THE ACQUISITION OF ENGLISH PROGRESSIVE ASPECT BY CHINESESPEAKING L2 LEARNERS: TOWARDS AN OT ACCOUNT 以汉语为母语的英语学习者习得英语进行体标记的优选论解释 A dissertation submitted by Wu Heping to Guangdong University of Foreign Studies in partial fulfillment of the requirement for the degree of DOCTOR OF PHILOSOPHY in Linguistics

Centre for Lingusitics and Applied Linguistics Guangdong University of Foreign Studies March, 2006

Declaration

I hereby declare that this thesis does not contain any material which has been accepted for the award of any othe higher degree or graduate diploma in any tertiary institution and that, to the best of my knowledge and belief, this thesis does not contain any material previously published or written by another person, except when due reference is made in the thesis.

Signed

Dated: March 16, 2006

i

Acknowledgements To write a dissertation is to embark on a solitary odyssey, yet its completion is obviously not possible without the personal and practical support of numerous people. There is one person above all others, Prof. Wu Xudong, my adivisor, who deserves my deepest gratitude and respect for his continued support, encouragement and commitment. It is inconceivable that the project could be accomplished without his gracious encouragement, rigiorious scholarship and skillful directions. He was always readily available for me, as he so generously is for all of his students. His oral and written comments are always extremely perceptive, helpful, and appropriate. I deeply appreciate how much time he devoted to providing skillful comments and helpful advice on each chapter and in every phase of my dissertation. I would like to extend my special thanks to Prof. Wang Chuming, for his scientific expertise in second language acquisition (SLA) research. His lectures and talks at SLA course and seminars and comments on the proposal of this dissertation have been a primary source of enlightenment in broadening my vision when conducting this project. I am very grateful to all the people who provided feedback and support at various phases of the dissertation: Prof. Dong Yanping, Prof. Wu Yi’an, Prof. Zheng Chao and Dr. Feng Wei. Their critical comments on the proposal or earlier versions of the project were enormously helpful for drafting and revising of this dissertation. Special thanks should be extended to all the reviewers of this dissertation, Prof. Ma Qiuwu, Prof. He Guangken, Dr. Hua Dongfan, Prof. Wang Chumin and Prof. Dong Yanping, all of whom offered constructive suggestions on revising this dissertation. My studies for Ph.D. degree would not have been the same without the social friendship and academic challenges provided by all my student-colleagues. I am particularly thankful to my friends Lu Renshun, Gao Yusong, Zhong Shunen, Xue Changmin, Ma Shuhong, Zhu Ye, He Xuelin, Wang Renqiang, Wang Min and Zuo Hongshan. We not only studied, relaxed, and had our weekly ascent of Mt.Bai ii

Yunshan together, but they were even willing to share their expertise with me and challenge the theorizing and experimentation of this dissertation and thus provided some very useful input. My communications with some scholars overseas like Lisa Davidson, Joan Chen-Main and Joseph Magliano contributed immensely to shaping my current understanding of Optimality Theory and other methodological issues. Joan ChenMain and Joseph Magliano also provided me with some reference materials that were otherwise unaccessible in mainland China. I wish to thank them for their generous help. There are several priorities in the life of a man in his middle age like me. But I have nice collegues and a loving family that share my professional and family responsibilities, and I could manage to put dissertation writing on top agenda for the better part of the past four years, owing to their support and understanding. I would like to extend my thanks to my collegues and friends Prof. Jiang Qiuxia, Professor Li Fayuan and Professor Cao Jin for shouldering my office duties. I owe a great deal of thanks to my wife who has been giving me unrelenting support throughout the years when I pursue my academic goal in life. Without her understanding and love, this project could not have been completed. Lastly, I would extend my special thanks to my parents for their pride and confidence in me.

iii

Abstract

This dissertation sets out to investigate, within the framework of Optimality Theory (OT, Prince & Smolensky, 1993, 2004), the problem of how Chinesespeaking learners of L2-English acquire the English progressive aspect, which poses a typical learning problem of mapping conflicting semantic, morphosyntactic and discourse constraints on an aspectual form. The overarching goal of this study is to provide a unified account of the development of L2 tense-aspect system in general, and the English progressive aspect in particular, by analyzing how these constraints interact and what impact their interactions have on L2 acquisition of tense-aspect morphology. Three OT-compatible constraints are identified, i.e., NOCOER, IDENT(TF) and MARKT, which pertain respectively to the semantic, discourse and morphosyntactic constraints on tense-aspect marking. These three constraints are ranked as IDENT(TF) >>MARKT >>NOCOER, with the discourse requirement for the temporal forms at the top of the scale in English norm. However, the tense-aspect system of L2 interlanguage, according to [M>>F] initial ranking hypothesis in OT, is assumed to be dominated by temporal features inherent in the verbal predicate when L2 learners make aspectual choices. It is further assumed that, with the growth of L2 proficiency, this semantic constraint is gradually demoted, triggered by L2 learners’ recognition of the dominant role played by discourse grounding function of the aspectual morphology. To assess the plausibility of these assumptions generated within OT framework, five specific hypotheses were investigated through three experimental tasks, i.e., an Acceptability Judgement Task (AJT), a Movie Clip Retell Task (MCRT) and an Aspect Sentence Interpretation Task (ASIT). Ninety Chinese EFL learners who

iv

represented three English proficiency levels, together with 16 English-speaking native controls, participated in the investigation. The results point to the effect of constraint reranking in L2 tense-asepct development. It is indicated that when learning the English tense-aspect system, L2 learners were initially influenced by the temporal features inherent in the verbs, showing a stronger inclination for natural combinations between aspect forms and situation types, like the progressive-activity and perfective-achievement mappings. It was not until the advanced stage of L2 development that their tense-aspect system was primarily dominated by discourse grounding status of the verbal predicates. In addition, an inhibitory effect was observed where L1 and L2 differ in aspect-situation compatibitlity, as in the cases of progressive marking on stative and achievement verbs. These findings empirically verify the validity of OT as an explanans for L2 acquisition of tense-aspect system, particularly the acquistioin of the progressive aspect. The contribution of this project to the study on L2 tense-aspect acquisition is that it provides a unified optimality-theoretic explanation for both early emergence and later development of both progressive and perfective aspect, and that it integrates the dual mapping process of both form-meaning and form-function mappings into one single theoretical framework, which are otherwise addressed in different and encapsulated modules, as is notably manifest in the studies testing Aspect Hypothesis and Discourse Hypothesis.

Key Words: tense-aspect system, second language acquisition, progressive aspect, Optimality Theory, constraint reranking

v

中文摘要

本研究在优选论理论框架下探讨以汉语为母语的英语学习者如何习得英语 进行体。 本文认为，进行体的二语习得是学习者将相互冲突的语义、句法形态 和语篇限制映射到同一种体态形式的过程。本研究的目的就在于通过分析这些 限制条件的相互作用以及对第二语言时体形态，特别是进行体形态习得的影 响，以期能为第二语言时体系统的发展提供一个统一的解释框架。 本研究确定了三个优选论限制条件：NOCOER, IDENT(TF)以及 MARKT，分 别对应时体标记的语义，语篇和形态句法限制。这三种限制条件在英语中的排 列次序为 Ident(TF)>>MarkT>>NoCoer, 其中语篇对时体形态的要求处于限制等 级最高位置。而根据优选论[M>>F]假说，我们假设中介语时体系统中最高的限 制条件则为动词内在时间语义特征。同时，我们假定，随着学习者二语水平的 提高，学习者逐渐认识到体态形式在标记语篇前/背景中的角色，语义限制逐渐 降低。 为了在优选论框架下检验这些假说的合理性，本研究提出了五个具体的研 究假设，并设计了三个研究工具来验证这些假设。这三个研究工具分别是可接 受性判断任务，电影片断重述任务和体态句解释任务。共有 90 名分属不同水平 组的英语学习者以及 16 个英语本族语人士作为对照组参加了本项研究。 本研究结果证实了第二语言时体习得过程中限制条件的重新排列。研究结 果表明，在学习英语时体系统时，二语学习者首先受动词内在时间特征的影 响，表现出较强的体态形式与情状类型之间的自然匹配，如进行体与活动体动 词、完整体(perfective)与达成体动词的匹配。到二语发展的高级阶段，其时体 系统主要受语篇前/背景地位的支配。另外，我们同时观察到，当一语和二语的 情状体－语法体匹配不一致时，如英汉语进行体标记状态动词和达成动词，时 体形态的发展会受到阻碍。这些研究发现从实证的角度证明了优选论在解释二 语时体习得，特别是进行体习得的有效性。 本研究对二语时体习得的贡献在于它提供了一个统一的框架，来解释进行 体和完整体的早期出现和后期发展，并把形式－意义匹配和形式－功能匹配置

vi

于同一个理论框架中考察，克服了前人研究把二者看成互不联系的模块（如体 态假说和语篇假说）的缺陷。

关键词：时体系统， 二语习得，进行体，优选论，限制条件重新排列

vii

CONTENTS Declaration...............................................................................................................................................i

Acknowledgements.................................................................................................................................ii

Abstract..................................................................................................................................................iv

中文摘要.................................................................................................................................................vi

Table of Figures...................................................................................................................................xiv

List of Tables........................................................................................................................................xvi

List of Abbreviations............................................................................................................................ xx Chapter 1 Introduction ...................................................................................................................... 1 1.1 Research Orientation................................................................................................................. 1 1.2 English Progressive Aspect: The Target of Research ............................................................... 1 1.3 Rationale ................................................................................................................................... 5 1.3.1 Decomposing the Learning Task of the English Progressive Aspect................................ 5 1.3.2 Previous Studies and Their Inadequacies.......................................................................... 7 1.4 Key Research Questions ........................................................................................................... 9 1.5 Structure of the Dissertation ................................................................................................... 10 Chapter 2 In Search of a Unified Theoretical Account for L2 Progressive Aspect Acquisition ...... 12 2.1 Previous Studies on L2 Aspect Acquisition: Empirical Findings and Theoretical Explanations ...................................................................................................................................................... 13 2.1.1 Form-meaning Mapping of L2 Tense-aspect Acquisition: Aspect Hypothesis............... 13 2.1.2 Form-function Mapping of L2 Tense-aspect Marking: Discourse Hypothesis ............... 17

viii

2.1.3 Narrative Structure and Lexical Aaspect: Conspiring Factors? ...................................... 18 2.2 Theoretical Accounts of Form-meaning Mapping in Tense-aspect Acquisition..................... 19 2.2.1 Language Bioprogram Hypothesis and Basic Child Grammar ....................................... 19 2.2.2 Prototype Account and Distributional Bias Argument.................................................... 19 2.2.3 Connectionist Account .................................................................................................... 20 2.3 Problems Existing in the Tense-aspect Acquisition Studies ................................................... 21 2.4 Optimality Theory as a Unified Framework ........................................................................... 22 2.5 Summary................................................................................................................................ 24 Chapter 3 Constraints on the Progressive Aspect Marking in English and Chinese......................... 25 3.1 Aspect and Other Related Notions.......................................................................................... 26 3.1.1 Aspect ............................................................................................................................. 27 3.1.2 Grammatical Aspect........................................................................................................ 27 3.1.3 Lexical Aspect................................................................................................................. 29 3.1.4 Technical Issues Related with Lexical Aspect ................................................................ 30 3.2. Interaction of Grammatical and Lexical Aspects................................................................... 45 3.3 Semantic Constraints on Progressive Marking: The Form-meaning Mapping ...................... 47 3.3.1 The Formal Approach ..................................................................................................... 48 3.3.2 The Functional Approach................................................................................................ 49 3.3.3 The PROG Operator........................................................................................................ 54 3.4 Discourse Constraints of Progressive Aspect Marking: The Form-Function Mapping .......... 57 3.5 Cross-linguistic Analysis of the Progressive Aspect in Chinese and English........................ 60 3.6 Summary................................................................................................................................. 63 Chapter 4 An OT Analysis of the Constraints on the Progressive and their Interaction.................. 65 4.1 Optimality Theory................................................................................................................... 65

ix

4.1.1 Basic Tenets of OT ......................................................................................................... 65 4.1.2 Conventional Notations................................................................................................... 66 4.1.3 Faithfulness and Markedness Constraints ....................................................................... 68 4.1.4 OT Account of Language Acquisition ............................................................................ 69 4.2 Conflicting Constraints on the Progressive Aspect Marking .................................................. 71 4.2.1 Transitivity Hypothesis (TH) .......................................................................................... 72 4.2.2 Discourse Representation Theory (DRT) and Aspect Coercion ..................................... 74 4.3 An OT Analysis of the Constraints on Aspect Marking and Their Ranking........................... 78 4.3.1. Typology of Constraints on Aspect Marking in English................................................ 78 4.3.2 Constraint Ranking of Aspect Marking in English ......................................................... 82 4.3.3 Testing the Ranking of Constraints on Aspect Marking ................................................. 84 4.4 Summary................................................................................................................................. 90 Chapter 5 L2 Acquisition of the Progressive Aspect: Towards an OT Account ................................ 92 5.1 An OT Account of L2 Tense-aspect Acquisition.................................................................... 92 5.1.1 Initial Ranking of Constraints on Tense-aspect marking ................................................ 93 5.1.2 Learning Algorithms in OT............................................................................................. 95 5.2 Hypotheses on L2 Acquisition of English Tense-aspect System by Chinese EFL Learners .. 98 5.2.1 Hypothesis on Aspect Coercion ...................................................................................... 98 5.2.2 Hypotheses on Constraint Reranking.............................................................................. 98 5.2.3 Hypotheses on Cross-linguistic Constraints.................................................................... 99 5.3 Variables Involved in Hypothesis Testing ............................................................................ 100 5.3.1 L2-English Proficiency Level ....................................................................................... 101 5.3.2 NOCOER Constraint ....................................................................................................... 101 5.3.4 Tense-aspect Marking ................................................................................................... 103

x

5.4 Summary............................................................................................................................... 103 Chapter 6

Research Design, Instrumentation, Data Collection and Data Analysis..................... 105

6.1 Research Design ................................................................................................................... 105 6.2 Participants ........................................................................................................................... 106 6.3 Instrumentation ..................................................................................................................... 108 6.3.1 Acceptability Judgment Test (AJT) .............................................................................. 109 6.3.2 Aspect Sentence Interpretation Task (ASIT) ................................................................ 112 6.3.3 Movie Clip Retell Task (MCRT) .................................................................................. 119 6.3.4 Summary of Test Instruments ....................................................................................... 123 6.4 Data Collection ..................................................................................................................... 124 6.4.1 Time and Venue ............................................................................................................ 124 6.4.2 Task Sequence............................................................................................................... 124 6.5 Data Scoring and Coding ...................................................................................................... 125 6.5.1 Scoring of ASIT Data.................................................................................................... 125 6.5.2 Scoring of AJT Data...................................................................................................... 127 6.5.3 Coding and Scoring MCRT Data .................................................................................. 127 6.6 Data Analysis........................................................................................................................ 128 6.7 Summary............................................................................................................................... 129 Chapter 7 Results for Research Hypotheses Testing ..................................................................... 130 7.1 Results for Hypothesis 1 ....................................................................................................... 130 7.1.1 Testing Hypotheses 1 through Movie Clip Retell Test (MCRT) Data.......................... 131 7.1.2 Testing Hypothesis 1 through Aspect Sentence Interpretation Task (ASIT) ................ 137 7.1.3 Testing Hypothesis 1 through Acceptability Judgment Task (AJT) ............................. 140 7.2 Testing Hypothesis 2 ............................................................................................................ 143

xi

7.2.1 Testing Hypothesis 2 through Movie Clip Retell Task (MCRT) .................................. 143 7.2.2 Testing Hypothesis 2 through Aspect Sentence Interpretation Task (ASIT) ................ 149 7.2.3 Testing Hypothesis 2 through Acceptability Judgment Task (AJT) ............................. 152 7.3 Testing Hypothesis 3 ............................................................................................................ 154 7.3.1 Testing Hypothesis 3 through Movie Clip Retell Task (MCRT) ................................. 155 7.3.2

Testing Hypothesis 3 through Aspect Sentence Interpretation Task (ASIT) ........... 157

7.4 Testing Hypothesis 4 ............................................................................................................ 164 7.4.1 Testing Hypothesis 4 through Acceptability Judgment Task (AJT) ............................. 165 7.4.2

Testing Hypothesis 4 through Movie Clip Retell Task (MCRT)............................. 169

7.5 Testing Hypothesis 5 ............................................................................................................ 171 7.5.1 Testing Hypothesis 5 through Acceptability Judgment Task (AJT) ............................. 172 7.5.2 Testing Hypothesis 5 through Movie Clip Retell Task (MCRT) .................................. 173 7.5.3 Testing Hypothesis 5 through Aspect Sentence Interpretation Task (ASIT) ............... 175 7.6 Summary............................................................................................................................... 177 Chapter 8 General Discussion......................................................................................................... 179 8.1 Developmental Route of English Tense-aspect Acquisition by Chinese EFL Learners ....... 180 8.1.1 A Summary of Major Findings ..................................................................................... 180 8.1.2 Characterizing L2 Acquisition of the English Aspect Markers by Chinese EFL Learners ............................................................................................................................................... 182 8.2 Implications of OT for the Developmental Route of L2 Tense-Aspect Forms ..................... 185 8.2.1 An OT Account of the Initial State ............................................................................... 185 8.2.2 The Triggering Factor in Tense-aspect Acquisition...................................................... 188 8.2.3 The Spread from Prototypical to Non-prototypical Situation-Aspect Mappings .......... 190 8.2.4 The Role of Statives in Progressive Aspect Marking.................................................... 192

xii

8.3 The Explanatory Power of OT for L2 Aspect Acquisition ................................................... 194 8.3.1 Temporal Semantics and Discourse Grounding, Two Conspiring Forces?................... 194 8.3.2 Nativism or Connectionism........................................................................................... 196 8.4 Summary............................................................................................................................... 198 Chapter 9 Conclusions, Implications, Limitations and Suggestions for Future Research .............. 201 9.1 Conclusions........................................................................................................................... 201 9.2 Implications .......................................................................................................................... 202 9.2.1 Implications for SLA Theory ........................................................................................ 202 9.2.2 Implications for L2 Tense-aspect Instruction................................................................ 204 9.3 Limitations............................................................................................................................ 205 9.4 Suggestions for Further Research ......................................................................................... 206 References ....................................................................................................................................... 208 Appendix I Progressive Aspect Acceptability Judgment Task (AJT) .............................................. 217 Appendix II Aspect Sentence Interpretation Task (ASIT) ................................................................ 220 Appendix IV Descriptive Statistic Data for the Figures ................................................................. 233

xiii

Table of Figures

Figure 1 A Preliminary Taxonomy of Aspect in Languages ................................................................. 30 Figure 2 Flowchart of Situation Type Identification Procedure ............................................................ 39 Figure 3 Reference of the Progressive Punctuals .................................................................................. 53 Figure 4 The functions of the [PROG] operator..................................................................................... 55 Figure 5 Semantic Referents of Progressive in Chinese and English.................................................... 62 Figure 6 Mapping of input to output in OT grammar........................................................................... 66 Figure 7 Conflicting Constraints on Aspect Marking in English ........................................................... 78 Figure 8 Categorical and Overlapping Ranking of Constraints along a Continuous Scale .................... 96 Figure 9 Schematic representation of the acceptability scale of AJT.................................................. 111 Figure 10 Experimental Set-up and Variables in ASIT...................................................................... 117 Figure 11 Means of Progressive and Perfective Marking for Five Situation Types in MCRT........... 132 Figure 12 Probability of Interpreting Aspect Sentence as Denoting Completed Events by all Participant groups in ASIT ........................................................................................................................... 138 Figure 13 Mean Scores for Acceptability of Progressive Marking in Five Situation Types by L2 Participants and the Native Controls........................................................................................... 141 Figure 14 Mean Scores for Progressive Aspect Marking by All Participant Groups on Each Situation Type in MCRT............................................................................................................................ 145 Figure 15 Mean Scores for Perfective Aspect Marking by All Participant Groups on Each Situation Type in MCRT............................................................................................................................ 146 Figure 16 Interpretation of Critical Aspect Sentences by Four Participant Groups ............................ 150 Figure 17 Acceptability of Aspect Marking on Five Subclasses of Stative Verbs by Four Participant Groups ........................................................................................................................................ 167

xiv

Figure 18 General Pattern of Constraints Re-ranking in L2 Tense-aspect Acquisition...................... 183

xv

List of Tables

Table 1 Semantic Features of Aspectual Classes ................................................................................... 34 Table 2 Diagnostic Tests for Different Temporal Features of Situation ............................................... 45 Table 3 Interaction of Lexical Aspect with Grammatical Aspect........................................................... 47 Table 4 Aspect Marking Scenarios under Different Semantic-Discourse Conditions........................... 85 Table 5 Violations of NoCoer Constraint on Different Situation Types .............................................. 102 Table 6 Grouping of Participants Based on Years of L2 English Learning and Scores on English Proficiency Test. ......................................................................................................................... 108 Table 7 Taxonomy and Token Numbers of the Verbs in AJT............................................................. 112 Table 8 Tokens of Telic Verbs in ASIT. .............................................................................................. 115 Table 9 Balancing Aspect, Test Statement and Filler Statement of One Passage across Arrangements of ASIT ........................................................................................................................................... 118 Table 10 A Summary of Instruments Developed for the Study ........................................................... 123 Table 11 Coding of the tense-aspect forms in MCRT.......................................................................... 128 Table 12 Two-way ANOVAs (Repeated Measures) of Results for MCRT ......................................... 134 Table 13 Pairwise Comparisons of the Mean Differences Between Different Situation Types Marked with the Progressive Aspect (LSD)............................................................................................. 134 Table 14 Pairwise Comparisons of the Mean Differences Between Different Situation Types Marked with the Perfective Aspect (LSD) ............................................................................................... 135 Table 15 Frequency Order of Perfective and Progressive Aspect Marking on Verbs of Five Situation Types by L2 Groups in MCRT ................................................................................................... 136 Table 16 Effect of Verb Situation Type on Aspect Interpretation in ASIT.......................................... 139 Table 17 Effect of Verb Situation Type on Acceptability Judgment of Progressive Marking: Results for One-Way ANOVA (Repeated Measures)................................................................................... 142

xvi

Table 18 Pairwise Comparisons of Means for Five Situation Types by Each L2 Group in AJT ......... 143 Table 19 Effects of Proficiency Level and Situation Type of Verb on Progressive and Perfective Marking in MCRT: Results for Two-way ANOVA (Mixed Design) ......................................... 146 Table 20 Pairwise Comparisons of the Progressive and Perfective Marking on Verbs of Five Situation Types in MCRT (LSD) ............................................................................................................... 147 Table 21 Interaction between Proficiency Level and Situation-Aspect Mapping as Shown in ASIT: Results for Two-way ANOVA (Mixed Design) ......................................................................... 150 Table 22 Pairwise Comparisons of the Aspect Sentence Interpretations by Participants of Lower Proficiency Group and the Other Three Groups ......................................................................... 151 Table 23 Means and Standard Deviations of AJT by L2 Participants and Native Controls ................. 152 Table 24 Effects of Proficiency Level and Situation Type on Acceptability of Sentences Marked with Progressive: Results for Two-way ANOVA (Mixed Design) .................................................... 153 Table 25 Pairwise Comparisons of Mean Scores between Low-Level Group and Three Participant Groups for AJT ........................................................................................................................... 154 Table 26 Aspect Marking for Two Discourse Grounding Status by 4 Participant Group in MCRT.... 155 Table 27 Effects of Proficiency Level on Aspect Marking of Discourse Grounding in MCRT: Results for One-way ANOVAs ............................................................................................................... 157 Table 28 Mean Scores and Standard Deviations for Aspect Sentence Interpretation by Four Participant Groups (ASIT ) ........................................................................................................................... 159 Table 30 Pairwise Comparison of the Mean Scores for Aspect Sentence Interpretation by Four Participant Groups (ASIT).......................................................................................................... 160 Table 31 Mean scores and Standard Deviations for Aspect Test Statement in Within- and After-text Position in ASIT ......................................................................................................................... 161

xvii

Table 32 Three-way ANOVA for Test Statement Interpretation Related to Aspect, Position and Proficiency Level (ASIT) ........................................................................................................... 162 Table 33 Pairwise Comparisons of Mean Scores for Aspect Sentence Interpretation by Four Participant Groups (ASIT) ............................................................................................................................ 163 Table 34 Mean Scores and Standard Deviations for Acceptability of the Progressive Marking on the Stative Verbs in AJT................................................................................................................... 165 Table 35 ANOVA for Acceptability Judgment of Five Subtypes of Stative Verb by Four Participant Groups ........................................................................................................................................ 167 Table 36 Pairwise Comparisons of Mean Scores for Five Subclasses of Stative Verbs in AJT........... 168 Table 37 Mean Scores and Standard Deviations for Progressive Aspect Marking on the Stative Verbs by Four Participant Groups in MCRT ........................................................................................ 170 Table 38 Pairwise Comparisons of the Mean Scores for Progressive Aspect Marking on the Stative Verbs by Four Participant Groups in MCRT .............................................................................. 170 Table 39 Mean Scores and Standard Deviations for Acceptability of the Progressive Marking on Achievement Verbs in AJT by Four Participant Groups ............................................................ 172 Table 40 Pairwise Comparisons of Mean Scores for Acceptability of Progressive Aspect Marking on Achievement Verbs in AJT......................................................................................................... 173 Table 41 Mean Scores and Standard Deviations for Progressive Aspect Marking on Achievement Verbs in MCRT by Four Participant Groups .............................................................................. 174 Table 42 Pairwise Comparisons of Mean Scores for Progressive Aspect Marking on Achievement Verbs in MCRT .......................................................................................................................... 175 Table 43 Mean Scores and Standard Deviations for Interpretation of Sentences Containing Achievement Predicates Marked with Progressive Aspect......................................................... 176

xviii

Table 44 Pairwise Comparisons of Mean Scores for Interpretation of Sentences Containing Achievement Predicates Marked with Progressive Aspect......................................................... 176

xix

List of Abbreviations ACC

Accomplishment Verbs

ACH

Achievement Verbs

ACT

Activity Verbs

AJT

Accetability Judgement Task

ASIT

Aspect Sentence Interpretation Task

BGR

Background

CD

Constraint Demotion

CET-4

College English Test Band 4

CHILDES

Child Language Data Exchange System

CL

Classifier

CLZ

Cloze

DRS

Discourse Representation Structure

DRT

Discourse Representation Theory

ESF

European Science Foundation

ETV

Externally-induced Telic Verbs

FGR

Foreground

GLA

Gradual Learning Algorithm

IMPF

Imperfective

ITV

Internally-induced Telic Verbs

L1

First Language

L2

Second Language

LSD

Least Significant D

MCRt

Movie Clip Retell Task

OT

Optimality Theory

PERF

Perfective

PROG

Progressive

PTV

Punctual Telic Verbs

SLA

Second Language Acquisition

SMF

Semelfactive Verbs

STA

Stative Verbs

TH

Transitivity Hypothesis

TVJT

Truth Value Judgement Task

VS

Vocabulary and Structure

xx

Chapter 1 Introduction

Chapter 1 Introduction

1.1 Research Orientation The overarching goal of this study is to provide a unified account of L2 acquisition of English progressive aspect by Chinese-speaking learners.

More

specifically, it attempts to (1) identify the linguistic and cross-linguistic constraints on progressive aspect morphology and then explore how these constraints interact with each other to yield the optimal aspect marking in general, and the progressive marking in English in particular, (2) examine the effect that these constraints and their interactions have on L2 learners’ acquisition of the progressive aspect, and more importantly, (3) explain, within a unified framework, the observed patterns of these constraints on L2 tense aspect acquisition in general, and on L2 progressive acquisition in particular. 1.2 English Progressive Aspect: The Target of Research The English language, unlike many other languages, has a fully grammaticalized progressive aspect. 1 Formally represented by a form of the verb to be followed by a verb inflectionally marked with an –ing suffix, the progressive aspect in English usually indicates a happening in progress at a reference time. However, it should be noted from the outset that the progressive construction “be+V-ing” does not always yield an on-goingness reading, nor does the meaning of on-goingness automatically require the use of the progressive construction.

In other words, the English

1 The overwhelming majority of progressive constructions in the world’s languages are periphrastic, having originated from various types of phraseological constructions devoid of a morphological form, often with the meaning of ‘be at doing something.’ (Binnick, 1991; Comrie, 1976; Dahl, 1999).

1

Chapter 1 Introduction

progressive aspect does not manifest a one-to-one form-meaning mapping. Thus, how an L2 learner learns to correctly establish the mappings between the forms and meanings of the progressive has become an issue of central interest in the studies of L2 acquisition of tense and aspect. As the primary concern of this study is the development of L2 English progressive morphology, the approach adopted is a form-oriented one 2 (BardoviHarlig, 1999, 2000). That is, we will investigate how the broad range of meanings and discourse functions associated with “on-goingness” are mapped onto this single aspectual form in L2-English learners, leaving aside for the moment the other periphrastic forms with similar semantic content. 3 The English progressive aspect has a plethora of meanings that typically resist an integrated linguistic account. Scheffer (1975: 21 - 42), for example, listed a dozen of its “basic meanings” in his monograph on the progressive aspect in English, ranging from “limited duration” to “enhancing the graphic or plastic effect”. Similarly, Comrie (1976: 33-38) also noted that the English progressive aspect “has an unusually broad range [of uses] and several idiosyncrasies …to militate against a general meaning being able to account for every single use of this form”. Quirk et al. decomposed the basic semantics of the progressive aspect into a number of dimensions, e.g., a happening described by the progressive aspect has a limited duration; the happening is not necessarily complete; and the happening has a relationship or identity with another simultaneous event, etc. (Quirk, Greenbaum,

2

An alternative school of L2 temporality studies, most of which was sponsored by the European Science Foundation (ESF) under the guidance of C. Perdue and W. Klein in late 1980s, took a semantic-oriented approach to L2 aspect acquisition (Dietrich, Klein, & Noyau, 1995). Unlike the form-oriented approach, the primary concern of this approach was how the full range of temporal reference system in L2 learners develops over time, including the pragmatic, lexical as well as morphological devices expressing temporality. 3 Etymologically, the English progressive form “to be V-ing” evolved from locative or spatial expressions roughly synonymous to to be at doing or to be on doing, which was later replaced by to be a-doing (of) (Bybee, Perkins, & Pagliuca, 1994; Comrie, 1976; Scheffer, 1975). In the present-day English, we still have some residual forms of these earlier versions of the progressive, such as to be under way, to be at work or to be asleep, all of which might as well be translated into the modern progressive morphology. These periphrastic forms, however, do not fall within the scope of the present study. 2

Chapter 1 Introduction

Leech, & Svartvik, 1985). In this sense, the acquisition of the English progressive aspect is a typical case of one-to-many mapping process, the working mechanism of which has not yet been fully explored in L2 acquisition research. For instance, philosophers, logicians and semanticists have long been intrigued by a property unique to the progressive aspect, i.e., the imperfective paradox (Asher, 1992; Dowty, 1977; Engelberg, 2001; Landman, 1992; Lascarides, 1988; Parson, 1990; Vlach, 1981). The sentences in (1) and (2) are otherwise identical except for their different entailment inferences. (1) a. John was pushing a cart. b. John pushed a cart. (2) a. John was writing a letter. b. John wrote a letter. The imperfective paradox lies in the observation that the entailment relationship between (1a) and (1b) can be inferred, while a similar inference cannot be made between (2a) and (2b). That is to say, John was pushing a cart entails John pushed a cart, but John was writing a letter does not entail John wrote a letter, hence logically paradoxical. For semanticists, the truth condition of (2a) is hard to warrant without recourse to a corresponding situation “in a possible world” (e.g., Dowty, 1977). Apart from the aforementioned “imperfective paradox” phenomenon, the progressive is recognized among linguists to be the most restricted with respect to its domain of application (Comrie, 1976; Lyons, 1977; Smith, 1997; Vendler, 1967). In particular, the verbs denoting a state or a punctual event are typically at odds with the semantics of the progressive aspect. For example,

(3) The boy was reaching the top of the tower. (4) I am understanding more about quantum mechanics.

3

Chapter 1 Introduction

Sentences (3) and (4) defy our immediate intuition of the well-formedness of the English progressive, owing to the semantic gap between the meanings of the verbs and the aspectual interpretation. These sentences, nonetheless, are not ungrammatical as long as certain felicitous conditions are satisfied. The punctual verb reach in (3), for example, develops a futurate interpretation and the progressive morphology renders the stative verb understand in (4) a flavor of dynamic vividness (Comrie, 1976; de Swart & Verkuyl, 1999; Smith, 1983; Smith, 1997). Sentences from (1) to (4) suggest that verbs with different temporal properties exhibit varying degrees of mapping compatibility with the progressive aspect. Some verbs like push in (1) combine naturally with the progressive marker, whilst logical paradox or semantic oddity arises when the aspectual form is mapped on verbs with semantic property different from push-like verbs, such as write, reach and understand exemplified above in (2), (3) and (4). Furthermore, the choice of progressive aspect is also susceptible to factors beyond the temporal semantics of the verbs. Cross-linguistic studies have shown that ‘‘the fundamental notion of aspect is not a local-semantic one but is discoursepragmatic” (Hopper, 1982: 5), and that aspectual distinctions contribute significantly to the discourse organization (Givon, 1982; Hopper, 1979; Hopper & Thompson, 1980; Smith, 2003; Wallace, 1982). A careful analysis of the English progressive aspect based on this line of thought has suggested that its function is to serve as discourse backgrounding that elaborates on, but does not push, the movement of a succession of events constituting the skeletal storyline of the narrative discourse (Glasbey, 1998; Hopper & Thompson, 1980). The two sentences in (5) below, for example, show that the discourse imposes an additional dimension of constraints on the progressive verbs and their non-progressive counterparts. In a narrative with a sequence of sentences describing various events, the simple past usually yields effect of sequential ordering of events, whereas the progressive has a simultaneous reading. In (5a), for example, letter-writing was going on when the phone rang, but in (5b),

4

Chapter 1 Introduction

there was a sequential order of temporality between the two events: Mary began to write the letter after the phone rang. (5) a. The phone rang. Mary was writing a letter. b. The phone rang. Mary wrote a letter.

The two events in (5a) are in the same temporal frame and the second clause backgrounds the first one, but in (5b), the two clauses belong to different time frame and the two are successive on the time axis. This difference in aspectual interpretation in discourse can be formally represented as (6a) and (6b) respectively: (6) a. ∃e1[Ring (phone, t1) ∧ ∃e2 [Write (Mary, letter, t2)] ∧t2 ⊆ t1]] b. ∃e1[Ring (phone, t1) ∧ ∃e2 [Write (Mary, letter, t2)] ∧ t1 < t2]]

Note that in (6b), the simple past moves the reference time of the storyline forward and the second event must be ordered after the first one: first ringing, then writing, represented as t1 < t2. But in (6a), the progressive aspect adds the second event as a further description of the situation going on simultaneously with the event introduced by the first sentence: the two events must overlap, represented as t2 ⊆ t1. The English progressive morphology, therefore, is subject to both semantic and discoursal constraints. However, these constraints do not always stay in good harmony and tension might occur between them because discourse and verbal semantics have independent, sometimes conflicting, demands on the aspectual form of the verb. The correct choice of the progressive aspect is presumably the outcome of the interaction between these two forces in conflict. 1.3 Rationale 1.3.1 Decomposing the Learning Task of the English Progressive Aspect The task of acquiring the progressive aspect is not as easy as it may appear. Even in L1 acquisition of the progressive aspect, which was supposed to be uniquely

5

Chapter 1 Introduction

effortless and error-free (Mapstone & Harris, 1985), errors of overusing or underusing the progressive aspect can be observed. The following examples, obtained from CHILDES 4 , provide us with the evidence of the learning difficulty that children around two years old experience when marking verb with aspectual morphology.

(7) a. Mother: what was granddad doing with the sand? Child: make a sand castle……I wrecked it all down. (Rachel, 2; 11, Belfast) b. Mother: what are you doing? Child: I made the cake. (Gary, 1; 6, Wells) (8) a. Mother: what did you do? Child: I was doing painting? (Becky, 2; 9, Manchester) b . Mother: you know what Mummy’ll do if you do? Child: running. (Anne, 2; 0, Manchester)

Note that in (7) the mothers’ what-questions constitute the obligatory context for eliciting responses with progressive aspect, while in (8) non-progressive responses are expected. The children’s responses, however, are just the other way round. This mismatch between the caretaker’s elicitations and the child’s responses in aspectual morphology is largely caused by the conflict of the verb’s temporal properties and its aspectual interpretation. For example, the verb make in (7), just like the verb write in (2) above, is a verb showing that something is accomplished, whereas the mother’s questions require that the responses be involved in an ongoing action. By the same token, the verbs do painting and run in (8a) and (8b), behaving like the verb push in (1) above, indicate that an activity was engaged, but the mother’s question was

CHILDES, the Child Language Data Exchange System, is an international database organized for the study of first and second language acquisition (MacWhinney, 1995; MacWhinney & Snow, 1985). 4

6

Chapter 1 Introduction

intended to elicit an answer suggesting the completion of the event, rather than its progression. These examples suggest that verbs of different types have different effect on the mapping between temporal semantics and aspectual morphology for the English-speaking children around two years old. The task confronting Chinese learners of the English progressive aspect is even more challenging. First, they need to distinguish between types of verbs with different temporal semantics and figure out their compatibility with the progressive aspect, in the absence of overt type markers. Second, they should learn the discourse constraints on progressive aspect and how these constraints interact with semantic constraints to yield the correct choice of progressive markers. As the form-function and the form-meaning mappings of the progressive aspect may stand in competition with each other, Chinese EFL learners have to find out, in the course of acquiring progressive aspect, to what extent form, meaning and function compete and interact with one another. Third, the fully-developed tense-aspect system that has already existed in Chinese EFL learners’ mind interacts in intricate ways with the new system to be acquired. In particular, as a tenseless language with rich aspectual markers, Chinese has more than one imperfective markers, with varying degrees of grammaticalization, to express temporal viewpoints similar to the English progressives, such as zai-, zheng-, –zhe and –ni (Chao, 1968; Dai, 1997; Li & Thompson, 1981; Wang, 1943; Xiao, 2002). This abundance of aspectual expressions in L1 imposes additional learning difficulty on Chinese learners of L2-English progressive aspect (Yang & Huang, 2004). 1.3.2 Previous Studies and Their Inadequacies Tense and aspect system has long been identified as one of the sources of learning difficulty for L2 learners. Some claimed that the temporal morphology tends to fossilize universally in second language acquisition (Seliger, 1978). After he tested the knowledge of the Imparfait/Passé Composé distinction by French near-native

7

Chapter 1 Introduction

speakers, Coppieters concluded that second language learners cannot attain the nativelike competence in L2 tense-aspect acquisition (Coppieters, 1987). Seeking the mechanisms underlying the emergence and development of this difficult category for language learners has thus been motivating the tense-aspect research in SLA, and second language researchers have shown a growing interest in finding out and further explaining the regularities in L2 tense-aspect morphology. In L2 tense-aspect studies, both the form-meaning mapping and form-function mapping processes were intensely examined, respectively labeled as Aspect Hypothesis and Discourse Hypothesis. The studies testing these hypotheses investigated the emergence and distribution of tense-aspect morphology with respect to the temporal properties of the verbs and the discourse structure (e.g., Andersen, 1991; Andersen & Shirai, 1996; Bardovi-Harlig, 1992, 1995b, 2000; Bardovi-Harlig & Reynolds, 1995; Li & Shirai, 2000; Robison, 1990; Salaberry, 1999; Shirai & Andersen, 1995; Slabakova, 2001, among others). It has been found in these studies that L2 temporal inflections emerge in a biased pattern that typically deviates from the target norm. It was also found that past tense or perfective marking tend to associate first with punctual and telic predicates, whereas progressive marking is mapped exclusively on verbs denoting process and activities by L2 learners. These studies have amassed reasonable evidence on the developmental patterns of L2 tense-aspect system. Nonetheless, there still remain inadequacies and limitations in them. First, in the majority of the existing studies, the two hyptheses on form-meaning mapping and form-function mapping were tested separately, as if these two mapping processes belong to isolated modules. Second, although the existing theories offer us plausible accounts on the early emergence of aspect forms in L1 and L2, it remains unknown as to what factors underlie the later development and ultimate attainment of L2 tense-aspect acquisition. Third, the existing studies drew heavily on the observations of the development of perfective aspect (e.g., the English simple past), but the research into the L2 development of the progressive aspect is disproportionately lacking. 8

Chapter 1 Introduction

Given this fragmentary nature of the current research, the present study represents an attempt at establishing a unified framework that delineates how the English progressive aspect is possibly acquired by L2 Chinese-speaking learners. More specifically, the aim of this unified framework is to explain the mechanisms underlying (i) both progressive and perfective aspect marking by L2 learners, (ii) both initial stage and later development of L2 aspect marking, and (iii) both semantic and discourse constraints on L2 aspect development. Only in this way can more convincing explanations be provided for the observed learning outcomes in the process of L2-English acquisition of the progressive aspect by Chinese-speaking learners. 1.4 Key Research Questions As stated in § 1.1, the primary goal of the present study is to examine how the English progressive aspect is acquired by adult Chinese-speaking learners of English. To achieve this goal, investigations should be made first to empirically observe the L2 acquisition of the progressive aspect. Then in the case where patterns of acquisitional behaviour are observed, describe these patterns, and finally explain them in a principled way. However, a linguistic and cross-linguistics analysis of the constraints on aspect marking in general, and on the progressive aspect marking in particular, is crucial for ensuring that the variables identified and data elicited in the study are valid and reliable, and furthermore, that the observed acquisitional patterns can be adequately explained. In other words, the validity and reliability of the study of L2 progressive aspect acquisition depend crucially on the soundness of a linguistic and cross-linguistic analysis of the constraints on aspect marking. A good point of departure for this study, based on the logic so delineated, is to establish a unified theoretical framework wherein the L2 acquisition of English progressive by Chinese learners can be explained. The next step is to observe the learning results of the learners at various proficiency levels. Finally, we need to explain, in a principled way, the observed learning results within the established 9

Chapter 1 Introduction

theoretical framework. In so doing, the general research aim outlined in § 1.1 is expected to achieve. More particularly, the implementation of the research approach stated above entails providing answers to the four key research questions listed below: (1) How can the linguistic and cross-linguistic constraints with regard to Chinese and English and their interactions governing the optimal selection of progressive aspect be adequately described within a unified theoretical framework? (2) What are the implications of these constraints and their interactive patterns for the acquisition of L2 progressive aspect by Chinese-speaking learners of English? (3) How might the L2 acquisition of English progressive aspect be described? (4) To what extent can the observed behaviors of acquisition be explained within the unified framework? 1.5 Structure of the Dissertation The rest of the dissertation is organized as follows. Chapter 2 gives a further justification why a unified framework tops the agenda of L2 tense-aspect research, after a critical review of the relevant literature in the past three decades. Also discussed in this chapter is why Optimality Theory (OT henceforth) is a candidate theory meeting our requirements for such a unified theory. Chapter 3 prepares the ground for an optimality-theoretic account of L2 progressive aspect marking and its acquisition. This chapter first introduces and defines some relevant tense-aspect notions consistently used in this dissertation. It then proceeds to analyze the linguistic and cross-linguistic constraints on the progressive aspect marking and how these constraints interact for optimal selection in English and Chinese language. Chapter 4 provides a brief introduction to the major tenets of OT and discusses how the constraints at various linguistic levels on aspect marking can be unified in OT framework.

10

Chapter 1 Introduction

Chapter 5 examines the implications of the OT account of the linguistic and cross-linguistic constraints on the tense-aspect marking for L2 acquisition of the progressive aspect. Five hypotheses to be tested in this study are presented, which make predictions about the initial ranking and how it develops as the L2 learner’s proficiency grows. Chapter 6 describes the research design, instruments, and the procedures that are adopted to test the research hypotheses. In addition, the ways in which data are collected, prepared and analyzed are presented. Chapter 7 reports the results of the empirical study with respect to the research hypotheses. In Chapter 8, the findings reported in Chapter 6 are discussed in the light of OT account of L2 tense-aspect acquisition discussed in Chapter 4. Chapter 9 presents the major conclusions of the present study, its pedagogical implications for TESOL, and suggestions for further research.

11

Chapter 2 In Search of a Unified Theoretical Account for L2 Progressive Aspect Acquisition

As has been explicated in the previous chapter, the temporal properties inherent in the verbs and the discourse functions have different, and sometimes competing demands on the progressive morphology of the verb. It has also been shown that these two mapping processes, i.e., the form-function mapping and the form-meaning mapping, impose learning difficulties on both L1 and L2 learners alike. A sound theory on tense-aspect acquisition should be able to give a consistent and unified explanation for these different demands on the progressive aspect and the dual mapping processes involved in its acquisition. The theoretical framework we are in search of should be a unified one in the sense that it should meet several important requirements. First, since tense-aspect acquisition is a dual mapping process of form-meaning and form-function mappings, 5 our unified account should be able to integrate these two mappings into one single theoretical framework. Second, this framework should have considerable generality and explanatory power in that it is capable of explaining the acquisition of the progressive aspect as well as that of the non-progressive aspect. Third, it should be able to explain both the initial state and the later development of tense marking. The logical first step towards such a unified account is to undertake a review of the empirical findings of, and theoretical explanations for, tense-aspect acquisition in the existing studies, with a view to assessing the extent to which the aforementioned requirements of such a unified theory are met by the current studies on tense-aspect acquisition. Only in so doing can we propose our own account, on the basis of a

5

See §1.3.1 for this point.

Chapter 2 In Search of a Unified Theoretical Account

critical examination of the inadequacies and limitations existing in the explanations available so far. Based on this delineation of the research agenda, the discussion of this chapter is oriented towards two goals, (1) an examination of the inadequacies and limitations of the existing theoretical accounts, and (2), how these inadequacies can be successfully handled by the unified framework to be proposed. This chapter is structured as follows. First, we will present a selective review of the studies on the acquisition of the tense-aspect morphology, in particular the studies on the effect of the linguistic constraints on L2 progressive acquisition, including their research paradigms and primary findings (§2.1), theoretical explanations (§2.2) and their inadequacies (§2.3). Next, an attempt will be made to justify the decision to compensate for these inadequacies by adopting Optimality Theory (OT henceforth) as an explanatory framework for the constraints on aspect acquisition (§2.4). §2.5 summarizes the major concents of this chapter. 2.1 Previous Studies on L2 Aspect Acquisition: Empirical Findings and Theoretical Explanations The review on the literature of L2 tense-aspect is bound to be selective because the evidence that have been amassed in this field in the past decades is ample and abundant. The review presented below is mainly organized around the two mapping processes central to the acquisition of tense-aspect marking, i.e., the form-meaning mapping and the form-function mapping, focusing particularly on L2 studies on the progressive aspect. 2.1.1 Form-meaning Mapping of L2 Tense-aspect Acquisition: Aspect Hypothesis As illustrated in the examples in§1.2, verbs with different temporal properties have different degrees of compatibility with the aspectual forms and so yield different aspect interpretations. This issue has been a central one in a series of studies now known as studies on Aspect Hypothesis.

