coffee drink-inf will</p><p>g. Kaffee trinkt Maria.</p><p>Coffee drinks Mary</p><p>‘Mary drinks coffee.’</p><p>h. *Kaffee Maria trinkt</p><p>coffee Mary drinks</p><p>i. Oft trinkt Maria Kaffee.</p><p>often drinks Mary coffee</p><p>j. *Oft Maria trinkt Kaffee.</p><p>often Mary drinks coffee</p><p>The position of the verb in German is accounted for in the following way.</p><p>According to standard analyses of German, VP and IP are head final, as shown</p><p>in (15) (e.g. Platzack 1986; Schwartz</p><p>and Vikner 1996; Thiersch 1978).5 Finite</p><p>verbs in main clauses undergo two movements: from V to I and then from I to C,</p><p>driven by strong features in C. Some other constituent (subject, object or adjunct)</p><p>raises to the Spec of CP, resulting in the V2 effect. In embedded clauses, the verb</p><p>cannot raise to C because this position is already filled by a complementizer, such</p><p>as dass (‘that’) in (14f); consequently, embedded clauses remain V-final. This is</p><p>shown in (16).</p><p>(15) CP</p><p>C</p><p>trinkti</p><p>IP</p><p>I'</p><p>VP I</p><p>ti</p><p>oft VP</p><p>NP</p><p>Kaffee</p><p>V</p><p>ti</p><p>Spec</p><p>Mariaj</p><p>C'</p><p>Spec</p><p>tj</p><p>14 1 Universal Grammar and language acquisition</p><p>(16)</p><p>CP</p><p>Spec</p><p>C</p><p>dass</p><p>IP</p><p>Spec</p><p>Maria</p><p>I'</p><p>VP I</p><p>trinkti</p><p>oft VP</p><p>NP</p><p>Kaffee</p><p>V</p><p>ti</p><p>C'</p><p>DPs provide a final example of word-order variation attributed to differences</p><p>in feature strength. On many current analyses, DPs contain a functional category</p><p>Num, located betweenD andNP, as shown in (17) (Bernstein 1993; Carstens 1991;</p><p>Ritter 1991; Valois 1991). Num has number features, as well as gender features in</p><p>some accounts (Ritter 1993).</p><p>(17) DP</p><p>D NumP</p><p>Num</p><p>[+/–pl]</p><p>NP</p><p>AdjP NP</p><p>|</p><p>N</p><p>In Romance languages such as French and Spanish, number features are strong</p><p>and nouns must raise overtly from N to Num for feature-checking purposes, over</p><p>any adjectives that may be present. This results in the noun adjective (N Adj) order</p><p>typical of Romance, as shown in the Spanish example in (18a). In English, on</p><p>the other hand, Num features are weak, nouns do not raise and the word order is</p><p>adjective noun (Adj N), as in (18b).</p><p>(18) a. la blusa roja</p><p>the blouse red</p><p>b. the red blouse</p><p>In other words, parallel to the situation with respect to the position of the verb</p><p>in the clause, crosslinguistic differences in the position of the noun in the DP are</p><p>determined by feature strength.</p><p>1.5 UG access: earlier approaches to UG and SLA 15</p><p>In summary, a variety of word-order differences are accounted for under the</p><p>assumption that the strength of features in functional categories varies, being ei-</p><p>ther strong or weak. Various word-order alternations between French and English</p><p>(including others that have not been mentioned here) can be accounted for by one</p><p>parametric difference between the two languages, namely the strength of V-related</p><p>features in I. A range of differences between languages like German and English</p><p>can be accounted for by two properties, the underlying position of the verb and the</p><p>strength of features in C. Differences in adjective placement between Romance</p><p>languages and Germanic languages can be accounted for in terms of the strength</p><p>of features in Num. These parametric differences will be discussed in greater detail</p><p>in later chapters, as we consider the extent to which the interlanguage grammar</p><p>exemplifies parameter settings distinct from those found in the L1 grammar.</p><p>As is the case with principles of UG, the formulation of the precise mechanisms</p><p>involved in feature strength and feature checking has changed over time. In this</p><p>book, the issues will be presented in a way which preserves the general insights</p><p>without being tied to technical details specific to any particular theory.</p><p>1.5 UG access: earlier approaches to UG and SLA</p><p>So far, we have considered UG as a system of principles and parameters</p><p>which provide constraints on grammars in the course of L1 acquisition, as well as</p><p>on adult native-speaker grammars. L2 learners face a task parallel to that of L1</p><p>acquirers, namely the need to arrive at a linguistic system which accounts for the</p><p>L2 input, allowing the learner to understand and speak the second language. Given</p><p>this apparent similarity, the question of whether UG also mediates L2 acquisition,</p><p>and to what extent, has been investigated and debated since the early 1980s. The</p><p>first decade of research on UG in L2 acquisition concentrated largely on the so-</p><p>called access issue, namely, whether or not UG remains available in non-primary</p><p>acquisition. (See White (1989) for an overview and discussion of the relevant</p><p>literature.) This research looked for evidence that L2 learners can (or cannot)</p><p>apply principles of UG, and set or reset parameters, as well as investigating the</p><p>extent to which the mother tongue (L1) was involved, for example through the</p><p>adoption of L1 parameter settings in interlanguage grammars. Hypotheses varied</p><p>as to whether L2 learners have no access, direct access or indirect access to UG.</p><p>All of these terms have turned out to be somewhat problematic.</p><p>One side of the debate, sometimes referred to as the no access position (for ex-</p><p>ample, by Cook 1988; Cook and Newson 1996; Epstein, Flynn andMartohardjono</p><p>1996), is represented by the Fundamental Difference Hypothesis (Bley-Vroman</p><p>1990) and related claims (Clahsen andMuysken 1986; Schachter 1988).According</p><p>16 1 Universal Grammar and language acquisition</p><p>to this view, child L1 and adult L2 acquisition differ in major respects. Proponents</p><p>claim that adult L2 acquisition is not constrained by UG, or that it is only con-</p><p>strained by UG insofar as universal properties can be accessed via the L1 grammar.</p><p>Indeed, the assumption that UG is at least partially implicated via the L1 suggests</p><p>that the term no access is a misnomer; hence, this view is sometimes also referred</p><p>to as partial access. Regardless of terminology, the crucial claim is that all the lin-</p><p>guistic mechanisms available to the L1 acquirer are no longer available to the L2</p><p>learner. In support, advocates of this position tried to show that learners are ‘stuck’</p><p>with principles and parameter settings exemplified in the L1 (e.g. Schachter 1989)</p><p>or that their grammars show no evidence for UG constraints at all (e.g. Clahsen</p><p>and Muysken 1986).</p><p>On the other side of the debate is the position that L2 learners indeed have</p><p>access to UG. In other words, interlanguage grammars show evidence of being</p><p>constrained by UG principles; at the same time, interlanguage grammars show</p><p>evidence of parameter settings other than those of the L1. Some proponents of</p><p>the UG access position argued that at no stage would the interlanguage grammar</p><p>actually exemplify L1 parameter settings (e.g. Flynn 1987). In other words, L2</p><p>learners arrive at relevant properties of the L2 independently of the L1 grammar.</p><p>Hence, this position was often referred to as direct access (e.g. by Cook 1988;</p><p>Cook and Newson 1996).</p><p>An alternative kind of account recognized the role of both the L1 and UG: L2</p><p>learners are indeed assumed to have access to principles and parameters of UG.</p><p>However, initially at least, access would be via the L1 grammar, with the possi-</p><p>bility of subsequent grammar restructuring and parameter resetting, in the light of</p><p>exposure to L2 input (e.g.White 1985b, 1989). This position is sometimes referred</p><p>to as indirect access (e.g. by Cook 1988; Cook and Newson 1996). However, as</p><p>pointed out by Thomas (1991b, 1993), it is just as appropriate to characterize this</p><p>perspective as involving direct access, since the learner is not restricted to UG</p><p>principles and parameter settings exemplified in the L1 grammar.</p><p>Terms like direct and indirect access have since been replaced with full and</p><p>partial access but these have proved to be equally problematic. As we shall see</p><p>in chapter 3, there is still disagreement as to whether or not full access to UG</p><p>implies absence of L1 effects on the interlanguage grammar. Epstein, Flynn and</p><p>Martohardjono (1996), for example, restrict the term full access to the position</p><p>that UG operates in interlanguage grammars independently of L1 representations.