13

Chapter 2 In Search of a Unified Theoretical Account

2.1.1.1 Aspect Hypothesis This hypothesis is generated on the observation that L1 and L2 learners, in the early stages of acquiring verbal morphology, use tense-aspect markers selectively according to the inherent lexico-semantic constraints on the verb to which the tenseaspect marker is attached (e.g., Andersen, 1991; Antinucci & Miller, 1976; Bloom, Lifter, & Hafitz, 1980; Brown, 1973; Shirai & Andersen, 1995; Weist, Wysocka, Witkoska-Stadnik, Buczowska, & Konieczna, 1984). For example, it was observed that in L1 acquisition of English, children initially used past marking on punctual and telic verbs much more frequently than on verbs denoting state and process, while the progressive –ing combined more often with activity verbs than with accomplishment and achievement verbs. Based on this observation of the biased distribution of aspectual morphology, a hypothesis was formulated on the emergence and progress of the grammatical aspects in L1 and L2 acquisition, hence Aspect Hypothesis. Interestingly, this hypothesis has been labeled in different ways, such as Defective Tense Hypothesis (Andersen, 1991), Primacy of Aspect Hypothesis (Robison, 1990) and Aspect-Before-Tense Hypothesis (Bloom et al., 1980), etc., depending on what the researcher assume to be the cause of this pattern. The “standard” formulation of the Aspect Hypothesis

6

consists of four

predictions concerning the distribution and development of aspect morphology in L1 and L2:

6

Aspect hypothesis makes predictions on the mapping between the aspect morphology and verbs of different temporal properties. The best known taxonomy of the verbs with different temporal semantics is the one made by Zeno Vendler, who distinguishes statives, which have no internal temporal structure whatsoever, such as be, have, feel, belong to; activities, which are homogeneous processes going on in time with no inherent endpoints or goal like run, walk, work, etc.; accomplishments, which involve a process going on in time and an inherent culmination point, after which the event can no longer continue, e.g., read a book, go to school, run a mile; and achievements, which also have an inherent culmination point, but in which the process leading up to this point is instantaneous and punctual, such as die, reach, find, etc (Vendler, 1976). A more detailed discussion of the verb’s temporal features inherent in the verbs will be given in §3.1.3. 14

Chapter 2 In Search of a Unified Theoretical Account

1) Learners first use past marking or perfective marking on achievement and accomplishment verbs, eventually extending its use to activities and stative verbs. 2) In languages that encode the perfective-imperfective distinction, imperfective past appears later than perfective past, and imperfective past marking begins with stative verbs and activity verbs, followed by accomplishment and achievement verbs. 3) In languages that have progressive aspect, progressive marking begins with activity verbs, and then extends to accomplishment or achievement verbs. 4) Progressive markings are not incorrectly overextended to stative verbs. (Andersen & Shirai, 1996: 531-532)

The predictions made by the Aspect Hypothesis are clear: there is a strong correlation between the development of the tense-aspect morphology and the temporal semantics of the verbs. Initially, perfective is marked on event verbs, imperfective on statives, and progressives on activities. That is to say, the natural combination is preferred by beginner learners. The initial stages of the developmental sequence of each aspect morphology have been extensively investigated across a broad spectrum of L1 backgrounds in small-scale case studies and large-scale experimental studies, and the spread of the simple past (perfective) is by far the most robustly attested stage in the distribution of interlanguage verbal morphology and compelling evidence has been obtained in support of the corresponding prediction made in Aspect Hypothesis (Andersen, 1991; Antinucci & Miller, 1976; Bardovi-Harlig, 1992; Bardovi-Harlig & Reynolds, 1995; Cai, 2002, 2003; Collins, 2002; Robison, 1990; Salaberry, 2000; Shirai & Andersen, 1995; Yang & Huang, 2004). In contrast, relatively few studies have been carried out to specifically test the predictions pertaining to distribution and development of the imperfective and progressive aspects in the learner language (Bardovi-Harlig, 2002). 2.1.1.2 Progressive Aspect Development in L2 The mention of developmental patterns of the progressive is invariably a byproduct of the studies on the simple past. Based on naturalistic oral or written production data, these studies show that the progressive aspect associates 15

Chapter 2 In Search of a Unified Theoretical Account

overwhelmingly with activities (Bardovi-Harlig, 1995b; Jin & Hendriks, 2003; Robison, 1995). It was also found that the acquisition of progressive aspect does not progress uniformly. In past-time contexts, the bare progressive marking –ing was found to emerge first, followed by the present progressive, and finally the past progressive (Bardovi-Harlig & Reynolds, 1995). In this way, while the use of the tense markers becomes increasingly target-like, the association of progressive and activities seemed to be fossilized and the spread of the progressive to other nonprototypical verb classes is surprisingly halting, an observation not compatible with the predictions of the Aspect Hypothesis (Robison, 1995). A similar pattern is shown in Bardovi-Harlig (1998) whose findings on the distribution and developmental pattern indicate that the learners’ use of the English progressive aspect robustly correlated with the activities, regardless of the task and the learners’ proficiency level. 2.1.1.3 Counter Evidence for Aspect Hypothesis There are also studies that presented evidence that appeared to run counter to the predictions made by the Aspect Hypothesis. Kumpf (1984), for example, studied the development of temporal morphology in a single Japanese learner who was found to use almost no verbal morphology except on states, consisting largely of be in the background, although it was argued that his sole subject may have been at too low a proficiency level to show productive use of verbal morphology (Bardovi-Harlig, 2000). Another challenge to the soundness of the Aspect Hypothesis comes from Rohde’s study of two children (Rohde, 1996). Unlike what is predicted in the Aspect Hypothesis, his study indicated that the progressive form did not show a distributional bias, having appeared with both activities and achievements, albeit the argument that this unexpected distribution might have resulted from the way the researcher calculated the occurrences of the aspect. More specifically, when the raw scores were converted into the percentages of the verbs in each category, the use of progressive with the achievement did not exceed the use of past with achievements. Robison (1995) also found unexpected uses of –ing with punctual events by the lowest group

16

Chapter 2 In Search of a Unified Theoretical Account

of six learners in his cross-sectional study 7 , despite the fact that most of the progressive punctual events were tokens of going to used in punctual sense, as in ‘nine or ten …he going to sleep’ (p. 357), a case that is otherwise categorized as an activity using the Vendlerian taxonomy of lexical aspect. Thus, it seems that reliable temporal characterization of verbs, large samples, and longitudinal observations are crucial to the validation of Aspect Hypothesis. 2.1.2 Form-function Mapping of L2 Tense-aspect Marking: Discourse Hypothesis Empirical studies were also conducted to explore whether learners are able to map temporal morphology upon the appropriate discourse function of grounding. An early study by Kumpf (1984) suggested that a relationship existed between the use of verbal morphology in L2 interlanguage and the narrative grounding. It was also found that “learners use emerging verbal morphology to distinguish foreground from background in narrative” (Bardovi-Harlig, 1994: 43), hence Discourse Hypothesis. Subsequent case studies and large-scale quantitative studies were conducted to test Discourse Hypothesis. Flashner (1989) found that the three Russian learners of English in his study distinguished foreground from background in oral narratives by marking the foreground predominantly with simple past, whereas the background verbs occurred mainly in base forms. Housen (1994) made a similar observation when he found that his subject, an American learner of Dutch, exhibited the following pattern when using tense-aspect morphology to mark the narrative structure: present perfect for the foregrounding information, whereas simple present and base forms for the backgrounding information. Later in a large-scale study, Bardovi-Harlig (1995b) found that Japanese learners of L2 English made frequent use of simple past in the foreground.

7 Unlike most of the studies testing the Aspect Hypothesis, Robison (1995) adopted a taxonomy of aspect different from the Vendlerian verb classes. He instead assessed three dimensions of lexical aspect, which are further divided into six aspectual categories, i.e., state, activity, durative event, punctual state, punctual activity and punctual event, the last of which is synonymous with the achievement in Vendler’s taxonomy.

17

Chapter 2 In Search of a Unified Theoretical Account

However, most of these studies were primarily concerned with the distribution of tense aspect morphology in the foreground of discourse, while the linguistic devices that learners used to mark background information, unfortunately, were seldom examined, if at all. In a word, the progressive aspect whose primary discourse function is backgrounding has been a marginal topic of research in the Discourse Hypothesis research paradigm. 2.1.3 Narrative Structure and Lexical Aaspect: Conspiring Factors? As can be seen from the above review, the testing of both Aspect Hypothesis and Discourse Hypothesis has to some extent indicated distributional and developmental patterns in L2 learners’ use of tense-aspect morphology. Unfortunately, these two hypotheses were in most cases tested independently, with only a few exceptions. Bardovi-Harlig (1998), for example, expanded the investigation to include an analysis of both narrative structure and lexical aspectual class in a single corpus comprising 74 narratives produced by adult learners of L2 English at ten different proficiency levels. The results suggest that both hypotheses are necessary to account for the distribution of verbal morphology in interlanguage. In other words, the narrative structure and the verb class are two “conspiring factors” influencing L2 tense-aspect acquisition. This finding, however, may well be accounted for in the framework of Transitivity Hypothesis (Hopper & Thompson, 1980), which shows, among a number of other components of transitivity, the aspectual forms and verb’s inherent temporal properties correlate highly with discourse grounding.

8

More crucial than a

distributional description is how these constraints interact and which is the driving force that propels L2 aspect development, the inherent verb semantics or the discourse structure? Does the grammatical and semantic prominence, as Hopper and Thompson (1980) argued, “derive from its characteristic discourse function” (p. 251) or vice versa? It remains an issue for further investigation as to whether this typological observation has explanatory power in L2 tense-aspect acquisition.

8

See § 4.2.1 for a more detailed study on this theory. 18

Chapter 2 In Search of a Unified Theoretical Account

2.2 Theoretical Accounts of Form-meaning Mapping in Tense-aspect Acquisition 2.2.1 Language Bioprogram Hypothesis and Basic Child Grammar A number of theoretical explanations have been proposed to account for the observed distribution and developmental patterns of tense-aspect morphology in L1 and L2 acquisition. One of the earliest proposals was Bickerton’s Language Bioprogram Hypothesis (Bickerton, 1984, 1999). Bickerton observed that the first generation of creole speakers created their own grammatical markers when they attempted to acquire the pidgin spoken by their parents and other community members. One remarkable finding was that creole languages have explicit progressive and imperfective markers which are lacking in their pidgin models. This observation led Bickerton to the claim that children are biologically endowed with such temporal distinctions as “punctual-non-punctual” and “state-process” contrasts. He suggested that children acquiring English first mark nonpunctuality by using –ing and then mark the punctuality by using irregular past tense forms. Similarly, Slobin (1981) proposed that the strong association of past marking with telic verbs and progressive marking with atelic verbs are due to the Basic Child Grammar, in which “temporal perspective” such as process and result are prelinguistic semantic categories that act as initial magnets for children’s grammatical aspect marking. But he did not make any explicit claims as to whether these categories are emergent or innate. 2.2.2 Prototype Account and Distributional Bias Argument Another more influential explanation for the observed pattern of tense-aspect marking in L1 and L2 is the prototype account proposed by Shirai and Andersen (Shirai & Andersen, 1995). The basic argument of this account is that learners initially discover the least marked prototypical member of each category (e.g., achievement and accomplishment for perfective, activity for progressive, and stative for imperfective), and then gradually extend the category boundary until they eventually acquire the target norm. The prototype for the category progressive, for example, can be categorized as ‘action in progress’. The lexical classes that exhibit 19

Chapter 2 In Search of a Unified Theoretical Account

this meaning are activities with the semantic features [+dynamic], [-punctual] and [telic] and they are predicted by Aspect Hypothesis to be the first class of verbs to be mapped with progressive forms. Closely akin to the prototype account is the distributional bias argument (Shirai & Andersen, 1995), which suggests that there is a distributional bias in the linguistic input that language acquirers are exposed to. More particularly, if language learners are consistently exposed to linguistic input with more –ing forms attached with activities and more past forms with achievements and accomplishments, they would start using verbal morphology in a way that it is distributed in a skewing fashion as is found in the input. Quite a number of studies lend support to this claim, as strong correlation was consistently found between the distribution of tense-aspect morphology across verb classes in learners’ production and the linguistic input directed to them (Dietrich et al., 1995; Shirai & Kurono, 1998). 2.2.3 Connectionist Account Li (2002) attempted to approach L2 tense-aspect acquisition in a connectionist framework. In these studies, he used the basic architecture of DISLEX, a multiple feature-map model of the lexicon, to simulate and model the acquisition of aspect. Self-organization in such neural networks is found to be the leading computational principles that can account for the psycholinguistic processes in the acquisition of lexical and grammatical aspect. Their simulation results demonstrated that a network supplied with the verbs from half the CHILDES database, organized in five stages, is capable of ‘learning’ the associations between verb class and aspect suffix. To interpret it differently, a neural network can produce developmental patterns that mirror children’s acquisition of aspect without recourse to innate rules or constraints (Li & Shirai, 2000). To sum up, studies of acquisition and development of tense-aspect system in L1 and L2 have accumulated substantial evidence, which, in turn, enabled the proposition of various theoretical explanations to account for the observed distributional and

20

Chapter 2 In Search of a Unified Theoretical Account

developmental patterns. It should be noted, however, that most of these theoretical accounts yielded from the “research-then-theory” studies, rather than from the “theory-then-research” studies (Larsen-Freeman & Long, 1991), and their explanatory power is restricted by the observational data obtained from these studies and thus lacks generalizability. 2.3 Problems Existing in the Tense-aspect Acquisition Studies Tense-aspect studies abound in the first and second language acquisition research literature, which enables us to develop a sophisticated understanding of the issue in question. However, some theoretical and empirical problems remain unsolved and fundamental questions unanswered if these studies are evaluated in the light of the three requirements that a unified account has to meet (See the beginning of this chapter). First, in the majority of the existing studies, the two hypotheses on form-meaning mapping and form-function mapping were tested separately, as if these two mapping processes were encapsulated modules unrelated with each other. Most of the theoretical accounts reviewed above focus on the biased distribution of a given tenseaspect morphology on a certain class of verbs, ignoring in the meantime the formfunction mapping between the temporal morphology and discourse functions. Only a few studies examined the distribution of verbal morphology across both temporal semantics and discourse function (e.g., Bardovi-Harlig, 1998; Cai, 2002; Housen, 1994), and it was recognized in these studies that semantic and discourse factors conspire to determine the development of tense-aspect forms. Nevertheless, the mechanism underlying the interaction remains unknown, and no accurate predictions on the developmental patterns involving these two mapping processes can be reasonably made. Second, although the theoretical explanations offered so far, in particular the prototype account and Language Bioprogram Hypothesis, appear to be plausible in their account of the emergence of the tense-aspect forms at early stage of tense-aspect 21

Chapter 2 In Search of a Unified Theoretical Account

acquisition and how the aspectual forms are biased towards specific types of verbs, e.g., perfective biased to achievements and progressive towards activities, they fail to identify the factors that trigger the spread from the default, prototypical mapping to the marked and non-prototypical mapping between aspectual forms and verb classes, let alone explain why the biased mapping at the early stage develops towards target norms the way it did. Regarding the connectionist account, its primary concern is on the simulation of the initial form-meaning mapping, given the limited input data for the learning program (Li & Shirai, 2000). Third, the perfective aspect (e.g., the English simple past) is the only aspectual category thoroughly investigated in studies testing Aspect Hypothesis and Discourse Hypothesis, yet studies on the L2 development of its imperfective counterparts (e.g., the English progressive) are disproportionately lacking (Bardovi-Harlig, 2002). Obviously, the soundness and generality of these hypotheses will always be a moot point unless the L2 development of a broader range of aspectual categories is examined. Given these problems, it is reasonable to conclude that current theories of tenseaspect acquisition cannot meet our requirements for a unified account of the mechanisms underlying tense-aspect acquisition. It is therefore necessary for us to set up a theoretical framework integrating different dimensions of tense-aspect acquisition. More specifically, this theory should be able to explain both semantic and discourse constraints on aspect marking, both initial emergence and later development of L2 aspect acquisition, and both perfective aspect and progressive aspect marking. Optimality Theory is exactly a theory appropriate for such a purpose, as will be shown below. 2.4 Optimality Theory as a Unified Framework Optimality Theory (OT), a theory of natural language and language acquisition, provides us with a candidate theoretical framework to unify the individual hypotheses concerning tense-aspect acquisition drawn from different linguistic sources. Initially 22

Chapter 2 In Search of a Unified Theoretical Account

proposed by Prince and Smolensky to compromise the universal constraints in phonology observed across human languages (Prince & Smolensky, 1993, 1997), OT is enjoying increasing popularity in areas of linguistic inquiry other than phonology, such as morphology (Russell, 1997), syntax (Legendre, Grimshaw, & Vikner, 1998; Sells, 2001), semantics (de Hoop & de Swart, 2000), pragmatics (Blutner & Zeevat, 2004), language change (Holt, 2003), and first language acquisition (Tesar & Smolensky, 1998a, 2000), though few attempts have so far been made to incorporate it in L2 acquisition studies . What follow are reasons why OT can serve as the unified theoretical framework we have been looking for. First, the tense-aspect system and its acquisition is a typical case of the tension between conflicting forces in human languages. The aspectual interpretation of the progressive, as was noted in Chapter 1, is sometimes at odds with the inherent temporal properties of the verbs. OT, as a theory formulated to deal with the conflicting forces in human languages, provides a linguistic framework wherein the conflict between the aspect markers and verb’s temporal semantics can be accounted for. Second, OT is also a theory of language acquisition. It regards language acquisition essentially as a process of constraint reranking, from the initial state to the target-like norm (Tesar & Smolensky, 1998a, 2000). As such, it has the potential for providing a principled account not only for the initial state and the later L2 development of tense-aspect morphology, but also for the factors triggering the constraint reranking during the course of L2 acquisition of the tense-aspect forms. Third, aspect is a multifaceted notion involving linguistic analysis at a number of levels, like semantics, discourse, and morphology and syntax (de Swart & Verkuyl, 1999; Tenny, 1994). This multidisciplinary nature of aspect in human languages makes it particularly hard to develop a comprehensive understanding of the linguistic phenomenon. This is partially the reason why a unified framework has been lacking in the study of tense and aspect and their acquisition. The studies on Aspect Hypothesis 23

Chapter 2 In Search of a Unified Theoretical Account

and Discourse Hypothesis, for example, have respectively captured one aspect of the essential characteristics of form-meaning mapping and form-function mapping in tense-aspect acquisition, yet the mechanisms underlying the interaction between these two mapping processes would remain unknown if they continue to be treated as belonging in two unrelated modules. Since OT takes a parallel and cross-modular approach to the constraints at different linguistic levels (Haji-Abdolhosseini, 2005; McCarthy, 2000), we are able to examine, within one single theoretical framework, the relationship as well as interaction between competing constraints on tense-aspect marking and its acquisition. 2.5 Summary The studies on the acquisition of tense-aspect studies in the past two decades have revealed robustly attested patterns of the aspect morphology development in L2 learners, and a number of theories have been proposed to account for the observed regularities. However, these theories, upon closer scrutiny, fail to provide a consistent account of the roles played by the verbal semantics and discourse functions in the early emergence and later development of L2 aspect acquisition. OT, a theory of language and language acquisition, is found to be a theoretical framework potentially capable of unifying disparate accounts of L2 tense-aspect acquisition. It is expected that an OT account will be able to shed new light on the interaction of competing forces in L2 tense-aspect acquisition and its underlying working mechanisms.

24

Chapter 3 Constraints on the Progressive Aspect Marking in English and Chinese

Chapter 3 Constraints on the Progressive Aspect Marking in English and Chinese Based on a review of a large body of research on L2 acquisition of tense-aspect morphology, we have proposed in the last chapter that OT might serve as a unified framework accounting for the observations in L2 progressive aspect acquisition. The next logical step is to investigate how tense-aspect marking in general, and the progressive aspect marking in particular, can be adequately described in this unified framework, what implications this OT account will have for the L2 acquisition of the tense-aspect markings, and whether this account can be empirically supported. Since OT is basically a constraint-based grammar, which sees language as a system consisting of conflicting but violable constraints (Kager, 1999; Prince & Smolensky, 1993), the first step to describe tense-aspect marking in OT framework is to identify the constraints on aspect marking and give a principled analysis on how these conflicting constraints interact with each other to yield the optimal tense-aspect forms. This step is in fact a response to the first key research question raised in § 1.3 , i.e., How might the linguistic and cross-linguistic constraints with regard to Chinese and English tense-aspect marking and their interactions governing the optimal selection of English progressive aspect in obligatory contexts be adequately described within a unified theoretical framework? Only in so doing is it possible to pave the way for establishing an OT-compatible ranking scale of the constraints and further making predictions on the effects of this scale on L2 aspect acquisition in general, and progressive aspect acquisition in particular. However, a sound description of the constraints on tense-aspect marking is not as straightforward as it might appear. Given their salience in the grammars of most natural languages and their intimate relationship with central cognitive categories of temporality, the linguistic phenomena subsumed under the label of aspect have unsurprisingly been pursued by linguists with a broad spectrum of academic interests since ancient times (e.g., Aristotle, 1933/1961), and different, sometimes divergent, 25

Chapter 3 Constraints on the Progressive Aspect Marking in English and Chinese

theories have been offered to describe how temporality is expressed in human languages. Owing to this abundance in theories and concepts, it is not uncommon to find that, on the one hand, different labels are often used to refer to the same phenomenon, while on the other hand, the same label is used by different linguists for drastically different concepts (Comrie, 1976). In order to warrant conceptual consistency in the present study, it is therefore crucial to define and clarify the key notions related to tense-aspect marking before we proceed to describe the linguistic and cross-linguistic constraints on progressive aspect marking. Thus, this chapter is devoted to the definition of the basic tense-aspect notions, and to a discussion of the (cross-)linguistic constraints on progressive aspect marking. This would prepare the grounds for a more detailed discussion on how these constraints interact and how they are ranked in an OT-compatible framework, which will be the focus of the next chapter. The organization of this chapter is as follows. The conceptual and technical confusions concerning the tense-aspect notions are clarified in §3.1.Then in §3.2, the interaction between grammatical and lexical aspect is examined. The semantic and discourse constraints on progressive aspect marking in English are explored respectively in §3.3 and §3.4. Finally in §3.5, the ways in which the progressive aspect is marked in English and Chinese are compared from a cross-linguistic perspective. §3.6 presents a summary of the chapter. 3.1 Aspect and Other Related Notions The field of tense-aspect studies is notoriously rich in terminology and conceptual confusion reigns in the sphere of temporal and aspectual semantics. Despite this abundance in terms, there are still occasional complaints that “the progress of theory formation in the field (of aspect studies) is hampered by a persistent lack of coined notions” (de Swart & Verkuyl, 1999: 3) and new notions are continuously introduced into the field, painting the picture more labyrinth-like. To make the matter more complicated, even such basic notions as tense and aspect and 26

Chapter 3 Constraints on the Progressive Aspect Marking in English and Chinese

their references are not interpreted unanimously. In English, for example, the verb took in the sentence He took some notes in the lecture, was interpreted by some linguists as referring to a simple past tense but by others as denoting a perfective aspect. It is therefore crucial to define these basic notions to ensure their consistent use in our study. 3.1.1 Aspect The notion of aspect, according to the oft-cited definition given by Bernard Comrie, refers to the “different ways of viewing the internal temporal constituency of a situation” (Comrie, 1976: 3). This now classic definition can be interpreted from two different perspectives: on the one hand, the situation referred to in a sentence has a ‘temporal constituency’ inherent in it, and on the other, human languages have ‘different ways of viewing’ the situation. In fact, this interpretation roughly corresponds to the dichotomy between situation aspect and viewpoint aspect (Smith, 1997). The former is also commonly known as lexical aspect, i.e., the temporal characteristics inherent in the lexical items describing the situation, and the latter, grammatical aspect, i.e., the linguistic devices marking aspectual distinctions by inflectional or periphrastic devices, depending on the degree of grammaticalization of aspectual devices. The use of the term aspect as a cover term for both grammatical and lexical aspect has contributed to much confusion and misunderstanding. To avoid such confusion, I reserve the term “aspect” invariably for grammatical aspect (e.g., the English progressive aspect). For the lexical semantic category of temporal classes, the more specific terms are preferred, like lexical aspect, situation type or temporal semantics. 3.1.2 Grammatical Aspect Grammatical aspect can be further categorized into perfective aspect and imperfective aspect. The former, to quote Comrie (1976) again, indicates “the view of a single situation as a whole” (p.16), while the latter “makes explicit reference to the

27

Chapter 3 Constraints on the Progressive Aspect Marking in English and Chinese

internal temporary structure of a situation” (p. 24). Put differently, a perfective sentence says that in a given period of time, an event of a certain type occurs in its entirety. Imperfective aspect, on the other hand, expresses propositions about states or ongoing processes, i.e., at a given time a state holds or a process is going on. English progressive aspect is considered a subset of imperfective aspect.9 Of course, as the term grammatical aspect suggests, grammatical devices are employed to refer respectively to the temporal constituency of an event as an entity or as a composite of individual phases.

(9) a. He was eating a cake. b. He ate a cake. (10) a. 他吃着一块蛋糕. 10 Ta chi-zhe yikuai dangao. He eat-IMPF one-CL cake. He was eating a cake. b. 他吃了一块蛋糕. Ta chi-le yikuai dangao. He eat-PERF one-CL cake. He ate (or has eaten) a cake. (9a) and (10a) are events with imperfective/progressive aspect, while (9b) and (10b) are those with perfective aspects. As is shown in these examples, the aspectual morphology (e.g., –ing in English) and aspectual particles (e.g., –le and –zhe in Chinese) are important grammatical devices to change the temporal structure of the predication. It is important to note that the simple past form like ate in (9b) is a conflation of two temporal dimensions: the simple past tense and the perfective aspect. As the focus

9

This view, however, does not go unchallenged. Dahl (1985) argued powerfully for the view that across languages progressive constructions had to be recognized as separate from Imperfective constructions. 10 The examples in Chinese will be represented in four ways in this dissertation: the Chinese characters, pinyin, literal word-for-word translation and sentence translation into English. In the glosses of this dissertation, CL=classifier, IMPF=imperfective aspect, PERF=perfective aspect, and PROG=progressive aspect. 28

Chapter 3 Constraints on the Progressive Aspect Marking in English and Chinese

of our study is on aspect, rather than on tense, the term perfective aspect will be consistently used to denote the simple past form in this study, unless the need arises for this form to refer explicitly to a past-time event. 3.1.3 Lexical Aspect Lexical aspect, also known as situational aspect, inherent aspect, or Aktionsart, refers to the inherent temporal properties of a verb or a predicate which describes the situation. Vendler’s quadripartition of verb phrases into aspectual classes is the most influential and best established categorical framework of verb classes in terms of the ‘time schemata’ inherent in the verbs (Vendler, 1967) 11. He distinguished between (1) statives, which have no internal temporal structure whatsoever, like the verbs to know and to be happy; (2) activities, which are homogeneous processes going on in time with no inherent endpoints or goal, such as the verbs to run and to work; (3) accomplishments, which involve a process going on in time and an inherent culmination point, after which the event can no longer continue, as is in the case of to build a house and to swim across the river; and (4) achievements, which also have an inherent culmination point, but in which the process leading up to this point is instantaneous and punctual, such as to die and to find. To this well-known taxonomy Smith (1997) added semelfactives, 12 which are instantaneous events with no built-in final boundary, like knock and cough. Smith argued that semelfactives resemble achievements in being punctual, but differ from them in that semelfactives encode no endpoint. When semelfactives are marked with progressive, they are interpreted as specifying a repeated event, like coughing or knocking, a feature not shared by other Vendlerian situation types.

11

Aristotle is credited with the earliest observation that such temporal meanings as “end” or “result” are inherent in verbs. In the Metaphysics 1048b, he distinguished between kineseis (movement) and energiai (actualities), a distinction corresponding roughly to accomplishments and activities in Vendler’s terminology (Dowty, 1979: 52). 12 Apart from Vendler and Smith’s categorization, another well-known tripartite aspect ontology was proposed by Mourelatos (1978). The three situation type identified by Mourelatos was States, Processes and Events. His classification was similar to Vendler’s in that states are statives and processes are Vendlerian activities. Mourelatos conflated Vendler’s achievements and accomplishments into one events category. 29

Chapter 3 Constraints on the Progressive Aspect Marking in English and Chinese

Since most of the existing L2 tense-aspect studies were conducted employing the Vendlerian verb classification system, the semelfactives were not adequately examined in L2 tense-aspect research. The present study will also make a tentative attempt to examine the role of the semelfactive in L2 progressive aspect development, alongside with the well-established Vendlerian taxonomy, by virtue of the aforementioned unique behavior of the semelfactive verbs with the progressive marker. Figure 1 summarizes the aspectual notions adopted in the present study.

Aspect

Grammatical Aspect （Viewpoint Aspect）

Imperfective

Lexical Aspect （Situation Type，Aktionsart）

Perfective States Activities Accomplishments Achievements Semelfactives

Figure 1 A Preliminary Taxonomy of Aspect in Languages

3.1.4 Technical Issues Related with Lexical Aspect When Vendler proposed his taxonomy of temporal semantics of the verbs, he was philosophically inspired, rather than linguistically and to a lesser extent, empirically, motivated. Since our study aims to propose an OT account for tenseaspect acquisition and we will put this account to empirical test, the philosophical

30

Chapter 3 Constraints on the Progressive Aspect Marking in English and Chinese

meditation on temporality in human language does not adequately serve our purposes. Therefore, some technical issues should be tackled to warrant empirical rigor before we can integrate this taxonomy into our OT analysis of tense-aspect marking and its L2 acquisition. Since the notion of lexical aspect plays a pivotal role in making hypotheses on tense-aspect acquisition, a finer-grained analysis is needed in order to distinguish unambiguously between different situation types. In this section, three technical issues related to different lexical aspects are discussed. They are: composition of lexical aspect, semantic construal of lexical aspect, and diagnostic tests. These issues have direct bearings on the designation of situation types and their instantiation, so it is important to deal with them before further investigation into the linguistic constraints on tense-aspect marking is possible. 3.1.4.1 Composition The first problem we encounter when designating a situation type such as achievement is its analytical unit: when we say a verb is an achievement, do we mean its bare form, or the verbal phrase, or the verbal predicate? In other words, what is lexical aspect composed of? When Vendler introduced his ontological verb classes, he hinted by the cited examples, but did not make it explicit, that the time schema 13 of a verb is determined by the verb, its nominal arguments and adverbial phrases. Strictly speaking, lexical aspect is not solely a semantic property of the verbal lexis. It would be misleading if lexical aspect is perceived as an attribute of the verb meaning per se, without acknowledging the contribution of the nominal arguments for the aspectual interpretation. The verb to run, for example, is an activity, but the predicate to run a mile is an accomplishment, because it has an obligatory endpoint. Henk Verkuyl is credited with the first explicit formalization of the aspect composition (Verkuyl, 1972, 1993, 1999). In order to capture the compositionality of

13

The notion of time schemata in Vender’s terminology is roughly synonymous to the more conventionally used concepts like lexical aspect, situation types or Aktionsart. 31

Chapter 3 Constraints on the Progressive Aspect Marking in English and Chinese

aspectual interpretation, he introduced two features, i.e., [+SQA] and [+ADD TO]. [+SQA], or [+Specified Quantity of A], simply reflects whether the NP’s denotation can be exhaustively counted or measured. In English, singular and definite plural NPs pertain to the positive value of [SQA], while mass nouns and indefinite plurals are characterized by its negative feature. 14 The [+ADD TO] property of a verb expresses dynamic progress, change, or nonstativity that distinguishes it from a stative verb, which has a minus value. According to Verkuyl, the (a)telicity of the predication is determined by the interaction of the [+ADD TO] feature of the predicate verb and [+SQA] feature of the arguments, as shown in (11) below.

(11) Composition of aspectual interpretation (Verkuyl, 1972, 1993) S [+ TELIC] NP [+ SQA]

VP V [+ADD TO]

NP [+ SQA]

The aspectual interpretation of these interactive features is governed by a simple Plus Principle: a telic reading of the predication is obtained only when all the values of these two features are assigned with positive values, hence plus [+]. If one or more features in a predication are assigned a minus value, the predication has an atelic interpretation, as illustrated in (12), (13) and (14) below.

(12) She played a sonata, three sonatas, some sonatas, a piece of music… [+SQA] [+ADD TO] [+SQA] = [+TELIC] (13) She played [+SQA] [+ADD TO] (14) She [+SQA]

music, sonatas, that (sort of) music… [-SQA] = [-TELIC]

wanted

a sandwich.

[-ADD TO]

[+SQA]

＝ [-TELIC]

14 The role played by nominal arguments in aspectual interpretation was also recognized by Tenny (1994) in his Aspectual Interface Hypothesis.

32

Chapter 3 Constraints on the Progressive Aspect Marking in English and Chinese

Following Verkuyl, the present study takes a compositional approach to aspectual interpretation, whereby lexical aspect is analyzed as a sentence-level notion. The verb, of course, contributes most to aspectual interpretation and is therefore analyzed as the nucleus of aspect and the verb plus its arguments is analyzed as the core of aspect.

15

The lexical aspect, in the light of this analysis, is the composite of

the temporal meanings of the verb plus other temporality-related constituents of the clause, and at the same time, the central role played by the verb in aspectual interpretation is acknowledged. The compositional approach to aspectual interpretation is a great help in determining the analytical units for aspectual interpretation. In particular, the Plus Principle is a useful linguistic apparatus for assessing the telicity status of a predicate. However, this principle cannot finely distinguish between other situation types. It cannot, for example, tell the accomplishments from the achievements, as both are judged to be telic by the Plus Principle. Semantic criteria, consequently, should be adopted to distinguish between verbs with different temporal properties. 3.1.4.2 Semantic Construal Lexical aspect, above all, is a semantic notion. Therefore, Vendler’s four verb classes can be differentiated and characterized by means of semantic analysis. Comrie (1976) identified three semantic distinctions of verbs with different temporal properties: stative vs. dynamic, telic vs. atelic, and punctual vs. durative as the semantic construal for the verbs of different situation types. They are summarized in Table 1.

15

Nucleus and core are notions in Van Valin’s Role and Reference Grammar. According to Van Valin (2002), there are three levels of syntactic units of the layered structure of the clause, namely, nucleus, core and clause, which correspond to the three levels of semantic units: predicate, predicate plus arguments, and predicate plus arguments as well as non-arguments. 33

Chapter 3 Constraints on the Progressive Aspect Marking in English and Chinese

Table 1 Semantic Features of Aspectual Classes

Statvies (STA) Activities (ACT) Accomplishments (ACC) Achievements (ACH) Semelfactives (SMF)

Dynamic + + + +

Telic + + -

Punctual + +

[+Dynamic] is a feature that distinguishes between two basic situations in human languages: events and states. This feature is generally given priority over other parameters to serve as the central criterion for the initial level distinction of situation aspect (Smith, 1997). According to Comrie (1976), the [+dynamic] contrast can be understood in terms of energy: Statives with the [-dynamic] predicates describe involuntary situations that need no energy to continue, while the predicates with the feature [+dynamic] denote situations that require a constant infusion of energy. A stative situation has no internal phases and involves no change. In other words, it remains steady for an undefined period unless a dynamic situation occurs to change it. Consequently, states do not happen, nor can they be done. They simply obtain or hold as an undifferentiated and homogeneous moment (e.g., believe). In contrast, a dynamic situation necessarily involves a change over time. The [+telic] contrast reflects the old distinction between Aristotle’s kinesis (accomplishments) and energia (activities or states) (Dahl, 1985: 80; see also Note 11 of this thesis). Many conflicting terminologies have been developed to express the telic/atelic distinction, such as bounded/unbounded distinction (Jackendoff, 1990), culminated/non-culminated contrast (Moens & Steedman, 1988), delimited/nondelimited pair (Tenny, 1994). While there is consensus that the [+telic] distinction is essential to aspectual classification, no uniform definition has been given yet. For Garey (1957), an action is telic if it tends “towards a goal” (p. 106). But this definition is somewhat misleading, since not all goal-oriented actions entail a temporal endpoint.

34

Chapter 3 Constraints on the Progressive Aspect Marking in English and Chinese

For example, the sentence she works for the benefit of the society has an explicitly stated goal, but does not have a well-defined endpoint. The contrast between telic and atelic situations can best be captured by a distinction made by Smith (1997), i.e., the distinction between the natural and arbitrary endpoints. Telic predicates have natural endpoints, and atelic events have arbitrary endpoints. Put differently, a telic predicate with the feature of [+telic] conceives a situation as having a well-defined endpoint or temporal boundary inherent in the situation denoted by the verb. An atelic predicate bearing the feature of [-telic], on the other hand, does not suggest such an endpoint unless it is arbitrarily terminated by external forces. While the event John ran can be terminated at any arbitrary moment, the running event is terminated naturally when John got to the park in the event denoted by the sentence John ran to the park. [+Punctual] as an aspectual feature relies on the contrast between a [+punctual] situation which does not last in time (or at least is not conceived as lasting in time without discernible intervals) and a [-punctual] situation which “last for a certain period of time (or at least is conceived of as lasting for a certain period of time) (Comrie, 1976: 41-42). Note that punctuality is a mental concept, hence punctualness is relative and can be of any specified temporal length. In contrast to these temporal features, Dowty (1979) assumed that all Vendlerian situation types are derived from a primitive stative predicate and a set of three aspectual operators BECOME, DO, and CAUSE 16. The four verb classes are then derived from underlying stative predicates combined with the appropriate aspectual operators: DO for activities, BECOME for achievements, DO and BECOME with the connective CAUSE for accomplishments. States directly correspond to stative predicates in the logical structure.

16

DO, BECOME and CAUSE are aspect calculus in Dowty’s terminology, roughly synonymous to aspectual operator in our study. 35

Chapter 3 Constraints on the Progressive Aspect Marking in English and Chinese

If Vn is a predicate and a1…an is (are) its argument(s), the four verb classes can be represented as the examples in (15).

(15) a. Statives:

Vn (a1…an)

e.g., The linen is white = [white (linen)] b. Activities: DO [a1, Vn (a1…an)] e.g., John swims = DO [John, swim (John)] c. Accomplishments: DO [a1, Vn (a1…an) CAUSE [BECOME [Vn (a1…an)]] e.g., John whitened the linen = DO [John, whiten (John, Linen)] CAUSE [BECOME [white (linen)]]

d. Achievements: BECOME [Vn (a1…an)] e.g., The old man died = BECOME [dead (old man)] (Dowty, 1979: 123-124) 17

(15a) shows that being white is a state, while the DO operator in (15b) indicates that there is someone by the name John who voluntarily performs the act of swimming, and the act is not caused by external forces whatsoever. The proposition denoted by the accomplishment verb, as exemplified by whiten in (15c), shows that the state of the linen from being non-white to being white is caused by some activity of John’s. In (15d), however, the proposition denoted by the achievement verb died is true if and only if the old man undergoes a change from a state of being alive to a state of being dead. In Dowty’s semantic construal of lexical aspect, DO operator represents mainly the volitionality and sentience of the subject, while BECOME

17

Dowty did not discuss the formal semantic representation of the semelfactive verbs. 36

Chapter 3 Constraints on the Progressive Aspect Marking in English and Chinese

represents a changed state and CAUSE functions as the causation process leading to the change of the state. The semantic construal of the lexical aspect proposed by Comrie and Dowty as sketched above is insightful in distinguishing between verbs of different situation types. However, as our study is an experimental one wherein lexical aspect is an important variable, these proposals on the semantic construal fall short of the rigor and empirical reliability required by the experimental purpose of testing hypotheses in tense-aspect acquisition studies, because designating a verb with one temporal value or another is prone to be subjective, intuitive and personal in the absence of objective criteria. For instance, a dynamic situation for one is very likely to be judged as stative by another, as is the case for posture verbs such as to lie and to sit. In view of this inadequacy, more reliable linguistic diagnostics should be sought to distinguish between different situation types with desirable validity and reliability in empirical acquisition studies. 3.1.4.3 Diagnostics Linguists have proposed a number of linguistic diagnostics to isolate the properties pertaining to the lexical aspect (e.g., Dowty, 1977; Tenny, 1994; Vendler, 1967). However, caution should be taken before these tests are put to use for empirical purposes. A typical case in point is the progressive test extensively used to distinguish between statives and achievements on the one hand, and activities and accomplishments on the other (Dowty, 1977; Vendler, 1967). This test, nonetheless, has been questioned by many linguists (Comrie, 1976; Verkuyl, 1989; Vlach, 1981). In the following examples, the sentences marked with the progressive aspect are all felicitous, even though they are either stative predicates or achievement predicates. The progressive aspect, therefore, does not serve as a reliable diagnostic test to distinguish between verbs of different situation types. (16)

a. He is being ill. (stative predicate) b. Max is being a fool. (stative predicate) c. John is knowing more and more about thermodynamics. (stative 37

Chapter 3 Constraints on the Progressive Aspect Marking in English and Chinese

predicate) d. She was winning the race. (achievement predicate) e. The old man is dying of cancer. (achievement predicate)

As noted by Andersen and Shirai (Andersen & Shirai, 1996; Shirai & Andersen, 1995), early studies of SLA were vague about how predicates were classified into aspectual categories. This inadequacy was recognized and rectified by later studies on L2 tense and aspect acquisition and a good number of diagnostic tests were proposed (e.g., Bardovi-Harlig, 1998, 2000; Robison, 1990, 1995; Shirai & Andersen, 1995). In particular, the ordered test scheme, proposed by Shirai and Andersen (1995) to distinguish between situation types step by step, was widely adopted in L2 tenseaspect studies (e.g., Bardovi-Harlig, 2000; Bardovi-Harlig & Reynolds, 1995; Cai, 2002). Specifically, the test is habitually used to first distinguish a stative from a nonstative. If a verb predicate passes the test, it will be regarded as a stative. If not, it goes to the next step, which is a test for activity and non-activity. The test proceeds stepwise until the situation type of the verb predicate is identified. The aspect diagnostic test system like the one proposed by Shirai and Andersen (1995) is a big step towards reliable categorization of the verb’s situation types. However, on closer examination, these tests are not exempt from problems. Take the aforementioned stative-nonstative test for example. This test says that a stative verb is interpreted as denoting a habitual event in the simple present. However, the following verbs, though obviously not statives, being an activity in (17) and a semelfactive in (18), have a habitual reading when marked with the simple present.

(17) The Earth moves around the Sun. (18) The old man coughs in winter.

Another problem for these diagnostics is that they are not able to cope with new situation types like semelfactives, since they have been devised mainly to distinguish between the four Vendlerian situation types. 38

Chapter 3 Constraints on the Progressive Aspect Marking in English and Chinese

Unlike the existing situation-based aspectual diagnostics, the present study adopts a feature-based test scheme to distinguish between different aspectual features, like [+telic], [+dynamic] and [+punctual] shown in Table 1 above. The assumption underlying this approach is that if the values of a predication’s semantic features are properly assigned, its situation type can be identified by examining the combinations of these features. For example, if the tests show that a verbal predicate has the temporal features [+dynamic], [-telic] and [-punctual], we can easily identify it as an activity verb by referring to the combination of semantic features as listed in Table 1. The diagnostic approach to be proposed is also ordered and stepwise, as is flowcharted in Figure 2 below. situation _ dynamic

STA

+ _ ACC

_ punctual

+

_ telic

_ punctual

+

+

ACH

SMF

ACT

Figure 2 Flowchart of Situation Type Identification Procedure

Recall that the lexical aspect in our study is not entirely a notion operating at the lexical level, but one at the sentential level, because we adopted the compositional approach to lexical aspect, as has been mentioned above in §3.1.4.1. In the light of this approach, the aspectual interpretation of a predicate is crucially dependent on the nucleus verb plus other temporal components, like the nominal arguments and 39

Chapter 3 Constraints on the Progressive Aspect Marking in English and Chinese

temporal adverbials. When a diagnostic test is administered in order to identify the situation type of a verbal predicate, it is the whole clause, rather than the nucleus verb alone, that is put to test for its aspectual interpretation. The tests for each temporal feature are discussed below. Dynamicity Tests Reading from top to bottom of Figure 2, the first task for identifying a sentence’s situation type is to see which parametric value of the feature [+dynamic] is set on the predicate. Only after the value of this feature is assigned on the verb can we proceed to check the other two features. In linguistic literature, a good number of tests have been suggested to set the dynamic predicates apart from stative ones. Among them the best known are the Comrie’s intuitive-based energy input criterion, Dowty’s agentive tests like cleft do structure, persuade to do test, and Do something deliberately test (Comrie, 1976; Dowty, 1979). According to Comrie’s energy input criterion (Comrie, 1976), dynamic and stative situations are different in terms of internal effort. Situations like to study for a degree and to typewrite a letter require provision of energy and are therefore [+dynamic]. In contrast, the predicate to look happy does not need input of energy, it is judged to have the feature [-dynamic]. This criterion, however, is intuition-based and devoid of empirical reliability. According to this criterion, to lie on the bed would be mistakenly treated as stative and to be at work as dynamic. Dowty (1979) proposed a number of felicitous conditions that differentiate between dynamic and non-dynamic situations. According to Dowty, the situations with the feature of [+dynamic] can occur in the progressives, or occur as complements of force and persuade, or as imperatives, or co-occur with the adverbs deliberately, carefully, or appear in Pseudo-cleft constructions. As shown in the following sentences, these tests can distinguish between the stative predicate to know the answer from the dynamic verb to build a house.

40

Chapter 3 Constraints on the Progressive Aspect Marking in English and Chinese

(19) a. *John is knowing the answer. 18 b. John is building a house. c. *John forced Harry to know the answer. d. John forced Harry to build a house. e. *John, know the answer! f. John, build the house! g. *John deliberately (carefully) knows the answer. h. John carefully (deliberately) build the house. i. *What John did was to know the answer. j. What John did was to build the house.

These linguistic tests are more empirically sound than Comrie’s intuition- based criterion. However, these tests are highly likely to be ambigious and confusing. On closer inspection, it can be found that some of these tests are more like criteria for agentivity than for dynamicity, i.e., they are tests that show whether an event involves the voluntary action of an agent or not. The non-agentive predicates in the following examples are all intuitively dynamic, but they fail to pass some of Dowty’s tests: (20) a. The leaves fell into the pond. b. ??John persuaded the leaves to fall into the pond. (21) a. The house exploded. b. ?? The house exploded deliberately. (22) a. Her heart beats faster. b. ?? Heart, beat faster! (23) a. The ice melted in the wintertime. b. ?? What the ice did was to melt in the winter time.