</p><p>In contrast, Schwartz and Sprouse (1996) propose the Full Transfer Full Access</p><p>hypothesis, arguing that there is nothing incompatible in the assumption that both</p><p>UG and the L1 grammar are implicated. Since the L1 is a natural language, there is</p><p>no a priori justification for assuming that a representation based on the L1 implies</p><p>lack of UG constraints, or restricted access to UG.</p><p>1.6 Methodological issues 17</p><p>As hypotheses about UG access developed, interest began to shift from over-</p><p>arching questions like ‘Is UG available?’ or ‘What kind of</p><p>UG access is there</p><p>in L2?’ to a closer examination of the nature of the interlanguage grammar, with</p><p>particular focus on whether interlanguage grammars exhibit properties charac-</p><p>teristic of natural language (e.g. du Plessis, Solin, Travis and White 1987; Finer</p><p>and Broselow 1986; Martohardjono and Gair 1993; Schwartz and Tomaselli 1990;</p><p>Thomas 1991a; White 1992c). As we shall see, this detailed focus on the gram-</p><p>matical properties of interlanguage grammars remains characteristic of current</p><p>research.</p><p>1.6 Methodological issues: ‘tapping’ linguistic competence</p><p>The research to be discussed in this book seeks to establish the nature</p><p>of the L2 learner’s linguistic competence, addressing in particular the question</p><p>of whether interlanguage grammars are UG-constrained. This raises the issue of</p><p>how one can in fact discover what the unconscious linguistic system consists</p><p>of. Linguistic competence is an abstraction; there is no way of directly tapping</p><p>that competence. Hence, researchers must resort to various kinds of performance</p><p>measures in order to determine, indirectly, the essential characteristics of mental</p><p>representations. This is true whether one is interested in adult native-speaker com-</p><p>petence, child L1 acquisition or child or adult non-native language acquisition.</p><p>A variety of methodologies have been developed over the years for investigating</p><p>linguistic competence, and data have been obtained using different experimental</p><p>techniques. It is, of course, the case that nomethodology allows one to tap linguistic</p><p>competence directly: in all cases, performance factors will be involved. Ideally,</p><p>performance data from various sources will converge. When results from different</p><p>tasks and different groups of learners show the same trends, this suggests that we</p><p>are indeed gaining insight (indirectly) into the nature of the underlying linguistic</p><p>competence.</p><p>Data can be broadly classified into three categories: production data, including</p><p>spontaneous and elicited production; comprehension data, including data obtained</p><p>from act-out and picture-identification tasks; and intuitional data, including data</p><p>from grammaticality judgments and truth-value judgments (see chapter 2), as well</p><p>as, more recently, a number of online techniques such as sentence matching (see</p><p>chapters 3 and 4).</p><p>A myth has developed in the field of L2 acquisition that researchers working in</p><p>the UG paradigm take grammaticality-judgment tasks to have some kind of priv-</p><p>ileged status, such that they provide a direct reflection of linguistic competence</p><p>(e.g. Carroll and Meisel 1990: 205; Ellis 1990: 388). This is a misconception: it</p><p>18 1 Universal Grammar and language acquisition</p><p>has always been recognized that judgment data are performance data, on a par</p><p>with other data (e.g. Cook 1990: 592; White 1989: 57–8). The appropriateness</p><p>of a particular task will depend on what the researcher is trying to discover. For</p><p>example, grammaticality-judgment tasks provide a means of establishing whether</p><p>learners know that certain forms are impossible or ungrammatical in the L2. Thus,</p><p>a grammaticality-judgment task can be used to find out whether sentences which</p><p>are ruled out by principles of UG are also disallowed in the interlanguage gram-</p><p>mar. Consider, for instance, the Adjunct Island Constraint (e.g. Cinque 1990), a</p><p>constraint which prohibits wh-phrases from being fronted out of adjunct clauses.</p><p>In order to establish whether L2 learners ‘know’ this constraint, one could ask</p><p>them whether or not sentences like those in (19) are grammatical:</p><p>(19) a. Who did you quit school because you hated?