It is very hard to conceive of situations exemplified in the (b) sentences in (20-23) in the real-life world. One has to resort to an imaginary world or a fairy land in order that these sentences make sense. This problem, therefore, should be borne in mind before these tests are put to use when teasing the stative situations apart from dynamic ones.

18 Recall in the examples of (16) that the progressive is not a valid test for dynamicity and stativity. This test is therefore discarded as a criterion for dynamicity of a situation in our study.

41

Chapter 3 Constraints on the Progressive Aspect Marking in English and Chinese

To solve the aforementioned problem, we propose a test as a supplement to Dowty’s tests. As was discussed in §3.1.4, a dynamic situation invariably involves a happening, or an occurrence, therefore, a predication with the feature of [+dynamic] should be able to give a felicitous answer to the question What happens/happened? All the dynamic situations in the following examples (24d-h) pass this test, but the statives (24a-c) fail. This test is tentatively referred to as the happening test. (24) –What is happening / happened? a. *–He knows the answer. b. *–She resembles her mother. c. *–The handbag belongs to her sister. d. –He sleeps in the bed. 19 e. –The house exploded. f. –The ice melts. g. –He runs to the park. h. –Her heart beats faster.

To sum up, no single linguistic test suffices to set the dynamic situations apart from non-dynamic ones. Dowty’s agentivity tests and the happening test proposed above may, in a complementary fashion, serve our present purpose. In this study, Dowty’s tests will be used when the situations are agentive, and the happening test will be used when the event is non-agentive, in order to determine whether a situation under consideration is dynamic or not. Thus, if a situation is found to have the feature of [-dynamic], as is shown in Figure 2, it is clearly a stative situation. However, when a situation turns out to have the [+dynamic] property, its situation type still cannot be readily identified unless two other related temporal features are checked, i.e., [+telic] and [+punctual]. Telicity Tests

Several behavioral tests have been suggested in the literature to

differentiate a telic situation from an atelic one. The entailment test proposed by Garey (1957: 195) is widely accepted. Garey asserts that the telicity value of a verb

19 Verbs of posture like sit, stand and lie are ambivalent in that they belong to a fuzzy category between stative and dynamic verbs. See also Dowty (1979: 185-186) and Quirk, et al. (1985: 201). These verbs are treated as dynamic in the present study because they all pass the happening test.

42

Chapter 3 Constraints on the Progressive Aspect Marking in English and Chinese

can be tested with the question “if one is verbing but interrupted while verbing, has one verbed?” With an atelic situation, like singing or playing, the answer is “yes”, while with a telic situation like drowning or climbing a mountain, the answer is “no”. To put it more generally, an atelic imperfective (e.g., the English progressive) entails its perfective counterpart (e.g., the English simple past), but a telic imperfective does not. As is shown in the following cases, there is an entailment relationship between (25a) and (25b) but not in (26a) and (26b). Therefore, the verbal predicate to push a cart is atelic, while the predicate to build a house is telic. (25) a. John was pushing a cart. b. John pushed a cart. (26) a. John was building a house. b. John built a house.

Another reliable telicity test is the well-documented in X time test. The telic sentences combine felicitously with adverbials of the type in X time, while atelic sentences are incompatible with it. Parallel to in X time test is the take X time to do test (Dowty, 1979). In the following examples, sentences (27a) and (27a’) do not pass the in X time test and take X time to do test, hence atelic. (27) a. *John walked in an hour. a’. *John took an hour to walk. b. John ate an apple in an hour. b’. John took an hour to eat an apple. c. He found the treasure in two days. c’. He took two days to find the treasure.

Punctuality Tests A closer look at the flowchart in Figure 2 shows that telicity test is an interim test in the sense that the situation type of a sentence, be it telic or atelic, remains unspecified unless its value of the feature [+punctual] is designated. That is, an atelic situation can be an activity or a semelfactive, and a telic situation can

43

Chapter 3 Constraints on the Progressive Aspect Marking in English and Chinese

be an achievement or an accomplishment, depending on the value of the temporal feature [punctual] on the verb. We should, therefore, proceed to test whether a situation is punctual or not, after its telicity type is identified. Dowty (1979) proposed an array of diagnostics for the feature of [+punctual]. These include spend X time doing test, complement of stop test, and the widely-used for X time test. As shown in the following examples, the punctual events cannot felicitously be combined with durative constructions like spend X time doing or for X time tests, neither are they acceptable with the stop doing construction. (28) a. *John spent a few minutes noticing the painting. b. *John noticed the painting for a few minutes. c. *John stopped noticing the painting. (29) a. * Jack spent two seconds breaking the cup. b. * Jack broke the cup for a few minutes. c. (?) Jack stopped breaking the cup. (30) a. Ted spent a few minutes drawing the circle. b. Ted drew the circle for five minutes. c. Ted stopped drawing the circle.

The present study adopts for X time and stop doing constructions as diagnostic tests for punctual situation. If a predication cannot be felicitously combined with these constructions, it is identified as a situation with the feature [+punctual]. Otherwise, it is judged as a durative and non-punctual verbal predication. Table 2 is a summary of the aforementioned diagnostic tests for different temporal features and how they behave with different temporal features of situations denoted by verbal predicates. These tests are applied, following the procedure flowcharted in Figure 2, whenever the need arises to distinguish between different situation types denoted by verbal predications, both in theoretical discussions and in empirical instrumentation development and data coding.

44

Chapter 3 Constraints on the Progressive Aspect Marking in English and Chinese

Table 2 Diagnostic Tests for Different Temporal Features of Situation

Diagnostic Test Agentivity Happening In X Time Take X Time For X Time Stop Doing

[+dynamic] Yes Yes n/a n/a n/a n/a

Temporal Semantic Feature [-dynamic] [+telic] [-telic] [+punctual] No n/a n/a n/a No n/a n/a n/a n/a Yes No n/a n/a Yes No n/a n/a n/a n/a No n/a n/a n/a No

[-punctual] n/a n/a n/a n/a Yes Yes

Note: Yes = the test construction can be felicitously combined with the temporal semantic feature; No = the temporal feature does not agree with the diagnostic construction; n/a = the test is not applicable to temporal feature concerned. 3.2. Interaction of Grammatical and Lexical Aspects Grammatical aspect and lexical aspect are related. There is, in fact, a ‘naturalness of combination’ (Comrie, 1976: 51) between imperfectivity and stativity, and between perfectivity and punctuality. For instance, perfective aspect combines naturally with achievement verbs, because by definition, perfective aspect presents a situation in its entirety and achievement verbs provide an ideal instantiation of such a viewpoint in that they depict punctual situations as single points without internal structure. The combination of perfective aspect with accomplishments is also a natural one, though to a lesser degree, since accomplishments incorporate an endpoint which makes it possible for the perfective aspect to view the situation externally as a single whole. The notion of completion, however, is not applicable to uses of the perfective with activity verbs, since activities encode situations with no inherent endpoint. In contrast to achievements, categories that have durations with the feature of [punctual] as indicated in Table 1, are compatible with an internal view, which is morphologically represented by imperfective or progressive aspect. Both general imperfective and progressive markers yield the action-in-progress interpretation when attached with activity and accomplishment verbs, focusing on the dynamic duration of the situation denoted by the verbs.

45

Chapter 3 Constraints on the Progressive Aspect Marking in English and Chinese

With respect to stative verbs, the picture is a little different. Stative verbs do not usually combine with progressive aspect: a large set of English stative verbs do not normally accept progressive marking, including psychological and cognitive verbs such as want, need, like, love, believe, and know. The incompatibility seems to be that since progressive aspect presents a situation as ongoing, it requires that situation have successive phases which is inherently dynamic, whereas the stative verbs indicate only homogeneous situations. Thus, progressive aspect combines naturally with activity and, to a lesser degree, accomplishment verbs, but not with stative verbs. There are, of course, exceptions, where the grammatical and lexical aspects defy the naturalness of combination. The progressive aspect, for example, does not conventionally correlate with the statives and achievements, because the values of some features of these two types of verbs are not compatible with the progressive semantics (see the detailed analysis in §3.1.4). Nonetheless, when such unnatural combination occurs, it can only have interpretation deviant from the default ongoingness or action-in-progress interpretation (Comrie, 1976: 46-50), like (31) a. Her uncle is dying. b. They are reaching the top of the mountain. c. He’s being stupid. d. Mary is resembling her mother more and more. As shown in (31a) and (31b), the progressive achievements mark the preliminary stage or a preparatory process of a punctual event, to the effect of expressions of “on the verge of” or “be going to”. In (31c) and (31d), the statives, when marked with the progressive aspect, present a temporary state of the verb predicate at the reference time, rendering them a dynamic reading. The interactive pattern between the grammatical and lexical aspect is summarized in Table 3.

46

Chapter 3 Constraints on the Progressive Aspect Marking in English and Chinese

Table 3 Interaction of Lexical Aspect with Grammatical Aspect

Imperfective

Stative state

Activity in-progress

Progressive

temporariness∗

in-progress

Perfective

inchoative∗

termination

Accomplishment in-progress in-progress imperfective paradox* completion

Achievement preliminary stage* preliminary stage∗ imperfective paradox* completion

Note: adapted from (Li & Shirai, 2000). The asterisk（*）is used to indicate unnatural combinations.

Implicit in Table 3 is the observation that the aspectual interpretation has a feature of type adaptation when combined with verbs with different situation types. Take the progressive aspect as an example. When it is combined with the activity verbs, it gains its most prototypical interpretation. However, when it is attached with other situation types, the aspectual interpretation varies, ranging from tempariness, imperfective paradox and preliminary stage of an event. The varied interpretations, obviously, are attributable to the temporal semantics inherent in the verbs, to which we will turn immediately below. 3.3 Semantic Constraints on Progressive Marking: The Form-meaning Mapping The progressive aspect, conventionally defined as a grammatical means of describing “a happening in progress at a given time” (Quirk et al., 1985), is conceived of as a subset of imperfective aspect 20 . That is, the progressive is a specimen of grammatical devices of ‘viewing a situation from within’ (Comrie, 1976: 24). There are two broad approaches to the semantics of the progressive in aspectual studies, i.e., the formal approach and the functional approach. This section explores the semantic constraints of the progressive aspect in languages, with particular

20 The view that progressive is a subtype of imperfective does not go unchallenged as human languages does not offer aspectual choices as neatly as is theoretically assumed. While some languages like French make a perfective-imperfective distinction, others (e.g., English) make a progressive-non-progressive one. Dahl (1985), for example, observed that across languages the perfective-imperfective contrast correlates strongly with past vs. non-past reference whereas progressive constructions are almost independent of time referent (Dowty, 1979) .

47

Chapter 3 Constraints on the Progressive Aspect Marking in English and Chinese

reference to English, following the formal and functional approaches to the semantics of the progressive. Based on these analyses, we will put forward our proposal which integrates various functions of the progressive into one single function whereby the progressive is analyzed as an operator that functions to alter the temporal features inherent in the verbs. 3.3.1 The Formal Approach The fundamental question that the formal approach to the meaning of the progressive aspect strives to answer is a truth-theoretic one, i.e., under which condition(s) does the meaning of the progressive have truth value? The bulk of the theoretic accounts along this line maintains that the truth conditions of a progressive sentence are defined in terms of the truth conditions of the corresponding nonprogressives (cf. Asher, 1992; Dowty, 1977; Engelberg, 2001; Landman, 1992; Lascarides, 1988; Ogihara, 1990; Parson, 1990; Vlach, 1981). The idea was that a progressive sentence is true at a time interval t if t is part of a larger, later-ending interval t’ in which the corresponding non-progressive sentence is true. This characterization predicts that a progressive sentence is only true if the corresponding non-progressive is also true. For instance, we can intuitively infer that if John was pushing a car is true, then its non-progressive counterpart John pushed a car is also true. However, this inference runs into a problem, i.e., “the imperfective paradox”, as has been exemplified in (2), repeated here as (32), where the entailment inference from (32a) to (32b) is not valid.

(32) a. John was writing a letter. b. John wrote a letter.

48

Chapter 3 Constraints on the Progressive Aspect Marking in English and Chinese

It is notoriously difficult to give a satisfying account of “the imperfective paradox” and other problems related to the semantics of the progressives 21 . Consequently, various approaches have been put forth to solve this problem, such as normality approach (Dowty, 1977), the reasonable-chance approach (Landman, 1992), and the continuity approach (Vlach, 1981). These approaches raise interesting questions concerning progressive aspect acquisition, e.g., Does the logical plausibility of these approaches to progressive semantics have cognitive validity? Do language learners infer the meaning of the progressives from their default non-progressive counterparts, as proposed by the formal semanticists 22? 3.3.2 The Functional Approach The functional approach, on the other hand, endeavors to come to grips with the semantic essence of the aspect. The grammarians adopting this approach have been striving to establish a unified and prototypical account of the meaning of the progressives. Johannes Scheffer (1975: 21-42), for example, based on an exhaustive review of the studies on the English progressive aspect in the first three quarters of the last century, presented a dozen of characterizations of “the basic function of the progressive”, among which are such notions as limited duration, uncompleted action, time of frame, simultaneity, synchronism, plasticity, and an event of slow cadence. This multiplication of terminology reflects on the one hand the different perspectives that researchers took when viewing the progressive aspect, and on the other hand, its “wide range of uses” (Comrie, 1976: 33-38). Among these semantic accounts of the progressive aspect, three analyses are gaining ground. These accounts respectively analyze the progressives as stativizer, as bound canceller, and as expander. 3.3.2.1 Progressive as Stativizer

21

Engelberg (2001) lists ‘seven problems out of many’ problems connected with the semantics of the progressive that are crucial for shaping its truth conditions (pp1-5). 22 The hypothesis that perfective is a default form and the progressive is a secondary derivation in English is supported by the aspect distribution data. Based on their SEU (Survey of English Usage) database, Quirk et al. observed that ‘[a] count of a large number of verb constructions has indicated that less than 5 percent of verb phrases are progressive, whereas more than 95 percent are non-progressive.’ (Quirk et al., 1985: 198) 49

Chapter 3 Constraints on the Progressive Aspect Marking in English and Chinese

One functionalist account of the semantics of the progressives analyses the progressive as stativizer. Vlach (1981) observed that progressives, like statives, convey discourse overlap, which is in stark contrast with the other Vendlerian verb classes:

(33) a. John was running when I arrived. (progressive) b. John was here when I arrived. (stative) c. John ran when I arrived. (process) d. John polished his shoes when I arrived. (accomplishment)

The verb predicates in (33a) and (33b) have an overlap reading, while (33c) and (33d) have a sequential reading. In a sequential reading, the two events involved in the matrix clause and in the when-clause take place consecutively. John’s running and polishing began at the time of, or slightly after, the speaker’s arrival. In an overlap reading, on the other hand, the event or state in question had been under way for some time extending up to the time of the speaker’s arrival, i.e., John’s running or being here overlaps with the speaker’s arrival on time axis. This observation shows that the progressives and statives have some shared temporal properties. The overlap reading of (34a) and (34b), but not (34c) further demonstrates that progressives perform the functions similar to those of statives.

(34) a. John opened his eyes. He was in a hospital room. b. John opened his eyes. The doctor was performing an operation on him. c. John opened his eyes. The doctor had performed an operation on him.

Another piece of evidence drawn by Vlach in support of his ‘progressive-asstativizer’ analysis is that statives do not take progressives — or, on the occasions when they do, they are being used in a non-stative and contingent sense. The fact that statives and progressives are in complementary distribution shows that the two may

50

Chapter 3 Constraints on the Progressive Aspect Marking in English and Chinese

fulfill the same function for different verb classes: there is no reason for progressives to be applied in sentences that are already stative. It should be noted, however, that not all statives have an overlap temporal interpretation in sentences subordinated by when-clauses. The sentence in (35) shows that the temporal interpretation of stative predicates is more complicated than we assume:

(35) Susan was happy when Smith arrived.

Here an inchoative reading is more appropriate than an overlap interpretation. It means that Susan became happy on his arrival: there is no need for her to have been happy prior to Smith’s arrival. Based on this counter evidence, we cannot reasonably conclude that progressives are statives, but rather, it shows that statives and progressives share something in common in that they both allow overlap readings. 3.3.2.2. Progressive as Bound Canceller The second functionalist account of the progressive semantics was proposed by Kabakciev (2000), who believes that ‘the real essence of the [English] progressive is …to cancel the boundedness of NPs functioning as subjects and objects in sentences of particular major semantico-syntactic patterns’ (174-175, my italics). According to Kabakciev, the function of the English progressive is ‘not only to mark the temporal non-boundedness of the action on the verb, but also to map it onto the referent of both the subject-NP and the object-NP in the sentence transforming them from bounded into non-bounded [referents].’ (175) Similarly, Kamp and Reyle (1993) also claimed that “the semantic effect of the progressive” is to describe an eventuality that “terminates in, but does not include, the culmination point” (p. 566). The object NPs in (36b) and (37a), for example, mark the temporal endpoint or the “culmination point” when marked with past tense: the event of eating a banana was brought to a natural end when the banana was eaten. By the same token, the

51

Chapter 3 Constraints on the Progressive Aspect Marking in English and Chinese

temporal endpoint of building a house was made visible by the perfective aspect built. However, no reference was made to the temporal bounds of these events when the sentences denoting them are marked with progressive aspect, which functions in these cases to cancel or eliminate the temporal bounds of the events.

(36) a. She was eating a banana. b. She ate a banana. (37) a. They built a house. b. They are building a house.

3.3.2.3 Progressive as Expander Of all the different basic meanings attributed to the progressive, that of duration is found most often (Scheffer, 1975). That is why the term “the expanded form” or the “expanded tense” is sometimes used to refer to the progressive aspect of English (Bache, 1985; Jespersen, 1933). This property makes it possible that the aspect can establish a temporal frame encompassing something else. This function of the progressive can be mnemonically labeled as expander. This is in particular the case with achievements, which are inherently punctual, instantaneous, and therefore unnatural when marked with the progressive in that this interpretation is not shared by the other progressive-situation mappings. When these verbs are progressivized, reference is made to the preliminary stage of the verbs before the culmination point. The punctual verbs like die, for example, allow no duration, but when they are used in the progressive aspect, the process prior to the action denoted by the verb becomes visible, as shown in (38).

(38) a. Macbeth was dying with volume turned down. b. As he was falling asleep, he remembered that he had not completed the article. c. John is winning the race. d. It seems that Owen is scoring another goal. e. He was stopping the car at the garage.

52

Chapter 3 Constraints on the Progressive Aspect Marking in English and Chinese

These punctual verbs, at first sight, seem immune to the progressive aspect, to the extent that their combination with the progressive aspect sometimes defies the immediate intuition of linguists (Dowty, 1979; Vendler, 1967). However, as shown in the examples above, there are many achievement verbs acceptable in the progressive, which yield a “slow-motion” or “film-strip” reading, where a normally instantaneous event is perceived as being “stretched” over time (Rothstein, 2004: 37). This function of the progressive achievement is schematized in Figure 3.

I Preparatory Process John is running abreast with other runners

FN Preliminary processes John is winning the race

Resultant state John has won the race

Figure 3 Reference of the Progressive Punctuals 23

As is illustrated in the figure, the event win the race, which would otherwise be punctual and instantaneous, is expanded as if it were being shown in a slow motion movie. 24 Note, however, that this analysis of the progressive as a temporal operator has not exhausted the multidimensional semantic constraints of the progressive aspect. As was pointed out by Quirk et al. (1985),

“The constraints of the progressive cannot, it seems, be explained entirely in terms of meaning. Since the use of the progressive aspect has been undergoing grammatical extension over the past few hundred years, it is likely that its use is still changing at the present day, and that its description at any one time cannot be totally systematic. This would explain the difficulties faced by those attempting to account in every respect for the conditions for the use of the progressive in terms of semantic generalization.” (p. 202)

24 In this figure, I stands for the initial point of the event and FN, the natural endpoint of the event, following the schematic representation of events proposed by Moens and Steedman (1988). 53

Chapter 3 Constraints on the Progressive Aspect Marking in English and Chinese

To sum up, there are two broad approaches to the semantics of the progressive. The formal approach implies that the inference of the meaning of the progressive aspect is based on the semantic content of its non-progressive counterpart. The functional approaches, however, analyze the meanings of the progressive as an operator, fulfilling the functions labeled variably as a stativizer, a bound canceller or an expander. As different classes of verbs call for different functioning of the aspect, the progressive aspect displays “an unusually wide range of uses” (Comrie, 1976: 33). 3.3.3 The PROG Operator Upon closer examination, we are able to see that the three functionalist analyses of the progressive aspect have a shared feature: the aspect does something to the verbal predicate when it is marked with the progressive. That is, the relation between aspect marker and the verbal predicate is one between operator and operandum (Bickel, 1997; Johanson, 2000). Therefore, the progressive aspect marker can be viewed as a [PROG] operator performing the semantic functions of stativizer, bound canceller, or expander. If we compare the effect of these functions on verbs of different situation types, we can see that the primary function that the progressive aspect fulfills is to alter the temporal features inherent in different situation types, as shown in Figure 4. Listed on the left columns are the temporal features inherent in the verbs, and on the right are the shifted features after [PROG] operates on the verb. Three points in Figure 4 are worth noting. First, the basic semantics of the progressive constitutes three temporal features, viz., [-dynamic], [-telic] and [punctual]. Put differently, if a verb predicate is progressively marked, it basically denotes a stative, incomplete and durative situation (Quirk et al., 1985). Second, there is a selective feature adaptation process involved when the three functions of the [PROG] operator are performed. The stativizer function, for example, is operated on the verbs with the [+dynamic] feature, and the expander function, on punctual verbs.

54

Chapter 3 Constraints on the Progressive Aspect Marking in English and Chinese

Similarly, if a verb is devoid of an endpoint, there is no need for the [PROG] operator to cancel its boundedness. This function is solely adapted to the telic verbs. Third, some verb types receive more operations of [PROG] than the others. As can be seen in the Figure, an achievement verb is subject to three [PROG] operations in order to be properly marked with the progressive, whereas a stative verb does not need even one single operation. It can perform similar functions of the progressive aspect without being marked by the progressive form, becasue basic semantic features are exactly the same as the aspectual interpretation of the progressives, i.e., [-dynamic], [telic], and [-punctual].

STA

[-dynamic]

[-telic]

[-punctual]

[-dynamic]

[-telic]

[-punctual]

ACT

[+dynamic]

[-telic]

[-punctual]

[-dynamic]

[-telic]

[-punctual]

ACC

[+dynamic]

[+telic]

[-punctual]

[-dynamic]

[-telic]

[-punctual]

SMF

[+dynamic]

[-telic]

[+punctual]

[-dynamic]

[-telic]

[-punctual]

ACH

[+dynamic]

[+telic]

[+punctual]

[-dynamic]

[-telic]

[-punctual]

[PROG]

Stativizer Bound Cancellor Expander

Figure 4 The functions of the [PROG] operator

This third point was supported by de Swart (1998) in her syntactic formulation of tense and aspect in the human languages, as shown in (39). In the light of her analysis, tense marker operates after all the aspectual operators have done their work. In addition, tense marker differs from aspect marker in that a tense marker operates only once while an aspect operator functions iteratively. The Kleene star * in (39) indicates zero, one or more aspectual operations.

55

Chapter 3 Constraints on the Progressive Aspect Marking in English and Chinese

(39) [Tense [Aspect*[Eventuality description]]] 25 (de Swart, 1998)

Some linguists noted that the functions of the [PROG] operator do not operate on a situation simultaneously and there is, instead, a priority sequence of operations. In support of his progressive-as-stativizer argument, Vlach (1981) claimed that process verbs have priority to other verbs types when marked with the progressive aspect. 26 If a situation is not a process, Vlach assumed, it is coerced into a process before the progressive marker is licensed. Consider the following examples: (40) a. John was building a house. b. [PAST [PROG [PROC [John build a house]]]] (41) a. John was running. b. [PAST [PROG [John run]]]

As mentioned above, Vlach maintained that the [PROG] operator always combines with a process sentence. In order to cope with progressivized telic sentence, Vlach posited the operator [PROC(ESS)], which serves to turn a telic sentence into a process sentence. As the progressive operator never combines directly with a telic sentence in Vlach’s analysis, the imperfective paradox never arises. Similarly, Moens (1987), following Vlach (1979), also claimed that the progressive aspect marker in English only combines with process sentences and the resulting sentence is stative. He described the transition from culminated processes (accomplishments) to processes as something that strips off the culmination, which is tantamount to the operation of “canceling boundedness” performed by the [PROG] operator.

25

de Swart’s eventuality description is synonymous to lexical aspects or situation type in our study. 26 Recall that process verbs, in Mourelatos’ aspect ontology, are like activity verbs (Mourelatos, 1978). See also Note 12 for more about Mourelatos’ classification of situation types. 56

Chapter 3 Constraints on the Progressive Aspect Marking in English and Chinese

Vlach and Moens’ analyses show that the essential function of the [PROG] operator is to turn a dynamic situation into a stative one. However, no situation type other than activity or process verb predicates is licensed to align naturally with the progressive aspect. The non-activity verbs, such as the telic or punctual verbs, should be coerced into activities in the first place before they can be marked with the progressive. According to this analysis, the bound canceller and the expander functions are secondary to the stativizer function of the [PROG] operator. Two questions, however, still remain: why is stativizer the most essential function of the progressive aspect? And, what are the determining factors responsible for the progressive aspect marking on the less canonical verb types, in spite of the difficulties recognized by Vlach and Moens? The answers to these questions, however, should be sought beyond the semantic constraints and this leads us to the field of discourse analysis for explanations. 3.4 Discourse Constraints of Progressive Aspect Marking: The Form-Function Mapping Apart from the semantic constraints, the progressive aspect is also subject to discourse constraints. Linguists have long observed that the use of the progressive is context-dependent. Hinrichs (1986: 69) used the notion context-dependency to describe the function of the progressive: “…in contrast to the simple form of the verb, the progressive form does not introduce a temporal context but depends for its interpretation on a temporal context independently established”. In English narratives, progressive morphology marks a verb predicate that overlaps another one. By contrast, non-progressive morphology pushes events forward and thus obtains a consecutive reading. (42) a. Sam was flying a kite (when/??and then)the string broke. b. Sam flew a kite (??when/and then)the string broke.

57

Chapter 3 Constraints on the Progressive Aspect Marking in English and Chinese

In sentence (42a), Sam’s flying of the kite is background information — the plot element in focus is that the string broke. In sentence (42b), where both clauses are in the perfective aspect, both the kite-flying and the string-breaking have more or less equal status as consecutive events of the plot. The preferred connective for the two clauses (when/and then) is effected by this narrative relationship. The serial reading of the perfective sentences and the overlap reading of the progressives are not confined to the time adverbial clause. In the parallel sentences without time adverbial markers, similar aspect interpretations remain, (43) a. John entered the president’s office. The president was writing a letter. b. He awoke to the sound of her screeching. She was shaking him, screaming at him. What is shared by the sentences in (42) and (43) is that the progressive sentences overlap or encompass the non-progressive sentences, making the non-progressive sentences more prominent in the narrative structure. Therefore, we can reasonably assume that the progressive and its non-progressive counterpart seem to have a division of labor in marking discourse functions in the narratives. Cross-linguistic studies show that ‘‘the fundamental notion of aspect is not a local-semantic one but is discourse-pragmatic” (Hopper, 1982: 5), and that such aspectual distinctions as event-state, punctual-non-punctual, telic-atelic contribute significantly to the discourse organization (Givon, 1982; Hopper, 1982; Hopper & Thompson, 1980; Wallace, 1982). What discourse functions does the progressive aspect actually perform? Based on her channel-theoretic analysis of the progressives, Glasbey (1996; 1998) gave a detailed formal analysis of the progressive functioning as discourse backgrounding. This analysis gained support from the cross-linguistic observation that discourse grounding plays a determining role in the aspect marking of the verbs (Ehrlich, 1987; Hopper, 1979, 1982; Hopper & Thompson, 1980; Wallace, 1982). Owing to their stative properties identified in the last section, the English progressive aspect is an

58

Chapter 3 Constraints on the Progressive Aspect Marking in English and Chinese

optimal candidate for discourse backgrounding, but not for discourse foregrounding, as is shown in the discourse fragment in (44): (44) He looked up at the pantry ceiling, which was shaking with the stamping and shuffling of feet on the floor above, listened for a moment to the piano and then glanced at the girl, who was folding his overcoat carefully at the end of a shelf. (James Joyce: The Dead)

In this episode, the main storyline “He looked up…listened…glanced…” is pushed forward by the sentences marked with the simple past, while the shaking of the pantry ceiling and the act of folding the overcoat, narrated in the progressive aspect, contribute little to the movement of the story. Their function is primarily to elaborate on the main events in the foreground narrated using the perfective aspect. Compared with the contribution of the “naturalness of combination” of the verbs, discourse grounding is more fundamental a function of aspect choice. Just as Hopper (1979) generalized:

“What morphosyntactic devices does a language possess? …Why does this language, and languages in general, have such a mechanism? … (F)rom a discourse viewpoint tense-aspect becomes intelligible. … If a language has a contrast between a perfective (completive, non-durative, punctual) aspect and other aspects, then part of the meaning of the perfective aspect, at least in narrations, is to specify major, sequential, foregrounded events, while part of the meaning of the contrasting nonperfective aspects, particularly an imperfective, is to give supportive backgrounding information.” (p. 239)

This generality reveals that the local natural-combination account of the aspect semantics is incomplete at best and our understanding of the aspect should primarily be rooted in discourse requirement for the temporal interpretation of the sentences. The use and interpretation of the progressives, apart from those semantic constraints identified by our progressive-as-operators analysis, is also subject to discourse constraints. In the light of this analysis, to acquire the progressive aspect in L2, a

59

Chapter 3 Constraints on the Progressive Aspect Marking in English and Chinese

learner is encountered with the task of acquiring the competing constraints at these two linguistic levels and the rules governing their ranking order. 3.5 Cross-linguistic Analysis of the Progressive Aspect in Chinese and English In addition to the L2 semantic and discourse constraints, adult learners have a full range of L1 tense-aspect constraints in position even at the initial stage of L2 progressive aspect acquisition. The L1 constraints on aspect marking also have a role to play in L2 aspect marking. Typologically, Chinese is an aspect-prominent language, and English is a tenseprominent one (Bhat, 1999). Unlike English, Mandarin Chinese is an aspect marking language devoid of grammaticalized tense distinctions. Furthermore, while the Chinese language is an aspectually rich language with a perfective-neutralimperfective tripartite structure (Smith, 1997), the English language is arguably a language that is aspectually impoverished (Giorgi & Pianesi, 1997) in which the deictic tense morphology sometimes conflates with aspectual markers, as is the case of simple past tense and perfective aspect in English language. The notion of imperfectivity (including progressivity) is typically expressed in Chinese by the imperfective markers zai- and –zhe, with zheng- and –ne as their supplementary variants (Chao, 1968), as is illustrated by the following construction: (45) Chinese imperfective construction and its varieties （正）（在）动词（着）（呢） (zheng) (zai) V

(zhe) (ne) 27

be doing

27

Note that the roles played by these progressive markers are different: while –zhe and zai- are more grammaticalized and therefore more obligatory in progressive construction, the other two function as progressive intensifiers. The present study only focuses zai- and –zhe, and conflates them into one general category. Following Smith, Li and Thompson, among others, both of these aspectual markers are treated as Chinese progressive expressions (Li & Thompson, 1981; Smith, 1994). Diachronic studies of the these two aspectual markers also show that the two are of the same origin (Xu, 1992). 60

Chapter 3 Constraints on the Progressive Aspect Marking in English and Chinese

Zai- and –zhe have a fuzzy functional division of labor, with the former similar to an English progressive and the latter to a stative imperfective, and particularly, to a resultative stative (Klein, Li, & Hendriks, 2000; Lu, 1942; Smith, 1994; Wang, 1943; Xiao, 2002). Progressives, whether in Chinese or in English, can prototypically co-occur with activities and accomplishments as both classes contain a durative process. However, English and Chinese show marked differences when the progressive correlates with the other two Vendlerian classes of verbs, i.e., statives and achievements. The English achievement verbs can be progressivized to profile a situation at its preliminary stage or anticipatory process, as the analysis of the progressive as an expander shows (See § 3.3.2), but their Chinese equivalents are unacceptable.

(46)

They are reaching the top of the mountain. *他们 在 到达 着 山顶。 *Tamen zai-(PROG) daoda –zhe (IMPF) shanding.

(47)

She was winning the race. *她 在 赢得 着 比赛。 *Ta zai-(PROG) yinde –zhe (IMPF) bisai.

On the other hand, the Chinese imperfective marker –zhe can be affixed to stagelevel stative sentences denoting transitory properties, while their English counterparts, if progressivized, sound unnatural, as in (48-49). When the English progressive aspect does appear in stative sentences, such sentences express transient and dynamic interpretations lacking in their Chinese translation, like (50-51)

(48)

他们彼此深爱着，依恋着。 Tamen bici shen’ai-zhe, yilian-zhe. They each other deeply love-IMPF rely-IMPF ?They are deeply loving and relying on each other.

(49)

我们的管理存在着问题。 Womende guanli cunzaizhe

61

wenti.

Chapter 3 Constraints on the Progressive Aspect Marking in English and Chinese

Our management exist-IMPF problem. ?Problems are existing in our management. (50)

He was really liking the play. ?她确实 在 喜欢着 这出戏 ?Ta queshi zai-PROG xihua –zhe (IMPF) zhechuxi.

(51)

She is resembling her father more and more. ？她 在 越来越 长得像着 她爸爸 ?Ta zai-(PROG) yuelaiyue zhangdexiang –zhe (IMPF) ta baba.

Based on typological observations like these, Shirai (1998) claimed that English progressive morphology and Chinese imperfective markers focus on different stages of an ongoing event. The English progressive can be used to denote an event at its preliminary stage and ongoing stage, the Chinese imperfectives, on the other hand, make reference to the ongoing and resultative state of the event. Figure 5 schematizes this difference.

I English progressive (achievements) Preliminary stage

FN(FA) Chinese -zhe

English Progressive Chinese zai- & -zhe (activities/accomplishments (statives) ongoing event resulative state

Figure 5 Semantic Referents of Progressive in Chinese and English 28

From the preceding cross-linguistic analysis, it can be concluded that, unlike English, Chinese is an aspect-dominant language. While little difference is shown in progressive aspect marking on activity and accomplishment verbs between English and Chinese, there are marked differences in progressive marking on stative and

28

Refer to Note 23 for a more detailed description of the notations in this scheme. 62

Chapter 3 Constraints on the Progressive Aspect Marking in English and Chinese

achievement situations. Chinese progressive achievement is not licensed, while its English counterpart has an anticipatory interpretation. By contrast, a subset of stative verbs in Chinese can be marked with the imperfective marker “-zhe”, whereas in English, the stative progressives are not conventionally acceptable.

3.6 Summary This chapter first introduced some basic tense-aspect notions and then dealt with some technical issues concerning the lexical aspect, a notion playing a vital role in the form-meaning mapping of aspect marking. It also discussed the relationship between grammatical aspect and lexical aspect. It was revealed that there existed both natural and unnatural combinations between the situation types and aspectual interpretations. In so doing, we prepared the grounds for further discussion about the constraints on aspect marking in general, and the progressive aspect marking in particular. The English progressive aspect is subject to semantic and discoursal constraints. Semantically, the English progressive marker operates to stativize a dynamic situation, to cancel its endpoint and/or to expand a punctual event. These functions are achieved via the implementation of the [PROG] operator, which functions primarily to alter the temporal properties inherent in verbs. The operation of [PROG], however, varies with the number of the features to be shifted. Naturalness of combination and imperfective paradox that manifest themselves in the progressive aspect, according to our analysis, are ascribed to varying degrees of compatibility of the progressive marker with verbs of different temporal features. From the discourse perspective, the primary function of the progressive is to stativize a situation, to license it as an optimal candidate for discourse backgrounding. The semantic implausibility caused by the progressive marking on non-prototypical verb types is resolved by meeting the discourse requirement for a situation. Put differently, the truth condition of the discourse is satisfied in the first place, at the cost of semantic truth condition. This discourse function can explain the non-prototypical 63

Chapter 3 Constraints on the Progressive Aspect Marking in English and Chinese

uses of the progressive aspect because the discourse requirement for the temporal properties of the verbs should be met in the first place. We have also examined the cross-linguistic constraints on the progressive aspect. While the temporal references of progressive activities and accomplishments are the same in English and Chinese alike, Chinese achievements and statives manifest different felicitous conditions from their English counterparts when marked with the progressive aspect.

64

Chapter 4 An OT Analysis of the Constraints on the Progressive and their Interaction In the previous chapter, we have clarified the prevalent conceptual and technical confusions in aspect studies. We have also identified the linguistic and cross-linguistic constraints on aspect marking. In so doing we have cleared the grounds for integrating the linguistic descriptions of aspect marking with an OT framework. This chapter, following the threads of the last chapter, further pursues the first key question stated in §1.3. It examines how the constraints on aspect marking identified in the previous chapter interact with each other and how the competition and ultimate resolution of these conflicting constraints can be adequately explained in an optimality-theoretic framework. This chapter consists of three parts. It starts with a concise overview of the basic tenets of Optimality Theory and its relevance to language acquisition studies (§4.1). It then proceeds to deal with the interactions of the conflicting constraints on the progressive aspect, thus paving the way for an OT account of the interaction of constraints on the progressive aspect (§4.2). It ends with ranking the conflicting constraints on progressive aspect marking in the OT framework (§4.3). 4.1 Optimality Theory 4.1.1 Basic Tenets of OT Optimality Theory was initiated by Prince and Smolensky as a new phonological model to deal with the interaction of violable constraints on the sound patterns governing human language (Prince & Smolensky, 1993, 1997). In recent years, OT is gaining increasing appeal among linguists working outside phonology, like morphology, syntax, natural language interpretation, pragmatics, language change, and language acquisition. At the heart of OT is the idea of universal constraints, which are nevertheless violable. According to OT, an output is ‘optimal’ in the sense

Chapter 4 An OT Analysis of the Constraints Interaction on the Progressive

that it incurs the least serious violation ranked in a hierarchy. For a given input, the grammar generates and then evaluates an infinite set of output candidates, from which it selects the optimal candidate, which is the actual output. Figure 6 is a schematic representation of these basic assumptions of OT (Kager, 1999: 8).

C1

Input

>> C2

>> C3

Candidate a Candidate b Candidate c Candidate d Candidate …

Output

Figure 6 Mapping of input to output in OT grammar

As is shown in Figure 6, evaluation takes place by a set of hierarchically ranked constraints (C1>>C2>>…Cn) 29, each of which may eliminate some candidate outputs, until a point is reached at which only one optimal output survives. 4.1.2 Conventional Notations A tableau is the conventional way of showing in a schematic form the gradient scale of constraints in OT analysis. (52) is a sample tableau showing the ranking of the constraints and how the conflicts between the constraints are resolved for the selection of the optimal output. (52) A sample tableau for constraints ranking

29

The notation “>>” means “more highly ranked” in OT analysis. 66

Chapter 4 An OT Analysis of the Constraints Interaction on the Progressive

C1 a.

candidate a

b.

candidate b

C2 *

*!

The tableau shows two (or any number of) output candidates vertically in random order, and two constraints C1 and C2 (or, again, any number of them) horizontally, in a descending ranking from left to right, i.e., C1>>C2. The cells contain violation marks “*” incurred by each candidate for the constraint heading the column. The optimal candidate is marked by the pointing finger “

”. In our example,

this candidate is (52a), which has no violations of the higher-ranked constraint C1, a constraint violated by its competitor (52b). Note that the optimal candidate (52a) is actually not unproblematic by itself: it has a violation of C2, although this flaw is insignificant to the outcome. Although the pattern of violation for C2 is the reverse of that for C1, this does not help candidate b. Its violation of C1 is already fatal, indicated by the accompanying exclamation mark (!) and the shading of the cells whose violation is no longer relevant. In short, candidate (a) is optimal as no candidate is available that fares better, satisfying both constraints at the same time. A violation of C2 is tolerated, as long as a higher constraint C1 can be satisfied (Kager, 1999: 13). Consider, for example, the difference between the simple English sentence It rains and its Italian counterpart piove—literally, “rains.” According to the OT analysis of Grimshaw and Samek-Lodovici (1998), at issue here is a conflict between two constraints—SUBJECT and FULL-INT(ERPRETATION): (53) Constraints on the subject: a. SUBJECT: Every sentence must have a subject. b. FULL-INT: Every element of a linguistic expression contributes to its interpretation. In English, the conflict is resolved in favor of SUBJECT: to provide a subject, it must appear, even though it has no referent and contributes nothing to the

67

Chapter 4 An OT Analysis of the Constraints Interaction on the Progressive

interpretation of the sentence, violating FULL-INT. In Italian, the conflict is resolved the other way round: no meaningless subject may appear, and FULL-INT prevails over SUBJECT. Therefore, the presence or absence of the subject in these two languages is subject to different ranking of these two constraints, as is shown in the constraints resolution tableau (54a) and (54b).

(54) Ranking constraints on the subject in English and Italian

a. English

SUBJECT

It rains. rains. b. Italian

* *! FULL-INT

Pivio Cio pivio

FULL-INT

SUBJECT *

*!

4.1.3 Faithfulness and Markedness Constraints Two basic types of constraints are distinguished in OT, i.e., faithfulness and markedness constraints, which are inherently conflicting. Faithfulness constraints require identity between the input and the output candidates and markedness constraints evaluate the form of the output candidate, favoring certain structural configurations over others. In the example cited above in (54), the constraint FULLINT is a faithfulness constraint, as it requires the correspondence of interpretation of the sentence as a whole and its individual elements. The constraint SUBJECT, on the other hand, is a markedness constraint, as it penalizes the more marked subjectless sentences. The roles played by faithfulness and markedness are essentially different. The markedness constraints militate against the marked and complex structures in languages and therefore favor those structures that are linguistically unmarked,

68

Chapter 4 An OT Analysis of the Constraints Interaction on the Progressive

cognitively economical or perceptively salient. The faithfulness constraints, by contrast, maximally preserve the linguistic properties of the input, functioning to neutralize the powers of the markedness constraints. Without faithfulness constraints, all distinctions made by the input forms would be reduced to some least marked outputs. And without markedness constraints, all the properties in the input form are copied in the output and language structures would be very complex. Interactions between faithfulness and markedness constraints are at the heart of OT explanation of the motivating force of

language diversity, language change, and language

acquisition (Kager, 1999; McCarthy, 2002; Prince & Smolensky, 1993). What follow in (55) and (56) are some typical specimens of faithfulness constraints and markedness constraints in phonology and syntax, together with their descriptions: (55) Examples of faithfulness constraints a. PARSE/MAX(X): Every X that occurs in the input must occur in the output. (Don’t delete X) b. FILL/DEP(X): Every X that occurs in the output must have occurred in the input. (Don’t insert X.) c. IDENT[F]: Where [F] is some feature: if a segment is specified [αF] in the input, it may not be [-αF] in the output. d. STAY: Trace is not allowed. (56) Examples of markedness constraints a. *VNASAL: Vowels must not be nasal. b. * STRUC: Penalize morphological marking c. OP-HD: A projection has a head.

4.1.4 OT Account of Language Acquisition As a theory of language acquisition, OT explicitly addresses the logical problem of language acquisition by demonstrating clearly that it is able to yield a learnable model of grammar (Boersma & Hayes, 2001; Ma, 2003; Tesar & Smolensky, 1998a, 1998b, 2000). More particularly, in OT, language acquisition is not viewed as a process of learning the well-formedness constraints of these structural descriptions. Rather, it is regarded as involving process of constraints reranking, when the ranking 69

Chapter 4 An OT Analysis of the Constraints Interaction on the Progressive

orders vary between initial ranking and ultimate ranking. In this process, the learner starts with the initial ranking of the constraints and their learning goal is to acquire their ultimate ranking, as is illustrated in (57) below,

(57) Constraints reranking involved in language acquisition Initial ranking Constraint 1>> Constraint 2 >> Constraint 3 …… Ultimate ranking Constraint 2 >>Constraint 1 >> Constraint 3

Apart from this principled account for language acquisition, OT also has a more specific working hypothesis on how this reranking takes place: markedness constraints must outrank faithfulness constraints in the initial state of grammar acquisition, i.e., markedness >> faithfulness, often abbreviated as [M>>F]. Broadly speaking, with markedness dominating faithfulness constraints in the initial ranking, the learner’s output contains only unmarked forms: violations of markedness constraints can only be optimal if higher-ranked faithfulness constraints require them, and in this initial state there are no higher-ranked faithfulness constraints coming into play (Gnanadesikan, 1995; Smolensky, Davidson, & Jusczyk, 2004). This hypothetical initial ranking is formulated as a general principle of OT, notably The Emergence of the Unmarked (McCarthy & Prince, 1994) . What light does this [M>>F] initial-ranking hypothesis shed on the acquisition of tense-aspect marking? As has been discussed in the last chapter, linguistic studies and language acquisition studies have identified at least two conflicting forces, i.e., the time schemata of different classes of verbs and discourse grounding. These two forces are assumed in OT to interact with each other for the user’s ultimate selection of the aspect marker. It is further assumed, in the light of OT hypothesis of the initial [M>>F] ranking, that the least marked form emerges first in the interlanguage of L2 aspect morphology and, with the demotion of the markedness constraints and promotion of the faithfulness constraints, more marked forms are selected and the tense-aspect

70

Chapter 4 An OT Analysis of the Constraints Interaction on the Progressive

marking system in L2 learners spreads from the propotypical marking to the less prototypical marking. In a word, OT is essentially a theory of constraint conflict, interaction and resolution. It explains explicitly such linguistic phenomenon as language universal, linguistic typology, language change, and language acquisition in terms of optimal ranking of two conflicting forces inherent in human languages, i.e., the markedness constraints and faithfulness constraints. Language acquisition is involved in a process of reranking the constraints from the initial [M>>F] configuration to the more marked structure whereby faithfulness constraints are permitted to dominate some markedness constraints. The first task of the present study is to ascertain, in the light of this framework, the initial ranking of the constraints underlying Chinese learners’ use of the English progressive aspect, and then how this ranking develops towards the targetlike ultimate ranking. 4.2 Conflicting Constraints on the Progressive Aspect Marking Recall that we have identified two factors affecting our choice of aspect markers, the temporal semantics of the verb and discourse grounding marking. However, further linguistic analysis is needed in order to identify the typology of these constraints and to incorporate these constraints into an OT-compatible framework. It should be noted from the outset that OT is a principled account for constraint interaction, and as such it is not able to specify a particular set of constraints on aspect marking. That is, it makes no claim a priori as to which constraints that the progressive-aspect marking is subject to and in what order they are ranked, both of which remain issues for further inquiry. As discussed in the last chapter, linguists have identified two levels of constraints on aspect marking, i.e., the discourse constraints and the semantic constraints. Since OT assigns no decisive role to any constraint on aspect marking, it is reasonable to assume that these constraints conflict and interact with each other, with some weighting more than others at a given stage of aspect acquisition. The 71

Chapter 4 An OT Analysis of the Constraints Interaction on the Progressive

Transitivity Hypothesis (Hopper & Thompson, 1980), Discourse Representation Theory (Kamp & Reyle, 1993), and in particular, theory of aspect coercion (de Swart, 1998), offer theoretical accounts of how constraints at different linguistic level interact with each other for optimal choice of aspectual forms. 4.2.1 Transitivity Hypothesis (TH) Traditionally, the notion of transitivity was understood as a global property of an entire clause, such that an activity is ‘carried-over’ or ‘transferred’ from an agent to a patient. Based on their cross-linguistic observation, Hopper and Thompson generalized in their Transitivity Hypothesis (TH henceforth) that the notion of transitivity is a cluster of properties, only one of which is the presence of an object of the verb, which is manifested in the conventional interpretation of the notion transitivity. In their study, ten scales were identified to which a clause can be ranked in the transitivity hierarchy.