</p><p>b. What did Tom fall when he slipped on?</p><p>If interlanguage grammars are constrained by UG, then learners are expected to</p><p>reject such sentences (while accepting corresponding grammatical ones).</p><p>Although grammaticality-judgment tasks suffer from a number of well-known</p><p>problems (see, for example, Birdsong (1989) and Schütze (1996)), in cases like the</p><p>above example they have advantages over other sources of data, such as sponta-</p><p>neous production. If L2 learners never produce sentences like (19), it would seem,</p><p>on the face of things, to provide support for the claim that interlanguage grammars</p><p>are UG-constrained. Unfortunately, however, failure to find certain sentence types</p><p>in production data is no guarantee that such sentences are in fact disallowed by</p><p>the grammar. There may be independent reasons why they fail to show up. The</p><p>use of methodologies such as grammaticality-judgment tasks, then, allows the</p><p>experimenter to investigate aspects of interlanguage competence which may not</p><p>otherwise be amenable to inspection.</p><p>It is important to recognize that there is no one methodology that is appropriate</p><p>for investigating all aspects of linguistic competence. For example, if questions</p><p>of interpretation are being investigated, grammaticality judgments will often be</p><p>totally uninformative. Consider the Overt Pronoun Constraint, as discussed in</p><p>section 1.3.1. As we have seen, certain Spanish and Japanese sentences involving</p><p>overt-pronoun subjects in embedded clauses and quantified phrases asmain clauses</p><p>subjects are ungrammatical under a bound variable interpretation, as in (7b) and</p><p>(8b). This contrasts with English, where the interpretation in question is possible.</p><p>If a researcher wanted to determine whether or not Spanish-speaking learners</p><p>of English know that a sentence like (3a), repeated here as (20), is possible, a</p><p>traditional grammaticality-judgment task would not be appropriate.</p><p>(20) Everyone thinks that she will win.</p><p>Topics for discussion 19</p><p>The problem is that this sentence is ambiguous for native speakers of English,</p><p>being grammatical on two different interpretations (i.e. with she taking everyone as</p><p>its antecedent orwith a discourse referent as its antecedent). If learners respond that</p><p>such sentences are grammatical, it would be impossible to tell which interpretation</p><p>of the sentence was being judged. In other words, one could not tell whether the</p><p>learner had acquired unconscious knowledge of the difference between Spanish</p><p>and English with respect to this property. In such cases, alternative methodolo-</p><p>gies are called for, which match sentences with potential interpretations. This is</p><p>often achieved by means of so-called truth-value-judgment tasks which require</p><p>the learner to assess the appropriateness of a sentence in relation to some context</p><p>(see chapter 2).</p><p>For such reasons, it is essential for the researcher to construct tasks that are</p><p>appropriate for the issue being investigated. Various different methodologies will</p><p>be described in greater detail in later chapters, including a consideration of their</p><p>appropriateness, as well as their advantages and disadvantages.</p><p>1.7 Conclusion</p><p>In conclusion, UG is proposed as a (partial) answer to questions such</p><p>as: What are natural language grammars like? What is the nature of linguistic</p><p>competence? How is it acquired? As far as the first language is concerned, the as-</p><p>sumption is that language acquisition would be impossible in the absence of innate</p><p>and specifically linguistic principles which place constraints on grammars, thus</p><p>restricting the ‘hypothesis space’ by severely limiting the range of possibilities</p><p>that the language acquirer has to entertain. In subsequent chapters, we will explore</p><p>the extent to which interlanguage grammars are similarly constrained. Research</p><p>will be considered which examines in detail the nature of interlanguage represen-</p><p>tations. As we shall see, claims are made for early grammars (the initial state), for</p><p>grammars during the course of development, as well as for the nature of the steady</p><p>state. We will contrast claims that interlanguage grammars are in some sense de-</p><p>fective (hence, not UG-constrained) with positions that argue that interlanguage</p><p>grammars are not impaired, showing, rather, properties characteristic of natural</p><p>languages constrained by UG.