(58) Transitivity Hierarchy (Hopper & Thompson, 1980: 252) High Transitivity

Low Transitivity

Participant

2 or more participants

1 participant

Kinesis

action

non-action

Aspect 30

telic

atelic

Punctuality

punctual

non-punctual

Volitionality

volitional

non-volitional

Affirmation

affirmative

negative

Mode

realis

irrealis

Agency

A high in potency

A

30

low

in

As has been shown in §3.1.4.2, the notion of telicity conventionally refers to the temporal endpoint inherent in a verb or a predication and it is obviously a property of lexical aspect. However, Hopper and Thompson’s terminology differs from this conventional interpretation of telic/atelic distinction. It can be inferred from the examples cited by Hopper and Thompson (1980) that the telic/atelic distinction represents the perfective/imperfective contrast in grammatical aspect.

72

Chapter 4 An OT Analysis of the Constraints Interaction on the Progressive

potency 31 Affectiveness of O O totally affected

O not affected

Individuation of O 32 O highly individuated

O nonindividuated

The hierarchy in (58) must be understood at the discoursal level. According to Hopper and Thompson (1980: 280), properties of high transitivity are correlated with discourse foregrounding, and those of low transitivity are closely akin to discourse backgrounding. Their claim is that unmarked situation of discourse foregrounding is collectively characterized by the high transitivity properties listed in the left column of the scales, while the unmarked situation for discourse backgrounding is represented by the properties listed in the right column. Central to TH is the concept of markedness. Hopper and Thompson, drawing on massive evidence from an array of typologically diverse languages, concluded that the discourse background clauses, which are low in transitivity, are usually more marked, morphosyntactically or semantically, than their high-transitivity counterparts. Among other things, a maximally unmarked use of the progressive aspect, atelic in Hopper and Thompson’s terminology, is to function as discourse backgrouding (low in transitivity); it is prototypically intransitive (one participant); it is semantically stative (non-action) and durative (non-punctual); it does not correlate with a change of state (not affected) and its object, if any, does not have a specified quantity or definite referent. Transitivity Hypothesis has close affinity with the basic tenets of Optimality Theory in two principled ways. On the one hand, TH provides us with a crossmodular account of the factors highly correlative with each other. As OT is committed to connectionism, evaluating constraints of different families in parallel, it provides us

31

Individuation is another notion that merits further clarification. According to Hopper and Thompson (1980), this term refers “both to the distinctness of the patient from the agent and to its dictinctness from its own background” (p.253). The highly individuated referents have the properties of animacy, being definite in quantity and reference, properties shared by Verkuyl’s [SQA], i.e., Specified Quantity of A, as was discussed in §2.1.4.1. A is the abbreviation for Agent. 32 O is the abbreviation for Object. 73

Chapter 4 An OT Analysis of the Constraints Interaction on the Progressive

with a sound explanation for the interaction of constraints at different linguistic levels (Haji-Abdolhosseini, 2005; Prince & Smolensky, 1997; Smolensky, 2000). On the other hand, transitivity and all the parameters contributing to it are all implicational and gradient by nature. These properties make it possible to incorporate the scales in transitivity hierarchy with the rankings of constraints in Optimality Theory. However, these correlative properties of the Transitivity Hypothesis are not readily translatable into the OT rankings of constraints. As has been discussed, what is exactly revealed in Transitivity Hierarchy is nothing more than the maximally unmarked alignment of the correlated properties contributing to the transitivity scale. Little is known yet, however, about the consequence of the combination of an unmarked property with a marked one, say, the interaction of a high-transitivity property with a low-transitivity one. In addition, these parameters are of equal weighting in Transitivity Hypothesis, but an OT analysis requires that the constraints be ordered, in a formally explicit manner, along a rank. Thus, we need to explore further the interactive patterns of the factors associated with tense-aspect markings before incorporating them into OT framework. 4.2.2 Discourse Representation Theory (DRT) and Aspect Coercion Recall that the progressive aspect marker was analyzed in §3.3.3 as fulfilling the functions of the [PROG] operator, which works to stativize a dynamic situation, to cancel the endpoint of an event, or to expand a punctual predicate. In addition, these functions are assumed to perform selectively on different situation types. Recall also that stativizing has greater priority over the other two functions. It is further proposed that the inherent temporal properties of the verb are altered by the [PROG] operator to meet the requirement of marking foregrounding/backgrounding distinction in a discourse. This proposal was presented, in a formally explicit manner, by Kamp and Reyle (1993) in their Discourse Representation Theory (DRT henceforth). In DRT, it is the discourse representation structure (DRS), but not individual sentences, that is

74

Chapter 4 An OT Analysis of the Constraints Interaction on the Progressive

semantically interpreted. According to DRT, the grammatical aspect operates to map one domain of eventualities to another. States, for example, can be shifted by grammatical aspects as events and to perform the functions thereof in the discourse. Progressive aspect, as analyzed above in §3.3.2.1, functions to turn an event into a state. Furthermore, DRS also establishes temporal relations like precedence and inclusion between sentences in a tensed discourse. If a sentence describes an event, its discourse reference follows the reference time of the preceding event. If it describes a state, it is included in the reference time established. Take the following discourse fragment (59), with its DRS represented as (60) (see next page)

(59) A man entered the pub. He was whistling a song. Bill served him a beer.

As shown in (60), the DRS for this discourse contains the event (e) of a man’s entering the pub, the state (s) of his being whistling a song, and another event (e’) of Bill’s serving him a beer. It specifies that s includes e, and e precedes e’. Noteworthy in (60) are three important points relevant to our understanding of the role aspects play in discourse. First, the progressive aspect maps a dynamic event (e) He whistled a song onto the state (s) of that event being in progress and integrates it into the discourse, while the simple past simply presents events (e and e’). Second, the temporal relations between events are precedence and succession (marked as < in DRS), but the relations between event and states are inclusion and overlapping (marked respectively as ○ and ⊆ in DRS). This ontological distinction between state sentences and event sentences, therefore, has important consequences on the discourse temporal structure, as summarized by Kamp and Reyle (1993): [S]tate-describing sentences tend to connect with the discourse of which they are part in a way other than sentences which describe events…[A] state-describing sentence typically situates the described state as surrounding the contextually salient “reference” time; in contrast, there exists a strong tendency to interpret event-sentences as describing events that follow the reference…Thus, whatever the distinction between events and states may be, correct interpretation of tensed discourse presupposes

75

Chapter 4 An OT Analysis of the Constraints Interaction on the Progressive

at a minimum the ability to tell state-describing and event-describing sentences apart. (Kamp & Reyle, 1993: 509-510)

(60) DRS for (59) 33

e t n x y t’ s u w e’ t” v u’ e⊆t t
x enter y s ○ t’ t’
s: `

PROG

e: u whistle w e’ ⊆ t” t”
e’:

v serve x r

According to DRT analysis, the state sentences and the event sentences have a fine division of labor. The function of the tense-aspect morphology is to mediate between the semantic constraints and discourse requirement when there is a conflict between them. This function of the aspect marker is better captured by de Swart’s recent proposal of “aspect coercion”, which “is governed by implicit contextual reinterpretation mechanisms triggered by the need to resolve aspectual conflicts” (de Swart, 1998: 360). Put differently, in order to satisfy the higher-level discourse requirement, aspect coercion operates to alter the inherent temporal properties of the

33

Listed on the top of the DRS boxes are discourse referents, or the universe of the DRS, where e = event, t = reference time, n = now, s = state, x,y,u,v,u are all nominal referents. Below the universe are displayed a set of DRS conditions. ○, ⊆ and < are all signs for temporal relations. “○” stands for overlapping,” ⊆” means temporal inclusion, and “<” means temporal precedence. 76

Chapter 4 An OT Analysis of the Constraints Interaction on the Progressive

verbs, for example, from state to event or vice versa. In this sense, coercion can be understood as an overarching term covering the three functions of the progressive operator [PROG], i.e., stativizer, expander, and endpoint canceller. 34 According to DRT and aspect coercion theory, the truth condition of the discourse as a whole must be achieved in the first place, at the cost of truth conditions of verbal semantics. The notion of aspect coercion is helpful in understanding the semantic nuances of aspect that would otherwise seem inexplicable. A typical case in point is the progressive stative like Susan is liking this play. As both progressive aspect and the stative predicate are able to fulfill the shared function of discourse backgrounding, there seems to be no need for the statives to be progressively marked. According to de Swart (1998), the stative sentence is first coerced from stative to dynamic, then back to stative again by a hidden coercive operator. This additional coercive operation from stative to dynamic can explain why a progressive stative is more ‘dynamic’ than its non-progressive counterpart. The vividness of the former is manifested by a comparison of the internal temporal structure between the two. (61) a. Susan likes the play. b. [PRES [Susan like the play]] (62) a. Susan is liking the play. b. [PRES [PROG [Csd [Susan like this play]]]] 35 Note that the roles played by Transitivity Hypothesis and DRT are different in our theorizing of the tense-aspect morphology. The former surveyed a broad, if not exhaustive, spectrum of variables contributing to discourse grounding among typologically diverse languages, whereas the latter provides a powerful formal apparatus accounting for the coercive functions of the aspect markers in discourse. These two theories form the basis for a more comprehensive theory of structural description and structural change, two crucial steps for OT analysis of the constraints on the structure under investigation.

34 35

See §3.3.3 for a more detailed discussion of this point. Csd means the coercion from stative to dynamic (de Swart, 1998). 77

Chapter 4 An OT Analysis of the Constraints Interaction on the Progressive

4.3 An OT Analysis of the Constraints on Aspect Marking and Their Ranking 4.3.1. Typology of Constraints on Aspect Marking in English Figure 8 visually presents the TH and DRT insights into the constraints on aspect marking at various linguistic levels and incorporates these constraints with OTcompatible faithfulness and markedness conditions. output

ASPECTUALIZAER

[-PROG]

+FAITH

+FAITH FOREGROUNDING

+M ARKED

-M ARKED

input

STATE

EVENT

-M ARKED

+M ARKED BACKGROUNDING

output

+FAITH

+FAITH [-PROG]

[+PROG]

Figure 7 Conflicting Constraints on Aspect Marking in English

As demonstrated in TH, high transitivity properties correlate with discourse foregrounding and low transitiviity properties correlate with discourse backgrounding. In the most faithful scenario, the verb is marked with the default aspectual form, and there is no need of the intervention of the [PROG] operator or aspectualizers 36 . However, when the temporal semantics of the input and output collide, as in the case of an event verb 37 performing the discourse function of backgrounding, the marked aspectual form (e.g., the progressive aspect) is employed to resolve this conflict, in

36

English has no grammaticalized form to mark the coercion from state to event (de Swart, 1998). Such aspectualizers like stop, finish, begin are used to denote the inchoativity of the states so that the statives obtain event-like features and perform the discourse functions of the event. 37 For sake of simplicity, The semantic nuances between subtypes of event verbs, i.e., activities, accomplishments, semelfactives and achievements are neglected and a dichotomous categorization of verbs into stative verbs and event verbs is adopted. 78

Chapter 4 An OT Analysis of the Constraints Interaction on the Progressive

order to satisfy the faithfulness condition of maintaining the correspondence between aspectual morphology and discourse function. Also shown in Figure 8 are two types of constraints on aspect marking and how they interact: the markedness constraints and the faithful constraints, which are respectively shown as [FAITH] and [MARKED]. Following OT, faithfulness constraints on aspect marking warrant the correspondence between the temporal interpretation of the tense-aspect morphology and discourse functions of grounding, whereas the markedness constraints evaluate whether an aspectually marked verb expresses its default temporal property or not. Reading Figure 8 from the middle to the bottom, if the input verb is an event verb and if it is required by the discourse to perform the function of discourse backgrounding, there occurs a conflict between the temporal semantics of the verb and its discourse function. In order for the marked scenario to be faithful in the output, the verb is inflected with a progressive marker. Similarly, if a stative verb is required to perform the discourse function of foregrounding (reading the Figure from middle to top), an aspectualizer is employed to make the marked situation faithful to the output condition. In the other two unmarked scenarios, i.e., the event verb performing the function of discourse foregrounding, and stative verb performing the function of discourse backgrounding, there is no need to change the aspectual forms of the verb in order for the output to be faithful. In the next two sections, we will identify what faithfulness and markedness constraints tense-aspect marking in English is subject to and how the identified constraints interact and compete for the optimal aspectual choice in the output. 4.3.1.1. Faithfulness Constraints on Aspect Marking An event verb, when marked with the progressive aspect, becomes a statedescribing verb, which faithfully match the requirement of discourse backgrounding. This match between the temporal interpretation of tense-aspect morphology and the temporal requirement for discourse grounding is translatable to a faithfulness

79

Chapter 4 An OT Analysis of the Constraints Interaction on the Progressive

constraint belonging to the IDENT family constraints in OT, which require that the input value be preserved in the output (McCarthy & Prince, 1995). (63) Ident(TF): Where [TF] is some temporal feature: if a tense-aspect form is specified [αTF] in the input, it may not be [-αTF] in the output.

This constraint says that the temporal interpretation of the aspectually marked verb in the input should be congruous with temporal requirement of the discourse in the output. For instance, the state-describing sentences, including events with progressive markers and stative predicates with no aspect marker, are not dynamic, not telic and not punctual, and they match the temporal properties of the sentences functioning as discourse backgrounding. They are, therefore, faithful candidates for backgrounding sentences and unfaithful candidates for foregrounding sentences in the discourse. 4.3.1.2 Markedness Constraints on Aspect Marking Figure 8 also teases apart the marked uses of the aspect from the default uses. The unmarked configuration for the constraints requires that the default values of verbal semantics stay intact in the output, which amounts to saying that no change occurs in semantic interpretation of the verb. In other words, in the maximally unmarked scenario, the sentence resists semantic coercions. This markedness condition can be tentatively expressed by an OT-compatible markedness constraint, NO COERCE (NOCOER): (64) NoCoer: Do not coerce the shift of the temporal semantics inherent in verbal predicate.

Coercion, as demonstrated by de Swart (1998) and Vlach (1991), operates sequentially and iteratively. For instance, the coercion from an achievement situation into a stative as a result of the [PROG] operator requires that the functions of

80

Chapter 4 An OT Analysis of the Constraints Interaction on the Progressive

expander and endpoint canceller precede the function of stativizer. 38 Following the convention of OT, each violation of the NOCOER constraints incurs one violation mark in the constraint evaluation tableau. Accordingly, when a candidate is coerced from achievement to stative, the NOCOER constraint is violated three times, which respectively functions to cancel the endpoint, to expand, and to stative the dynamic situation. It should be noted that these two constraints, IDENT(TF) and NOCOER, were they correctly ranked, could produce optimal output with proper aspect markers. However, the correct tense marking remains unwarranted. The absence of constraints on the tense marking has two undesirable consequences. First, it overgenerates sentences with the correct aspectual interpretation devoid of proper tense markers. Second, the soundness of the constraints ranking is called into question because, in human languages, it is common to see that aspect markers are interwoven with tense markers in intricate ways (Bache, 1985; Bybee et al., 1994; Comrie, 1976; Dahl, 1985, 1999; Smith, 1997). In English, for example, the perfective aspect marker and the simple past tense marker are conflated into one morphological form. In order to capture the effect of tense on aspect marking, it is necessary to have a third constraint on tense marking, i.e., MARKT, in addition to IDENT(TF) and NOCOER. (65) MarkT: Every sentence must be properly located on the time axis by a tense marker.

(65) means that every sentence must be properly marked with a tense, which locates a sentence in the past, present or future. This constraint is violated if a mismatch arises between the tense marker and the time referent triggered by some contextual or pragmatic factors, as is in the case of simple present describing historical past or simple past expressing politeness (Quirk et al., 1985).

38

See also §3.3.3 for this point. 81

Chapter 4 An OT Analysis of the Constraints Interaction on the Progressive

4.3.2 Constraint Ranking of Aspect Marking in English As has been mentioned, the three constraints in (63), (64) and (65) are inherently conflicting. The markedness constraints, i.e., NOCOER and MARKT, require that the output be economical and unmarked, whereas the faithfulness constraint IDENT[TF] requires that temporal semantics of the output be faithfully copied from the input. It is assumed in this study that when a language learner selects one aspectual form rather than other alternatives, this form is the optimal choice of the aspect marker as a response to the competition of these constraints along a given ranking. Therefore, it is essential to develop a proper ranking of these constraints in order to evaluate all the possible outputs of these constraints and correctly predict which candidate output survives the competition and becomes the optimal choice. Logically, there are six possible rankings of the three constraints just proposed on aspect marking, as shown in (66), which have different consequences for the wellformedness of the output. (66) Ranking typology of constraints on aspect marking a. MarkT >> Ident[TF] >>NoCoer b. MarkT >> NoCoer>> Ident[TF] c. NoCoer >> Ident[TF]>> MarkT d. NoCoer >>MarkT >> Ident[TF] e. Ident[TF] >>NoCoer>> MarkT f. Ident[TF] >> MarkT >> NoCoer

In (66a) and (66b), where MARKT ranks highest among the constraints, the output would be a sentence with correct temporal deixis devoid of proper aspectual interpretation. In (66c) and (66d), the aspectual information is expressed by the semantics of the verb and only the unmarked forms are favored. It is obvious that the rankings in (66a-d) do not conform to the constraints ranking in the English language since the aspectual forms in these rankings are determined by inherent semantics of

82

Chapter 4 An OT Analysis of the Constraints Interaction on the Progressive

the verbs, not by the discourse requirement for the temporal features of the verbs. In (66e) and (66f), the correspondence of the verbal semantics and discourse grounding tops the rankings and becomes the predominant constraint for aspectual marking. In the light of the Transitivity Hypothesis, the ultimate purpose of aspect marking is not local-semantic, but to make discourse contrasts between foregrounding and backgrounding events (Hopper, 1982; Glasbey, 1998). Similarly, it was proposed in DRT that the contrast between state-describing sentences and event-describing sentences is primarily motivated by discourse requirement for the semantic input of the verbs. These theories lend support to our assumption that the discourse constraint expressed by IDENT(TF) ranks higher than the other constraints in tense-aspect marking. If the assumption that the constraint IDENT(TF) tops the aspect marking ranking scale is correct, then which constraint occupies the second highest position, NOCOER or MARKT? Or to frame the question differently, are verbs tense-marked in order to coerce an alternative temporal interpretation, or are they inflected with tense markers as a result of the semantic coercion? Which is the motivating force between the two? Tense morphology sometimes alters aspectual interpretation of the verbs. A typical case in point is the English simple past tense that describes completed actions in reference to the present time. Nonetheless, a conflict arises when the simple past marks stative verbs or homogeneous process verbs without inherent endpoint. This conflict is manifest in the following continuation test sentence (Smith, 1997). (67) a. He was ill and he may still be ill. b. He ran along the street and he may still be running. c. *He built a house and he may still be building it. d. *He died and he may still be dying.

The verb predicates with inherent endpoint like (67c) and (67d), when marked with the simple past, militate against a present continuation test. In contrast, stative or

83

Chapter 4 An OT Analysis of the Constraints Interaction on the Progressive

activity verbs lack of an inherent endpoint, as shown in (67a) and 67b], are still open to incomplete interpretations even when marked with the simple past. The sentences in (67) shows that the temporal interpretation of tense and that of the inherent temporal features of verbs compete with each other, with the semantics derived from tense markers prevailing over the inherent temporal properties of verbs. Based on this analysis, we can tentatively rank the MarkT constraint higher than the NOCOER constraint and thus come up with the following ranking of the constraints underlying aspect choice in English. (68) Constraint Ranking on English Aspect Marking IDENT(TF)>>MARKT>>NOCOER

This is a hypothetical ranking of the constraints on aspect marking. Its soundness needs to be tested before further hypotheses are generated on the aspect acquisition based on Optimality Theory. As no linguistic theory available gives an explicit account of the ranking of these constraints, it is necessary to test it against various possible scenarios, to see whether optimal output can be yielded, as will be done immediately below. 4.3.3 Testing the Ranking of Constraints on Aspect Marking Table 4 below is a simplified version of Figure 8, which specifies the optimal output of aspectual forms in various combinatorial scenarios of the semantic and discourse input. It can be seen from the table that the state-backgrounding and eventforegrounding combinations are unmarked and natural, and so there is no need of the intervention of the [PROG] operator. In contrast, the state-foregrounding and eventbackgrounding scenarios are inherently conflicting, which calls for the external coercive forces, like aspectualizer or [PROG] operator. Each of these four scenarios will be examined below against linguistic evidence in English.

84

Chapter 4 An OT Analysis of the Constraints Interaction on the Progressive

Table 4 Aspect Marking Scenarios under Different Semantic-Discourse Conditions

Event (ACT, ACC, ACH, SMF) State

Backgrounding (BRG) [+PROG] [-PROG]

Foregrounding(FGR) [-PROG] Aspectualizer

4.3.3.1 Event-backgrounding Combination Event-denoting verbs are inherently dynamic, but backgounding function requires that the situation be a stative one. Thus there is a tension within the eventbackgrounding combination, which cannot be satisfactorily resolved without the semantic coercion brought about by the [PROG] operator. Tableaux (69) and (70) respectively evaluate the candidate outputs when the event verbs function as discourse backgrounding, and show why the progressive aspect is the optimal selection in these scenarios. (69) Achievement verb functioning as discourse backgrounding 39 [reach (x,y), x = he, y = the top of the tower, tense = past , grounding = BGR]

IDENT(TF)

MARKT

He was reaching the top of the tower

***

He reached the top of the tower

*！

He reaches the top of the tower

*！

He is reaching the top of the tower

NOCOER

*!

The candidate He was reaching the top of the tower violates the constraint NOCOER three times in that the achievement verb reach is inherently punctual, telic and dynamic, whereas the progressive aspect has the temporal interpretation of durative, incomplete and stative. But this sentence does not violate the higher ranked IDENT(TF) and lower ranked MARK(T) constraints. In contrast, its non-progressive

39 The clause indicating the grounding status is not given in the examples for the sake of expository simplicity. Instead, we use abbreviations like FGR (foregrounding) and BGR (backgrounding) to indicate the grounding status of the clause.

85

Chapter 4 An OT Analysis of the Constraints Interaction on the Progressive

counterparts violate the constraint IDENT(TF) which dominates all the other constraints in the rank (69). Tableau (70) shows that the candidate with the progressive marker is the optimal selection because the constraint it violates ranks lower than IDENT(TF), a constraint that the other two candidates violate.

(70) Activity verb functioning as discourse backgrounding [run along the street (x), x = he,

IDENT(TF)

tense = past , grounding = BGR]

MARKT

He was running along the street

NOCOER *

He ran along the street

*！

He is running along the street

*!

He runs along the street

*！

In this scenario, as the candidate output He was running along the street violates the fewest and lowest ranked constraints, it naturally becomes the optimal selection. In contrast, He ran along the street also violates NOCOER constraint in that the perfective aspect has the temporal interpretation of completeness of action, a feature that the activity predicate run along the street does not fit. However, as the aspectual interpretation of the perfective aspect does not match the temporal requirement for the discourse backgrounding, it does not survive the higher ranked constraint IDENT(TF), hence, a lost competitor in the evaluation. 4.3.3.2 State-backgrounding Combination Stative verbs and discourse backgrounding constitute a harmonious combination, and, according to DRT, there is no need for the intervention of additional operation of aspect morphology. In this scenario, semantic coercion does not occur. Tableau (71) shows that this is indeed the case.

(71)

Stative functioning as discourse backgrounding

86

Chapter 4 An OT Analysis of the Constraints Interaction on the Progressive

[like (x, y), x = Susan, y = the play,

IDENT(TF)

tense = present, grounding = BGR]

MARKT

NOCOER

Susan likes the play Susan is liking the play

****

Susan was liking the play

*!

Susan liked the play

*!

In all these four scenarios, as the verbs are either progressive or stative, none of them violate the IDENT(TF) constraint. The non-progressive sentence wins the competition because it survives all the constraints. The progressive stative sentence Susan is liking the play violates the constraint NOCOER four times because, according to de Swart (1998), there is a hidden coercion process involved wherein the stative predicate is at first coerced into a dynamic one and then back into a stative one.40 This double coercion can explain why this sentence has a dynamic vividness in it and why it is not an optimal choice for discourse backgrounding. 4.3.3.3 Event-foregrounding Combination Like the state-backgrounding combination, the combination of event with foregrounding is natural. However, the naturalness of combination varies with the situation type denoted by the verb predicate. For the telic and punctual event situations, the need for semantic coercion does not arise because the inherent temporal properties of these situation types are in alignment with discourse temporal requirement for discourse foregrounding, as shown in the example of (72).

(72) Accomplishments for discourse foregrounding

[destroy (x, y), x = he, y = a house,

40

IDENT(TF)

See § 4.2.2 for a more detailed discussion on this point. 87

MARKT

NOCOER

Chapter 4 An OT Analysis of the Constraints Interaction on the Progressive

tense = past , grounding = FGR] He destroyed a house He was destroying a house

*！

He destroys a house

*!

He is destroying a house

*！

For atelic event situations like activities and semelfactives, however, the combination is less natural and more marked, as shown in (73) below, because the sentences fit for discourse foregrounding should have well-defined endpoints in order that the actions take place successively, without overlapping with each other.

(73) Activities for discourse foregrounding (in the past tense)

[run(x), x=he, tense=past,

IDENT(TF)

grounding = FGR] He ran along the street

MARKT

NOCOER **

He was running along the street

*！

He runs along the street

*！

He is running along the street

*！

Note that the grounding status of the sentence He runs along the street in (73) is questionable. Unlike the sentence He ran along the street, which has an arbitrary endpoint coerced by the perfective aspect, the sentence in the simple present tense does not terminate the situation. It is, therefore, not an optimal candidate for discourse foregrounding, which requires that the situation have a well-defined endpoint. This conflict between the discourse requirement and the temporal interpretation of the tense-aspect morphology can be resolved if an external coercion such as an

88

Chapter 4 An OT Analysis of the Constraints Interaction on the Progressive

aspectualizer, e.g., stop, begin, start, and finish, is introduced into the predicate, as shown in (74).

(74) Activities for discourse foregrounding (in the present tense) [[run(x), x = he, tense = present, grounding = FGR]]

IDENT(TF) MARKT

He ran along the street

NOCOER

*!

He was running along the street

*

He runs along the street

*

He is running along the street

*

He stops (starts) running along the street

*

Tableau (74) shows that the aspectualizer to stop or to start coerces an endpoint to the activity and thus violates the NOCOER constraint. In so doing, however, the higher ranked constraint IDENT(TF) is satisfied. This feature licenses He stops (starts) running along the street as a better choice for discourse foregrounding than He runs along the street. 4.3.3.4 States-foregrounding Combination The state-foregrounding pairing is an unnatural one, as there is an inherent conflict between the temporal properties of the statives and the discourse requirement for temporality. This conflict, in most cases, is irresolvable without the semantic coercion of aspect operators or aspectualizers. Consider for instance the following example in (75). The aspectualizer start coerces the temporal interpretation of the stative verb to feel into a dynamic and telic situation. But this coercion makes the predicate a more optimal candidate for discourse foregrounding. That is, the stative verb contributes to moving on the storyline when combined with an aspectualizer. Otherwise, the foregrounding interpretation would not be easy to recognize.

89

Chapter 4 An OT Analysis of the Constraints Interaction on the Progressive

(75) Statives for discourse foregrounding (in the present tense)

[[be happy(x), x = Mary, tense = past , discourse = FGR]]

IDENT(TF)

She was very happy

*！

She was being very happy

*！

She started to feel very happy.

MARKT

NOCOER

**

To recapitulate, we have tested the validity of the constraint ranking for aspect marking in English language as described in (68) by evaluating all four possible scenarios of the semantic-discourse combinations to see whether they yield the optimal temporal morphology. The outcomes of our tests indicate that the ranking in (68) is a plausible one in that the optimal outputs it yields in each scenario are theoretically derivable from linguistic proposals like TH and DRT, and empirically congruous to the observational data of tense-aspect marking in English language. This ranking, therefore, is a good point of departure for us to make further hypotheses on the emergence and development of aspect marking in L2 aspect acquisition in general, and progressive aspect in particular. 4.4 Summary In this chapter, we have demonstrated that the constraints on aspect marking at different linguistic levels are inherently interactive and conflicting. Drawing on the proposals of Transitivity Hypothesis and Discourse Representation Theory, we have identified three OT-compatible constraints on aspect marking in English. The faithfulness constraint IDENT(TF) requires that the temporal interpretation of the tense-aspect morphology be identical to the temporal requirement for discourse grounding status. On the other hand, the two markedness constraints, MARKT and NOCOER require that the temporal morphology and inherent temporal semantics of the verbs preserve their default and unmarked values. These constraints are ranked in the 90

Chapter 4 An OT Analysis of the Constraints Interaction on the Progressive

order of IDENT(TF) >> MARKT >> NOCOER and this ranking is put to test against the full set of combinations of temporal semantics and discourse grounding. In all four possible scenarios, the optimal candidates yielded by this ranking are congruous to the use of tense-aspect morphology in English, hence the justification of its soundness. The dominance of discourse constraints over semantic ones further supports our assumption that English tense-aspect marking is primarily motivated by discourse requirement for the temporal interpretation of the predicate. So far we have answered the first key research question in §1.4. In the next chapter, we will examine the implications of an OT account of tense-aspect marking on its L2 acquisition.

91

Chapter 5 L2 Acquisition of the Progressive Aspect: Towards an OT Account In the preceding two chapters, we have analyzed the linguistic constraints on aspect marking, and we have also tried to explain, within the OT framework, how these constraints interact with each other, and how an optimal output is yielded after the candidates are evaluated against a ranking of competing constraints on tenseaspect marking. Since OT is a theory of language and a theory of language of acquisition as well, it is assumed that the constraints on aspect marking and their ranking as analyzed in the last chapter will have some significant implications for L2 acquisition of aspect marking. The present chapter addresses the possibility of extending the OT account of aspect marking to the filed of L2 aspect acquisition, as a response to the second key research question stated in §1.4. That is, what could be the implications of an OT account of the constraints and their interactive patterns so analyzed for the acquisition of L2 progressive aspect by Chinese-speaking learners? The answer to this question would prepare the grounds for making further hypotheses on the emergence and development of L2 progressive aspect. This chapter is structured as follows. First, we present our OT account of L2 tense-aspect acquisition. In this part, our focus is on a principled account of the L2 development from the initial ranking to the ultimate ranking of the constraints on tense-aspect marking. Second, based on this OT account of aspect acquisition, we propose our own hypotheses on the acquisition of the progressive aspect marking by Chinese-speaking learners of L2 English. 5.1 An OT Account of L2 Tense-aspect Acquisition As mentioned in §4.1.4, OT views language learning as a process of determining the target ranking relation over a given set of constraints, starting from

Chapter 5 L2 Acquisition of the Progressive Aspect: Towards an OT Account

their initial ranking. Also in Chapter 4, we have formulated the dominant order of the constraints on tense-aspect marking in English along the ranking of IDENT(TF)>> MARKT>> NOCOER, based on the insights obtained from Tansitivity Hypothesis and Discourse Representation Theory. This is, of course, the ultimate ranking order of constraints that L2 English learners are expected to attain. If, however, OT is to be adopted as an overarching linguistic framework for explaining the development of language competence in general, it entails that L2 acquisition of tense-aspect marking should be seen as a gradual process of constraint reranking from the initial state to the target-like ranking. But two questions remain. First, what is the L2 learner’s initial ranking of constraints on aspect marking? Second, what is the leading principle underlying the gradual development from initial ranking to the ultimate ranking? Both questions are approachable within the OT account of language acquisition. More specifically, the answer to the first question lies in the OT assumption of [M>>F] initial ranking of constraints, and the second, in the learning algorithms proposed in the OT framework. 5.1.1 Initial Ranking of Constraints on Tense-aspect marking Recall that in OT, markedness constraints and faithfulness constraints are inherently conflicting because they have competing demands. The faithfulness constraints require that the features of the input be preserved faithfully in the output, while the markedness constraints demand that the more marked structure should not appear in the output. Depending on the relative ranking of faithfulness and conflicting markedness constraints, the input features may or may not surface. If faithfulness dominates markedness, i.e., [F>>M], then F-dependent features surface in the output. If, conversely, markedness dominates faithfulness, i.e., [M>>F], then F-dependent contrasts are neutralized in the output, in favor of the M-respecting structures. That is, the unmarked structures are favored in the beginning language learner’s output. In OT, it has been proposed, and empirically confirmed as well, that language learners start

93

Chapter 5 L2 Acquisition of the Progressive Aspect: Towards an OT Account

with an [M>>F] initial state of the structure to be learned (e.g., Gnanadesikan, 1995; Smolensky et al., 2004). Regarding the typology of constraints on aspect marking, we have identified one faithfulness constraint, IDENT(TF) and two markedness constraints, MARKT and NOCOER. If we accept the OT hypothesis of the [M>>F] initial ranking, then the learners’ initial state of the tense-aspect constraints ranking order can be tentatively formulated as follows:

(76) NoCoer, MarkT>>Ident(TF)

This ranking indicates that in the learner’s initial state of tense-aspect system, the two markedness constraints outrank the faithfulness constraint. The outcome of this ranking is that the tense-aspect markers are used to mark the temporal semantics or the temporal deixis, not the discourse grounding status. Note that we do not define the ranking relation between the two markedness constraints, MARKT and NOCOER. However, as the subjects in our study are Chinesespeaking English learners, it is reasonable to assume that the initial state of L2 English tense-aspect marking is affected by the constraints on Chinese temporal system, which, as has been mentioned, is typically devoid of tense markers but rich in aspectual expressions (Chao, 1968; Dai, 1997; Wang, 1943). This absence of tense markers is assumed to block their dominance in the initial ranking of the constraints. Based on this cross-linguistic observation, (76) can be rewritten as (77)

(77) NoCoer >> IDENT(TF), MarkT,

This ranking says that the constraints on tense markers are initially inert in the learner’s tense-aspect system and their temporal morphology is primarily determined by temporal semantics of the verb. As this ranking is based on the OT assumption of

94

Chapter 5 L2 Acquisition of the Progressive Aspect: Towards an OT Account

[M>>F] initial state and the cross-linguistic observation between the tense-aspect marking in Chinese and English, it serves as the working hypothesis of the initial state of L2 English tense-aspect system acquisition by Chinese-speaking learners. The question that logically follows is, in what way does the learners’ tenseaspect system develop from the initial NOCOER>> IDENT(TF), MARK(T) ranking to the ultimate and target-like IDENT(TF)>>MARKT>>NOCOER ranking? The answer to this question can be sought from the learning algorithms proposed in OT. 5.1.2 Learning Algorithms in OT Two influential proposals concerning the constraint reranking were proposed in OT: Constraint Demotion (CD) and Gradual Learning Algorithm (GLA) (Boersma & Hayes, 2001; Ma, 2003; Tesar & Smolensky, 1998a, 1998b, 2000). While both algorithms agree that language learning is error-driven and the initial ranking is altered when the input data conflict with the current ranking hypothesis, there are marked differences between these two proposals as to how this reranking takes place. The basic idea behind CD is that the learner gradually approaches the target ranking of the constraints by demoting constraints that are observed by the learner to be violated in the target output form. When a learner observes a target language form, s/he computes the optimal output that its current constraint ranking would produce. If the learner’s predicted output matches the actual form, no action takes place; but when the learner’s predicted output fails to match the target form, constraint demotion occurs. In this sense, CD views language acquisition as an error-driven process wherein the constraint demotion process is repeated until the learner’s constraint ranking orders match the target form (Tesar & Smolensky, 1998a, 2000). Like CD, GLA also assumes that language learning is error-driven and constraint reranking takes place where there is a mismatch between the input and the learner’s current constraints ranking hypotheses. What is different about GLA is that it assumes that the constraints are ranked on a continuous scale. Rather than a whole-sale reranking as is assumed in CD, GLA executes only small violations to the constraints’

95

Chapter 5 L2 Acquisition of the Progressive Aspect: Towards an OT Account

locations along the scale, with higher-ranked constraints slightly demoted and lowerranked ones slightly promoted (Boersma & Hayes, 2001). Figure 8 is a schematic illustration of two constraints (C1 and C2) ranked on the continuous scale.

Categorical ranking C1

Overlapping ranking

C2

C1

C2

Figure 8 Categorical and Overlapping Ranking of Constraints along a Continuous Scale

As can be seen from the figure, the two constraints act as if they were associated with ranges of values, instead of single points. If the ranges of the two constraints do not overlap, the ranking scale shows ordinary categorical ranking, i.e., C1>>C2. But if the ranges overlap, ranking becomes variable and both the common ranking C1>>C2 and the rare ranking C2>>C1 occur in the learner’s language. Obviously, optionality and variation of the ranking relations occur where two constraints have overlapping ranges (Boersma & Hayes, 2001). In the GLA, a learning step triggered by an observed error in the existing ranking does not lead to radical change in the constraint ranks. Rather, a slight adjustment of the constraint ranks is made, promoting the constraints violated by the erroneous winner, and demoting the constraints of the observed output. GLA is an illuminating learning algorithm in accounting for the L2 tense-aspect acquisition. According to the GLA assumption, the constraints reranking from the

96

Chapter 5 L2 Acquisition of the Progressive Aspect: Towards an OT Account

initial NOCOER>>IDENT(TF), MARK(T) ranking to the ultimate IDENT(TF)>> MARKT>> NOCOER ranking scale does not take place in a leap but involves a gradual learning process. This is highly probable because the constraint NOCOER involves different types of temporal semantic shifts and it operates iteratively and selectively (de Swart, 1998; see also §3.3.3 for this point). Take for instance the aspect coercion triggered by the progressive morphology. As discussed in § 3.3.2, aspect coercion is an umbrella term which involves, in the case of the progressive marking, altering the temporal semantics from dynamic to stative, from telic to atelic, and from punctual to durative. When a learner observes an input data involving the constraint NOCOER whose position in the rank is incongruous with his/her existing ranking hypothesis, s/he makes an attempt to demote the constraints involved to make a new hypothesis about the ranking order of the constraints. However, the strength of the higher ranked constraint NOCOER varies from one situation type to another. The activity verbs, for instance, are more susceptible to progressive marking than achievement. In the former case, there is only one violation of the constraint NOCOER (dynamic to stative), but in the latter case, three violations are involved (dynamic to stative, punctual to durative, telic to atelic). Following GLA algorithm, the present study assumes that the demotion of the constraint NOCOER is gradual and stepwise, depending on the number of violations involved. To wind up the discussion of this section, the present study assumes, following the OT [M>>F] assumption, that the markedness constraint NOCOER ranks higher than the faithfuness constraint IDENT(TF) in the initial ranking of constraints on tenseaspect marking. It is further assumed, following the Gradual Learning Algorithm, that the spread of initial ranking to the ultimate ranking involves a gradual demotion of the higher ranked NOCOER or promotion of the lower ranked IDENT(TF). These OTcompatible analyses pave the way for making testable predictions and working hypotheses on the L2 development of progressive aspect marking.

97

Chapter 5 L2 Acquisition of the Progressive Aspect: Towards an OT Account

5.2 Hypotheses on L2 Acquisition of English Tense-aspect System by Chinese EFL Learners From the preceding OT account of the constraints on aspect marking and its acquisition, we can envisage how an aspectual form develops in Chinese EFL learners: the tense-aspect morphology starts out from the most unmarked assumptions on aspect marking, which is then gradually neutralized by discourse constraints, until the ranking of the constraints in the learner has ultimately become native-like. This general prediction of the L2 developmental pattern of tense-aspect system can be decomposed into the following hypotheses, all of which will be empirically tested in the subsequent chapters. 5.2.1 Hypothesis on Aspect Coercion Verbs typically resist shifts of inherent temporal properties, as is defined by the constraint NOCOER. Since the constraint NOCOER can be violated iteratively, it is reasonable to assume, following Gradual Learning Algorism, that the demotion of this constraint does not take place in a leap, but gradually from one verb class to another. We can further assume that more violations of the constraint give rise to greater acquisitional difficulty. Based on these assumptions, we can formulate the following hypothesis with regard to the NOCOER constraint: Hypothesis 1 (H1): Aspect morphology involving more shifts of the inherent temporal features are more difficult to acquire than those with fewer violations for the learners of the same proficiency level. 5.2.2 Hypotheses on Constraint Reranking Recall that we have a faithfulness constraint on tense-aspect marking: IDENT(TF), which states that the temporal interpretation of the aspectually marked verb should be congruous with temporal requirement of the discourse. In particular, the perfective aspect performs the function of discourse foregrounding and imperfective aspect, including the progressive event verbs and statives in the simple forms, fulfills the

98

Chapter 5 L2 Acquisition of the Progressive Aspect: Towards an OT Account

function of discourse backgrounding. Recall also that in the initial stage, IDENT (TF) is dominated by the semantic constraint NOCOER, which gradually demotes as the learner’s L2 tense-aspect system develops. It can thus be predicted that the NOCOER constraint gradually slackens or softens with the growth of the participants’ L2 proficiency, as a result of the promotion of the IDENT(TF) constraint. Lower-level learners will find prototypical combinations of the aspect and situation types more readily acceptable and easier to interpret and produce than the less prototypical combinations. In contrast, higher-level learners were more likely to integrate use, processing, and interpretation of the progressive in the discourse, owing to the effect of IDENT(TF). Thus, two predictions can be made as follows. Hypothesis 2 (H2): The effect of NOCOER is greater among low level L2 learners than among advanced L2 learners. Hypothesis 3 (H3): The effect of IDENT(TF) is greater among advanced L2 learners than among low level L2 learners. 5.2.3 Hypotheses on Cross-linguistic Constraints As discussed in § 3.5, Chinese language is rich in aspectual expressions, with varying degrees of grammaticalization. Recall also that we conflated two most frequently used progressive marker zai- and –zhe into the same category, following Li & Thompson (1981) and Smith (1994). Our cross-linguistic investigation shows that English and Chinese manifest marked differences when the progressive correlates with statives and achievements. While Chinese imperfective marker –zhe can be felicitously affixed to the stage-level stative sentences, the English progressive statives sound unnatural. By contrast, the English progressive achievements describe a situation at its preliminary stage, yet their Chinese counterparts are unacceptable (See §3.5 for the detailed discussion). If it is true that first language is part of the initial state of second language acquisition (Slabakova, 2000b), we can assume that in the process of acquiring the English progressive aspect, the Chinese learners, in particular those at the initial stage, would overuse the progressive statives and underuse the

99

Chapter 5 L2 Acquisition of the Progressive Aspect: Towards an OT Account

progressive achievements. More particularly, the following two hypotheses can be generated. Hypothesis 4 (H4): Chinese learners of L2 English, at least those at their initial stage of learning, tend to accept and produce more progressive statives than native speakers of English do. Hypothesis 5 (H5): Chinese learners of L2 English, especially those at their initial stage of learning, are not inclined to accept and produce the progressive achievements, compared with the native speakers of English. 5.3 Variables Involved in Hypothesis Testing If we take a closer look at the five hypotheses formulated above, we would be able to extract the variables involved in testing them empirically. The first hypothesis is chiefly concerned with NOCOER constraint, and the other four hypotheses make predications about how L2 learners of different English proficiency levels would perform on two of the three important constraints on aspect marking. Therefore, there are four major variables involved in our study, i.e., learners’ level of English proficiency, NOCOER constraint, IDENT(TF) constraint, and the developmental stage of L2 tense-aspect marking. Recall that we have identified three constraints on tense aspect marking in §4.3.1, i.e., NOCOER, IDENT(TF) and MARKT. Since the focus of our study is on the emergence, development and interaction of the first two constraints, the role played by MARKT is neutralized as a constrant variable in all experimental tasks of this study whereby only one tense marker is invariably used

for each

instrument. Owing to the scope of the present study, the empirical investigation of the role played by MARKT in L2 tense-aspect study is reserved for further investigation.41

41 Yang and Huang (2004) examined the impact of the absence of tense markers on the acquisition of the aspect morphology.

100

Chapter 5 L2 Acquisition of the Progressive Aspect: Towards an OT Account

5.3.1 L2-English Proficiency Level Four out of the five hypotheses formulated above addresses the development of L2 progressive aspect among learners of different proficiency levels. This, obviously, involves between-group comparison of the performance between learners at different L2 proficiency levels. The L2 proficiency level of the learners, therefore, would be an important independent variable in our study. That is, each participant was assigned to one of the three L2 proficiency groups in the light of certain predetermined criteria, namely, the advanced level, the intermediate level, and the low level. Noteworthy is that our study also makes predictions in Hypotheses 4 and 5 about the cross-linguistic differences in L2-English progressive aspect acquisition. Testing these two hypotheses involves comparison between the performance of native English speakers and non-native English learners. Therefore, apart from the L2 English learners of different proficiency level, we also need to include native speakers of English language as controls in our experiments, so that a target norm of the ultimate constraint ranking can be established and the performance differences between native and non-native speakers of English can be compared. 5.3.2 NOCOER Constraint Four out of the five hypotheses formulated above pertain, directly or indirectly, to the NOCOER constraint, which is concerned the form-meaning mapping of tenseaspect morphology. Recall that the constraint NOCOER requires that the temporal semantics inherent in a verb not be altered or shifted by external coercion. However, this constraint is sometimes at odds with the aspectual interpretation of the progressive or perfective aspect, resulting in violations of the constraint. According to our analysis of the function of the [PROG] operator, the semantics of an English progressive predicate is construed as a happening that is stative, durative and incomplete. 42 As a consequence, a violation of NOCOER constraint inevitably occurs wherever a mismatch exists between the inherent temporal properties of the verb and

42

See §3.3.2 for this point. 101

Chapter 5 L2 Acquisition of the Progressive Aspect: Towards an OT Account

the semantic interpretation of the progressive. Table 5 demonstrates how many violations of NOCOER constraint can be incurred by the tension between lexical and grammatical aspect.

Table 5 Violations of NOCOER Constraint on Different Situation Types

STA ACT ACC ACH SMF

[-dynamic, -telic, -punctual] a [+dynamic, -telic, -punctual] [+dynamic, +telic, -punctual] [+dynamic, +telic, +punctual] [+dynamic, -telic, +punctual]

Progressive [-dynamic, -telic, -puctual] **** * ** *** **

Perfective (Simple past) [+dynamic, +telic, +puctual] *** ** * *

Note: One violation mark “*” indicates one violation of the NOCOER constraint. a. According to de Swart (1998), when a stative predicate is progressivized, it undergoes a hidden coercion from stative to event, then back to stative again. The four violation markers do not show the accurate number of violations involved. They are tantamount to saying that a progressive stative incurs more violations of NOCOER constraint than other situation types marked by the progressive aspect. See also §3.3.3 for a more detailed discussion on this point.

The effect of the NOCOER constraint, therefore, can be operationally measured by the number of constraint violations involved. H1, for example, predicts that more violations of the NOCOER constraints give rise to greater learning difficulty. This prediction will be validated if, for example, we can obtain evidence that shows the stative-progressive mapping, which involves four violations of the constraint, is more difficult to acquire than the achievement-progressive mapping, which involves three violations of the constraint. 5.3.3 IDENT (TF) Constraint In H3, we predict that the effect of IDENT(TF) is greater among higher-level L2 learners than among lower-level learners. IDENT(TF) constraint, therefore, becomes another important variable in tasks designed to test H3. Recall that IDENT(TF) constraint states that the tense-aspect marking and the discourse grounding status

102

Chapter 5 L2 Acquisition of the Progressive Aspect: Towards an OT Account

should be kept congruous with each other. The evidence for the effect of this constraint in experimental tasks lies in L2 learners’ ability to correctly map aspectual forms onto discourse grounding functions. It should be noted, however, that the way in which the discourse constraint is measured depends on the nature of elicitation task that learners are required to perform. In an interpretation task, for example, they will be asked to interprete the function played by a tense-aspect marker in the discourse, while in a production task, they will be required to use appropriate tense-aspect forms in the given discourse contexts. 5.3.4 Tense-aspect Marking The effect of this variable can be investigated in two possible ways. Either it is manipulated and the participants are required to respond under different experimental conditions, or other relevant variables are manipulated and then observation is made of whether the obligatory tense-aspect forms are produced. More specifically, when the former method is adopted, the tense-aspect marking is the dependent variable and, by manipulating variables like discourse grounding and verb’s situation types, the learners’ tense-aspect marking performance can be observed. By contrast, when the latter method is adopted, it is the variable of tense-aspect marking that is manipulated, thus enabling one to observe the effect of this manipulation on the learners’ judgment or interpretation of sentences of different situation types. Noteworthy is that the focus of our study is the English progressive aspect. In this study we will set up conditions to observe the interaction of the progressive aspect with other aforementioned variables. However, we also include in our study the non-progressive aspect marking, in particular the perfective aspect, in order to examine whether the generalizations formulated from our OT account of progressive aspect marking can be extended to other tense-aspect forms. 5.4 Summary In this chapter, based on the [M>>F] assumption about the initial ranking in OT and the cross-linguistic observation that Chinese language is devoid of tense markers 103

Chapter 5 L2 Acquisition of the Progressive Aspect: Towards an OT Account

but rich in aspectual markers, we assume that the NOCOER>>IDENT(TF), MARKT ranking order will be the initial ranking of the constraints on tense-aspect marking for Chinese-speaking L2-English learners. It is also assumed, following Gradual Learning Algorithm, that the constraint reranking involved in L2 tense-aspect acquisition will not take place in a leap, instead, it will be a gradual error-driven process wherein L2 learners start with the initial ranking, compare this ranking with the target form, and gradually demote the higher ranked constraints until their ranking matches the ultimate ranking of the target norm. On the basis of these two assumptions, five hypotheses are generated, respectively concerning the acquisition of NOCOER constraint, the constraint reranking as well as the cross-linguistic constraints. In order to test these hypotheses, we single out four chief variables for investigation, i.e., the L2 learners’ proficiency level, the NOCOER constraint pertaining to form-meaning mapping of tense-aspect marking, the IDENT(TF) constraint concerning form-function mapping of tense-aspect marking, and the tense-aspect marking per se, in particular, the progressive aspect marking. Investigating the interaction of these variables will contribute to our understanding of the emergence and development patterns of L2 tense-aspect marking in general, and of the progressive aspect marking in particular.

104

Chapter 6 Research Design, Instrumentation, Data Collection and Data Analysis

This chapter, together with the next one, attempts to provide the answer to the third key research questions, i.e., how might the L2 acquisition of the English progressive aspect be described? In order for this question to be properly addressed, an empirical investigation should be conducted to collect data that can adequately describe the emergence and development of the progressive aspect in L2 learners. However, some measures must be taken in the first place to warrant the reliability and validity of the collected data. Based on the above consideration, the primary concern of this chapter is a methodological one. More specifically, we will describe in detail the experimental design, the participant sampling, the data elicitation instruments, as well as the data collection procedures and the ways in which the collected data are prepared and coded. More particularly, in §6.1, the research design to be adopted in the present study is described. §6.2 presents a description of the participants in this study and how they were sampled and grouped. Following this is a detailed account of the instruments used in the study (§6.3). §6.4 presents the task administration procedures, and finally in §6.5 the scoring and coding of the participants’ performance on the elicitation tasks will be described. 6.1 Research Design Since the purpose of our study is to investigate and explain the development of the progressive aspect in L2 learners, the most ideal investigative approach would be a longitudinal one. However, owing to the constraints of time and resources on this researcher, a cross-sectional approach will be adopted, in which the results achieved at a single time from L2 learners at different levels of proficiency are taken as an

Chapter 6 Research Design, Instrumentation, Data Collection and Data Analysis

indicator of the development of the aspect marking over time and interpreted as such (Larsen-Freeman & Long, 1991). In §5.3, we have identified four variables involved in testing the hypotheses formulated in §5.2. In a cross-sectional research design, the roles played by these four variables are different, with L2 proficiency level as a between-group variable and the other three variables as within-group ones. By examining how the participants in different proficiency groups perform on the within-group variables, we will be able to observe and describe the patterns of the emergence and development of tense-aspect marking in L2 learners. 6.2 Participants Since the cross-sectional research design involves comparisons of the performance between participants of different proficiency levels, L2-English proficiency becomes the most important defining feature for sampling and grouping participants. Ninety Chinese-speaking learners of L2 English participated in the study. Among them were 30 Senior II middle school students, who were in their fifth year of classroom English learning, 30 university sophomores majoring in political science or biology, who had learned English for eight years at the time of investigation, and 30 English major seniors who were in their fourth year of intensive English learning at the English department in addition to six years of English learning in the middle school. These students were assigned into three L2-English proficiency groups, i.e., Low-level Group, Intermediate Group, and Advanced Group. Apart from their L2-English learning experiences in formal educational institution, another criterion, i.e., performance on a proficiency test, was employed in the selection of participants. The proficiency test adopted was part of a prototype paper for CET-4 (College English Test Band 4), which had been used in a five-year validation study of CET and was proved to be a highly valid and reliable measure of

106

Chapter 6 Research Design, Instrumentation, Data Collection and Data Analysis

Chinese L2 English learner’s proficiency (Yang & Weir, 1998).

43

Another reason for

adopting this prototype paper was that it did not have a wide circulation among the participants of our study: of the 3,000 copies of the book containing this test paper that had been published, only one was available in the university where this study was conducted. Since this copy was exclusively reserved for English teachers, the chances of its exposure to the participants were very slim. As the full version of the standardized test required longer testing time and more complicated testing procedures, two out of the total five parts of the test were administered to test the participants’ L2 proficiency level: a 30-item Vocabulary and Structure (VS) component and a 20-item Cloze (CLZ) component. These two parts had been found to be good predictors of the overall L2 proficiency of learners because both were correlated highly with the total scores of the test.

44

Since all the test items were in the

multiple-choice format, the scoring was to a large extent objective. Each correct choice was awarded one point and no point was given to incorrect choice. Thus, the maximum score was 50 points. All the test takers were required to complete the test within 35 minutes. Table 6 below summarizes the personal profiles of the participants in each group, including the range, means and standard deviations of each groups’ scores on the proficiency test. Results of One-way ANOVA indicate that the three groups are significantly different from each other in L2 English proficiency (F = 442.109; df=2, 87; p= .000) . The results of post hoc test (Scheffé) shows that Advanced Group scored significantly higher than both Intermediate Group (Mean difference = 9.17, p=.000) and Low-level Group (Mean Difference = 20.57, p=.000), and Intermediate Group scored significantly higher than Lower-level group (Mean Difference = 11.40,

43

The internal reliability index of the test was reported to be above .90, and 92 percent of the teachers responding to the questionnaire thought that the result of the test is congruous with the test takers’ real English proficiency level (Yang & Weir, 1998: 186). 44 The correlation coefficient between the total score and VS is .8022, and between the total score and CL was .7076 (Yang & Weir, 1998: 60). 107

Chapter 6 Research Design, Instrumentation, Data Collection and Data Analysis

p= .000). It can therefore be concluded that the participants sampled represented three distinct levels of L2-English proficiency.

Table 6 Grouping of Participants Based on Years of L2 English Learning and Scores on English Proficiency Test.

Learner Profile Groups Low (n=30) Intermediate(n=30) Advanced (n=30)

Education Middle School Senior II Non-English Major Sophomores English Major Seniors

English Proficiency Test EFL Years 5 8 10

Range

Mean

S.D

18-29 31-39 41-48

22.63 34.03 43.20

3.30 2.59 1.99

Since our hypotheses also involve native-nonnative comparison, 16 native speakers of English served as controls. They were teaching English in three universities in a northwestern city in China at the time of the present study. 6.3 Instrumentation Most of the existing L2 tense-aspect acquisition studies relied heavily on the evidence obtained from productive tasks, with an unpronounced association of language production on a par with language acquisition. This association, however, is a misleading one for three reasons. To begin with, even though the non-target forms in the learner language can help researchers to detect the errors of commission, little is known about the absence of errors, i.e., the errors of omission. This absence of evidence, however, should not be taken as the evidence of absence. Errors of omission are as much, if not more than, revealing as errors of commission. Furthermore, the learner’s ability to produce the form does not warrant its acquisition, since there is also a likelihood that the learner might have derived a different interpretation of the form from the target norm even when the same morphology appears in the learner’s speech production. It is necessary, therefore, to

108

Chapter 6 Research Design, Instrumentation, Data Collection and Data Analysis

tease apart different aspects of acquisition when generalization is made on the basis of the learner’s naturalistic production data. More importantly, while the evidence from naturalistic production tasks reveals the natural distribution of the tense-aspect morphology in the learner language, it conceals the patterns of some less frequently occurring morphology. In English, for example, the gross distribution of the progressive aspect is “less than 5 percent” in natural discourse, as is indicated by evidence from language survey studies (Quirk et al., 1985; Scheffer, 1975). It is inconceivable, then, that the distributional patterns of the progressive aspect will surface within a limited fragment of narrative discourse, given its sparse occurrence in naturalistic speech. Techniques that are specifically devised to elicit data of linguistic categories with low rate of occurrence should be utilized to find out patterns that are otherwise opaque. Based on these considerations, it is reasonable that alternative instruments and tasks be adopted so that the developmental patterns of L2 tense-aspect morphology are well manifest when examined from different perspectives. Three instruments are developed for this purpose in the present study. The first is an Acceptability Judgment Test (AJT), which is intended as a measure of the learner’s perceived compatibility of different classes of verbs with the progressive aspect. The second is an Aspect Sentence Interpretation Test (ASIT), which is designed to elicit the participants’ interpretation of sentences in narrative discourse when they are marked alternately with progressive and perfective aspects. The third is an elicited production task in which the participants were required to watch 10 movie clips, read the movie retell cloze passages, and then mark the removed verbs with proper tenseaspect morphology. This task is labeled as Movie Clip Retell Task (MCRT) in our study. 6.3.1 Acceptability Judgment Test (AJT) Judgment tasks have not been fully exploited in SLA tense-aspect research. However, as work by Slabakova and Montrul (2002) shows, such tasks can have

109

Chapter 6 Research Design, Instrumentation, Data Collection and Data Analysis

promising applications for understanding interlanguage temporal semantics in a way only hinted at by the data of production tasks. AJT is an endeavor to explore the use of judgment instrument in L2 tense-aspect research. This test was designed to investigate the participants’ knowledge about the compatibility between the English progressive aspect and verbs with different temporal properties. As is defined by the NOCOER constraint, verbs with different temporal schemata exhibit varying degrees of compatibility with the progressive aspect, depending on the number of the violations involved. An activity verb, for example, involves one shift of its inherent temporal semantics, and so it is more compatible with the progressive aspect than an achievement, which involves three violations of the NOCOER constraint. 45 If the participants were constrained by NOCOER on marking verbs with appropriate aspects, they would show different degrees of acceptability when verbs of different temporal properties are marked with the progressive aspect. This test contained 52 English sentences. The predicate verbs of all the sentences were marked with the progressive aspect. The participants were asked to judge the degree of acceptability of each sentence. Their responses were measured on a 5-point Likert scale ranging from -2 to 2. More specifically, the meaning of each point on the scale was explained, both in English and their Chinese equivalent, so that the participants could locate each sentence on the scale on the basis of their own judgment. What follows are the explanatory labels for the five points on the Likert scale -2 = totally unacceptable -1 = mostly unacceptable 0 = sounds odd but acceptable 1 = mostly acceptable 2 = completely acceptable

45

See §5.3.2 for a more detailed discussion on this point. 110

Chapter 6 Research Design, Instrumentation, Data Collection and Data Analysis

The corresponding relationship between the points on the scale and their interpretations was schematically represented in Figure 9 below. This scale diagram helped the participants to recognize that acceptability is a continuum and they should keep the whole spectrum of compatibility scale in mind when completing the task.

-2

-1

0

Totally unacceptable

1

2

Completely acceptable

Figure 9 Schematic representation of the acceptability scale of AJT

Also implicit in Figure 9 was the assumption that the acceptability scale of AJT was interval by nature. 46 In other words, the distance between any two adjacent points on the scale is constant, suggesting that a judgment made on one sentence is exactly equivalent to the same judgment made on any other sentences (Sorace, 1996). In addition, the interval scale provides not only the information that a given sentence is more acceptable than another when marked with the progressive aspect, but also, and more crucially, how much more acceptable it is. The 52 verbs in AJT were sampled respectively from Levin (1993), a resource book on the semantic and syntactic properties of the English verbs, and Quirk, et al. (1985), which provides a good taxonomy of the situation types and a detailed discussion of the compatibility of each situation type with the English progressive aspect. This is an attempt to ensure that the sampled verbs had an exhaustive coverage

46 Sorace (1996) distinguishes between four basic types of scales on which subjects are asked to respond in experiments concerned with the measurement of stimuli, i.e., category scale, ordinal scale, interval scale, and ratio scale.

111

Chapter 6 Research Design, Instrumentation, Data Collection and Data Analysis

of all situation types in the English language. Table 7 presents the taxonomy construct of AJT and specimens for each category.

Table 7 Taxonomy and Token Numbers of the Verbs in AJT

Situation Type Stative

Activity Accomplishment Achievement Semelfactive

Subclass Intellectual Relational Emotional Bodily sensation Property Agentive Non-agentive Agentive Non-agentive Agentive Non-agentive Agentive Non-agentive

Number 4 4 4 4 4 4 4 4 4 4 4 4 4

Example Peter is realizing his fault The box is containing candies Cindy is liking jazz music Ted’s back is aching Stephen is being a tall boy George is dancing at the club The engine is running well Dick is painting the wall The weather is getting warmer Nick is stopping the car The house is exploding in the fire Tony is jumping in the backyard Her eyes are blinking

After the verbs were sampled and the sentences constructed, they were checked against the vocabulary list provided by the English textbook that the participants of the beginner group were using (Liu, Brooks, Sjoquist, & Rickley, 2004). The sentences in AJT that contained words not appearing in the textbook were adapted to include the words that appeared in the textbook.

Besides, all the sentences

are invariably constructed with the simple present tense to minimize the effect of MarkT on participants’ judgement. Statistics shows that AJT is a reliable instrument for judguing acceptability of sentences marked with the English progressive aspect, the reliability α= .9034. (For details of AJT, refer to Appendix I ) 6.3.2 Aspect Sentence Interpretation Task (ASIT) ASIT is a derived form of Truth Value Judgment Task (TVJT), a research technique designed to measure comprehension of sentences in discourse. In TVJT, the subject’s task is simply to respond “Yes” or “No” depending on whether they judge

112

Chapter 6 Research Design, Instrumentation, Data Collection and Data Analysis

the test sentence to be true or false in the discourse. This method proves particularly useful in assessing language acquirers’ knowledge of constraints on the meanings that can and cannot be assigned to sentences in the target norm (Crain & Thornton, 1998). One great advantage of this method is that it enables experimenters to manipulate both the grammatical form and its semantic interpretation of the test sentence. Recent years witness an increasing use of this approach in investigating L2 tense-aspect acquisition, in particular in the studies done by Slabakova and her colleagues (e.g., Montrul & Slabakova, 2002; Montrul & Slabakova, 2003; Slabakova, 2003; Slabakova & Montrul, 2002). ASIT, our version of TVJT, was designed to examine the participants’ interpretation of sentences with different aspects when they were embedded in narrative discourse. In particular, this task assessed the effect of lexical and grammatical aspects on the participants’ interpretation of the completeness or incompleteness of an event in a discourse. In this task, we observed how the interplay of discourse functions, tense-aspect markers, and the temporal properties inherent in the verbs influenced the way in which the participants interpreted the sentences with different aspectual markers and different temporal semantics. 6.3.2.1 Components of ASIT ASIT has 12 short narratives, and each of them is so structured that it contains same types of information. What follows is a sample passage. Introduction Ever since he was a child, Gabe was afraid of heights. He decided to get rid of it once and for all. His plan was to spend the night on the local fire tower. He took his friends along with him that night. Aspect sentence Gabe was reaching the top of the tower. Or, Gabe reached the top of the tower. Post-aspect sentences

113

Chapter 6 Research Design, Instrumentation, Data Collection and Data Analysis

He was never more scared. His friends tried to calm him down. Gabe tried his best to listen to them. Conclusion The next weekend he rode the rollercoaster for the first time. Critical test statement Gabe got to the top of the tower that night.

□ True □ False

Filler statement Gabe spent the night on the tower with his friends.

□ True □ False

As can be seen from the above example, the passage consisted of four types of information, i.e., Introduction, Aspect sentence, Post-aspect sentences, and Conclusion. More specifically, the Introduction contained three to five sentences that introduced the main characters and the context of the story. It is followed by a critical aspect sentence, which is marked either with the progressive aspect, or with the perfective aspect (the simple past). After the aspect sentence are two or three “postaspect” sentences describing events that were concurrent with or subsequent to the event in the aspect sentence. It should be noted that these post-aspect sentences are constructed in a way that they do not explicitly indicate whether the event in the aspect sentence is completed or not. For each aspect sentence, there was a critical test statement, located either subsequent to the aspect sentence or at the end of the text. These critical test statements invariably stated that the event denoted by the aspect sentence was completed, irrespective of the aspectual morphology of the aspect sentence. In the example shown above, the aspect sentence was Gabe was reaching the top of the tower, or, Gabe reached the top of the tower, but the critical test statement was the same for both aspect sentences, i.e., Gabe got to the top of the tower that night. The participants performing this task were required to determine the truth or falsehood of the critical test statement, based on their understanding of the aspect sentence. In the sample passage shown above, for example, the aspect sentence was

114

Chapter 6 Research Design, Instrumentation, Data Collection and Data Analysis

Gabe reached (or was reaching) the top of the tower, and the participants had to judge whether the critical test statement Gabe got to the top of the tower that night was true or not. It was hypothesized that the participants capable of integrating aspectual information into discourse interpretation would judge the perfective aspect sentence as true and the alternative progressive aspect sentence as false. Therefore, the answer “yes” to the test statement means that the participants perceived the event in question as a completed one and the answer ‘no”, an uncompleted one. 6.3.2.2 Writing Aspect Sentences The verbs in the aspect sentences were all telic, but they differed in the internal properties of telicity. 47 More specifically, we examined in this task how the interpretation of the aspect sentences in discourse was influenced by three types of telic verbs, i.e., the externally induced telic verbs (ETV), the internally induced telic verbs (ITV) and the punctual telic verbs (PTV).

Table 8 Tokens of Telic Verbs in ASIT.

Situation Type

Telicity Type ITV

ACC ETV

ACH

PTV

Tokens the icecream melted (was melting) the big day approached (was approaching) he drowned (was drowning) the water freezed (was freezing) drew (was drawing) some money from the bank drove (was driving) from Montana towards New York danced (was dancing) to his favorite song changed (was changing) the flat tire reached (was reaching) the top of the tower found (was finding) some coins in his pocket shot (was shooting) the tiger won (was winning) the bet

47 Note that only telic verbs can be addressed by test statements pertaining to the completed events, and other atelic verb types cannot be so addressed.

115

Chapter 6 Research Design, Instrumentation, Data Collection and Data Analysis

As was illustrated by the examples given in Table 8, by externally induced telicity, we meant that the telicity had an overt NP or PP telicity marker indicating the temporal endpoint of the event, like drive from Motana to New York and change the flat tire. In contrast, internally induced telic verbs were devoid of these markers, like the verb freeze. These two types of telic verbs constitute one Vendlerian verb class — the accomplishment verb (ACC). The punctual telic verbs, by definition, correspond to the achievement verbs (ACH). The property of telicity would have an effect on the participants’ interpretation of the aspect sentence in the narrative discourse. However, the effect was assumed to vary with the number of violation marks incurred by the NOCOER constraint. 48 In this task, the critical test statement was manipulated to alternate between the within-text position and after-text position. In one arrangement of the task, the test statement was given right after the aspect sentence. In the other arrangement, the test statement was postponed to the end of the narrative discourse. The assumption underlying this design was that learners more sensitive to aspectual distinctions would respond differently to the test statements located in different positions in the text. They would give more correct responses to the within-text test statements than to test statements postponed to the end of the test, owing to their limited online processing capacity. In contrast, the locations of the test statement are assumed to make no difference in the responses given by learners less sensitive to aspectual distinctions in discourse (Magliano & Schleich, 2000). Figure 10 below is a schematic representation of the experimental set-up of the ASIT task. It is obvious that there are three within-subject variables involved in ASIT, i.e., the grammatical aspect, the lexical aspect, and the location of the test statement.

48

See §5.3.2 for a more detailed discussion on this point. 116

Chapter 6 Research Design, Instrumentation, Data Collection and Data Analysis

Aspect Sentence Grammatical Aspect

Perfective

Lexical Aspect

…

Test Statement Location

Progressive

ITV Within-text

ETV After-text

…

PTV …

Figure 10 Experimental Set-up and Variables in ASIT

6.3.2.3 The Format of ASIT It should be noted that the aspect variable and position variable yielded four possible combinations，i.e., perfective-within-text, perfective-after-text, progressivewithin-text, progressive-after-text. To study the effect of these two major variables on the interpretation of the aspect sentences in discourse, it seems plausible to divide the participants into four groups and assign each group to the condition of one combination of these variables. A group comparison of the participants’ performance in each experimental condition would then reveal the effect of these two variables. However, upon a closer examination of the nature of the task, this design was not adopted because the participants may easily work out the purpose of the task, given the prospect that they would have to judge whether the event denoted by the same aspect sentence was completed on the basis of the test statement at the same position throughout the task performance, to the detriment of the internal validity of the study. To minimize the practice effect, each group of the participants was exposed to all four different arrangements of the three within-subject variables as shown in Figure 10. Table 9 below illustates how the aspects, test statements and filler statements in one passage were balanced in each arrangement.

117

Chapter 6 Research Design, Instrumentation, Data Collection and Data Analysis

Table 9 Balancing Aspect, Test Statement and Filler Statement of One Passage across Arrangements of ASIT

Arrangement 1

Arrangement 2

Arrangement 3

Arrangement 4

Aspect

progressive

perfective

progressive

perfective

Test statement

within-text

within-text

after-text

after-text

Filler statement

after-text

after-text

within-text

within-text

In addition, to mitigate the sequencing effect of the passages on aspectual interpretation, the passages for each arrangement were randomly ordered. Note that each passage contained a filler statement about the explicit content of the stories. Unlike the critical statement, this filler statement did not address the aspectual status of the sentences, as shown in the sample passage. The position of the filler statement alternated with that of the test statement in complementary distribution. That is to say, when the filler statement was embedded within the text, the test statement would be placed at the end of the narrative and vice versa. The filler statements had two primary functions to fulfill in this task. On the one hand, they were included to avoid the prominence of the aspect sentences and the critical test statements, thus reducing the possibility of the participants’ awareness of the purpose of the task. On the other hand, these filler statements constituted one of the reliability measures of the task. More specifically, if the participants of different proficiency groups show no significant difference in responding to these statements, this could be taken as an indicator that the participants have the same level of comprehension of the passages. This way the effect of the variables for investigation could be unambiguously evaluated. Indeed, the results of one-way ANOVA showed that no significant mean differences were found in the responses to the fillers by the four participant groups, F(3, 102) = 1.822, p = 0.148. Besides, four other measures were taken to minimize the threat to the internal validity of the research. First, the length of all the passages was kept constant, ranging 118

Chapter 6 Research Design, Instrumentation, Data Collection and Data Analysis

between 76 and 79 words, with a mean of 77.87 and a standard deviation of 0.987. The effect of passage length was therefore negligible. Second, all the passages were displayed on computer monitor screens on a sentence-by-sentence basis. The sentences previously shown were irreversible, and so the participants had to construct the meaning of the passages without recourse to their idiosyncratic reading strategies to make inference of the aspect sentences. Third, the interval between sentence displays was preset by the computer program, according to the length of the sentence involved, so that the participants at the same proficiency level processed the material with the same pace. Fourth, Chinese equivalents were given to the words that the lowlevel participants may not know. The judgement of whether a Chinese equivalent should be given was based on the English textbooks the low-level participants had used (Liu et al., 2004). Finally, all the passages are narrated in the past tense to neutralize the effect of MarkT on the participants’ interpretation of the aspect sentences. For the details of ASIT passages, see Appendix II. 6.3.3 Movie Clip Retell Task (MCRT) This task was designed to examine the effect of lexical aspect and discourse grounding on aspect marking. More specifically, when performing the task, the participants were required to watch a short clip of the British television comedy Mr. Bean, then read a written description of the scenes of the clip with the verbs in their base forms, and finally provide correct verb forms. Altogether there were ten short clips. What follows is an example of the task. When Mr. Bean was waiting for his meal in a restaurant, he became very impatient （不耐烦）. He took the knife on his table and thrust it into the air. The woman sitting next to his table was startled (惊愕). She ____ (shoot) Mr. Bean a disapproving （不赞同） look to stop his ill manner. Mr. Bean ____ (appear) very unhappy. He____ (stop) playing with the knife reluctantly. Instead, he ____ (tap) on the glasses and cups when the woman was drinking. He ______ (make) such a great noise that most of the dinner guests at the restaurant could not go on with their dinner. They _______ (express) their annoyance (不满，恼火) silently but Mr. Bean did not seem to care.

119

Chapter 6 Research Design, Instrumentation, Data Collection and Data Analysis

For details of all ten passages of MCRT, see Appendix III. 6.3.3.1 Advantages of MCRT Movie retell task, especially when conjoined with cloze passage, has several advantages over naturalistic production task, which, as we have noted at the begining of this chapter, predominated the research paradigm of the previous L2 tense-aspect studies. Firstly, as story plot of the movie clip is the same across all participants, the context of the story and the context where the verbs appear are manipulated as constant factors. Therefore, it would not be hard for the researcher to judge whether the participants’ choice of verb forms is appropriate to the context or not. Secondly, the researcher is in a good position to proportionately balance the target verbs in terms of their situation types and grounding status, and the responses given by the participants are comparable across groups with desirable statistic rigor. 6.3.3.2 Video Segments and Story Construction Mr. Bean was selected as the prompt materials because we noted an increasing interest in this British television comedy in SLA research (e.g., Gass, Mackey, Alvarez-Torres, & Fernández-García, 1999; Skehan & Foster, 1999; Sullivan & Caplan, 2004). As Skehan & Foster (1999:103) explained, Mr. Bean sketches were ideal speech elicitation materials because they are short, almost mimed, widely appealing, and devoid of cultural bias. This would make the production task amusing, motivating, and engaging. The video segments were selected according to three criteria. First, the selected segments should involve at least two characters, whose actions were more likely to be simultaneous or overlapping than the actions taken by one single character in a monodrama. The assumption was that retelling simultaneous or overlapping events called for more uses of the progressive markers. Secondly, each clip should contain only one scene, because scene shifts are very likely to engender a shift of narrative perspective and grounding status of the events involved. A backgounding event in one scene is very likely to be foregrounded in another, depending on the narrator’s

120

Chapter 6 Research Design, Instrumentation, Data Collection and Data Analysis

perspective. Thirdly, the story of the clips should be transparent and easy to follow. The episodes involving complicated series of actions and culturally specific interpretations were not considered. Ten segments were thus sampled. They were clipped, reedited and synthesized, with the aid of a video editing program, Ulead VideoStudio 7.0.

49

The synthesized

materials also contained subtitled bilingual directions to the participants. For the events shown in each video clip, a narrative passage was written by the researcher and revised by a visiting English-speaking American teacher working in a university in Northwest China. To minimize the effect of MARKT constraint on aspect marking, all the stories are narrated in the past tense. 6.3.3.3 Balancing the Target Verbs After the selection of video segments and the construction of stories for each clip, the target words were then identified. As our purpose was to investigate the effects of situation types of verbs and their discourse grounding status on the tense-aspect marking, the verbs were selected in a way that their situation types and discourse grounding status were balanced. From the ten passages 40 bare-form target verbs were chosen whose distribution was well balanced by five situation types (i.e., statives, activities, accomplishments, achievements, and semelfactives) and two discourse grounding status (i.e., foregrounding and backgrounding), with each combinatorial scenario having four tokens. 6.3.3.4 Identifying the Situation Types of the Verbs Following the diagnostic procedures discussed in § 3.1.4.3, the temporal semantics of the target verbs were determined. First, their temporal primitives like telicity, dynamicity and punctuality were identified. This was followed by the establishment of their situation types on the basis of the values set to the selected verbs (See Table 2 on this point). For instance, when determining the situation type of the verbal predicate she shot Mr. Bean a disapproving look, we first successively

49

This program is downloadable at http://www.ulead.com/vs/trial.htm. 121

Chapter 6 Research Design, Instrumentation, Data Collection and Data Analysis

applied the dynamicity tests, telicity tests, and punctuality tests before the combinatorial feature set [+dynamic, +telic, +punctual] was obtained. Then the verb predicate was further identified as an achievement situation. The following are the details about the procedure of identification. (79) She shot Mr. Bean a disapproving look. a. What she did was to shoot Mr. Bean a disapproving look. (Agentivity Test) b. –What happened? (Happening Test) c. She shot Mr. Bean a disapproving look. d. She shot Mr. Bean a disapproving look in a minute. (In X Time Test) e. It took her a minute to shoot Mr. Bean a disapproving look. (Take X Time Test) f. *She spent a minute shooting Mr. Bean a disapproving look. (Spend X Time Test) g. *She shot Mr. Bean a disapproving look for a minute. (For X Time Test)

In our study, the work of designating the target verbs with proper situation types was accomplished by the researcher and a native speaker of English. Both coders adopted the same procedure as proposed above. Inter-rater reliability on the token verbs was 81.75% (agreement on 35/40 coding decisions). Disagreement was resolved by discussions. 6.3.3.5 Identifying the Grounding Status of the Target Verbs Reinhart (1984) put forth three criteria for sorting out the foregrounding events from backgrounding ones in a narrative discourse, i.e., the chronological sequencing, punctuality, and completeness. However, since two of these criteria, i.e., punctuality and completeness, overlapped with temporal semantics of the verbs, they were excluded from the present research. In other words, only the criterion of chronological sequencing (Reinhart, 1984) was adopted to set foregrounding events apart from backgrounding events. This criterion stated that the temporally ordered clauses whose

122

Chapter 6 Research Design, Instrumentation, Data Collection and Data Analysis

order matches the order of the events could serve as discourse foregrounding and the other events served the function of discourse backgrounding. Based on this criterion, the clauses that carried the storyline of the narrative discourse were coded as foregrounding clauses and those that were out of sequence with regard to the foregrounding events were coded as backgrounding events. In the above sample passage, for instance, the skeletal storyline was “Mr. Bean take the knife—thrust the knife into the air—woman be startled—she shoot Mr. Bean a disapproving look—Bean unhappy—stop playing with the knife—tap on the glasses— dinner takers annoyed”. These clauses were coded as foregrounding and the other clauses were coded as backgrounding. This coding was collaboratively completed by the present researcher and a native speaker of English. Interrater reliability was 89% (71/80 agreement decisions), and again, disagreement was resolved by discussion.

6.3.4 Summary of Test Instruments Table 10 presents a summary of the instruments developed for the present study. As is shown in the table, three instruments were constructed that vary in the medium of message conveyance, linguistic units involved, situation types and grammatical aspects of the target verbs. In so doing we are able to examine the factors that might have an effect on the L2 development of the tense-aspect marking from a broad spectrum of perspectives.

Table 10 A Summary of Instruments Developed for the Study

Tasks

Media

AJT ASIT MCRT

text text text+movie clip

Linguistic unit sentence discourse discourse

Situation types

Aspects

STA, ACT, ACC, ACH, SMF ITV, ETV, PTV (ACH) STA, ACT, ACC, ACH, SMF

Progressive Progressive, perfective Progressive, perfective, perfect, etc.

123

Chapter 6 Research Design, Instrumentation, Data Collection and Data Analysis

6.4 Data Collection 6.4.1 Time and Venue The data were collected between the eleventh week and the thirteenth week of the first semester of the 2003-2004 academic year. A proficiency test was administered to a larger sample of students two weeks prior to the data collection sessions. The students meeting our subject-selection requirements were then invited to take part in an “English language learning”experiment.

50

They were not aware of the

nature and purpose of the experiment until all the three tasks were completed. The participants performed the tasks during weekends when they were free from regular classroom obligations. The venues were language labs, each of which was equipped with one main control center and 24 14-inch terminal monitors. The directions and prompt materials of ASIT and MCRT were shown on the monitor screens. The instructions and sentences of AJT were given to the participants on a paper version. The native controls performed these tasks at their residence. All the participants were rewarded a small token gift for their participation.

6.4.2 Task Sequence In order to minimize the confounding effect caused by task sequence (Elmes, Kantowitz, & Roediger III, 1999) , AJT, the task with the most explicitly stated data collection purpose was delayed until the other two tasks were completed, whereas ASIT, in which the nature and purpose of the task were blind to the participants, was the first task given to the participants. Before the formal administration of ASIT, a sample passage was first shown to demonstrate how the task should be performed. When the researcher was sure that all the participants understood what they were required to do, the test began. Then instructions, passages, test statements and filler statements appeared sentence by

50

See § 6.2 for the requirements of subject sampling. 124

Chapter 6 Research Design, Instrumentation, Data Collection and Data Analysis

sentence on the computer screen. The participants read the sentences on the screen and gave responses to the test statements and filler statements on the task sheet. MCRT was the task following ASIT. The procedures of administration were similar to those adopted in ASIT. That is, prior to the commencement of the test, there was a demonstration showing how the task should be performed. When this researcher was sure that all the participants knew exactly what they were supposed to do, the test began. The participants first watched the movie clips one by one. After each clip was shown, they had to inflect verb with tense-aspect forms, based on their understanding of the movie clip they had just seen, for base forms of the verbs in a booklet containing the corresponding cloze passage that described the scenes of the watched movie clip. They were allowed a ten-minute break between the fifth and sixth clip show. This was to alleviate the participants from the fatigue caused by long sessions of task performance. This task lasted around fifty minutes to one hour, depending on the proficiency level of the participants. AJT was the last task that the participants accomplished. Unlike ASIT and MCRT, the directions and the test sentences of AJT were printed on paper and the participants were required to supply their answer in the bracket before each sentence after the directions for the task were explained in Chinese. In the course of the experiments, the participants were allowed to consult the researcher only when they could not understand what they had to do. They were not allowed to consult either other participants or reference books. 6.5 Data Scoring and Coding This section describes how the data obtained from the three tasks were coded and scored for further statistical analysis. 6.5.1 Scoring of ASIT Data Recall that the participants had to give two different responses when performing ASIT, i.e., the response to the aspect sentences and that to the critical test statement. Also recall that our concerns were different for these two types of responses. In the 125

Chapter 6 Research Design, Instrumentation, Data Collection and Data Analysis

former case, we were interested in whether the aspect sentences were interpreted as complete or incomplete, whereas in the latter case, our primary concern was the participants’ understanding of the explicit information conveyed in the stories. These two types of responses, therefore, were scored differently, and naturally, the mean scores for these two types of responses had to be interpreted differently. As mentioned in §6.3.2.1, the proposition of the critical test statement invariably pertained to the completion of the event as described in the corresponding aspect sentence, thus a “true” response indicated that the participants interpreted the verb predicates of the aspect sentence as having culminated to its endpoint, whereas a “false” response suggested that the participants interpreted the aspect sentence as an unfinished action. Each “true” response was awarded “1” point, and each “false” response, “0” point. It should be noted, however, that a “true” response does not necessarily mean a correct response. Therefore, the mean scores for the participants’ responses should be interpreted as the general tendency or probability of interpreting the aspect sentence as denoting a completed action. The purpose of the filler statements, as mentioned above, was to tap the participants’ understanding of the content message of the story. Therefore, a response, be it “true” or “false”, was awarded 1 point if it accorded with the contents of the passage and 0 point if it did not. The mean scores for the filler statements, therefore, indicated the means of correct responses and participants’ understanding of the stories. As shown in §6.2.2, the means for the filler statements were not statistically different between participants of different L2 proficiency levels. So it is safe to assume that despite the difference in general English proficiency, the participants had the same level of understanding of the contents of the passages in ASIT. That is to say, the effect of general comprehension of the passages as an extraneous factor was neutralized in this task.

126

Chapter 6 Research Design, Instrumentation, Data Collection and Data Analysis

6.5.2 Scoring of AJT Data As has been made clear in § 6.2.1, in AJT the participants were asked to judge on a scale the degrees of acceptability of 52 sentences marked with the progressive aspect. The judgments were made on a 5-point scale from -2 (totally unacceptable) to 2 (completely acceptable). As it was assumed that the scale in AJT was interval by nature, the point the participants gave on the scale equals to the point the participants were awarded. The higher score in this task suggests greater acceptability of the test sentence. 6.5.3 Coding and Scoring MCRT Data Recall that the primary goal of MCRT was to elicit the participants’ responses to the target verbs in the cloze passages. Not surprisingly, the informants’ responses were diverse and sometimes hard to categorize. However, as the cloze passages were narrative texts describing the scenes of Mr. Bean, simple past tense and the past progressives constituted the bulk of the tense-aspect forms that the participants used. It should be noted that our coding was based on the participants’ inclination to mark a verb with a tense-aspect form, not the correct tense-aspect inflections. Take for example the simple past form of the verb run. If the participant chose to mark it as ran, it would be coded as a token of the simple past form. But in the cases where other incorrect variants such as runned or runed were used, they were still treated as belonging in the same category as ran and coded as such. This is because these forms indicate that the participants were inclined to mark the verb perfectively, despite their linguistic deficiency. The same coding scheme was adopted for the progressive aspect. A closer examination of the participants’ responses had revealed that they used three variants of progressive aspect in their production data, i.e., the past progressive, the present progressive, and the tenseless progressive (the bare V-ing form devoid of a finite tense marker). Obviously, these forms indicated the participants’ efforts at marking the verbs with the progressive aspect, and so they were conflated into one category.

127

Chapter 6 Research Design, Instrumentation, Data Collection and Data Analysis

Simple present and present/past perfect were coded as two independent categories. Simple present constituted the base form and the third-person singular form of the verb. In the same vein, present perfect, past perfect, and the past participles were coded as the perfect class. Other miscellaneous inflectional forms were categorized as an “unclassifiable” category, which included passives, modal verbs, forms indicating futurity, and noun forms derived from the verbs, like action for act. Following this coding framework, five broad categories of the tense-aspect forms were identified. They were the progressive, the simple past, the simple present, the perfect and the unclassifiable. Table 11 summarizes this coding scheme.

Table 11 Coding of the tense-aspect forms in MCRT

Categories Past Progressive Perfect Present Unclassifiable

Tokens ran, runed, runned, etc. sitting, was sitting, is sitting, siting, was taking, takeing, etc. written, has written, have written, had written, written, etc. hurries, hurry, hurrys, etc. was done, would take, had to go, action (for act), behavior (for behave), etc.

6.6 Data Analysis As was made clear above, the five research hypotheses were tested by means of three elicitation tasks. Since the variables and the relationship between them vary with tasks and hypotheses, it is hard to conceive of one single data analysis technique to meet the needs of our study. However, a close scrutiny of the five hypotheses reveals that they all involve comparing differences between groups, such as the differences existing between verbs of different situation types (as in H1), between participants of different L2 proficiency levels (as in H2 and H3), and between the native and non-

128

Chapter 6 Research Design, Instrumentation, Data Collection and Data Analysis

native speakers of English language (as in H3 and H4). Naturally, statistical techniques of comparing means would be the most appropriate data analysis methods. 6.7 Summary In this study, 90 L2 learners of three proficiency groups were sampled, together with 16 native speakers of English serving as controls in this study. Three instruments were devised to elicit data of aspect production, judgement and interpretation in L2. Also discussed in this chapter are techniques of data collection, data scoring and data analysis. These methodological measures were taken in this study to guarantee the validity and reliability of our investigation.

129

Chapter 7 Results for Research Hypotheses Testing This chapter evaluates the extent to which the five hypotheses formulated in § 4.5 are confirmed by the data obtained in the empirical investigation. This would provide the answer to the third key research question raised in § 1.4, i.e., how might the acquisition of English progressive aspect be described? As mentioned in Chapters 5 and 6, three instruments were developed to empirically test the hypotheses. It should be noted, however, that there was no one-toone mapping between instruments and hypotheses. That is to say, each instrument was intended to test more than one hypothesis, or one hypothesis was tested by the data obtained from more than one instrument. In so doing the predictions made in these hypotheses can be put under greater scrutiny through the triangulation of data from a production task, an interpretation task, and a judgment task. This chapter is organized as follows. § 7.1 to § 7.5 present the results for testing Hypothesis 1 to Hypothesis 5. More specifically, each section provides the following types of information: a restatement of the hypothesis, the empirical manifestation of the hypothesis, the variables involved, the statistical technique adopted, the statistical results obtained, and, finally, the conclusion on whether, and if so, to what extent, the hypothesis is empirically confirmed. §7.6 summarizes results for hypotheses testing. 7.1 Results for Hypothesis 1 Recall that Hypothesis 1 (H1) stated that L2 acquisition and processing difficulty increases with the number of violations of the NOCOER constraint, i.e., for the learners of the same L2 proficiency level, greater acquisitional or processing cost results from more violations of the NOCOER constraint. As mentioned in §5.3.2, a violation of NOCOER constraint occurs wherever there is a mismatch between the inherent temporal properties of the verb and the

Chapter 7 Results for Testing of Research Hypotheses

aspectual interpretation, and the effect of the NOCOER constraint can be operationally measured by the number of constraint violations involved. It was further shown in Table 5 that the number of constraint violations varies with the situation type of the verb marked with different aspect morphology. Put differently, more violation of the NOCOER constraint is predicted to gives rise to greater acquisitional or processing difficulty on the part of the language learners. Therefore, the indication that H1 is true would be that the order of acquisitional or processing difficulty conforms to the number of violations of the NOCOER constraint incurred by a specific aspect marker. More particularly, we predict that L2 learners, whatever their proficiency level, would all follow the order of STA>ACH>ACC, SMF>ACT 51 when acquiring progressive aspect, and by analogue, an STA>ACT>ACC, SMF>ACH order when acquiring perfective aspect. 7.1.1 Testing Hypotheses 1 through Movie Clip Retell Test (MCRT) Data Recall that in MCRT the participants were asked to use proper tense-aspect forms of the verbs given to complete a cloze passage modeled after short episodes of the British comedy video Mr. Bean. The purpose of this task is to examine the participants’ ability to mark the given verbs with proper tense-aspect markers in discourse context. It is perceivable that testing H1 through MCRT data involves an examination of two variables: i.e., the aspect marking and the situation type of the verbs. The variable aspect marking consisted of two levels: perfective marking and progressive marking, 52 and the variable situation type, five levels: stative, activity, accomplishment, achievement, and semelfactive.

51This

acquisitional order is ranked in descending order. The symbol “>” stands for “more difficult to acquire or process than…” and the symbol “,” means no significant differences between the two situation types involved, given that the number of the violations of the NoCoer constraint is the same. 52 Recall that in § 6.5.3 we coded the tense-aspect forms that participants used in MCRT into three categories: i.e., the perfective, the progressive and the unclassifiable. However, in testing the hypotheses, we do not include the unclassifiable forms that the participants produced in this task, such as the simple present, the present perfect and the modalities, because these forms are, in the first place, not of our primary concern in testing the hypotheses. Secondly, they constitute a marginal percentage of the tense aspect forms that the participants used (only 4.03%, compared with 66.84% for the perfective and 29.13% for the progressive aspect). 131

Chapter 7 Results for Testing of Research Hypotheses

As stated, the confirmation of H1 can be ascertained only when there is evidence showing that more violations of the constraint NOCOER give rise to greater processing difficulty. In terms of performance on MCRT, we would observe more frequent use of the aspectual forms with fewer violations of the constraint among L2 learners, and, alternatively, less frequent use of the aspectual forms with more violations of the constraint. More specifically, Chinese L2 groups would demonstrate a frequency order of STA < ACH < ACC, SMF < ACT for progressive marking, and an STA < ACT < SMF, ACC < ACH order for perfective aspect marking. 53 Figure 11 presents the means obtained by the three L2 participant groups in progressive and perfective marking of five situation types of English verb. 54

3.5 3

Mean Frequency

2.5 Low

2

Intermediate 1.5

Advanced

1 0.5 0 ACT

ACC

SMF

ACH

STA

ACT

PROG

ACC

SMF

ACH

STA

PERF

Figure 11 Means of Progressive and Perfective Marking for Five Situation Types in MCRT

53

The symbol “<” stands for “less frequently used than…”, and the symbol “,” means “no significant differences between the situation types involved”. 54 For details of the descriptive statistics, see Table A Appendix IV. 132

Chapter 7 Results for Testing of Research Hypotheses

It can be seen from the figure that, as predicted, there was an STA < ACH < SMF < ACC < ACT order of frequency (from the least frequent to the most frequent) for progressive aspect marking across all L2 groups (except for the stative progressive for low-level participants), which means that the stative-progressive combination was the most difficult to produce and the activity-progressive is the least difficult, since the most frequently used language items are in most cases the easiest to acquire. By contrast, the perfective aspect marking showed a grossly different pattern. For all L2 groups, the achievement verb was the category most frequently marked with the perfective aspect and the activity verb, the category least frequently marked with the perfective verbs. In between were the three other situation types, which showed a mixed order of compatibility for learners in different L2 groups. However, if we exclude for the moment the statives from our consideration 55, we can see that all L2 groups followed the same ACT < ACC < SMF < ACH order of frequency with the perfective aspect, which appears to suggest that the effect of NOCOER constraint gradually slackens with the decrease of its violation. To further examine whether the observed differences between means for different verb types in each aspect are statistically significant, three two-way ANOVAs (repeated measures) were performed, one for each L2 group. The results are presented in Table 12. As is shown in the table, for each L2 group, the two main effects and the interaction were all statistically significant, indicating that while the L2 participants showed marked difference in their use of progressive and perfective aspect, this may be due to the situation type of verbs.