</p><p>Topics for discussion</p><p>� A number of researchers have suggested that negative evidence is in</p><p>fact available in L1 acquisition. For example,</p><p>Hirsh-Pasek, Treiman and</p><p>Schneiderman (1984) report that mothers of 2-year-olds are significantly</p><p>20 1 Universal Grammar and language acquisition</p><p>more likely to repeat (and sometimes rephrase) children’s ill-formed ut-</p><p>terances than their well-formed utterances. Does the availability of such</p><p>feedback in fact solve the logical problem of language acquisition?</p><p>� The claim that there are domain-specific universal linguistic principles</p><p>constraining grammars is, of course, contested. For example, O’Grady</p><p>(1987, 1996, 1997, 2003) proposes that language acquisition should be</p><p>accounted for in terms of more general cognitive principles which are</p><p>not unique to language. Others place far greater emphasis on statistical</p><p>properties of the input, in some cases downplaying or denying a role for</p><p>innate constraints, for example, connectionist models such as Parallel</p><p>Distributed Processing (Rumelhart and McClelland 1987) or the Compe-</p><p>tition Model (Bates and MacWhinney 1987). (For an overview of recent</p><p>research which assumes a major role for statistical learning as well as</p><p>innate constraints, see Newport and Aslin (2000).) How can one choose</p><p>between these very different kinds of account (i.e. what kinds of argumen-</p><p>tation and data are relevant)? For relevant discussion in the L2 context,</p><p>see Gregg (2003).</p><p>� To what extent are functional categories universally realized and what</p><p>are the implications for theories of L2 acquisition? There is consider-</p><p>able disagreement as to whether or not languages differ in the functional</p><p>categories that they instantiate. See Bobaljik and Thráinsson (1998),</p><p>Thráinsson (1996) and Webelhuth (1995) for useful discussion.</p><p>� The problem of teleology. The task of the language acquirer (L1 or L2) is</p><p>to ‘construct’ a grammar that accommodates the linguistic input, allowing</p><p>the learner to provide structural representations to utterances. The task</p><p>should not be seen as having to acquire a grammar that matches the</p><p>grammar of adult speakers of the language in question.Why is it important</p><p>to make such a distinction?</p><p>Suggestions for additional reading</p><p>� It will be presupposed that the reader has some familiarity with the</p><p>concepts and mechanisms assumed in current generative grammar. The</p><p>Government and Binding framework is presented in Haegeman (1991),</p><p>Minimalism in Radford (1997). Papers in Webelhuth (1995) provide a</p><p>useful overview of both frameworks and the connections between them.</p><p>� Arguments for an innate and specifically linguistic basis to first language</p><p>acquisition can be found in Chomsky (1999), Crain and Thornton (1998),</p><p>Pinker (1994), amongst others. Useful overviews of L1 acquisition</p><p>Suggestions for additional reading 21</p><p>theories and findings within this framework can be found in Goodluck</p><p>(1991) and O’Grady (1997), as well as in several of the chapters in Bloom</p><p>(1994) and in Ritchie and Bhatia (1999).</p><p>� For detailed presentation of methodologies appropriate for research on</p><p>first language acquisition, see Crain and Thornton (1998) and McDaniel,</p><p>McKee and Cairns (1996). For detailed discussion of the pros and cons of</p><p>grammaticality-judgment tasks, see Birdsong (1989) and Schütze (1996).</p><p>� Other recent books looking at L2 acquisition can be seen as complemen-</p><p>tary to this book. Hawkins (2001a) provides an excellent introduction to</p><p>L2 acquisition of syntax and morphology within a generative linguistic</p><p>perspective. Herschensohn (2000) adopts a more technical Minimalist</p><p>approach to L2 acquisition.