55

The stative verb marked with the perfective aspect incurred different number of violations when it performed different discourse functions. When the verb functions as discourse foregrounding, it incurs three violations of the NOCOER constraint, as shown in Table 5. However, when the verb functions as discourse backgrounding, it is the optimal choice, without violating the constraint at all, as was noted in § 3.3.3. The dual roles played by the perfective statives give rise to mixed results in aspect marking by learners of different proficiency levels. 133

Chapter 7 Results for Testing of Research Hypotheses

Table 12 Two-way ANOVAs (Repeated Measures) of Results for MCRT

Aspect Situation Type Aspect×Situation

Lower F 180.023 13.653 46.298

df 1 4 4

p .000 .000 .000

Intermediate F df p 109.87 1 .000 7.868 4 .000 20.967 4 .000

Advanced F df 139.620 1 26.632 4 32.731 4

p .000 .000 .000

To find out where the significant differences between means for situation types lie, post hoc tests (LSD) were conducted at p = . 05. The results are presented in Tables 13 and 14 respectively.

Table 13 Pairwise Comparisons of the Mean Differences Between Different Situation Types Marked with the Progressive Aspect (LSD)

STA vs ACT STA vs ACC STA vs ACH STA vs SMF ACT vs ACC ACT vs ACH ACT vs SMF ACC vs ACH ACC vs SMF ACH vs SMF

Low -1.217*** -.433*** .167 .017 .783*** 1.383*** 1.233*** .600*** .450** -.150

Intermediate -1.200*** -1.017*** -.133 -.483*** .183 1.067*** .717*** .883*** .533** -.350**

Advanced -1.317*** -1.183*** -.283** -.550*** .133 .1.033*** .767*** .900*** .633*** -.267*

*** p<.001, ** p<.01, *p<.05

Table 13 shows that, first of all, activity was the situation type that was most frequently marked with the progressive aspect, as indicated by the significant mean differences between this situation type and the other four situation types across the three L2 groups, the exceptions being the difference between activities and accomplishments for the intermediate (Mean Difference =.183, p=.266) and advanced group (Mean Difference =.133, p = .211). This is followed by accomplishment verbs, which were more frequently marked with the progressive aspect than achievements,

134

Chapter 7 Results for Testing of Research Hypotheses

semelfactives and statives were across all three L2 groups, as shown by the significant mean differences between the accomplishments and these three situation types. The situation of progressive marking on semelfactives, however, was obviously more complicated. More specifically, intermediate- and advanced-level participants’ use of progressive markers on semelfactives was significantly more frequent than their use of progressive markers on achievements and statives, but the frequency of low-level participants’ use of progressive markers on semelfactives did not differ significantly from that on statives (Mean Difference = .017, p=.869) and on achievements (Mean Difference = .150, p=.184). Finally, the advanced participants’ use of progressive aspect on achievement predicates was significantly more frequent than that on statives, but the mean differences between the use of these two types of verbs did not reach significant level among the intermediate- ( Mean Difference = -.133 , p = .265) and low-level participants (Mean Difference = .167 , p= .096).

Table 14 Pairwise Comparisons of the Mean Differences Between Different Situation Types Marked with the Perfective Aspect (LSD)

ACT vs ACC ACT vs ACH ACT vs SMF ACC vs ACH ACC vs SMF ACH vs SMF

Low -.1.217*** -1.783*** -1.600*** -.567*** -.383* .183

Intermediate -.217 -.983*** -.750*** -.767*** -.533** .233

Advanced -.150 -.900*** -.817*** -.750*** -.667*** .083

Note: We do not include the data for the statives in this table for the reasons given in Note 57. *** p<.001, ** p<.01, *p<.05

Table 14 shows that the use of the perfective aspect on activity verbs was significantly lower than that on accomplishments, achievements and semelfactives across all L2 groups, the only exception being that the mean difference between activities and accomplishments did not reach the level of significance among 135

Chapter 7 Results for Testing of Research Hypotheses

intermediate participants (Mean Difference = -.217, p = .177) and advanced participants (Mean Difference = -.150, p = .174). Also shown in the table was the significantly less frequent use of perfective marker on accomplishments than that on semelfactives and achievements by all three L2 groups. Finally, the frequency of perfective marking on semelfactives did not differ significantly from that on achievement verbs by all three L2 groups. To recapitulate, the patterns that emerged from the post hoc tests are summarized in Table 15. This table reveals a tendency of aspect marking on verbs of different situation types that basically conforms to our predictions. It should be noted, however, that the non-significant stative vs. activity mean differences for low and intermediate groups in progressive aspect marking and the non-significant achievement vs semelfactive mean differences for all groups in perfective aspect marking should not be taken as counter-evidence for H1. This is because we have assumed a Gradual Learning Algorithm (GLA) in which the demotion of a high-ranked constraint does not take place in a leap, and there is often an overlapping range between constraints. 56 The exceptions can thus be accounted for by GLA assumed in OT.

Table 15 Frequency Order of Perfective and Progressive Aspect Marking on Verbs of Five Situation Types by L2 Groups in MCRT

Low Intermediate Advanced

Progressive SMF, STA, ACH < ACC < ACT ACH, STA < SMF < ACC, ACT STA < ACH < SMF < ACC, ACT

Perfective ACT < ACC < SMF, ACH, ACT, ACC < ACH, SMF ACT, ACC < ACH, SMF

Note: The symbol “<” means “less frequently used than”, and “,” means the difference did not reach statistical significance level. To conclude, H1 is to a large extent verified by the MCRT data if we take the GLA learning hypothesis into consideration and exclude for the moment the perfective aspect marking on the stative predicates.

56

See §5.1.2 for a detailed discussion on this learning algorithm proposed in OT framework. 136

Chapter 7 Results for Testing of Research Hypotheses

7.1.2 Testing Hypothesis 1 through Aspect Sentence Interpretation Task (ASIT) Recall that ASIT was designed to test whether the participants are able to correctly interpret the critical aspect sentences embedded in a narrative discourse. As the progressive aspect typically refers to an incomplete event and the perfective aspect views an event in its entirety (Comrie, 1976), we expect that learners with adequate aspectual knowledge are able to distinguish between the meanings denoted by these two grammatical aspects, as revealed by the participants’ responses to the critical test question. However, in the situation where learners are not able to make the correct distinction because of their insufficient knowledge about English aspect, they are more likely to choose as the correct answer the critical test question that contains the verb involving fewer violations of the NOCOER constraint. Thus, although the correct responses to the following two test statements, which all denote complete events, should be “false” because the progressive aspect as used in the related aspect sentence denotes an incomplete event, learners are expected to interpret (80), rather than (81), as denoting a complete event. This is because the achievement verb in (80) involves more violations of the NOCOER constraint than the accomplishment verb in (81) when both are marked with the progressive aspect. In other words, learners of the same proficiency level would score higher in (80) than in (81), meaning that they are more inclined to iinterprt achievement prediates as denoting complete events, regardless of its aspectual marker. (80)

a. aspect sentence: Gabe was reaching the top of the tower. (achievement) b. test statement: Gabe got to the top of the tower.

(81)

a. aspect sentence: Stephanie was changing the tyre of the car. (accomplishment) b. test statement: Stephanie managed to change the flat tire.

137

Chapter 7 Results for Testing of Research Hypotheses

Therefore, testing H1 using ASIT data entails an examination of the effect of one variable across all participant groups, i.e., the verb’s situation type, which is further divided into two broad categories: the achievement and the accomplishment. More specifically, by examining how the participants of the same L2 proficiency level perform ASIT, we can observe the effect of NOCOER constraint on the interpretation of the aspect sentence. Figure 12 displays the mean scores for ASIT for each of the four participant groups. 57 Recall that the scores represent the probability of interpreting a verbal predicate as denoting a complete event.

0.9 0.8 0.7 Probability

0.6 0.5

ACH ACC

0.4 0.3 0.2 0.1 0 Low

Intermediate Higher Proficiency Group

Native

Figure 12 Probability of Interpreting Aspect Sentence as Denoting Completed Events by all Participant groups in ASIT

As is indicated in the figure, for low-level participants, there was about 80 percent probability of interpreting an achievement predicate as denoting complete events in narrative discourse, irrespective of their aspect markers. The probability of

57

For details of statistical data for this figure, see Table B in Appendix IV. 138

Chapter 7 Results for Testing of Research Hypotheses

interpreting an achievement predicate as a complete event was also very high for the intermediate and advanced participants, reaching 70 percent and above. In comparison, the probability of interpreting an accomplishment as denoting a complete event was lower, and this was true for all L2-English groups. However, unlike the L2 participants, the native controls did not seem to be influenced by the situation type of verbs when they interpreted the aspect sentences, as the chance of interpreting achievement and accomplishment predicates as complete events was almost the same. Paired-sample t-tests were further conducted to examine whether the observed mean differences in aspect sentence interpretation were statistically significant. The results were summarized in Table 16.

Table 16 Effect of Verb Situation Type on Aspect Interpretation in ASIT

Low Intermediate Advanced Native

t 6.553 2.743 1.923 .205

df 29 29 29 15

p (two-tailed) .000 .010 .064 .840

The test results show clearly that the mean differences between the interpretation of achievement and accomplishment predicates were statistically different for lowlevel and intermediate groups, indicating that the participants were susceptible to NOCOER constraint at the early stages of L2 aspect acquisition. As for the participants in the advanced group, however, the mean differences did not reach statistical significance (p = .064), although there was a clear tendency that they were more likely to interpret achievement predicates as denoting complete events. This suggests that the participants were behaving more native-like as their L2 tense-aspect system developed.

139

Chapter 7 Results for Testing of Research Hypotheses

To sum up, the participants in the low-level and intermediate groups, and to a lesser extent, those in the advanced group, were more likely to interpret the achievement predicate as completed events, indicating that L2 learners, especially those at the earlier stage of L2 development, rely primarily on temporal semantics, rather than aspectual morphology, features of the verbs when interpreting sentences marked with different aspects in discourse. This suggests that Chinese learners were subject to the NOCOER constraint in their acquisition of the L2-English tense-aspect system, thus confirming H1. 7.1.3 Testing Hypothesis 1 through Acceptability Judgment Task (AJT) AJT was designed to tap the participants’ knowledge about the compatibility between the English progressive aspect and verbs having different temporal properties. In this task, the participants were asked to judge on a five-point Likert scale the acceptability of progressive-marked sentences containing verb predicates of different situation types. Testing H1 by AJT involves an investigation into the variable of situation type across all participant groups. This entails an examination of the mean scores obtained by the participants on each situation category. If H1 is true, we would observe that the participants at each proficiency level are more likely to accept the progressivesituation type combinations with fewer violations of NOCOER constraints, such as the activities marked with the progressive, while at the same time reject the combinations involving more violations of the constraint, as was in the cses of achievements and statives. More specifically, there would be a compatibility order of ACT > ACC, SMF > ACH > STA for predicates marked with the progressive aspect. 58 Note that this order has derived from our discussion on the ranking order of the NOCOER constraint on aspect marking.

59

Therefore, by comparing the mean scores on each situation type,

we are able to tell whether H1 is confirmed or not.

58

The notation “>” stands for “more compatible than” and the scale is ranked in descending compatibility order. 59 See §5.3.2 for a discussion on this ranking order. 140

Chapter 7 Results for Testing of Research Hypotheses

Figure 13 presents the mean scores obtained by the three L2 groups as well as by the native speaker group when judging the acceptability of the progressive-marked sentences containing the five situation types. 60

2.5 2

Mean Score

1.5 ACT ACC SMF ACH STA

1 0.5 0 Lower

Middle

Higher

Native

-0.5 -1 -1.5 Level

Figure 13 Mean Scores for Acceptability of Progressive Marking in Five Situation Types by L2 Participants and the Native Controls

It can be seen from the figure that the participants of each L2 proficiency group demonstrated an identical acceptability ranking order of the progressive aspect, i.e., ACT > ACC > SMF > ACH > STA, with activity verb the most acceptable verb category to mark with progressive aspect, and achievement and stative verbs among the least acceptable situation types with the progressive marker, as shown by the means below or around “0” for these two situation types.

60

61

By contrast, the native

For detailed statistical data for this figure, see Table C of Appendix IV. Recall that “0” stands for “sounds odd but acceptable” in AJT scale. Scores below “0” is interpreted as “mostly unacceptable” or “totally unacceptable”. See §6.2.1 for a detailed discussion on this point.

61

141

Chapter 7 Results for Testing of Research Hypotheses

controls’ pattern of behavior in this task was different, suggesting that they are less susceptible to the effect of NOCOER constraint than the L2 English learners. Three one-way ANOVAs (repeated measures) were run to examine whether the observed mean differences reached statistically significant level. The results are presented below in Table 17.

Table 17 Effect of Verb Situation Type on Acceptability Judgment of Progressive Marking: Results for One-Way ANOVA (Repeated Measures)

Low Intermediate Advanced

F 75.234 155.987 222.054

df 4, 116 4, 116 4, 116

p .000 .000 .000

As shown in Table 17, the effect of the situation type was significant across all L2 groups, so post hoc tests (LSD) were performed to further examine where the significant mean differences for each L2 group lie. Table 18 presents the results of the pairwise comparisons. As is shown in the table, with the exception of the mean difference between activities and accomplishments by the low-level participants (MD = .128, p=.076), the pairwise comparisons of the mean differences between different situation types by participants of all L2 groups were statistically significant. More particularly, the participants of each L2 proficiency group demonstrated an identical acceptability ranking order of the progressive aspect, i.e., ACT > ACC > SMF > ACH > STA, with activity verb the most acceptable verb category to mark with progressive aspect, and they typically reject the progressive aspect marking on achievement and stative verbs.

142

Chapter 7 Results for Testing of Research Hypotheses

Table 18 Pairwise Comparisons of Means for Five Situation Types by Each L2 Group in AJT

Level STA vs ACT STA vs ACC STA vs ACH STA vs SMF ACT vs ACC ACT vs ACH ACT vs SMF ACC vs ACH ACC vs SMF ACH vs SMF

Lower -1.650*** -1.522*** -.338** -.737*** .128 1.313*** .913*** 1.185*** .785*** -.400**

Intermediate -2.253*** -1.962*** -.941*** -1.387*** .292*** 1.313*** .867*** 1.021*** .575*** -.446**

Advanced -2.423*** -2.140*** -.853*** -1.790*** .283*** 1.287*** .336*** 1.287*** .350*** -.938***

**p < .01, ***p < .001

This result, together with the compatibility order derived from descriptive statistics, matches the empirical manifestation for H1. We can, therefore, conclude that H1 is to a large extent confirmed by AJT data. 7.2 Testing Hypothesis 2 Hypothesis 2 (H2) predicts that the effect of NOCOER constraint is stronger among low-level learners than among higher proficiency learners. That is to say, when performing the three data-elicitation tasks, the low-level participants would exhibit greater susceptibility to semantic markedness of the verb in aspect marking than their intermediate and advanced counterparts. The empirical manifestations for this hypothesis and the statistical techniques adopted to analyze the participants’ test performance, however, differ from task to task. 7.2.1 Testing Hypothesis 2 through Movie Clip Retell Task (MCRT) MCRT was a controlled production task (i.e., movie retelling) in which the participants were asked to mark the verbs of different situation types with proper tense-aspect morphology. The confirmation of H2 by means of MCRT data would manifest itself in the low-level participants’ preference for prototypical aspect-

143

Chapter 7 Results for Testing of Research Hypotheses

situation combinations when marking the verbs, because these combinations involve fewer violations of the NOCOER constraint. More particularly, these participants would tend to mark achievement verbs with the perfective aspect, or mark activity verbs with the progressive aspect, regardless of the discourse grounding status of these verbs. This tendency would be reflected in higher scores obtained by low-level participants on prototypical mappings like achievement-perfective and activityprogressive combinations than those obtained by the intermediate and advanced participants and the native controls. In the meantime, owing to the strong effect of the NOCOER constraint, the low-level participants would score lower than other participant groups on non-prototypical mappings. Using MCRT data to test H2, therefore, involves the examination of two independent variables, i.e., the participants’ proficiency level and the verb’s situation type. We can tell whether H2 is confirmed by comparing the mean scores for lowlevel participants and those for the other two L2 groups on aspect marking of each situation type. Figure 14 presents the mean scores of progressive aspect marking by all groups of participants on each situation type in MCRT.

62

As shown in the figure, the low-

level participants appeared to be more inclined to mark activity verbs with progressive aspect than the intermediate and advanced participants. It is also shown in the figure that their scores for progressive marking on the other three categories of event verbs, i.e., the accomplishments, the achievements, and the semelfactives, were lower than those obtained by the intermediate and advanced participants. These observations were congruous to our prediction that the low-level participants would score higher on the prototypical mapping between situation type and aspect marking, owing to the stronger effect of NOCOER constraint on this group of L2 learners. However, their score for progressive marking on the statives was surprisingly higher than the other two proficiency groups.

62

For details of statistic data for this figure, see Table D of Appendix IV. 144

Chapter 7 Results for Testing of Research Hypotheses

2.5

Mean Score

2

Low Intermediate Advanced Native

1.5

1

0.5

0 ACT

ACC

ACH

SMF

STA

Situation Types

Figure 14 Mean Scores for Progressive Aspect Marking by All Participant Groups on Each Situation Type in MCRT

Figure 15 is a visual presentation of the mean scores for perfective aspect marking by the participants of all four groups on each situation type of the verb. 63 3.5 3

Mean Score

2.5 Low Intermediate Higher Native

2 1.5 1 0.5 0 ACT

ACC

SMF

ACH

STA

Situation Type

63

For details of statistic data for this figure, see Table E of Appendix IV. 145

Chapter 7 Results for Testing of Research Hypotheses

Figure 15 Mean Scores for Perfective Aspect Marking by All Participant Groups on Each Situation Type in MCRT

It can be seen from the figure that the low-level participants, behaving differently from their performance in progressive aspect marking, scored lower for the perfective aspect marking on the activity verb, while their scores for perfective aspect marking on the other three types of verbs, i.e., achievements, accomplishments, and semelfactives, were invariably higher than the other two proficiency groups. These findings were in agreement with our prediction that the low-level participants were more likely to select the most prototypical aspect-situation mapping, due to the strong effect of NOCOER constraint. However, the results for perfective aspect marking on the statives presented a mixed picture: the score of the low-level participants was nearly the same as the advanced participants, lower than the intermediate participants and higher than the native controls. Two two-way ANOVAs (mixed design) were performed respectively to examine whether there was a significant interaction effect between the between-subject variable proficiency level and the within-subject variable situation type of verb on either perfective aspect marking or progressive aspect marking. The result is presented in Table 19.

Table 19 Effects of Proficiency Level and Situation Type of Verb on Progressive and Perfective Marking in MCRT: Results for Two-way ANOVA (Mixed Design)

Souce Situation Level Situation×Level

F 90.702 10.309 5.868

Progressive df 4 3,102 12, 408

p .000 .000 .000

146

F 47.161 2.074 47.161

Perfective df 4 3, 102 12, 408

p .000 .108 .000

Chapter 7 Results for Testing of Research Hypotheses

The test results show that the interaction between the situation type and the participants’ proficiency level reached significance level for both perfective marking and progressive marking. Simple main effects analysis (LSD) was performed to further identify the significant group differences. Table 20 presents the pairwise comparisons of the mean differences for progressive aspect marking and those for perfective aspect marking between low-level group and the three other groups on each situation type.

Table 20 Pairwise Comparisons of the Progressive and Perfective Marking on Verbs of Five Situation Types in MCRT (LSD)

Level

Aspect

Low vs. Intermediate Low vs. Advanced Low vs. Native

PROG PERF PROG PERF PROG PERF

STA .000 -.217 -.008 .000 -.512** .312

Verb Type ACT ACC .283 -.317* -.700*** .300* .008 -.567*** -633*** .433** .502** -.750*** -1.119*** .567**

ACH -.003 .010 -.267* .250 -1.006*** .790*

SMF -.233 .150 -.383* .150 -.669** .419*

*** p< .001, ** p< .01, *p< .05 As is shown in the table, the low-level group did not score significantly higher than two other L2 groups on prototypical aspect-situation mappings, i.e., progressive aspect marking on activity verbs and perfective marking on achievement verbs, the mappings that incur the fewest number of violations of NOCOER constraint, suggesting that they were not more sensitive to NOCOER constraint than the two higher proficiency groups. Their greater sensitivity was shown only when compared with the native speakers, as indicated by their significantly higher scores for both aspect-situation mappings. By contrast, the low-level group scored lower than the other three groups for the most non-prototypical aspect-situation mappings, i.e., the mappings that incur the greatest number of violations of NOCOER constraint, like activity-perfective and

147

Chapter 7 Results for Testing of Research Hypotheses

achievement-progressive combinations.

64

Table 20 shows that the mean differences

between the scores of the low-level participants and those of the other three groups reached significance level on both of these two non-prototypical mappings, the only exception being the mean differences between the low-level and intermediate in progressive aspect marking on the achievement verbs (Mean Difference = .003, p = .459 ). As was mentioned earlier, confirmation of H2 through MCRT requires evidence from two sources: the comparison of the mean scores on the prototypical aspectsituation mapping and the non-prototypical aspect-situation mapping between lowlevel participants and the other three groups. The general tendency shown by the evidence from descriptive statistics was in alignment with our predictions that lowlevel participants would use the prototypical mappings more frequently and nonprototypical mappings less frequently than the participants from the other three groups. In addition, the evidence obtained from inferential statistical analysis also gives support to our predictions on the non-prototypical scenarios. However, the mean differences in prototypical mappings among the low-level, intermediate, and advanced participants were not statistically significant. This shows that, on the one hand, the L2 proficiency of the low-level participants in our study may not be low enough: 65 they might have passed the initial stage of over-reliance on prototypical mapping. On the other hand, the spread from prototypical mapping to nonprototypical mapping has been conceptualized as a gradual process of constraint reranking from one stage of L2 proficiency to another, according to Gradual Learning Algorithm (GLA) in OT.

66

For these two reasons, the statistical insignificance

shown in post hoc tests in the case of prototypical mappings should not necessarily be

64

Note that the statives were excluded from consideration for reasons given in Note 55 and the linguistic analysis pertaining to Hypothesis 4. 65 Recall that the low-level participants were Senior Two middle school students and they had learned English for five years at the time when the experiment was carried out. See §6.2 for more information about the participants in our study. 66 See §5.1.2 for a more detailed discussion on GLA. 148

Chapter 7 Results for Testing of Research Hypotheses

taken as counter-evidence of H2. This issue will be further discussed in the next chapter. To sum up, H2 is confirmed by evidence from MCRT on the conditions that the statives were excluded from consideration and Gradual Learning Algorithm was assumed. 7.2.2 Testing Hypothesis 2 through Aspect Sentence Interpretation Task (ASIT) Recall that ASIT was designed to test whether the participants were able to correctly point out whether the event described in the critical aspect sentences embedded in a narrative discourse was complete or not. If H2 is true, we would then expect that, owing to the strong effects of NOCOER constraint, the low-level participants would be more likely to interpret the achievement predicates marked with progressive aspect as completed events, and at same time less likely to interpret the accomplishment predicates marked with perfective aspect as completed events. Testing H2 through ASIT data, therefore, involves an investigation of possible effects of interaction between proficiency level and situation-aspect mapping on the participants’ interpretation of the nature of the events as described in the critical test statements. Figure 16 presents the mean scores obtained by 4 participant groups for the two situation-aspect combinations where more violations of the NOCOER constraint were involved, i.e., in the progressive-achievement mapping and perfective-accomplishment mapping. 67 As indicated in the figure, the low-level participants were more likely to interpret the achievement predicates marked with progressive aspect as denoting completed events (Mean = .800, SD = .31) than intermediate (M = .650, SD = .33), advanced participants (Mean = .476, SD = .39), and native controls (Mean = .375, SD = .39). By contrast, they were least likely to interpret accomplishments marked with perfective as denoting completed actions among the four participant groups. This suggests that when interpreting the critical aspect statements, the low-level

67

For details of statistic data for this figure, see Table F of Appendix IV. 149

Chapter 7 Results for Testing of Research Hypotheses

participants tended to be guided by the temporal properties inherent in the verbs, rather than by the aspectual distinctions.

1 0.9 0.8 0.7 Mean

0.6 0.5

PROG-ACH PERF-ACC

0.4 0.3 0.2 0.1 0 Low

Intermediate Advanced Level

Native

Figure 16 Interpretation of Critical Aspect Sentences by Four Participant Groups

To further examine whether the interaction between the two variables under investigation was statistically significant, a two-way ANOVA (mixed design) was run. The results are shown in Table 21.

Table 21 Interaction between Proficiency Level and Situation-Aspect Mapping as Shown in ASIT: Results for Two-way ANOVA (Mixed Design)

Source Mapping Level Mapping×Level

F 10.355 .527 20.890

df 1 3, 102 3, 102

150

p .002 .665 .000

Chapter 7 Results for Testing of Research Hypotheses

As is shown in Table 21, the interaction between participants’ proficiency level and situation-aspect mapping reached significant level (F= 20.890, df= 3, 102, p=.000). Post hoc tests were run to further examine the details of interaction. Table 22 presents the pairwise comparisons of the mean differences of the scores between low-level group and three other groups on each situation-aspect combination.

Table 22 Pairwise Comparisons of the Aspect Sentence Interpretations by Participants of Lower Proficiency Group and the Other Three Groups

Lower vs. Intermediate Lower vs. Higher Lower vs. Native

PROG-ACH .150 .333** .425***

PERF-ACC -.075 -.192* -.412**

*** p < .001, ** p < .01, * p < .05

Table 22 shows that the low-level participants did not differ significantly from the intermediate-level participants in their interpretation of the critical test statements. However, they scored significantly higher than advanced participants and native controls when interpreting the critical test statements that contained achievement verbs marked with progressive aspect, indicating a stronger inclination to treat the achievement-progressive mappings as descriptions of completed events. On the other hand, they scored significantly lower than these two participant groups when interpreting the critical test statements that contained accomplishment verbs marked with perfective aspect, suggesting a weaker inclination to treat accomplishmentperfective combinations as descriptions of completed events. All these appear to indicate a stronger effect of NOCOER constraint on low-level and intermediate participants than on advanced-level participants and native controls. Thus, H2 was to a great extent confirmed by the ASIT data.

151

Chapter 7 Results for Testing of Research Hypotheses

7.2.3 Testing Hypothesis 2 through Acceptability Judgment Task (AJT) In AJT, the participants were asked to decide on the degree to which verbs of different situation types were compatible with the progressive aspect marker. More specifically, the participants were expected to make judgment on the acceptability of sentences marked with progressive on a 5-point Likert scale ranging from -2 to 2, which was literally interpreted as an acceptability continuum from “Totally Unacceptable” to “Completely Acceptable”. Recall that the NOCOER constraint is defined as “Do not coerce the shift of the temporal semantics inherent in verbal predicate”. Each violation of this constraint is assumed to lower the acceptability of the sentence. That is, the greater the effect of NOCOER constraint, the less acceptable the sentence marked with progressive. Thus, if H2 is true, the low-level participants would be less likely to accept sentences marked with progressive than native speakers of English and the L2 participants of higher proficiency level. Empirically, this entails a comparison between the mean score of the low-level group and that of each of the other three participant groups on the acceptability of the progressive marking on verbs of five situation types. Table 23 summarizes the results.

Table 23 Means and Standard Deviations of AJT by L2 Participants and Native Controls

Level Low Intermediate Advanced Native

STA Means SD -.592 .56 -.953 .57 -.773 .58 .001 .38

ACT Means SD 1.058 .63 1.300 .37 1.650 .33 1.892 .11

ACC Means SD .931 .39 1.008 .38 1.367 .30 1.889 .11

ACH Means SD -.254 .58 .013 .53 .079 .67 1.378 .56

SMF Means SD .146 .54 .433 .49 1.017 .49 1.142 .51

As shown in the table, the low-level participants scored lower than three other groups on all situation types except on the statives, where they scored higher (Mean

152

Chapter 7 Results for Testing of Research Hypotheses

=-.592, SD=.56) than intermediate (Mean=-.953, SD=.57) and advanced participants (Mean=-.773, SD=.58), but lower than the native controls (Mean=.001, SD=.38). Noteworthy in Table 23 was that the effect of the constraint was so strong that lowlevel participants found it difficult to accept the progressive marking on achievement and stative verbs, with the mean scores on both situation types below the zero point. This is tantamount to an unacceptable judgment of the aspect marking on the situation types under consideration, because the literal interpretation of a score below zero is “almost unacceptable” or “totally unacceptable” in AJT. 68 A two-way ANOVA (mixed design) was run to see whether the interaction between the two variables involved, i.e., proficiency level and situation type, was statistically different. The results are displayed in Table 24. It can be seen from the table that the interaction was statistically significant, i.e., F (12, 404) = 8.965, p =.000. To further examine the details of interaction, post hoc test (LSD) was performed. The results are presented in Table 25.

Table 24 Effects of Proficiency Level and Situation Type on Acceptability of Sentences Marked with Progressive: Results for Two-way ANOVA (Mixed Design)

Source Situation type Level Situation Type×Level

68

F 420.776 33.685 8.965

df 4 3, 101 12, 404

See §6.3.1 for a detailed discussion on this point. 153

p .000 .000 .000

Chapter 7 Results for Testing of Research Hypotheses

Table 25 Pairwise Comparisons of Mean Scores between Low-Level Group and Three Participant Groups for AJT

Lower vs. Intermediate Lower vs. Higher Lower vs. Native

STA .362* .182 -.615**

ACT -.242 -.592*** -.833***

ACC -.008 -.436*** -.958***

ACH -.242 -.333* -1.625***

SMF -.288 -.871*** -.996***

*** p<.001, ** p<.01

As indicated in the table, there was no significant difference between low-level participants and intermediate participants on all situation types except the statives. However, low-level participants scored significantly lower than the advanced participants on activity, accomplishment and semelfactive predicates, and the mean differences between low-level participants and native controls were significant on all situation types. This suggests that, compared with learners of higher proficiency, especially advanced L2 learners and native speakers, the low-level participants had a significantly lower level of acceptability of the progressive aspect on all dynamic situations. H2 is confirmed, therefore, for the four dynamic situations, but not for the statives. 7.3 Testing Hypothesis 3 H3 states that the effect of IDENT (TF) is greater among higher level participants than among lower level participants. Recall that IDENT (TF) is a constraint on aspect marking which says that the temporal interpretation of aspectually marked verbs in the input should be congruous with temporal requirement of the discourse in the output. Thus, to investigate the effect of this constraint requires cross-group comparisons of the participants’ performance on the two discourse-based tasks, i.e., movie clip retell task and aspect sentence interpretation task. Therefore, the confirmation of H3 would empirically manifest itself in the advanced participants’ greater sensitivity to the discourse requirement of the verb in aspect marking than that 154

Chapter 7 Results for Testing of Research Hypotheses

exhibited by the intermediate and low-level participants when performing tasks at the discoural level, such as MCRT and ASIT.

7.3.1 Testing Hypothesis 3 through Movie Clip Retell Task (MCRT) Recall that MCRT is a controlled production task in which the participants were asked to mark the verbs of different situation types with proper tense-aspect morphology when retelling the movie scenes they had just seen. If this prediction is true, we would observe stronger tendency among advanced participants to mark verbs with aspect that satisfies the discourse requirement. That is to say, compared with intermediate and low-level participants, their use of progressive aspect for discourse backgrounding and perfective aspect for discourse foregrounding would be significantly more frequent, and by contrast, their use of perfective aspect for discourse backgrounding and progressive for discourse foregrounding would be significantly less frequent than the other two groups. Table 26 below presents the means and standard deviations of aspect marking for two discourse grounding status by 4 participant groups.

Table 26 Aspect Marking for Two Discourse Grounding Status by 4 Participant Group in MCRT

Foreground Level Lower Intermediate Advanced Native

PERF Mean 2.887 3.360 3.620 3.650

SD .40 .53 .26 .44

Background

PROG Mean .620 .333 .113 .175

SD .37 .39 .16 .29

PERF Mean 2.167 1.840 1.353 1.013

SD .62 .60 .49 .17

PROG Mean 1.493 1.900 2.487 2.913

SD .71 .45 .48 .26

The table shows that advanced participants used more progressive markers in discourse backgrounding situations (M = 2.487, SD = .48) than low-level (M = 1.493, SD = .71) and intermediate learners (M = 1.900, SD = .45). When it came to the marking of discourse foregrounding, their use of perfective aspect was more frequent 155

Chapter 7 Results for Testing of Research Hypotheses

(M = 3.620, SD = .26) than that by the low-level (M = 2.887, SD = .40) and intermediate (M = 3.360, SD = .53) participants. On the other hand, the advanced participants tended to use perfective markers less frequently (M = 1.353, SD = .49) than lower-level (M = 2.167, SD = .62) and intermediate (M = 1.840, SD = .60) participants in backgrounding situations. They also tended to use progressive markers less frequently (M = .113, SD = .16) than lower-level (M = .620, SD = .37) and intermediate (M = .333, SD = .39) participants in the foregrounding situations. To further examine whether the observed differences in mean scores were statistically significant, four one-way ANOVAs were performed, one for each aspect under each grounding situation. In addition, post hoc tests were conducted where Fvalues reached the significance level. The results are presented in Table 27. This table shows that all F-values reached the significance level at p < .05. Post hoc tests revealed that the mean differences of advanced vs. low and advanced vs. intermediate were statistically significant for both foregrounding and backgrounding situations. More specifically, in foregrounding situations, the advanced group’s scores for perfective marking were significantly higher than those obtained by the low-level and the intermediate groups. Similarly, in backgrounding situations, their scores for progressive marking were also significantly higher. By contrast, they scored significantly lower than low-level and intermediate participants on perfective marking in backgrounding situations and on progressive marking in foregrounding situations. The advanced learners behaved like native speakers when aspect-marking the foregrounding information, but, compared with the native speakers, their aspect marking performance in backgrounding situations was much less native-like, because they preferred to mark backgrounding information with perfective aspect rather than with progressive aspect. These results suggested that while advanced participants were more sensitive than low-level and intermediate participants to discourse constraints regarding the foregrounding and backgrounding information, their sensitivity was not entirely native-like when marking the backgrounding events with appropriate asepct. But 156

Chapter 7 Results for Testing of Research Hypotheses

since their behaviours in aspect marking the discoursal information conformed to a great extent to what has been predicted, it can be concluded that H3 is supported by the participants’ performance on MCRT.

Table 27 Effects of Proficiency Level on Aspect Marking of Discourse Grounding in MCRT: Results for One-way ANOVAs

F-value (df = 3,102) p-value Advanced vs. Low Advanced vs. Intermediate Advanced vs. Native

7.3.2

Foreground PERF PROG 19.452 13.976 .000 .000 .733*** -.507*** .260* -.220** .003 -.006

Background PERF PROG 21.619 32.564 .000 .000 -.813*** .993*** -.487*** .587*** .341* -.426*

Testing Hypothesis 3 through Aspect Sentence Interpretation Task (ASIT) In ASIT, the participants were required to interpret aspect sentences embedded in

narrative discourse. Recall that for each aspect sentence in the task, there is one critical statement pertaining to the completion of the event denoted by the aspect sentence. Recall also that a higher score for the performance on the task indicated a greater likelihood that the event involved was interpreted as a completed one. Since the perfective aspect typically makes reference to a completed event while the progressive aspect expresses an action in progress, we would expect that the participants sensitive to the effect of IDENT (TF) would score higher on perfective sentences and lower on progressive sentences in ASIT. That is to say, compared with lower-level and intermediate participants, advanced participants would be more likely to score higher on perfective sentences and lower on the progressive sentences. Another piece of evidence that can be used to support H3 comes from the comparison of the mean scores for the aspect in within-text position and after-text position. Recall that in ASIT position is the variable intended for examining the

157

Chapter 7 Results for Testing of Research Hypotheses

participants’ ability to integrate aspectual information in text interpretation. Recall also that in ASIT the critical test statements were alternated between within-text position and after-text position to tap the participants’ online processing of the aspectual distinctions in discourse. The basic assumption underlying this design was that the participants would be more accurate in responding to the test statement given immediately after the aspect sentence, i.e., the within-text test statement, than to the test statement postponed to the end of the text, i.e., the after-text statement, if they were sensitive to the aspectual distinctions in discourse (Magliano & Schleich, 2000). If H3 is true, then advanced participants would score significantly lower when responding to the within-text test statements for critical sentences marked with progressive aspect than when responding to the after-text test statement. On the other hand, they would score significantly higher when responding to the within-text test statements for critical sentences marked with perfective aspect than when responding to the after-text statement. In contrast, low-level participants, being insensitive to the aspectual distinctions in discourse, would show no difference in responding to the same test statements located in two different positions of the text. Testing H3 through ASIT data, therefore, requires two types of evidence. The first comes from the comparison of mean scores obtained by advanced participants and those obtained by other groups of participants for temporal interpretation of both perfective and progressive aspect marking. The second comes from group comparison of the mean scores for temporal interpretation of the aspect sentences addressed by critical test statements in different positions of the text. In the former case, two variables are involved, i.e., the participants’ proficiency level and aspect of sentence predicate, and in the latter case, three variables are involved, i.e., the participants’ proficiency level, aspect of sentence predicate, and the position of the test statements in the texts. 7.3.2.1 Testing H3 through ASIT Data: First Type of Evidence

158

Chapter 7 Results for Testing of Research Hypotheses

Table 28 presents the mean scores and standard deviations for aspect sentence interpretation by four participant groups.

Table 28 Mean Scores and Standard Deviations for Aspect Sentence Interpretation by Four Participant Groups (ASIT )

PROG Low Intermediate Advanced Native

Mean .622 .650 .506 .312

PERF SD .21 .20 .26 .25

Mean .617 .650 .794 .917

SD .20 .18 .20 .21

As shown in the table, participants of low- level and intermediate groups scored similarly on temporal interpretation of progressive aspect and perfective aspect i.e., .622 vs. .617 for low-level participants, and .650 vs. .650 for intermediate-level participants. The advanced-level participants, however, scored higher on test statements for perfective aspect sentences (Mean = .794, SD = .20) than on the test statements for progressive aspect sentences (Mean = .506, SD = .26). Their performance in this task was more approximate to that of the native speakers than the other two L2-English procifiency groups. To investigate whether the observed differences in the aspectual interpretation were statistically significant, a two-way ANOVA (mixed design) was performed. The results are presented below in Table 29. Both the main effect of Aspect and the interaction of Aspect×Level are significant, suggesting that while the sensitivity to the aspectual distinction did exert significant influence on the participants’ interpretation of the test statements, this may be due to their different proficiency levels. To further examine the details of interaction effect, simple main effects analysis (LSD) was further performed. The results are displayed in Table 30.

159

Chapter 7 Results for Testing of Research Hypotheses

Table 29 Effects of Sentence Aspect and Proficiency on Temporal Interpretation of Test Statement in ASIT: Results for Two-way ANOVA (Mixed Design) Source Aspect Level Aspect×Level

F 65.296 .381 23.491

df 1 3, 102 3, 102

p .000 .767 .000

As is shown in the table, low-level participants and intermediate participants did not differ significantly in interpreting both progressive and perfective sentences, suggesting that these two groups of participants were equally insensitive to the aspectual distinction in discourse. Advanced participants, however, scored significantly lower than low-level and intermediate participants when interpreting the aspect sentences marked with the progressive aspect, but they scored significantly higher than these two L2-English groups in interpreting sentences marked with the perfective aspect. In other words, the advanced participants were more likely to interpret the events in progressive sentences as incomplete and those in perfective sentences as complete. This suggests that advanced participants were more sensitive to the aspectual distinctions in discourse than their lower level counterparts, just as has been predicted.

Table 30 Pairwise Comparison of the Mean Scores for Aspect Sentence Interpretation by Four Participant Groups (ASIT)

F-Value (df = 3,102) p-Value Low vs Intermediate Advanced vs. Low Advanced vs. Intermediate Advanced vs. Native

PROG 9.269 .000 -.028 -.117* -.144* .193**

*** p < .001, ** p < .01, * p < .05

160

PERF 12.323 .000 .033 .178*** .144** -.122*

Chapter 7 Results for Testing of Research Hypotheses

Significant differences also existed between the mean scores obtained by the advanced participants and those obtained by the native controls, indicating that, despite their advanced proficiency level, these Chinese EFL learners had not reached native-likeness in interpreting temporal messages expressed the sentences marked with different aspect sentences. This result, however, does not affect the above conclusion. 7.3.2.2 Testing Hypothesis 3 through ASIT Data: Second Type of Evidence As has been mentioned, another piece of evidence in support of H3 in ASIT comes from the comparison of the mean scores for the test statements located at two different positions of the text. Table 31 shows four participant groups’ mean scores and standard deviations for the test statements in these two positions.

Table 31 Mean scores and Standard Deviations for Aspect Test Statement in Withinand After-text Position in ASIT

PROG

Low Intermediate Advanced Native

Within-text Mean SD .633 .24 .678 .30 .411 .38 .208 .24

PERF After-text Mean SD .611 .32 .622 .26 .600 .21 .417 .17

Within-text Mean SD .633 .22 .644 .28 .856 .27 .958 .33

After-text Mean SD .600 32 .656 .28 .733 .25 .875 .11

It can be seen from the table that when interpreting the test statements marked with progressive aspect, the probability for advanced participants to judge those in within-text position as describing incomplete events was lower (Mean = .411, SD = .38) than that of judging those in after-text position (Mean = .600, SD = .21). In contrast, when interpreting the test statements marked with the perfective aspect, the probability of judging those in within-text position was higher (Mean = .856, SD

161

Chapter 7 Results for Testing of Research Hypotheses

= .27) than that of judging those in after-text position (Mean = .733, SD = .25). In other words, advanced participants’ correct interpretation of test statements in withintext position appeared to be more frequent than that of test statements in after-text position and their performance is more approximate to the patterns exhibited by the native speakers. However, low-level and intermediate participants did not seem to behave differently when interpreting the test statements located in these two different positions of the text. A three-way ANOVA was run to examine whether the observed differences were statistically significant. The results are presented in Table 32.

As shown in the

Table, in addition to the significant main effect of Aspect and Level, the two-way interaction of Aspect and Level and the three-way interaction of Aspect, Level and Position are also significant, suggesting that while the sensitivity to aspectual distinction in discourse and proficiency level may jointly influence the participants’ interpretation of test statements, this may be due to the position of statements in the text. To further examine this interaction effect, simple main effects analysis (LSD) was performed.

Table 32 Three-way ANOVA for Test Statement Interpretation Related to Aspect, Position and Proficiency Level (ASIT)

Source Aspect Position Level Aspect×Level Position×Level Aspect×Position Aspect×Position×Level

df 1 1 3, 102 3, 102 3, 102 1 3, 102

F 65.296 1.507 9.007 23.491 1.858 3.662 3.300

162

p .000 .222 .000 .000 .141 .058 .023

Chapter 7 Results for Testing of Research Hypotheses

Table 33 presents the results for pairwise comparisons of the mean scores for aspect sentence interpretation in two positions of the discourse by the four participant groups.

Table 33 Pairwise Comparisons of Mean Scores for Aspect Sentence Interpretation by Four Participant Groups (ASIT)

Level Low Intermediate Advanced Native

Position Within-text vs After-text Within-text vs After-text Within-text vs After-text Within-text vs After-text

PROG .022 .056 -.189** -.208*

PERF .033 .011 .122* .083

*** p < .001, ** p < .01, * p < .05

As is indicated in the table, when responding to test statements marked with progressive aspect, the advanced participants scored significantly lower on the withintext statements than on their after-text test counterparts; but when responding to test statements marked with perfective aspect, they scored significantly higher on the within-text statements than on the after-text ones. This tendency to respond more accurately to the within-text test statements suggests that advanced participants were more capable of integrating the temporal messages conveyed by the aspect sentence into discourse interpretation, an ability largely absent among the low-level and intermediate participants, as indicated by the non-significant differences between their responses to the within-text statements and those to the after-text statements. This is a clear indication that advanced participants were more sensitive to the aspectual distinctions in discourse interpretation, an inclination that low-level and intermediate participants did not seem to have developed. 69

69 Also revealed in this table is that the native controls scored significantly lower on the test statements marked with progressive aspect in the within-text position than on those in the aftertext position, while their scores on the test statements marked with perfective aspect did not show

163

Chapter 7 Results for Testing of Research Hypotheses

This finding conforms to our predictions that advanced participants were more likely to give correct temporal interpretations to the test sentences located within the text than after the text. Therefore, H3 was further supported. Taking both pieces of evidence together, we can conclude that H3 is confirmed by the ASIT data. 7.4 Testing Hypothesis 4 Recall that Hypotheses 4 predicts the effects of the cross-linguistic constraints on aspect marking. More specifically, it states that Chinese EFL learners, at least those at their initial stage of L2 learning, were more likely to accept and produce more progressive statives than native speakers of English. This hypothesis is formulated on the basis of the cross-linguistic observation that Chinese progressive marker –zhe is compatible with a subset of stative verbs (Li & Thompson, 1981), while the English statives are typically subject to the violations of NOCOER constraint with progressive aspect 70 . If this hypothesis is true, it will be observed that Chinese EFL learners, especially those at lower proficiency levels, find it less difficult to accept, interpret, and produce stative-progressive combinations than English native speakers do. In the present study, this evidence will be sought primarily by comparing (i) the performance of elicitation tasks by Chinese EFL participants with that by native speakers of English, and (ii) performance of elicitation tasks by low-level participants with that by intermediate and advanced participants. Since two out of the three tasks (i.e., AJT and MCRT) involved aspect marking on the statives, only the results for the participants’ performance on these two tasks will be reported.

significant differences between the two positions. This, however, should not be taken as evidence that the native speakers were not sensitive to the aspectual distinctions in the discourse. If we examine the mean scores given by native controls, as shown in Table 31, we can see that it was highly probable that the native controls invariably interpreted the perfective aspect sentences as denoting completed events, whether the test sentence appeared in the within-text position (M = .985, SD = .33) or after-test position (M = .875, SD = .11). 70 See §3.5 for a detailed discussion on this point. 164

Chapter 7 Results for Testing of Research Hypotheses

7.4.1 Testing Hypothesis 4 through Acceptability Judgment Task (AJT) In AJT the participants were asked to judge the compatibility of verbs of different situation types with the progressive aspect in English. Recall that acceptability is a scalar notion in our study whereby higher score indicates greater acceptability. Therefore, if H4 is confirmed, we would then expect to observe that the participants in the native speaker group score significantly lower than the L2 participant groups do on the items testing the compatibility of the progressives with the statives, owing to the cross-linguistic differences between English and Chinese on marking the statives with the progressive aspect. In addition, as the influence of the mother tongue on tense-aspect marking is assumed to be stronger in the early stage of L2 development (Slabakova, 2000a), we would also expect that low-level participants score significantly higher on the statives, but lower on other situation types than intermediate and advanced participants do: the evidence of overuse of the progressive marking on statives as a result of mother tongue influence. Table 34 presents the results for the participants’ performance on AJT.

Table 34 Mean Scores and Standard Deviations for Acceptability of the Progressive Marking on the Stative Verbs in AJT

Low Intermediate Advanced Native

Mean -.592 -.953 -.773 .023

SD .56 .57 .58 .38

It can be seen from the table that, contrary to our prediction, the native controls scored higher than the three other L2 groups did. However, as predicted, the low-level group scored higher than intermediate and advanced groups.

165

Chapter 7 Results for Testing of Research Hypotheses

The fact that native speakers did not perform as predicted, however, should not be immediately taken as counter-evidence for H4, because stativity is an umbrella notion consisting of a wide range of subclasses of the stative verbs. Recall that, following Quirk, et al. (1985),

71

five subtypes of stative verbs were included in AJT,

i.e., intellectual statives, emotional statives, property statives, relational statives, and sensational statives. A closer examination of the data from these subtypes would help reveal more about the participants’ perceived compatibility of the progressive verbs with stative verbs. Figure 17 presents results for the five subtypes. 72 As can be seen from the figure, compared with the other three L2 participant groups, the native speaker group scored lower on relational statives and property statives, but higher on the other three subclasses of the statives. The figure also shows that low-level participants scored higher than intermediate and advanced participants on all the subtypes except for sensational statives, where they scored higher than intermediate participants but lower than advanced participants.

2 1.5

Mean Score

1 0.5 0 Property

Relational

Intellectual

Emotional

Sensational

-0.5 -1 -1.5 -2 Subtypes of Statives

71 72

See §6.3.1, and in particular, Table 7 for details of this point. For details of statistic data for this figure, see Table G of Appendix IV. 166

Low Intermediate Advanced Native

Chapter 7 Results for Testing of Research Hypotheses

Figure 17 Acceptability of Aspect Marking on Five Subclasses of Stative Verbs by Four Participant Groups

To further investigate whether the observed differences exhibited by four participant groups when judging the five subtypes of stative verb are statistically significant, a two-way ANOVA (mixed design) was performed. The results are presented below in Table 35.

Table 35 ANOVA for Acceptability Judgment of Five Subtypes of Stative Verb by Four Participant Groups

Source Stative Level Stative×Level

F 74.351 11.259 26.441

df 4 3, 101 12, 404

p .000 .000 .000

The interaction is significant, suggesting that proficiency level and subtypes of statives jointly exert influence on the participants’ judgment of the compatibility of progressive aspect with statives. To further examine the details of interaction effect, simple main effects analysis was performed. Table 36 presents the results for pairwise comparisons of the mean scores obtained by the four participant groups for the five subtypes of stative verbs. As shown in the table, the native controls scored significantly lower than intermediate and advanced participants on relation and property verbs (as indicated by the minus signs of the mean differences), but they scored significantly higher than all three L2 participant groups on the other three subclasses. Low-level participants, on the other hand, scored significantly higher than (i) intermediate participants on

167

Chapter 7 Results for Testing of Research Hypotheses

emotional, property, and sensational statives, (ii) advanced participants on property statives, and (iii) the native controls on relational statives.

Table 36 Pairwise Comparisons of Mean Scores for Five Subclasses of Stative Verbs in AJT

Low vs. Intermediate Low vs. Advanced Low vs.Native Native vs. Intermediate Native vs. Advanced

Intellectual .267 .258 1.717*** 1.983*** 1.975***

Relational .092 .075 .658* -.567* -.642*

Emotional .517* .158 -1.083* 1.600*** 1.242***

Property .367* .592*** .842*** -.475* -.250*

Sensational .567** -.117 -1.775*** 2.342*** 1.658***

*** p < .001, ** p < .01, * p < .05

To reveal the possible regularity underlying the participants’ perception of compatibility of progressive aspect with stative verbs, Smith’s (1997) categorization of stative verbs has to be drawn upon. According to Smith (1997), stative verbs fall into two broad categories: individual-level statives and stage-level statives, which are analogous to Dowty’s dichotomy between “interval stative verb” and “momentary states” (Dowty, 1979: 180). Property and relational statives belong in the former category, which typically denotes a more or less permanent property of individuals and so cannot occur with progressive aspect. Emotional, intellectual and sensational statives belong in the latter category, which denotes a temporal state and so can felicitously occur with the progressive aspect (Filip, 1999; Smith, 1997). In the light of this distinction, it can be seen that native control did perform to a large extent as predicted in H4, because they scored significantly lower than all three L2 participant groups on individual-level statives, i.e., the property and relational statives, except for the comparison between the native controls and advanced participants. By contrast, they scored significantly higher than all the L2 participant groups on stage-level statives, i.e. the emotional, intellectual and sensational statives.

168

Chapter 7 Results for Testing of Research Hypotheses

Although low-level participants did not perform as predicted (i.e., scoring significantly higher on individual-level statives and lower on stage-level statives than the other two L2 participant groups), a closer examination of their scores on the statives as a whole and their scores on other four situation types indicates that they scored higher on the statives but lower on other situation types than intermediate and advanced participants (see Table 23 in §7.2.3), suggesting a relatively lenient judgment on the stative-progressive combination. Taking these two pieces of information together, we can conclude that H4 is partially confirmed by AJT data in that only the evidence from the participants’ performance on individual-level statives, but not the stage-level ones, supported this hypothesis. 7.4.2

Testing Hypothesis 4 through Movie Clip Retell Task (MCRT) In MCRT, the participants were required to mark the verbs in discourse with the

proper tense-aspect forms after they watched clips of video shows of the British comedy Mr. Bean. The verbs were selected from the narrative describing the story of the shows and were balanced across discourse grounding status and situation types. If H4 is true, we would then observe that, (i) owing to the cross-linguistic differences between English and Chinese, the Chinese EFL learners mark stative verbs with the progressive aspect more frequently than native speakers of English, and (ii) among the Chinese EFL learners, those at lower level of proficiency mark stative verbs with progressive aspect more frequently than those at higher level of proficiency. Empirically, this prediction can be tested by comparing (i) the task performance of the native controls and three other L2 participant groups, and (ii) the task performance of the three L2 participant groups. Table 37 displays the mean scores and standard deviations for progressive aspect marking on the stative verbs in MCRT by participants of all groups. As shown in the Table, low-level participants scored higher than both intermediate and advanced

169

Chapter 7 Results for Testing of Research Hypotheses

participants, and intermediate participants scored lower than advanced participants. In addition, the native controls scored higher than all three L2 participant groups.

Table 37 Mean Scores and Standard Deviations for Progressive Aspect Marking on the Stative Verbs by Four Participant Groups in MCRT

Level Low Intermediate Advanced Native

Mean .817 .550 .633 .844

SD .38 .42 .43 .30

To further examine whether the observed differences in mean scores were statistically significant, one-way ANOVA was performed. It was found that the differences were significant (F = 3.229, df = 3, 102, p = .026). To determine which group means differ significantly from each other, post hoc test (LSD) was conducted at p<.05. Table 38 presents the results of pairwise comparisons of the mean scores when the stative verbs were marked by the progressive aspect.

Table 38 Pairwise Comparisons of the Mean Scores for Progressive Aspect Marking on the Stative Verbs by Four Participant Groups in MCRT

Low Intermediate Advanced

Intermediate .267*

Advanced .183 -.083

Native -.027 -.294* -.210

As shown in the table, among the three L2 participant groups, only low-level participants scored significantly higher than intermediate participants. The mean

170

Chapter 7 Results for Testing of Research Hypotheses

difference between low-level group and advanced group, and that between low-level group and native speaker group did not reach the significance level at p < .05. Besides, intermediate participants scored significantly lower than the native speaker participants. Thus, H4 was only marginally confirmed. To further explore the possible causes of the native speakers’ higher scores on progressive aspect marking of stative verbs, which was contrary to the prediction made in H4, the types of the statives verbs used in MCRT were closely examined. It was found that the majority of the target verbs belong in the category of the aforementioned stage-level statives (Smith, 1997). More specifically, among the eight verbs used in MCRT, seven were stage-level statives, i.e., bother, wonder, express, amuse, prefer, decide and concentrate, which, according to Dowty (1979) and Filip (1999), can occur felicitously with the progressive aspect in English. This seems to explain why the native controls scored unexpectedly higher than the Chinese participants in MCRT. To sum up, Hypothesis 4 was not fully confirmed by the MCRT data. This incongruity between theoretical hypothesis and empirical finding is partly attributable to the fact that we failed to take into account the distinction between stage-level statives and individual-level statives, especially when we decided on the target verbs for MCRT, where the number of these two subtypes of the statives verbs was thrown out of balance. 7.5 Testing Hypothesis 5 Hypothesis 5 is also concerned with the cross-linguistic constraints on L2 tenseaspect acquisition, which is based on the observation that Chinese does not license the progressive marking on the achievement verbs, while the English language does. This hypothesis predicts that Chinese learners of L2 English, especially those at their early stage of L2 learning, would not be inclined to accept and produce the progressive achievements, when compared with native speakers of English. If this is true, we would then be able to observe that Chinese EFL learners, in particular those of lower 171

Chapter 7 Results for Testing of Research Hypotheses

proficiency level, experience greater difficulty than the native speakers do in interpreting and producing stative verbs marked with progressive aspect. In the present study, such evidence can be sought by observing how the native speaker participants behave differently from the L2 participants, especially those at lower proficiency level, when performing the three elicitation tasks. 7.5.1 Testing Hypothesis 5 through Acceptability Judgment Task (AJT) Recall that in AJT the participants were asked to judge the degree of acceptability of the sentences containing progressive predicates of various situation types. If H5 is true, it will be observed that native controls are more likely to accept the progressive marking on achievement verbs, and that the Chinese participants, in particular those of lower proficiency level, are more likely to reject the progressive marking on achievement verbs. Recall also that the acceptance of progressiveachievement combination was indicated by scores higher than 0, while the rejection of this combination was indicated by scores below 0. Table 39 presents the mean scores and standard deviations for acceptability of the progressive aspect marking on achievement verbs by the four participant groups.

Table 39 Mean Scores and Standard Deviations for Acceptability of the Progressive Marking on Achievement Verbs in AJT by Four Participant Groups

Low Intermediate Advanced Native

Mean -.254 -.013 -.079 1.37

SD .58 .53 .67 .56

As is shown in the Table, the native speaker group scored higher than 0 (Mean = 1.375, SD = .56), indicating that they tended to accept progressive aspect marking on

172

Chapter 7 Results for Testing of Research Hypotheses

achievement verbs. In contrast, all three L2 participant groups scored lower than 0, suggesting that they tended to reject the progressive-achievement combination. To find out if the observed differences between the above group means were statistically significant, one-way ANOVA was performed. It was found that the difference was significant (F = 27.042, df = 3, 101, p = .000). To determine which group means differ significantly from each other, post hoc test (LSD) was conducted. The results are presented in Table 40.

Table 40 Pairwise Comparisons of Mean Scores for Acceptability of Progressive Aspect Marking on Achievement Verbs in AJT

Low Intermediate Advanced

Intermediate -.242

Advanced -.333* -.092

Native -1.629*** -1.388*** -1.296***

***p<.001, **p<.01, *p<.05

As is indicated in the table, the native controls scored significantly higher than the three groups of Chinese participants. Also shown in this table was that the lowlevel group scored significantly lower than the advanced group. This suggests that Chinese learners of English, particularly those in their early stage of English tenseaspect acquisition, tended not to accept the progressive-marked achievement verbs. The prediction made in H5 is therefore confirmed by the AJT data. 7.5.2 Testing Hypothesis 5 through Movie Clip Retell Task (MCRT) As mentioned, in MCRT, the participants were required to mark the verbs embedded in a passage with proper tense-aspect forms based on their understanding of the events described in the video show of the British comedy Mr. Bean. Thus, the evidence that can be used to confirm H5 would be that the native speaker participants mark achievement verbs with progressive aspect more frequently than the Chinese

173

Chapter 7 Results for Testing of Research Hypotheses

participants. Empirically, this entails a comparison of the mean scores for progressive aspect marking on achievement verbs obtained by the 4 participant groups. Table 41 displays the mean scores and standard deviations of each group for the progressive marking on achievement verbs. As can be seen from the table, the likelihood of progressive aspect marking on achievement verbs increased with the participants’ English proficiency. The native speakers of English scored highest (M = 1.656, SD = .24), followed by advanced participants (M = .917, SD = .49), intermediate participants (M = .683, SD = .43) and low-level participants (M = .650, SD = .42) respectively.

Table 41 Mean Scores and Standard Deviations for Progressive Aspect Marking on Achievement Verbs in MCRT by Four Participant Groups

Mean Low Intermediate Advanced Native

.650 .683 .917 1.656

SD

.42 .43 .49 .24

To find out if the observed differences between the above group means were statistically significant, one-way ANOVA was performed. It was found that the difference was significant (F = 23.187, df = 3, 102, p = .000). To determine which group means differ significantly from each other, post hoc test (LSD) was conducted. The results are presented in Table 42. As can be seen from the table, the native controls scored significantly higher than all three groups of Chinese participants. Among the three L2 participant groups, the advanced group scored significantly higher than the other two groups, but the intermediate group did not score significantly higher than the low-level group. Therefore, H5 is to a large extent confirmed by the MCRT data.

174

Chapter 7 Results for Testing of Research Hypotheses

Table 42 Pairwise Comparisons of Mean Scores for Progressive Aspect Marking on Achievement Verbs in MCRT

Low Intermediate Advanced

Intermediate -.033

Advanced -.267* -.233*

Native -1.005*** -.973*** -.739***

***p<.001, **p<.01, *p<.05

7.5.3 Testing Hypothesis 5 through Aspect Sentence Interpretation Task (ASIT) Recall that in ASIT, the participants were asked to interpret aspect sentences embedded in narrative discourse. Also recall that for each aspect sentence, there was one critical statement addressing the completion of the event denoted by the aspect sentence, and that a higher score obtained for task performance indicated a greater probability that the event under concern was interpreted as a completed one. Since progressive aspect typically refers to an incomplete event, it was predicted that the participants sensitive to the form-function mapping of progressive (i.e., mostly those at higher level of proficiency) would tend not to interpret the event denoted by a progressive aspect sentence as a completed one, and thus would obtain a lower score, regardless of situation types of the verbal predicates. That is to say, if H5 is true, it will be observed that native speaker participants score lower on the sentences containing achievement verbs marked with progressive aspect. By contrast, the L2 participants, especially those of low-level proficiency, would score higher than the native speakers on the same type of sentences, since they are less sensitive to the aspectual information conveyed by the progressive aspect marker, due to the fact that the progressive-achievement mapping is not licensed in Chinese. Testing H5 through ASIT data, therefore, entails a comparison of the mean scores for the aspectual interpretation of achievement predicates marked with progressive

175

Chapter 7 Results for Testing of Research Hypotheses

aspect by all groups of participants. Table 43 presents the relevant results for each group.

Table 43 Mean Scores and Standard Deviations for Interpretation of Sentences Containing Achievement Predicates Marked with Progressive Aspect

Mean Low Intermediate Advanced Native

.800 .650 .467 .375

SD

.31 .33 .39 .39

It can be seen that the native controls scored the lowest among the four participant groups (M = .375, SD = .39), followed by the advanced group (M = .467, SD = .39) and the intermediate group (M = .650, SD = .33) respectively. In contrast, the low-level group scored the highest (M = .800, SD = .31). To find out if the observed differences between the above group means were statistically significant, one-way ANOVA was performed. It was found that the difference was significant (F = 7.070; df =3, 102; p=.000). To determine which group means differ significantly from each other, post hoc test (LSD) was conducted. The results are presented in Table 44.

Table 44 Pairwise Comparisons of Mean Scores for Interpretation of Sentences Containing Achievement Predicates Marked with Progressive Aspect

Low Intermediate Advanced

Intermediate .150

Advanced .333*** .183*

***p<.001, **p<.01, *p<.05 176

Native .425*** .275* .092

Chapter 7 Results for Testing of Research Hypotheses

The table shows that both low-level and intermediate participants scored significantly higher than the native controls and advanced participants. But the mean difference between low-level and intermediate participants, and that between advanced and native speaker participants did not reach the significance level. This is a clear indication that participants of lower proficiency levels tended to interpret the progressive achievement predicates as denoting completed events, but such a tendency is to a large extent absent in the participants of more advanced proficiency levels. Therefore, H5 is basically confirmed by the ASIT data. 7.6 Summary In this chapter, we have empirically tested the five hypotheses formulated in §5.2 through the triangulation of data from a production task, an interpretation task and a judgment task data. The first hypothesis predicts that more violation of NOCOER constraint gives rise to greater acquisitional or processing difficulty for progressive and perfective aspect marking. In particular, we predicted that there would be an ACT>ACC, SMR>ACH>STA acquisitional order for the progressive aspect and an ACH>ACC, SMF>ACT>STA order for the perfective aspect. This hypothesis has been fully confirmed by ASIT data and to a large extent by AJT data. This order was also conditionally validated in MCRT in that the predicted order emerged if we exclude the perfective aspect marking on the statives and assume the Gradual Learning Algorithm proposed by Boersma and Hayes (2001). H2 predicts that the effect of NOCOER constraint is stronger among low-level learners than among learners more proficient in L2 English. This hypothesis predicts that low-level L2 learners favor prototypical mapping between aspect marking and the situation types of verbs. The hypothesis was fully supported by ASIT data, and it was largely supported by AJT and MCRT data on the condition that the statives situations were not taken into consideration and that Gradual Learning Algorithm was assumed. 177

Chapter 7 Results for Testing of Research Hypotheses

It was hypothesized in H3 that the effect of IDENT(TF) is greater among higher level participants than among lower level participants. Put differently, this hypothesis predicted that advanced participants are more susceptible to discourse requirement of the verb in aspect marking than intermediate and low-level groups of L2 English learners. This hypothesis was fully confirmed by the data obtained from MCRT and ASIT. Both H4 and H5 were generated on the basis of cross-linguistic differences between Chinese and English progressive aspect marking. It was hypothesized in these two hypotheses that the Chinese L2 learners, particularly those at low proficiency level, would overuse the progressive aspect on the statives and underuse it on the achievement, owing to L2 influence on progressive aspect marking. H4, the hypothesis pertaining to progressive marking on the statives, was conditionally supported by the data from AJT in that only the evidence from the individual-level statives, but not that from the stage-level ones, renders full support to this hypothesis. This hypothesis was not fully supported by MCRT data, partly because of the imbalance of the target words on two subtypes of the statives. H5 addressed the progressive marking on the achievement verbs and it was fully confirmed by data obtained from all the three experimental tasks. Thus, the five hypotheses were confirmed, to varying degrees, by the evidence from the three experimental tasks.

178

Chapter 8 General Discussion This chapter addresses the fourth key research question raised in §1.3, i.e., to what extent can the observed behaviors of L2 tense-aspect acquisition be explained in a principled way within the unified framework? To put it more accurately, this chapter discusses L2 acquisition of the English progressive aspect within the OT framework proposed in Chapter 4 and Chapter 5, in relation with the linguistic and empirical findings presented in the preceding chapters and in the extant literature. This chapter is organized as follows. In § 8.1, the developmental route followed by Chinese EFL learners in their acquisition of the English tense-aspect is delineated on the basis of a summary of the major findings. § 8.2 discusses the major findings, focusing on what new light they shed on our understanding of L2 tense-aspect acquisition, with a particular reference to the OT account of language acquisition. These issues include an OT account of the initial state, the triggering factor in tense-aspect acquisition, the spread from prototypical to nonprototypical mapping, and the role of stative verbs in progressive aspect marking. § 8.3 presents a discussion on the explanatory power of OT for L2 tense-aspect acquisition. § 8.4 presents a summary of this chapter.

Chapter 8 General Discussion

8.1 Developmental Route of English Tense-aspect Acquisition by Chinese EFL Learners 8.1.1 A Summary of Major Findings --- OT, a theory on the resolution of conflicting forces in human language, has been tentatively proposed as a candidate theoretical framework explaining L2 progressive aspect acquisition. This is the first attempt in L2 studies to identify, in the unified OT framework, the constraints on the acquisition of L2 tense-aspect system, in particular of the progressive aspect. Unlike many linguistic theories available, OT evaluates in parallel the constraints belonging to different linguistic categories and dimensions. This feature makes OT a powerful apparatus of explaining linguistic phenomenon involving multifaceted and cross-modular “interface” property like aspect marking (Kager, 1999: 25). Optimality Theory enables us to examine the conflicts and interactions between these cross-modular constraints and evaluate them in one single hierarchy, i.e., IDENT (TF) >> MARK(T) >> NOCOER, which is validated by the test of all the four scenarios of situation-grounding combinations. This effort to theorize the constraints on tense-aspect marking and their effect on L2 acquisition is to a large extent successful, as has been demonstrated by the results of empirical investigation, --- The natural combinations between aspect form and temporal semantics of the verb, i.e., the default unmarked form-meaning mappings involving fewer violations of the NOCOER constraint, e.g., the activity-progressive and achievement-perfective mappings, are found to be acquired earlier, judged as more acceptable, and interpreted more correctly in narrative discourse amongst L2-English learners. This tendency is more evident among the participants who are at the initial stage of L2 tense-aspect acquisition, suggesting that L2 learners of the tense-aspect morphology start with the most unmarked assumptions about the form-meaning mapping, which are gradually developed into the more marked and less prototypical mappings as their interlanguage system progresses towards the target norm. This finding gives support to the 180

Chapter 8 General Discussion

predictions and research findings of the Aspect Hypothesis on the initial stage of L2 tense-aspect acquisition (Andersen, 1991; Bardovi-Harlig, 1992; Bardovi-Harlig & Reynolds, 1995; Cai, 2002; Jin & Hendriks, 2003; Li & Bowerman, 1998; Robison, 1990, 1995; Shirai & Andersen, 1995; Shirai & Kurono, 1998). --- The effect of discourse constraints becomes stronger at the later stage of L2 tense-aspect development. It is shown in ASIT, for example, that advanced participants were more sensitive to the discourse requirement for the temporal interpretation of the aspectual forms and they can interpret more correctly the discourse grounding status of the verbs marked with different aspects, an ability that low-level participants have not developed. The same tendency manifested itself in the participants’ performance on MCRT, a controlled production task, wherein advanced participants showed a greater inclination than low-level and intermediate participants did to mark backgrounding events with the progressive aspect and foregrounding events with perfective aspect. This indicates a greater ability among advanced L2 learners to integrate the information of discourse grounding in aspect interpretation and aspect marking, thus pointing to a strong effect of constraints re-ranking occurring at the later stage of tense-aspect acquisition. --- L1 influence on tense-aspect marking was observed in situations where English and Chinese showed differences in mapping compatibility between progressive aspect and particular situation types, i.e., the achievement and the stative. Owing to cross-linguistic differences, Chinese EFL learners, especially those at the initial stage of tense-aspect acquisition, were found to reject progressive aspect marking on the achievement verbs in both MCRT and AJT, when compared with the performance of the native controls in the same tasks. It was also shown that Chinese EFL learners were more likely to accept the progressive aspect marking on stage-level stative verbs than the native speakers, since the Chinese language licenses progressive marking on this subset of stative verbs. No effect of Chinese influence, however, was observed where Chinese and English share mapping compatibility between grammatical aspect and the verb’s situation type. The above findings are largely in 181

Chapter 8 General Discussion

agreement with those of the studies examining the role of L1 in tense-aspect acquisition across typologically different native languages, like French (Collins, 2002), Spanish and Bulgarian (Slabakova, 2000a), and Chinese (Cai, 2002). Thus, the role of L1 in L2 tense-aspect marking appears to be limited to the aspect-situation scenarios where L1 and L2 show cross-linguistic differences. — In the standard formulation of the Aspect Hypothesis, the progressive markings are predicted not to be “incorrectly overextended to stative verbs” (Andersen & Shirai, 1996: 532). The empirical findings of our study, however, suggest that this prediction is too simplistic to make valid generalizations about the role played by stative verbs in the progressive aspect acquisition. More specifically, we found that Chinese EFL learners, especially those at the initial stage of tenseaspect acquisition, were more tolerant to the progressive mapping on the individuallevel statives than the native speakers of English. Statives also appeared to be a source of incongruities between our research hypotheses and empirical findings, as shown by the failure of perfective marking in MCRT to follow the hypothetical order unless the statives were excluded from consideration. All these indicate a mixed role played by stative verbs in aspect marking, which makes it impossible for any single hypothesis to make accurate predictions about its emergence and development in the course of L2 acquisition. 8.1.2 Characterizing L2 Acquisition of the English Aspect Markers by Chinese EFL Learners The research findings summarized above indicate a pattern of emergence and development of L2 tense-aspect forms in general, and progressive aspect in particular, which is characterized by constraints re-ranking in OT framework. Schematically, this general pattern can be represented as Figure 18. At the initial stage, the NOCOER constraint tops the ranking of constraints and the learner’s L2 grammar only permits the most unmarked and prototypical output, i.e., the least coerced aspectual forms, such as the perfective marking on the achievements

182

Chapter 8 General Discussion

and progressive marking on the activities. The empirical evidence from our study suggests that the situation-aspect mappings involving more violations of the NOCOER constrain are more difficult to acquire, interpret and produce than those involving fewer violations of the constraint.

Initial Stage:

NOCOER (L1) >> IDENT(TF), MARKT

Intermediate Stages:

NOCOER (L1) >> IDENT (TF)>>MARKT

IDENT(TF) >>NOCOER (L1) >> MARKT

IDENT (TF)>>MARK(T)>>NOCOER (L1)

Advanced Stage:

IDENT (TF) >> MARKT >> NOCOER (L1/L2)

Near-native Stage: IDENT (TF) >> MARKT >> NOCOER (L2/L1)

Figure 18 General Pattern of Constraints Re-ranking in L2 Tense-aspect Acquisition

It is also found that, at this early stage of L2 tense-aspect development, the effect of L1 influence is strong where the learner’s L1 and L2 exhibit typological difference in compatibility between grammatical aspects and situation types, i.e., between progressive aspect on the one hand, and stative or achievement verbs, on the other. The research findings from our study suggest that low-level L2 participants are more tolerant to the stative-progressive mapping, but less likely to accept the achievementprogressive mapping than the participants of higher English proficiency, owing to cross-linguistic differences between English and Chinese. 183

Chapter 8 General Discussion

The intermediate stage witnesses a gradual demotion of the NOCOER constraint and promotion of the IDENT (TF) constraint. At this stage, having noticed a discrepancy between their current grammar and the linguistic input, the learners start to make adjustments to their initial constraints ranking order. This leads to a gradual spread from the prototypical to the non-prototypical mapping, and the prioritization of the discourse requirement for the temporal interpretation of the aspect morphology. The effect of L1 influence is less strong than it was in the initial stage. When L2 learners are at the advanced stage of tense-aspect acquisition, their performance in aspect marking and aspectual interpretation is almost native-like. At this stage, the ranking order of the constraints is dominated by the discourse requirement for the temporal interpretation of the verb’s aspect morphology. At the same time, the NOCOER constraint is demoted to the lowest position and its effect on aspectual choice is minimized. Besides, the effect of L1 influence still exists where L1 and L2 show typological differences in situation-aspect mapping but it is much weaker than in the initial stage, as shown by advanced participants’ greater sensitivity to the aspectual distinctions in discourse than their lower level counterparts and their greater capability of marking verbs with tense-aspect morphology appropriate for their discourse grounding status. This general tendency of L2 tense-aspect development demonstrates the constraint re-ranking involved in L2 tense-aspect acquisition and how the initially higher-ranked NOCOER constraint is gradually demoted as L2 learners’ tense-aspect progresses towards the target norm. This developmental route, in turn, validates the status of OT as a coherent framework for the observed pattern of the participants’ behaviours in the acquisition of L2 tense-aspect system. The implications of OT for L2 acquisition studies, together with the new issues it raises, will be evaluated and examined in the next section.

184

Chapter 8 General Discussion

8.2 Implications of OT for the Developmental Route of L2 Tense-Aspect Forms What new light does an OT account throw on studies on the emergence and development of L2 tense-aspect morphology in general, and the progressive aspect in particular, apart from the general tendency of constraint re-ranking sketched above? How does OT help to resolve some of the long-standing controversies in L2 tenseaspect acquisition, like the issue of the initial state of tense-aspect acquisition? These questions will be addressed in the following three sections. More particularly, we are going to discuss four issues crucial in L2 tense-aspect studies and how these issues are approached by OT. These issues include the initial state of tense-aspect acquisition, the triggering factor, the spread from prototypical to non-prototypical mappings between situation types and grammatical aspects, and the role of stative verbs in progressive aspect marking. 8.2.1 An OT Account of the Initial State The problem of what constitutes the initial state of L2 acquisition is a vexed issue of central importance in SLA studies. From the perspective of Universal Grammar, Schwartz & Sprouse (1996) proposed a Full Transfer/Full Access model according to which the final state of L1 acquisition is the initial state of L2 acquisition. In other words, the learner’s default hypotheses about the L2 input derive directly from the UG principles and parameter values of the L1. This proposal concerning the initial state of L2 acquisition has also been examined in L2 tense-aspect acquisition. Slabakova (2000), for example, claims on the basis of her empirical study that the differences in the performance of learners are directly traceable to their native language, thus supporting the full-transfer/full-access hypothesis (Schwartz & Sprouse, 1996). If Full Transfer/Full Access argument is true, it would predict that the full set of L1 constraints are present in the initial state of L2 acquisition. However, this prediction is at odds with the empirical data obtained from our study. While this model offers some plausibility in explaining the participants’ performance on the 185

Chapter 8 General Discussion

semantic constraints on tense-aspect marking, i.e., L2 initial state of tense-aspect acquisition mirrors the L1 parameter setting at the semantic level, the data are incongruous with the Full Transfer/Full Access argument if we examine the evidence beyond the semantic level. More particularly, the tense-aspect markers in both English and Chinese have shared functions of marking discourse grounding status of the verbs, but this shared discourse function does not seem to facilitate the acquisition of the discourse constraints on tense-aspect acquisition and the L2 participants in our study have to learn the discourse constraints from the scratch. In other words, L2 participants, particularly those low-level participants, do not seem to have access to the full set of constraints on tense-aspect marking. Only the semantic, but not the discourse constraints, are accessed and transferred. A closer examination of the evidence presented by Slabokavo (2000a) also reveals that she did not examine the full set of constraints on aspect marking, and the discourse constraints were not given due consideration in her study. By contrast, OT assumes that the task for language learners is to learn the correct ranking of the constraints specific to a particular language (Boersma & Hayes, 2001; Tesar & Smolensky, 1998b, 2000). The key idea is that learning involves constraint demotion. In OT, one influential proposal for implementing this strategy is that learners start from an [M>>F] initial state. This initial constraint ranking describes a ranking bias present at the beginning of learning, with all markedness constraints above all faithfulness constraints, and the most unmarked form emerges first in the language learner’s output (McCarthy & Prince, 1994; Prince & Smolensky, 1997; Smolensky et al., 2004). This [M>>F] initial ranking is the working hypothesis of the initial state of L2 tense-aspect acquisition in the present study. According to this hypothesis, the markedness constraint NOCOER proceeds the faithfulness constraint IDENT(TF) in the process of L2 tense-aspect acquisition. This hypothesis has been confirmed by the empirical observation that L2 learners of English tense-aspect system prefer the most

186

Chapter 8 General Discussion

unmarked situation-aspect scenario in all the three tasks and effect of discourse grounding emerges at the later stage of aspect acquisition. The findings of our study also show that the influence of L1 on L2 aspect acquisition is restricted to the cases where two languages differ in aspect-situation mapping at the semantic level. No L1 effect, facilitative or impeding, has been observed at the discourse level. That is to say, despite their competence in L1 to mark verbs of different situation types with appropriate aspect markers according to the grounding status of these verbs in the context of discourse, L2 learners in our study failed to develop the ability to vary aspectual markers with discourse requirement until they are at the latter stage of interlanguage development. Expanding on this point, an OT perspective on second-language acquisition may offer a partial solution to the vexed issue of the initial state of second language acquisition. In the light of this perspective, the markedness constraint dominates the faithfulness constraint in L2 learner’ initial grammar and the unmarked canonical form-meanings emerge first in L2 learners’ interlanguage. The evidence from our study suggests that this initial form-meaning mapping is susceptible to cross-linguistic differences between L1 and L2. The family of faithful constraints, on the other hand, occupies a relatively low position in the initial constraint ranking and L1 influence upon this family of constraints does not appear to be very significant at the initial stage of L2 tense-aspect acquisition. The [M>>F] claim on the initial ranking, together with the empirical evidence obtained from our study, indicates that the initial state of L2 tense-aspect learners is jointly determined by the markedness relations between the verb’s grammatical aspect and the inherent temporal features, and by the cross-linguistic difference in the initially higher ranked constraint. On the other hand, the initially lower-ranked discourse constraints play a secondary and marginal role in the initial state of L2 tense-aspect acquisition.

187

Chapter 8 General Discussion

8.2.2 The Triggering Factor in Tense-aspect Acquisition The studies on L2 tense-aspect acquisition, in particular those on testing the claims of Aspect Hypothesis, gave us plausible accounts of the early emergence of L2 tense-aspect acquisition, such as the prototype account (Shirai & Andersen, 1995). However, no existing theories can convincingly explain the later development of the early prototypical mapping between the aspectual forms and the temporal semantics of the verbs. To be more exact, we do not have a valid account of the factors triggering the spread from the early unmarked mapping to the later stage of nonprototypical mapping. Identification of the triggering factors responsible for language development is an issue of central importance in language acquisition theories. For instance, it is well known that in Universal Grammar (UG), language acquisition is perceived to be determined by the innate language acquisition faculty and triggered by the linguistic input. This idea of a trigger is formalized by Gibson and Wexler (1994) in their Trigger Learning Algorithm (TLA), which selects one grammar from the finite number of valued parameters. Thus, when faced with variation exemplifying conflicting parameter setting in the input (a trigger), the TLA updates a parameter and guides the learner’s grammar gradually to the target form. But TLA does not specify which parameter is most likely to be modified and reset at a given stage of language acquisition. Slabakova (2001) examined the triggering factor responsible for L2 acquisition of aspect morphology, i.e., the telicity parameter, in a UG-based framework. Following Snyder (1996; 2001), she claims that a cluster of aspect-related constructions like Verb Particles, Secondary Resultative Predicates, and Double Objects in English could be the linguistic trigger necessary for acquiring aspect. However, her study shows that 25 out of the 122 subjects were found to have acquired aspect but not the related constructions (Slabakova, 2001: 194), suggesting that

188

Chapter 8 General Discussion

aspect-related constructions may not be a necessary and sufficient trigger for L2 tense-aspect acquisition. The inadequacies of the UG-based accounts appear to be overcome by the learning algorithms proposed in OT, such as Constraint Demotion (Tesar & Smolensky, 1998a, 2000) and Gradual Learning Algorithm (Boersma & Hayes, 2001), which claim that language learning is a process involving constraint re-ranking. In this process, the learner gradually approaches the target ranking of the constraints by demoting the constraints that are regarded as being violated in the target output form. This constraint re-ranking, according to OT account, is triggered by a mismatch between the learner’s ranking and the target ranking of a given constraint. More specifically, in L2 tense-aspect marking, by comparing the initial ranking NOCOER>>IDENT(TF), MARKT and the target constraints ranking IDENT(TF) >>MARK(T) >>NOCOER, the learner gradually demotes the higher-ranked NOCOER constraint. This demotion of NOCOER constraint, therefore, is triggered when the learner observes that IDENT (TF) and MARK (T) rank higher than NOCOER in the target form. The empirical data from our study have provided the evidence of the triggering effect of the IDENT (TF) constraint. Specifically, our hypotheses were (i) the effect of NOCOER constraint was stronger among lower-level participants (H2), and (ii) that of IDENT (TF) constraint was stronger among higher-level participants (H3). These two hypotheses have been to a large extent confirmed, suggesting that the higher-ranked IDENT(TF) in the target form motivates, or triggers, the demotion of NOCOER constraint initially high-ranked in the L2 interlanguage. We can, therefore, reasonably conclude that the constraint IDENT (TF) in the input data serves as a triggering factor in the re-ranking process of the constraints on tense-aspect marking. From this it can be further inferred that the spread from prototypical aspect-situation mapping to the non-prototypical mapping is motivated by the discourse requirement for the temporal interpretation of the aspect morphology.

189

Chapter 8 General Discussion

As a matter of fact, the triggering effect of the discourse context in tense-aspect acquisition is also recognized in studies on L1 tense-aspect acquisition. In her study based on picture-story matching comprehension task, for example, Angeliek van Hout (2005) found that L1-Polish children’s knowledge of discourse rules do not develop abreast with their knowledge of the verbal semantics in tense-aspect acquisition, a finding similar to the one described to our study. Stoll (2001) also observed that the development of L1-Russion aspect system is context-driven after the children have gone through the initial stage characterized by individual verb meanings. The correct mapping between aspect and discourse grounding is found to occur later only when children’s narrative competence has developed sufficiently. This observation leads Stoll to proposing Hypothesis of Context-Driven Learning of Aspect, a hypothesis translatable to the triggering effect of discourse constraint IDENT(TF) in the present study. 8.2.3 The Spread from Prototypical to Non-prototypical Situation-Aspect Mappings Aspect Hypothesis predicts that L2 development of tense-aspect morphology starts with prototypical mappings between situation types and aspect morphology which gradually spread to non-prototypical mappings (Andersen & Shirai, 1996). There exists conflicting evidence for this prediction. While the findings of some studies lend support to this prediction, counter evidence exists in others. For instance, some studies show that the correlation between progressive and activities fossilizes and the predicted spread of the progressive to other non-prototypical verb classes is halting (e.g., Robison, 1995). The theoretical analysis and empirical findings in our study support the prediction made by Aspect Hypothesis. The participants’ performance on all three elicitation tasks showed that the spread from prototypical to non-prototypical mapping was gradual as their L2 proficiency advanced towards the target norm. This process of gradual spread from prototypical to non-prototypical mappings can be explained by Gradual Learning Algorithm (GLA) proposed by Boersma and

190

Chapter 8 General Discussion

Hayes (2001) as a constraint ranking algorithm for learning Optimality-theoretic grammars. This algorithm assumes that each OT constraint has a ranking value along a continuous scale which allows for overlapping between two adjacent constraints in the rank. According to this proposal, a learning step triggered by an observed error in the existing ranking does not lead to radical change in the constraint ranks. Rather, a slight adjustment of the constraint ranks is made, resulting in optionality and variations in the learner’s output. This algorithm suggests that in L2 tense-aspect acquisition, the demotion of the initially high-ranked NOCOER constraint does not take place in a leap and the process is a gradual and continuous one. Specifically, verbs of different situation types involve different number of violations of the NOCOER constraint when marked with different grammatical aspects.

73

For instance, when marked with the English

progressive verb, an achievement verb involves three violations of the constraint. The cancellation of these three violations is assumed by GLA to be gradual and stepwise, but not immediate. This GLA account of constraint re-ranking in tense-aspect acquisition is supported by the findings obtained from testing Hypothesis 1, which indicate that the gradual spread from prototypical to non-prototypical mappings is primarily determined by the number of violations of the NOCOER constraint involved. As fewer violations of the constraint are involved in prototypical mappings, they are acquired earlier and judged to be more compatible by L2 learners than those involving more violations of the constraint. However, as the spread process is gradual and stepwise, variations and overlappings may occur when the constraint demotes from one stage to another in L2 tense-aspect development. This GLA account can explain why, in testing Hypothesis 1, some pairwise comparisons between verbs of different situation types are not statistically significant when they are marked with different grammatical

73

See §5.3.2 for a more detailed discussion on this point. 191

Chapter 8 General Discussion

aspects, even though the general spreading order is closely correlated with the number of violations of the NOCOER constraint. 8.2.4 The Role of Statives in Progressive Aspect Marking According to Aspect Hypothesis, L2-English learners will not extend progressive aspect marking onto stative verbs (Andersen & Shirai, 1996). However, this prediction is not supported by our linguistic theorizing on the role of statives in progressive marking, neither is it verified by the data elicited in the empirical part of our study. More particularly, aspect marking on stative verbs does not seem to be constrained by the gradual but ordered spread from prototypical to non-prototypical mappings between verb aspect and situation types, as shown by the failure of the participants’ perfective marking in MCRT to follow the predicted order of spread unless the statives are excluded from consideration. 74 This idiosyncratic behaviour of statives in aspect marking is mainly attributable to the following three sources. First, a stative verb is ontologically different from an event verb (Kamp & Reyle, 1993). These two classes of verbs are distinguished from each other by the presence or absence of the temporal feature [Dynamic], which, as noted in § 3.1.4, is generally given priority over other temporal parameters like [Telic] or [Punctual], and serves as the central criterion for the initial level distinction of situation aspect (Comrie, 1976; Smith, 1997). This ontological difference between stative verbs and event verbs can partly explain why the participants responded to them differently when performing our experimental tasks. Second, the stative is not a homogeneous entirety, since it can be decomposed into several subclasses (Quirk et al., 1985). As shown by the empirical data obtained from AJT, stative verbs belonging in different subclasses are perceived by the participants to have different degrees of compatibility with the progressive aspect. The dichotomy between individual-level and stage-level predicates, first proposed by Carlson (1977) and followed by Dowty (1979) and Smith (1997), proved to be a

74

See § 7.1.1 for this point. 192

Chapter 8 General Discussion

helpful linguistic framework capable of accounting for the empirical observation in AJT, which shows that native speakers of English are more likely to accept the cooccurrence of progressive with the individual-level statives than with the stage-level statives. In MCRT, however, this categorization of stative verbs was not maintained for technical reasons in the development of research instrument, 75 and so the results have thus become only partly explicable. Third, Chinese and English languages exhibit cross-linguistic differences in the progressive aspect marking on the statives. While a subset of Chinese stative verbs can be felicitously marked by the progressive markers (Li & Thompson, 1981; Smith, 1997), the English statives, in particular the individual-level statives denoting property or relation, typically reject progressive marking. To complicate the situation, the two progressive markers in Chinese, i.e., the preverbal zai- and postverbal –zhe may overlap or differ in semantic interpretation and morphsyntactic distribution when used to mark stative verbs. This characteristics of the Chinese progressive markers poses an additional many-to-one mapping problems for Chinese-speaking learners of English progressive aspect. Influenced by their L1, Chinese EFL learners, especially those at their initial stage of learning, may experience a mapping difficulty when marking the English progressive on stative verbs. If, for example, their initial hypothesis about the English progressive is built on their knowledge of zai-, they would tend to restrict their use of the progressive markers since zai- has a narrower range of uses than the English progressive aspect. On the other hand, if their initial assumption is based on the postverbal progressive marker zhe-, they would mark the stative verbs with progressive markers more frequently than native speakers of English, because –zhe has a broader range of uses than the English progressive aspect, just as our cross-linguistic analysis in §3.5 shows. The general tendency observed in

75 Recall MCRT is a controlled production task wherein the text is modeled after the movie clips of the British comedy Mr. Bean and the stative verbs, like the other four situation types, are balanced across the grounding. Given these limitations, it is not feasible to balance the subclasses of the stative verbs when choosing the target verbs.

193

Chapter 8 General Discussion

our experimental tasks has basically verified this prediction, although the pattern is not neat enough to make robust generalizations on marking statives with the progressive marker. Piecing together these linguistic and cross-linguistic attributes of statives verbs, we can conclude that the role played by the stative verbs in aspect marking is a mixed one and no single hypothesis suffices to make accurate predictions about the emergence and development of the aspect marking on the statives. 8.3 The Explanatory Power of OT for L2 Aspect Acquisition We have, in the preceding section, discussed the empirical findings of our study within the OT framework. The focus of this section is to demonstrate the theoretical power of OT for L2 tense-aspect acquisition. More particularly, we are going to present a coherent explanation for two disputable issues in studies on L2 tense-aspect acquisition, i.e., the roles respectively played by temporal semantics of the verb and its discourse grounding status, and the connectionism-nativism controversy reigning in L2 tense-aspect acquisition. 8.3.1 Temporal Semantics and Discourse Grounding, Two Conspiring Forces? The temporal semantics of verbs and their discourse grounding status have been recognized as two predominant factors, among others, which are responsible for L2 acquisition of tense-aspect morphology.