</p><p>2</p><p>Principles of Universal Grammar</p><p>in L2 acquisition</p><p>2.1 UG and the logical problem of L2 acquisition</p><p>As discussed in chapter 1, UG is motivated on learnability grounds: the</p><p>subtle and abstract knowledge attained by native speakers goes far beyond the input</p><p>that they receive as young children. In L2 acquisition, learners are faced with a</p><p>similar task to that of L1 acquirers, namely the need to arrive at a system accounting</p><p>for L2 input. In addition, L2 learners are also faced, at least potentially, with a</p><p>logical problem of language acquisition, in that there are abstract, complex and</p><p>subtle properties of grammar that are underdetermined by the L2 input (Schwartz</p><p>andSprouse2000a, b;White 1985a, 1989). If it turns out that theL2 learner acquires</p><p>abstract properties that could not have been induced from the input, this is strongly</p><p>indicative that principles of UG constrain interlanguage grammars, parallel to the</p><p>situation in L1 acquisition. This is true even if the linguistic competence of L2</p><p>learners differs from the linguistic competence of native speakers. In otherwords, it</p><p>is not necessary for L2 learners to acquire the same knowledge as native speakers</p><p>in order to demonstrate a poverty-of-the-stimulus situation in L2 acquisition; it</p><p>is sufficient to show that L2 learners acquire complex and subtle properties of</p><p>language that could not have been induced from the L2 input.</p><p>However, L2 learners already have ameans of representing language, namely the</p><p>grammar of the mother tongue. Thus, it might be that there is, in fact, no underde-</p><p>termination problem: if L2 learners demonstrate the relevant kind of unconscious</p><p>knowledge, it might be the case that they are drawing on the L1 grammar, rather</p><p>than on UG itself, as argued, for example, by Bley-Vroman (1990) and Schachter</p><p>(1990).</p><p>Thus, the strongest case for the operation of principles of UG in interlanguage</p><p>grammars can be made if learners demonstrate knowledge of subtle and abstract</p><p>linguistic properties which could neither have been learned from L2 input alone</p><p>nor derived from the grammar of the mother tongue. In other words, there should</p><p>be underdetermination not only with respect to L2 input but also with respect to</p><p>the L1 grammar. Furthermore, one must also be able to rule out the possibility</p><p>of learning on the basis of explicit instruction or by means of general learning</p><p>22</p><p>2.1 UG and the logical problem of L2 acquisition 23</p><p>principles (not specifically linguistic). For such reasons, L2 researchers try to</p><p>identify situations involving a poverty of the L2 stimulus, where the available L2</p><p>input togetherwith existing grammatical knowledge cannot account for acquisition</p><p>unless one assumes that interlanguage grammars are constrained by UG (Schwartz</p><p>and Sprouse 2000a, b; White 1989, 1990).</p><p>In summary, to demonstrate convincingly that interlanguage grammars are con-</p><p>strained by principles of UG, the following conditions should hold:</p><p>i. The phenomenon being investigated must be underdetermined by the L2</p><p>input. That is, it must not be something that could be acquired by observa-</p><p>tion of the L2 input, including statistical inferencing based on frequency</p><p>of occurrence, on the basis of analogy, or on the basis of instruction.</p><p>ii. The phenomenon should work differently in the L1 and the L2. That is, it</p><p>must be underdetermined by the L1 grammar as well. In this way, transfer</p><p>of surface properties can be ruled out as an explanation of any knowledge</p><p>that L2 learners attain.</p><p>2.1.1 The Overt Pronoun Constraint in L2</p><p>Let us reconsider the Overt Pronoun Constraint (Montalbetti 1984) (see</p><p>chapter 1, section 1.3.1) in the context of L2 acquisition. Recall that, in [+null</p><p>subject] languages, overt pronoun subjects of embedded clauses cannot receive</p><p>a bound variable interpretation, hence cannot take quantified expressions or wh-</p><p>phrases as antecedents, in contrast to null pronoun subjects. The sentences in</p><p>(1) to (2) (repeated from chapter 1) illustrate the contrast in question. In the Span-</p><p>ish sentence in (1a), the overt pronoun él cannot have the quantifier nadie as its</p><p>antecedent, whereas this interpretation is possible with the null pronoun in (1b).</p><p>Similar facts obtain for the Japanese sentences in (2).</p><p>(1) a. *Nadiei cree [que éli es inteligente]</p><p>Nobodyi believes that hei is intelligent</p><p>b. Nadiei cree [que proi es inteligente]</p><p>Nobodyi believes that (hei) is intelligent</p><p>(2) a. *Darei ga [karei ga kuruma o katta to] itta no?</p><p>Whoi nom hei nom car acc bought that said Q</p><p>‘Who said that he bought a car?’</p><p>b. Darei ga [proi kuruma</p>
- Sabendo disso, imagine hipoteticamente que você acaba de assumir a operação de um restaurante self-service e sua primeira tarefa é detalhar o modelo de transformação deste restaurante, ou seja, detalh
- O fragmento relata a apresentação geral de um restaurante self-service, em nossa disciplina de Administração da Produção vimos a importância de compreender o modelo de transformação Saben
- A refeição fora de casa é uma vertente dentro do que se pode chamar de terceirização dos serviços familiares, acompanhadas pelo surgimento de outros serviços, como a venda de comida congelada, en
- b) Descreva o funcionamento da coleta de esgoto sanitário
- Considerando o que nosso livro aborda acerca da coleta e do tratamento de esgoto, descreva de forma simples, observando os textos apresentados e respondendo em consonância com o conteúdo exposto por a
- Como o próprio nome já diz, o esgoto doméstico é aquele que sai das nossas casas, o industrial é oriundo das indústrias, e o pluvial é aquele que provém das águas da chuva que percorrem as cidades
- Neste exercício, o estudante é desafiado a refletir, pensar, visualizar com atenção, ler e compreender para buscar e organizar seus conhecimentos, sintetizando a identificação dos sistema de tratament
- O estudante desenvolve novas competências ao conhecer o processo utilizado para coleta e tratamento do esgoto sanitário e o impacto na área urbana, esse conhecimento gera melhor enquadramento
- Esta etapa detecta e compreende o uso da lógica cognitiva do pensamento pela busca da compreensão histórica em que a humanidade se depara com um estilo de vida social pautado inici
- Apresentação de reflexão sobre o Planejamento Urbano e Meio Ambiente, haverá questionamentos quanto ao processo de coleta e tratamento de esgoto nos municípios, bem como sua importância para o sistema
- ATIVIDADE 1 - GIMOB - PLANEJAMENTO URBANO E MEIO AMBIENTE - 54_2024
- Com base no resumo apresentado, avalie as afirmacoes a seguir: I- O autor analisa um contexto educacional específico com uso da lousa interativa co...
- Com base no que foi estudado acerca do método, assinale a alternativa CORRETA.a) ( ) Este método era muito usado no exército, pois os soldados d...
- Agora, assinale a alternativa CORRETA:a) ( ) As sentenças I, III e IV estão corretas.b) ( ) Todas as sentenças estão corretas.c) (x) A senten...
- Acerca do Audiolingual Method, assinale a alternativa CORRETA:a) (x) O Audiolingual Method parte dos pressupostos behavioristas, portanto, o estí...
- Agora, assinale a alternativa CORRETA:a) ( ) As sentenças I, II e III estão corretas.b) ( ) As sentenças I, III e IV estão corretas.c) (x) As...
- Agora, assinale a alternativa que apresenta a resposta CORRETA:a) ( ) V – F – F – F.b) (x) F – V – V – F.c) ( ) V – F – F – V.d) ( ) F – F ...
- Read the text and choose the correct answer: What is the main idea of the passage?a) The dingo caused changes in Australia’s balance of nature.b...
- Assinale a alternativa que não contempla uma etapa da sequência didática.a) ( ) Definição do tema.b) ( ) Sondagem inicial.c) ( ) Cálculo ...
- Assinale a alternativa que não apresenta três importantes recursos para uma aula de língua inglesa eficaz.a) ( ) Textos de diversos gêneros, liv...
- Com base nos estudos que fizemos acerca das sequências didáticas, assinale a alternativa INCORRETA:a) ( ) Na sequência didática, o planejamento ...
- Acerca de language learning e language acquisition, marque T para a(s) sentença(s) verdadeira(s) e F para a(s) falsa(s):( ) A language learning ...
- Qual é a característica apresentada por um software típico de “instrução programada” que o torna incompatível com a proposta dos PCN?a) (x) Uma s...
- Dentre os trechos transcritos dos Parâmetros Curriculares Nacionais Língua Estrangeira Ensino Fundamental, aquele que relaciona diretamente a apren...
- interpretação e produção de texto
- Como fazer uma Redação