However, these two factors often have

competing demands on the aspectual forms of the verbal predicate. For instance, when a verb is dynamic, punctual and telic, the most prototypical aspect morphology congruous with these temporal features is the perfective aspect. Nevertheless, if the same verb functions as discourse background in a narrative text, the progressive aspect is more desirable. The competition between these two mapping processes, i.e., the form-meaning mapping and the form-function mapping poses a challenge to L2 learners when acquiring tense-aspect morphology. A sound theory of L2 tense-aspect acquisition should give an adequate explanation for how these two factors interact at different developmental stages of L2 tense-aspect morphology. 194

Chapter 8 General Discussion

However, before the present study, the roles that these two factors respectively play in L2 tense-aspect acquisition were to a large extent unspecified. The majority of the studies examine either one of these two factors, with a view to testing Aspect Hypothesis or Discourse Hypothesis (e.g., Andersen, 1991; Andersen & Shirai, 1996; Bardovi-Harlig, 1994, 1995b; Bardovi-Harlig & Reynolds, 1995; Bloom et al., 1980; Li & Bowerman, 1998; Montrul & Slabakova, 2002; Robison, 1990, 1995; Salaberry, 1999, 2000; Shirai & Andersen, 1995; Shirai & Kurono, 1998; Weist et al., 1984). Only a few studies address both factors (Bardovi-Harlig, 1998; Cai, 2002; Housen, 1994; Yang & Huang, 2004). However, these studies tended to perceive the semantic factor and the discourse factors as two “conspiring factors” in L2 development of tense-aspect morphology, without specifying their intrinsic relationship (BardoviHarlig, 1998). Our empirically-supported OT theorizing on the roles played by these two factors has indicated that these two mapping processes not only conspire, but interact and compete with each other to yield optimal forms of progressive aspect marking at different stages of L2 development. At the early stage, by virtue of the initial [M>>F] constraint ranking, L2 learners’ interlanguage tense-aspect system is primarily guided by the naturalness of form-meaning mapping, i.e., the mapping between the tenseaspect morphology and temporal primitives of the verbs. At the more advanced stage, L2 learners recognize the role played by the discourse requirement for the temporal interpretations of the verbs marked with different aspectual forms, and so they gradually promote the faithfulness constraint IDENT[TF] to the topmost position of the OT ranking of constraints on tense-aspect marking. At this stage, form-function mapping process gradually dominates the tense-aspect system of their interlanguage. Note that these two competing mapping processes tend to display stage-specific characteristics in L2 development of the tense-aspect marking in general, and the progressive aspect marking in particular. More accurately, the tense-aspect morphology is initially bootstrapped by semantic markedness of the verbs involved, and later it spread from the prototypical to non-prototypical mapping, triggered by L2 195

Chapter 8 General Discussion

learners’ recognition of the leading role played by discourse grounding in tense-aspect marking. The proposal that semantic constraint and discourse constraint have a division of labor at different stages of L2 tense-aspect development is a step forward from the aforementioned “conspiracy account”, which failed to differentiate the specific roles played respectively by these two factors in L2 tense-aspect acquisition. One reason why such a proposal was lacking in earlier studies is that these studies were mainly descriptive and data-driven in nature, without being guided by a coherent theory of language and language learning. Therefore, the competition and interaction between two mapping processes in aspect acquisition, i.e., the form-meaning mapping and form-function mapping escaped direct scrutiny. However, the present study, following the [M>>F] hypothesis about the initial ranking of OT constraints, is able to reveal the underlying mechanism responsible for L2 tense-aspect development at different stages of L2 development. This advantage over the current accounts is a clear demonstration of the superior explanatory power of OT for L2 tense-aspect acquisition. 8.3.2 Nativism or Connectionism There exist two board theoretical explanations for the emergence and development patterns of L2 tense-aspect acquisition, i.e., the nativist approach and the connectionist approach. Both have inherent strong points and inadequacies. Earlier nativist proposals include Bickerton’s Language Bioprogram Hypothesis (Bickerton, 1984, 1999) and Slobin’s Basic Child Grammar Hypothesis (Slobin, 1981, 1985). According to Bickerton, two semantic contrasts are pre-programmed, so they emerge early in child language, i.e., the state-process distinction, and the punctualnon-punctual distinction. Similar to Bickerton, Slobin argues that children acquiring L1 come to the language acquisition task with a prestructured “semantic space,” containing a universal set of semantic notions, like process and result. More recently, Olsen (1997) proposed her Subset Principle Explanation to account for the biased

196

Chapter 8 General Discussion

mapping between aspect and situation types. Following the Subset Principle (Berwick, 1985; Gibson & Wexler, 1994), it is argued that children assume the most restrictive one-to-one mapping as their initial hypothesis, i.e., imperfective is mapped onto dynamic, durative predicates, and perfective is mapped onto telic predicates. By contrast, the connectionist approach maintains that that language emerges through an interaction between domain-general learning mechanisms and the environment without crediting the child with innate knowledge of domain-specific rules. The child’s biased mapping between aspect form and verb’s temporal semantics, according to this view, reflects his/her ability to extract statistical patterns from the input data and to use the patterns productively in building his/her grammar. Li Ping and his colleagues demonstrate that a network (DIXLEX) supplied with the verbs from half the CHILDES database, organized in five stages, is capable of “learning” the verb aspect-situation mappings (Li, 2002; Li & Shirai, 2000; Magliano & Schleich, 2000). Both nativist approach and connectionist approach have some inherent inadequacies in accounting for the emergence and development of the L2 tense-aspect acquisition. While the nativist account has strong explanatory power for the initial representation of the tense-aspect acquisition, the role of learning by experience and environmental support is often discounted by nativist theories (Li & Shirai, 2000). On the other hand, the connectionist learning network such as DIXLEX is very likely to overgenerate impossible strings unlikely to occur in language learner’s grammar, if the constraints on the linguistic properties, representations and rules are excluded from considerations (Slabakova, 2002). Therefore, there arises a need to reconcile nativist account and connectionist account of second language learning (Hulstijn, 2002). OT, a theory that respects the legacy of both nativist and connectionist approaches in attempting to incorporate the optimization in grammatical theory into optimization in neural networks (Prince & Smolensky, 1997; Smolensky, 2000), is a likely candidate to fulfill this purpose. On the one hand, OT is basically nativist in assuming that the well-formed constraints 197

Chapter 8 General Discussion

universal to all human languages are preprogrammed in the mind of human beings. On the other hand, OT is deeply rooted in connectionism (Prince & Smolensky, 1993) since it emphasizes that these constraints interact with each other in a parallel and cross-modular fashion, with the higher ranked constraints having greater weight and priority to opt for the optimal output. In this sense, OT as a hybrid theory integrating the strengths of nativist and connectionist approaches offers some promise to resolve the connectionism-nativist dispute on tense-aspect acquisition. Although the present study is committed to neither of these two approaches, our linguistic theorizing and empirical findings embrace some of the theoretical claims in both of them. On the one hand, we assume that such temporal constraints as telicity, punctuality and dynamicity are prepreliguistic categories. 76 Since these temporal features lack overt linguistic markers, it is inconceivable that they are learned through probabilistic patterns in the input. On the other hand, the constraint re-ranking as assumed by OT is a typical case of error-driven connectionist learning process wherein the leaner tries in parallel various aspect-situation and aspect-grounding mapping relations and the learner grammar proceed from the initial ranking of constraints to the ultimate target ranking. Therefore, our study of the L2 tense-aspect development is, to some extent, a response to the recent calls in the L2 research community for an integrative thinking in SLA studies (Hulstijn, 2002; O'Grady, 2003). 8.4 Summary In this chapter, we first summarized the major findings of the hypotheses testing in our study. This was then followed by a delineation of the developmental route of

76 The human beings’ need to register and detect change in the environment is fundamental for their evolution and survival. Since moving objects are either likely to be dangerous or potential food, rapid appropriate action is demanded, while stationary objects can generally be ignored (Gregory, 1997). This evidence shows that such temporal notions like dynamic change or stativity might have a prelinguistic and biological basis.

198

Chapter 8 General Discussion

English tense-aspect acquisition by Chinese EFL learners. Finally, four issues crucially related to L2 acquisition of the progressive were discussed within the framework of OT, in comparison with relevant previous research in the field of SLA. These four issues were: the initial state of tense-aspect acquisition, the triggering role played by the discourse grounding, the spread from prototypical to non-prototypical situation-aspect mapping, and the idiosyncrasy of stative verbs in the acquisition of progressive aspect. Our study shows that the canonical form-meaning mappings involving fewer violations of the NOCOER constraint are acquired earlier, processed with greater ease and judged as more acceptable than those involving more violations of the constraint. This finding supports the OT assumption that markedness constraints outrank the faithfulness constraints in the initial state of language acquisition. Constraint re-ranking central to OT account of language acquisition was observed at a later stage when advanced participants demoted the NOCOER constraint below IDENT (TF) constraint. This shows that these participants were primarily guided by the discourse requirement for aspectual interpretation, rather than by the temporal semantics inherent in the verbs. This finding suggests that the “conspiracy” account of the roles of the semantic and discourse constraints is at best incomplete. Furthermore, it is indicated in our study that the spread from the early unmarked mapping to the later stage of non-prototypical mapping is triggered by the faithfulness constraint IDENT(TF), which is initially ranked lower than the markedness constraint in L2 tense-aspect interlanguage. That is, L2 learners were found not to use the nonprototypical aspect-situation combinations until they began to recognize the role played by the discourse requirement for the temporal interpretation of the verb’s aspect. This finding is akin to what is predicted by Stoll’s (2001) Hypothesis of Context-Driven Learning of Aspect, but it does not support Slabakova’s claim that the trigger is the aspectually related constructions (Slabakova, 2001). Statives were found to behave quite differently from the event verbs. This led to most of the incongruities between the research hypotheses and empirical findings. The 199

Chapter 8 General Discussion

idiosyncratic behaviours of stative verbs seem to be attributable to their ontological difference from the event verbs, to their more fine-grained categorization, and to the typological difference between English and Chinese. Therefore, the role played by stative verbs in aspect acquisition is a mixed one and the prediction by Aspect Hypothesis that the progressive markings would be incorrectly overextended to the stative verbs is shown to be too simplistic. These findings have validated the status of OT as an appropriate explanatory framework for the L2 development of the English progressive aspect. Our OT theorizing and empirical findings can also be extrapolated to the development of other aspectual forms, in particular the perfective aspect. Finally, we have demonstrated that OT is a coherent theory with explanatory adequacy capable of settling long-standing controversies in L2 tense-aspect acquisition studies. Since OT is able to differentiate the specific roles of semantic and discourse factors in L2 tense-aspect acquisition, and to reconcile between the nativist and connectionist approaches to L2 tense-aspect acquisition, it fulfills its role as a unified account of L2 development of tense-aspect morphology.

200

Chapter 9 Conclusions, Implications, Limitations and Suggestions for Future Research This chapter winds up the dissertation. It first draws conclusions on English progressive aspect acquisition by Chinese EFL learners by summarizing the answers to all four key research questions stated in §1.4 in a unified OT framework. This is followed by a discussion of the implications of the research findings for L2 tenseaspect research and instruction. In §9.3, the limitations of this study are pointed out, and finally, recommendations for further research into L2 tense-aspect acquisition are suggested. 9.1 Conclusions The linguistic constraints on tense-aspect marking are inherently conflicting. These competing demands on the temporal interpretations of the tense-aspect forms engender learning difficulties for L2 acquisition of these forms. The empirical part of this study shows that the L2 learners’ tense-aspect morphlogy is initially characterized by naturalness of combinations between situation and aspect, like progressive-activity mapping and perfective-achievement mapping. Only at advanced stage of L2 development does the prototypical mapping spread to less prototypical combinations between aspectual forms and the temporal features inherent in the verbs. This gradual spreading process is found to be triggered by advanced L2 learners’ recognition of the dominant role played by the discourse grounding status in tense-aspect marking. The effect of L1 on L2 tense-aspect acquisition is made manifest where the two languages differ in mapping compatibility between aspect forms and temporal features inherent in the verbs, and the effect is stronger at intial stage of tense-aspect acquisition. These conflicting constraints and their impact on L2 tense-aspect development are approached with OT model of human language, which assumes language acquisition to be a process of constraints reranking from initial [M>>F] ranking to the

target ranking. In the light of this assumption, the temporal semantics inherent in the verbs assumes a dominant role in the initial L2 grammar which permits learners to choose the aspectual markers in natural alignment with the temporal features of the verbs. More conflicts between the temporal features and the interpretation of the aspect markers give rise to greater learning difficulties. It is at the advanced stage that L2 aspect marking is dominated by discourse requirement for the temporal interpretation of tense-aspect forms, hence constraints reranking. Our OT theorizing of L2 tense-aspect acquisition takes a big step forward from the existing fragmentary and isolated accounts of L2 tense-aspect acquisition in that it provides a consistent and systematic explanation for the linguistic and cross-linguistic constraints on tense-aspect marking, their ranking relationships and their implications for L2 acquisition of aspect forms. The predications generated within OT framework and their validation in three experimental tasks has further demonstrated the explanatory power of the OT model for L2 acquisition of the English tense-aspect system, in particular the progressive aspect. Unlike other theories of L2 tense-aspect development, OT can explain, in a principled and unified way, the initial state of L2 tense-aspect acquisition and the factor triggering its later development, the developmental patterns of progressive aspect and perfective aspect . It also specifies the roles played by the semantic constraint and discourse constraint in L2 development of both progressive and perfective aspect within one single theoretical framework, which are otherwise considered as two ‘conspiring’ factors. Finally, OT offers a way to reconcile between the nativism and connectionism, two dominant but often conflicting theories on L2 acquisition of tense-aspect morphology as well as other linguistic categories. 9.2 Implications 9.2.1 Implications for SLA Theory OT has fundamentally renewed the outlook of linguistic inquiry since its advent and it contributes a great deal to our rethinking of the theoretical problems in such 202

areas of linguistic inquiry as phonology, morphology, syntax, semantics, pragmatics, linguistic typology as well as first language acquisition. However, few attempts have so far been made to incorporate it in L2 acquisition studies. The linguistic theorizing and empirical findings from this study indicate that the application of OT to second language learning represents a new and potentially productive line of inquiry. To begin with, taking a cross-modular approach to grammar, OT is capable of evaluating in parallel the linguistic constraints at different linguistic levels within a single theoretical framework. In particular, claims about interface between grammar and discourse are even more explicitly pronounced within OT. For instance, the regularities underlying the L2 tense-aspect acquisition, as observed in the present study, would not have emerged if OT was disallowed from the outset, because the constraints we have identified belong to different ‘modules’ in other linguistic framework, like morphosyntax, semantics and discourse. This interface approach proves highly successful in resolving a wide range of linguistic controversies, particularly those concerning discourse-syntactic-semantic interface, like typological differences in word order (Choi, 1996), passivization (Dingare, 2001), subject selection (Aissen, 1999),

and anaphor resolution (de Hoop & de Swart, 2000).

Therefore, OT grammar has some potential advantages over other linguistic frameworks for investigating how these categories are acquired by L2 learners. Furthermore, although there is nothing new about the claim that initial L2 grammar is characterized by prototypical and unmarked configuration of linguistic constraints (Eckman, 1977), the innovative part of OT is that it sees language acquisition as a process of constraint re-ranking from the initial ranking to the target ranking, by virtue of the faithfulness constraints competing with the markedness constraints, which are assumed to be dominant over the faithfulness constraints in the initial stage of language acquisition. This fundamental underlying principle of OT provides the theoretical basis for profiling distinctive stages of L2 tense-aspect development wherein the initially dominating markedness constraint NOCOER

203

gradually demotes until it becomes the least dominant constraint and the initially lower ranked faithfulness IDENT[TF] constraint tops the ranking. 9.2.2 Implications for L2 Tense-aspect Instruction Tense-aspect system has long been recognized to be a major source of L2 learning difficulty (Coppieters, 1987; Seliger, 1978). Not surprisingly, it is an area where L2 learners receive intensive classroom instructions. However, research on the effect of tense-aspect instruction shows that the learners receiving instructions on tense-aspect exhibit the same stages of temporal development observed in the learners without receiving instructions, even though the tutored learners show more immediate gains in tense-aspect development than untutored learners (Bardovi-Harlig, 2000; Harley, 1989). Following OT accounts of language acquisition, we have hypothesized, and further confirmed empirically, that L2 learners will experience a process of constraint re-ranking in the course of L2 tense-aspect acquisition. In particular, the findings of our study suggest that L2 learners are more likely to be affected by the prototypical mappings between grammatical aspects and situation types at the initial stage, and the discourse constraints do not play a dominant role until later when learners are at the advanced stage of L2 tense-aspect acquisition. This finding points to the importance of exposing L2 learners, at an earlier stage of L2 tense-aspect development, to the discourse-based or contextualized input which contains non-prototypical situationaspect mapping, such as activity-perfective or achievement-progressive mappings. Although this approach to tense-aspect instruction is unlikely to alter the natural route of L2 tense-aspect development, due to varying coercive force of the situation types on aspect marking, it functions effectively to accelerate the rate of L2 tense-aspect acquisition, as has been confirmed by the study conducted by Bardovi-Harlig (Bardovi-Harlig, 1995a). Due to cross-linguistic differences between English and Chinese progressive marking, stative verbs and achievement verbs have been found to pose additional

204

learning difficulties for L2 acquisition of the English progressive aspect by Chinesespeaking learners. The findings from our study suggest that Chinese L2 learners, in particular those at low proficiency level, are inclined to overuse progressive marking on statives and underuse progressive aspect on achievement verbs. Also shown in our study is that this L1 influence remains until L2 learners reach the advanced stage of L2 development. This cross-linguistic difference must be taken into account when pedagogical scheme is devised for tense-aspect teaching. For instance, bi-directional translation exercises might be helpful in bringing these cross-linguistic differences into the awareness of L2 learners. 9.3 Limitations In spite of our endeavor to give a consistent and unified explanation for the L2 acquisition of the English progressive aspect by Chinese-speaking learners, the study is susceptible to some theoretical or empirical limitations. First, although we identified three OT constraints on English tense-aspect marking, i.e., NOCOER, MARKT, and IDENT[TF], only two were empirically investigated, leaving the third, i.e., MARKT unexamined, due to limitations of experimental design. The second limitation is concerned with the sampling of participants. Since our tasks were administered in a 30-seat language lab, we could sample no more than 30 students for each experimental group. In addition, the English proficiency level of low-level participants, having learned English for five years at school, is not low enough, even though they scored significantly lower than the other two participant groups in our proficiency test. 77 This flaw in participant sampling might be detrimental to the internal validity of our study. Finally, as we pointed out earlier in §7.4.2, the categorization of statives (Filip, 1999; Quirk et al., 1985; Smith, 1997) was not maintained in Movie Clip Retell Task

77

See §6.2 for details of partipant sampling. 205

(MCRT), due to technical reasons of experimental design, and this flaw in experimental design led to some inexplicable results for MCRT. 9.4 Suggestions for Further Research There are a number of areas touched upon in our study that are worthy of further pursuing. First, although we have identified three OT constraints on English aspect marking and their ranking relations, only two of them have been empirically examined. In addition, there are an array of other variables that tense-aspect marking in human languages is susceptible to (de Swart & Verkuyl, 1999; Hopper & Thompson, 1980). For instance, the roles played by direct object and temporal adverbial in aspect marking have been extensively explored in linguistic studies (Filip, 1999; Hinrichs, 1986; Verkuyl, 1972, 1993). However, the effects of these factors on L2 tense-aspect marking are seldom investigated, let alone incorporating these factors into OT framework. Further studies are needed to explore the impact of these factors on L2 acquisition of tense-aspect marking. Second, the role of tense marker, together with its absence in Chinese language in L2 aspect acquisition, should be further investigated. We have examined the role played by the tense marker in the ranking of OT constraints in English aspect marking. Nonetheless, for one reason or another, we have not conceived of an ingenious method to empirically investigate the effect of this constraint, i.e., MARKT, on L2 acquisition of the progressive aspect. To complicate the situation, Chinese language is rich in aspect markers but devoid of tense inflections. However, the appropriate research method for investating the role played by tense markers, and its absence in Chinese language in L2 tense-aspect acquisition is yet to be invested. Third, the impact of statives on progressive aspect marking merits greater scrutiny. The findings from our study indicate that the stative verb plays a unique and mixed role in progressive aspect marking and its acquisition. On the one hand, the progressive aspect marking on the statives is susceptible to the greatest number of violations of the NOCOER constraint, and therefore, its acquisition is the most difficult 206

for L2 learners of English tense-aspect system. On the other hand, the Chinese language allows the stative-progressive mapping and our study shows that Chinese L2 learners, in particular those at low proficiency level, tended to incorrectly use more progressive markers on stative verbs than native speakers did. More studies are to be conducted on the role of the stative in progressive aspect acquisition. Finally, since OT has proved to be a highly effective linguistic instrument for solving interface problems in language, having taken a parallel and cross-modular approach to human language, it opens a new vista for investigation into some crossmodular issues in SLA that are otherwise hard to handle in other linguistic theories, such as passivization, topicalization, anaphor resolution, and the use of article, all of which require parallel analysis of linguistic constraints on grammar and discourse. Whether OT and its language learning algorithms prove valid in describing and explaining acquisition of these linguistic categories is an issue worthy of further exploration.

207

References Aissen, J. (1999). Markedness and subject choice in Optimality Theory. Natural Language and Linguistic Theory, 17, 673-711. Andersen, R. W. (1991). Developmental sequences: The emergence of aspect marking in second language acquisition. In T.Huebner & A. Charles (Eds.), Crosscurrents in Second Language Acquisition and Linguistic Theories. Amsterdam/Philadelphia: John Benjamins. Andersen, R. W., & Shirai, Y. (1996). The primacy of aspect in first and second language acquisition: The pidgin-creole connection. In W. C. Ritchie & T. K.Bhatia (Eds.), Handbook of Second Language Acquisition. San Diego, CA: Academic Press. Antinucci, F., & Miller, R. (1976). How children talk about what happened. Journal of Child Language, 3, 167-189. Aristotle. (1933/1961). Metaphysics (H. Treddenick, Trans.). Cambridge, Massachusette: Heinemann and Harvard University Press Asher, N. (1992). A default, truth-conditional semantics for the progressive. Linguistics and Philosophy, 15, 463-508. Bache, C. (1985). Verbal Aspect: A General Theory and its Application to Present-Day English. Odense: Odense University Press. Bardovi-Harlig, K. (1992). The relationship of form and meaning: A cross-sectional study of tense and aspect in the interlanguage of learners of English as a second language. Applied Psycholinguistics, 13, 253-278. Bardovi-Harlig, K. (1994). Reverse-order reports and the acquisition of tense: beyond the principle of chronological order. Language Learning, 47, 375-422. Bardovi-Harlig, K. (1995a). The Interaction of pedagogy and natural sequence in the acquisition of tense and aspect. In F. R. Eckman & D. Highland & P. W. Lee & J. Mileham & R. R. Weeber (Eds.), Second Language Acqusition Theory and Pedagogy. Mahwah, New Jersey: Erlbaum. Bardovi-Harlig, K. (1995b). A narrative perspective on the development of tense/aspect system in second language acquisition. Studies in Second Language Acquisition, 17, 263-291. Bardovi-Harlig, K. (1998). Narrative structure and lexical aspect: Conspiring factors in second language acquisition of tense-aspect morphology. Studies in Second Language Acquisition, 20, 471-508. Bardovi-Harlig, K. (1999). From morpheme studies to temporal semantics: tense-aspect research in SLA. Studies in Second Language Acquisition, 21, 341-382. Bardovi-Harlig, K. (2000). Tense and Aspect in Second Language Acquisition: Form, Meaning and Use. Malden: Blackwell. Bardovi-Harlig, K. (2002). Tracking the illusive imperfect in adult L2 acquisition: Refining the hunt. Paper presented at the NSF-funded Workshop on the Syntax,Semantics and Acquisition of Aspect (May 24-26, 2002), University of Iowa.

Bardovi-Harlig, K., & Reynolds, D. W. (1995). The role of lexical aspect in the acquisition of tense and aspect. TESOL Quarterly, 29, 107-131. Berwick, R. (1985). The Acquisition of Syntactic Knowledge. Cambridge, MA. : MIT Press. Bhat, D. N. S. (1999). The Prominent of Tense, Aspect and Mood. Amsterdam/Philadelphia: John Benjamins. Bickel, B. (1997). Aspectual scope and the difference between logical and semantic representation. Lingua, 102, 115-131. Bickerton, D. (1984). The language bioprogram hypothesis. Behaviorial and Brain Science, 7, 173-188. Bickerton, D. (1999). Creole language, the Langauge Bioprogram Hypothesis, and langauge acquisition. In W. C. Richie & Tej.K.Bhatia (Eds.), Handbook of Child Langauge Acquisition. San Diego, CA: Academic Press. Binnick, R., I. . (1991). Time and the Verb: A Guide to Tense and Aspect. New York: Oxford University Press. Bloom, L., Lifter, K., & Hafitz, J. (1980). Semantics of verbs and development of verb inflection in child language. Language, 56, 386-412. Blutner, R., & Zeevat, H. (Eds.). (2004). Optimality Theory and Pragmatics. Houndmills, Basingstoke, Hampshire: Palgrave Macmillan. Boersma, P., & Hayes, B. (2001). Empirical tests of the Gradual Learning Algorithm. Linguistic Inquiry, 32(1), 45-86. Brown, R. (1973). A First Language: The Early Stages. Cambridge,Mass.: Harvard University Press. Bybee, J., Perkins, R., & Pagliuca, W. (1994). Evolution of Grammar: Tense, Aspect, and Modality in the Languages of the World. Chicago and Loncon: The University of Chicago Press. Cai, J. (2002). The Effects of multiple Linguistic Factors on the Simple Past Use in English Interlanguage. Unpublished PhD Dissertation, Luoyang Foreign Language University. Cai, J. (2003). Jixuwen yupian jiegou yu yingyu guduyuzhong yiban guoqushi bianyi de guanxi (The relationship between narrative structure and the simple past variation in English interlanguage--A test of the discourse hypothesis). Xiandai Waiyu (Modern Foreign Languages, 26(1), 59-68. Carlson, G. (1977). Reference to Kinds in English. New York: Garland. Chao, Y. R. (1968). A Grammar of Spoken Chinese. Berkeley and Los Angeles: University of Californica Press. Choi, H.-W. (1996). Optimizing Structure in Context: Scrambling and Information Structure Unpublished Ph.D. Dissertation, Stanford University. Collins, L. (2002). The roles of L1 influence and lexical aspect in the acquisition of temporal morphology. Lanuage Learning, 52(1), 43-94. Comrie, B. (1976). Aspect: An Introduction to the Study of Verbal Aspect and Related Problems. Cambridge: Cambridge University Press. 209

Coppieters, R. (1987). Competence differences between native and near-native speakers. Language, 63, 544-573. Crain, S., & Thornton, R. (1998). Investigating in Universal Grammar. Cambridge, Massachusetts: The MIT Press. Dahl, O. (1985). Tense and Aspect System. Oxford: Basil Blackwell. Dahl, O. (1999). Aspect: Basic principles. In K. Brown & J. Miller (Eds.), Concise Encyclopedia of Grammatical Categories. Amsterdam: Elsevier. Dai, Y. (1997). Xiandai Hanyu Shiti Xitong Yanjiu [A Study on the Tense-Aspect System in Modern Chinese]. Hangzhou: Zhejiang Jiaoyu Chubanshe [Zhejiang Education Press]. de Hoop, H., & de Swart, H. (Eds.). (2000). Optimality Theoretic Semantics. OTS Working Papers: Utrecht Institute of Linguistics OTS. de Swart, H. (1998). Aspect shift and coercion. Natural Language and Linguistic Theory, 16, 347-384. de Swart, H., & Verkuyl, H. (1999). Tense and Aspect in Sentence and Discourse. Available: http://www.folli.uva.nl/CD/1999/esslli99/courses/swart-verkuyl.html. April 23,2003]. Dietrich, R., Klein, W., & Noyau, C. (Eds.). (1995). The Acquisition of Temporality in a Second Language. Amsterdam/Philadelphia: John Benjamins. Dingare, S. (2001). The effect of feature hierarchies on frequencies of passivization in English. Rutgers Optimality Archive (ROA). Available: http://roa.rutgers.edu/files/467-0901/4670901-DINGARE-0-3.PDF [2004, May 4]. Dowty, D. R. (1977). Toward a semantic analysis of verb aspect and the English 'imperfective' progressive. Linguistics and Philosophy, 1, 45-77. Dowty, D. R. (1979). Word Meaning and Montague Grammar. Dordretcht: Kluwer. Eckman, F. R. (1977). Markedness and contrastive analysis. Language Learning, 27, 315-330. Ehrlich, S. (1987). Aspect, foregrounding and point of view. Text, 7(4), 363-376. Elmes, D. G., Kantowitz, B. H., & Roediger III, H. L. (1999). Research Methods in Psychology (Sixth ed.). New York: Brooks/Cole Publishing Comapany. Engelberg, S. (2001). The Semantics of the Progressive. Paper presented at the 2001 Conference of the Australian Liguistic Society, Sydney. Filip, H. (1999). Aspect, Eventuality Type and Nominal Reference. London: Routledge. Flashner, V. E. (1989). Transfer of aspect in the English oral narrative of native Russian speakers. In H. Dechert & M. Raupach (Eds.), Transfer in Language Production. Amsterdam: John Benjamins. Garey, H. (1957). Verbal Aspect in French. Language, 33(91-110). Gass, S., Mackey, A., Alvarez-Torres, M. J., & Fernández-García, M. (1999). The effects of task repetition on linguistic output. Language Learning, 49, 549-581. Gibson, E., & Wexler, K. (1994). Triggers. Linguistic Inquiry, 25, 405-457.

210

Giorgi, A., & Pianesi, F. (1997). Tense and Aspect: From Semantics to Morphosyntax. New York: Oxford University Press. Givon, T. (1982). Tense-asypect modality: The creole prototype and beyond. In P. Hopper (Ed.), Tense-aspect: Between Semantics and Pragmatics (pp. 115-163). Amsterdam: Benjamins. Glasbey, S. (1996). The progressive: a channel-theoretic analysis. Journal of Semantics, 13, 331361. Glasbey, S. (1998). Progressives, states and backgrounding. In S. Rothstein (Ed.), Event and Grammar (pp. 105-124). Dordrecht: Kluwer Academic Publishers. Gnanadesikan, A. (1995). Markedness and Faithfulness Constraints in Child Phonology. Available: http://roa.rutgers.edu/files/67-0000/67-0000-GNANADESIKAN-0-0.PDF [2004, October 26]. Gregory, R. L. (1997). Eye and Brain: The Psychology of Seeing. Oxford: Oxford University Press. Grimshaw, J., & Samek-Lodovici, V. (1998). Optimal Subjects and Subject Universals. In P. Barbosa (Ed.), Is the Best Good Enough. Cambridge, MA: MIT Press. Haji-Abdolhosseini, M. (2005). Modularity and Soft Constraints: A Study of Conflict Resolution in Grammar. Unpublished Doctoral Dissertation, University of Toronto, Toronto. Harley, B. (1989). Functional Grammar in French Immersion: A Classroom Experiment. Applied Linguistics, 10(331-359). Hinrichs, E. (1986). Temporal anaphora in discourse Linguistics and Philosophy, 9, 63-82. Holt, D. E. (2003). Optimality Theory and Language Change. Dordrecht: Kluwer Academic Publishers. Hopper, P. J. (1979). Aspect and foregrounding in discourse. In T. Givon (Ed.), Syntax and Semantics: Discourse and Syntax. New York: Academic Press. Hopper, P. J. (1982). Aspect between discourse and grammar: An introductory essay for the volume. In P. Hopper (Ed.), Tense-aspect: Between Semantics and Pragmatics (pp. 3-18). Amsterdam: Benjamins. Hopper, P. J., & Thompson, S. A. (1980). Transitivity in grammar and discourse. Language, 56(2), 251-299. Housen, A. (1994). Tense and aspect in second langauge learning. In C. Vet & C. Vetters (Eds.), Tense and Aspect in Discourse (pp. 257-291). Berlin: Mouton de Gruyter. Hulstijn, J. H. (2002). Towards a Unified Account of the Representation, Processing and Acquisition of L2 Knowledge. Second Language Research, 18(3), 193-223. Jackendoff, R. (1990). Semantic Structure. Cambridge, MA. : MIT Press. Jespersen, O. (1933). Essentials of English Grammar. London: George Allen and Unwin. Jin, L., & Hendriks, H. (2003). The development of aspect marking in L1 and L2 Chinese. Research Centre for English and Applied Linguistics, University of Cambridge. Available: http://www.rceal.cam.ac.uk/Publications/Working/Vol9/Jin.pdf [2004, March, 12].

211

Johanson, L. (2000). Voipoint operators in european langauges. In O. Dahl (Ed.), Tense and Aspect in the Langauges of Europe. Berlin: Mouton de Gruyter. Kabakciev, K. (2000). Aspect in English: A "Common-Sense" View of the Interplay between Verbal and Nominal Referent. Dordrecht: Kluwer Academic Publishers. Kager, R. (1999). Optimality Theory. Cambridge: Cambridge University Press. Kamp, H., & Reyle, U. (1993). From Discourse to Logic: Introduction to Model-theoretic Semantics of Natural Languages, Formal Logic and Discourse Representation. Dordrecht: Kluwer. Klein, W., Li, P., & Hendriks, H. (2000). Aspect and Assertion in Mandarin Chinese Natural langauge and Linguistic Theory 18, 723-770. Kumpf, L. (1984). Temporal systems and universality in interlanguage: a case study. In F. R. Eckman & L. H. Bell & D. Nelson (Eds.), Universals of Second Language Acquisition. Rowley, MA: Newbury House. Landman, F. (1992). The progressive. Natural Language Semantics, 1(1), 1-32. Larsen-Freeman, D., & Long, M. H. (1991). An Introduction to Second Language Acquisition Research. London: Longman. Lascarides, A. (1988). A Formal Semantic Analysis of the Progressive. Unpublished PhD. Dissertation, University of Edinburgh, Edinburgh. Legendre, G., Grimshaw, J., & Vikner, S. (Eds.). (1998). Optimality Theoretic Syntax. Cambridge, Mass.: MIT Press. Levin, B. (1993). English Verb Classes and Alternations: A preliminary Investigation. Chicago: The University of Chicago Press. Li, C. N., & Thompson, S. A. (1981). Mandarin Chinese: A Functional Reference Grammar. Berkley: University of California Press. Li, P. (2002). Emergent sematnic structure and language acquisition. In H. Kao & C. K. Leong & G. D. Guo (Eds.), Cognitive Neuroscience Studies of the Chinese Language (pp. 79-98). Hong Kong: Hong Kong University Press. Li, P., & Bowerman, M. (1998). The acquisition of lexical and grammatical aspect in Chinese. First Langauge(18), 311-350. Li, P., & Shirai, Y. (2000). The Acquisition of Lexical and Grammatical Aspect. Berlin and New York: Wouton de Gruyter. Liu, D., Brooks, D., Sjoquist, R., & Rickley, S. L. (Eds.). (2004). New Senior English for China: Student's Book 2. Beijing: Renmin Jiaoyu Chubanshe [People's Eduction Press]. Lu, S. (1942). Zhongguo Wenfa Yaolue [The Essence of Chinese Grammar]. Hongkong: Shangwu Yingshu Guang. Lyons, J. (1977). Semantics. Cambridge: Cambridge University Press. Ma, Q. (2003). OT Yufa de Kexunxing Yanjiu [Researches on Learnability of OT Grammar]. Waiguoyu [Jounal of Foreign Languages](5), 18-26.

212

MacWhinney, B. (1995). Computational Analysis of Interactions. In P. Fletcher & B. MacWhinney (Eds.), The Handbook of Child Language. Cambridge, Massachussets: Blackwell. MacWhinney, B., & Snow, C. (1985). The Child Language Data Exchange System. Journal of Child Language, 12, 271-296. Magliano, J. P., & Schleich, M. C. (2000). Verb Aspect and Situation Models. Discourse Processes, 29(2), 83-112. Mapstone, E. R., & Harris, P. L. (1985). Is the English present progress unique? Journal of Child Language, 12, 433-441. McCarthy, J. J. (2000). Harmonic Serialism and Parallelism. In M. Hirotani & A. Coetzee & N. Hall & J.-y. Kim (Eds.), Proceedings of the North East Linguistics Society 30. Amherst, MA. McCarthy, J. J. (2002). A Thematic Guide to Optimality Theory. Cambridge: Cambridge University Press. McCarthy, J. J., & Prince, A. (1994). The emergence of the unmarked: Optimality in prosodic morphology. Paper presented at the The North East Linguistic Society 24, Amherst, MA. McCarthy, J. J., & Prince, A. S. (1995). Faithfulness and Reduplicative Identity. In J. Beckman & L. W. Dickey & S. Urbanczyk (Eds.), Papers in Optimality Theory. Amherst, Mass.: Graduate LIngusitic student Association. Moens, M. (1987). Tense, Aspect and Temporal Reference. Unpublished Ph.D. dissertation, University of Edinburgh. Moens, M., & Steedman, M. (1988). Temporal ontology and temporal reference. Computatinal Linguistics, 14(2), 15-28. Montrul, S., & Slabakova, R. (2002). The L2 acquisition of morphosyntactic and semantic properties of the aspectual tenses preterite and imperfect. In A. T. Perez-Leroux & J. Liceras (Eds.), The Acquisition of Spanish Morphosyntax. Dordrecht: Kluwer Publishers. Montrul, S., & Slabakova, R. (2003). Competence similarities between native and near-native speakers: An investigation of the preterite/imperfect contrast in Spanish. Studies in Second Language Acquisition, 25, 351-398. Mourelatos, A. (1978). Events, Processes and States, Syntax and Semantics Vol. 14, Tense and Aspect. New York: Academic Press. O'Grady, W. (2003). The Radical Middles: Nativism without Universal Grammar. In C. J. Doughty & M. H. Long (Eds.), The Handbook of Second Language Acquisition. Malden, MA. : Blackwell. Ogihara, T. (1990). The semantics of the progressive and the perfect in English, [Html]. Centre for Cognitive Science. Available: faculty.washington.edu/ogihara/papers/deliv90.pdf [2003, October 15]. Olsen, M. B. (1997). A Semantic and Pragmatic Model of Lexical and Grammatical Aspect. New York: Garland. Parson, T. (1990). The progressive in English: Events, states and Processes. Linguistics and Philosophy, 12.

213

Prince, A., & Smolensky, P. (1993). Optimality Theory: Constraint Interaction in Generative Grammar. New Brunswick, NJ: Rutgers University Center for Cognitive Science. Prince, A., & Smolensky, P. (1997). Optimality: From neural networks to Universal Grammar. Science, 275(14), 1604-1610. Prince, A., & Smolensky, P. (2004). Optimality Theory: Constraint Interaction in Generative Grammar. Oxford: Blackwell. Quirk, R., Greenbaum, H., Leech, G., & Svartvik, J. (1985). A Comprehensive Grammar of the English Language. London and New York: Longman. Reinhart, T. (1984). Principles of Gestalt Perception in the Temporal Organization of Narrative Text. Lingusitics, 22, 779-809. Robison, R. E. (1990). The primacy of aspect: Aspectual marking in English interlanguage. Studies in Second Language Acquisition, 12, 315-330. Robison, R. E. (1995). The Aspect Hypothesis revisited: A cross-sectional study of tense and aspect marking in interlanguage. Applied Linguistics, 16, 344-370. Rohde, A. (1996). The aspect hypothesis and he emergence of tense distinctions in naturalistic L2 acquisition. Linguistics, 34(1115-1137). Rothstein, S. (2004). Structruring Events: A Study in the Semantics of Lexical Aspect. Oxford: Oxford University Press. Russell, K. (1997). Optimality Theory and morphology. In D. Archangeli & D. T. Langendoen (Eds.), Optimality Theory: An Overview. Oxford: Blackwell. Salaberry, R. M. (1999). The development of aspectual distinctions in classroom L2 Spanish. Applied Linguistics, 20, 151-178. Salaberry, R. M. (2000). The Development of Past Tense Morphology in L2 Spanish. Amsterdam and Philadelphia: John Benjamins. Scheffer, J. (1975). The Progressive in English. Amsterdam: North-Holland Publishing Company. Schwartz, B., & Sprouse, R. (1996). L2 Cognitive States and the Full Transfer/Full Access Model. Second Language Research, 12(1), 40-72. Seliger, H. (1978). Implications of a multiple critical periods hypothesis for second language learning. In W. C. Richie (Ed.), Second Language Research: Issues and Implications (pp. 11-19). New York: Academic Press. Sells, P. (Ed.). (2001). Formal and Empirical Issues in Optimality-theoretic Syntax. Stanford: CSLI Publications. Shirai, Y. (1998). Where the progressive and the resultative meet imperfective aspect in Japanese, Chinese, Korean and English. Studies in Language, 22(3), 661-692. Shirai, Y., & Andersen, R. W. (1995). The acquisition of tense-aspect morphology: A prototype account. Language, 71, 743-762. Shirai, Y., & Kurono, A. (1998). The acquisition of tense-aspect marking in Japanese as a second language. Language Learning, 48(2), 245-279.

214

Skehan, P., & Foster, P. (1999). The influence of task structure and processing conditions on narrative retellings. Language Learning, 49, 93-120. Slabakova, R. (2000). L1 transfer revisited: the L2 acquisition of telicity marking in English by Spanish and Bulgarian native speakers. Lingusitics(38), 739-770. Slabakova, R. (2001). Telicity in the Second Language. Amsterdam: John Benjamins. Slabakova, R. (2002). Recent research on the acquisition of aspect: an embarrassment of riches? Second Language Research, 18(2), 172-188. Slabakova, R. (2003). Semantic evidence for functional categories in interlanguage grammars. Second Language Research, 19(1), 42-75. Slabakova, R., & Montrul, S. (2002). On the semantics of viewpoint aspect and its L2 acquisition. In Y. Shirai & M. R. Salaberry (Eds.), Tense-Aspect Morphology in L2 Acquisition (pp. 363398). Amsterdam: John Benjamins. Slobin, D. I. (1981). The origins of grammatical encoding of events. In W. Deutsch (Ed.), The Child's Construction of Language. London: Academic Press. Slobin, D. I. (1985). crosslinguistic Evidence for the Language-making Capacity. In D. I. Slobin (Ed.), The Crosslinguistic Study of Language Acquisition, Vol 2, Theoretical Issues (pp. 1157-1257). Hillsdale, New Jersey: Lawrence Erlbaum. Smith, C. (2003). Modes of Discourse: The Local Structure of Discourse. Cambridge: Cambridge University Press. Smith, C. S. (1983). A theory of aspectual choice. Language, 59, 479-501. Smith, C. S. (1994). Aspectual viewpoint and situation type in Mandarian Chinese. Journal of East Asian Linguistics, 3, 107-146. Smith, C. S. (1997). The Parameter of Aspect (2nd ed.). Dordrecht: Kluwer Academic Publishers. Smolensky, P. (2000). Grammar-based connectionist approaches to language Cognitive Science, 23, 589-613. Smolensky, P., Davidson, L., & Jusczyk, P. (2004). The Intial and final states: Theoretical implications and experimental explorations of richness of the base. Available: http://roa.rutgers.edu/files/428-1200/428-1200-SMOLENSKY-0-0.PDF [2005, February 22]. Sorace, A. (1996). The Use of acceptability judgments in second language acquisition research. In W. C. Ritchie & T. K.Bhatia (Eds.), Handbook of Second Language Acquisition. San Deigo, CA: Academic Press. Stoll, S. E. (2001). Acquisition of Russion Aspect. Unpublished PhD Dissertation, University of California, Berkeley. Sullivan, J., & Caplan, N. A. (2004). Beyond the Dictogloss :Learner-Generated Attention to Form in a Collaborative, Communicative Classroom Activity. Available: http://www.msu.edu/user/caplan/tesol04/ [2005, May 10]. Synder, W. (1996). The Acquisitional Role of the Syntax-morphology Interface: Morphological Compounds and Syntaxtic Complex Predicates. In A. Stringerfellow (Ed.), Proceedings of the 20th Annual Boston University Conference on Language Development (pp. 728-735). Somerville, MA. : Cascadilla Press. 215

Synder, W. (2001). On the Nature of Syntactic Variation: Evidence from Complex Predicates and Complex Word-formation. Language, 77, 324-342. Tenny, C. L. (1994). Aspect Roles and the Syntax-semantics Interface. Dordrecht: Kluwer Academic Publishers. Tesar, B., & Smolensky, P. (1998a). Learnalibility in Optimality Theory. Linguistic Inquiry, 29(2), 229-268. Tesar, B., & Smolensky, P. (1998b). Learning Optimality-Theoretic Grammars. Lingua, 106, 161-196. Tesar, B., & Smolensky, P. (2000). Learnalibility in Optimality Theory. Cambridge, MA.: MIT Press. van Hout, A. (2005). Imperfect Imperfective. In P. Kempchinsky & R. Slabakova (Eds.), Aspectual Inquiries. Dordrecht: Springer. Van Valin, R. D. (2002). A Brief Overview of Role and Reference Grammar. Available: http://wings.buffalo.edu/linguistics/research/rrg/RRGpaper.pdf [2005, February 21]. Vendler, Z. (1967). Linguistics in Philosophy. Ithaca and London: Cornell University Press. Verkuyl, H. (1972). On the Compositional Nature of the Aspects. Dordrecht: Reidel. Verkuyl, H. (1989). Aspectual Classes and Aspectual Composition. Linguistics and Philosophy, 12(39-94). Verkuyl, H. (1993). A Theory of Aspectuality: The Interaction between Temporal and Atemporal Structure. Cambridge: Cambridge University Press. Verkuyl, H. (1999). Aspectual Issues: Studies on Time and Quantity. Stanford, California: CSLI Publications. Vlach, F. (1981). The semantics of the progressives. In P. Tedeschi & A. Zaenen (Eds.), Syntax and Semantics 14: Tense and Aspect. New York: Academic Press. Wallace, S. (1982). Figure and ground: The interrelationships of linguistic categories. In P. J. Hopper (Ed.), Tense-aspect: Between semantics and pragmatics (pp. 201-226). Amsterdam/Philadelphia: John Benjamins Publishing Company. Wang, L. (1943). Zhongguo Yufa Lilun [Chinese Grammar Theory]. Beijing: Zhonghua Shuju. Weist, R. M., Wysocka, H., Witkoska-Stadnik, K., Buczowska, E., & Konieczna, E. (1984). The defective tense hypothesis: On the emergence of tense and aspect in Child Polish. Journal of Child Language, 11, 347-374. Xiao, Z. (2002). A Corpus-based Study of Aspect in Mandarin Chinese Lancaster University. Xu, D. (1992). Hanyulide "zai" and "zhe" ["Zai" and "Zhe" in Manderin Chinese] Zhongguo Yuwen [Chinese Lingua](6), 453-461. Yang, H., & Weir, C. (1998). Daxue Yingyu Siliuji Kaoshi Xiaodu Yanjiu [Validation Studies of the National College English Test]. Shanghai: Shanghai Waiyu Jiaoyu Chubanshe [Shanghai Foreign Language Education Press]. Yang, S., & Huang, Y. (2004). The impact of the absence of grammatical tense in L1 on the acquisition of the tense-aspect system in L2. IRAL, 42(49-70). 216

Appendix I Progressive Aspect Acceptability Judgment Task (AJT)

注意：请阅读下列句子，判断它们在英语中的可接受程度，并将相应的数字填 入句子前的括号中。 2=completely acceptable

2=完全可以接受

1=almost acceptable

1=几乎可以接受

0=sounds odd but acceptable

0＝勉强可以接受

-2

-1

0

Totally Unacceptable

1

2

Completely Acceptable

(

)

1

Alice’s face is turning red

(

)

2

Brown is walking slowly along the street

(

)

3

Carl is eating with his parents

(

)

4

Chickens are having wings

(

)

5

Cindy is liking jazz music

(

)

6

Cindy is understanding the lecture

(

)

7

Davis is swimming across the pool

(

)

8

Dick is painting the wall

(

)

9

Evan is supposing that she lied

(

)

10

Gill is tapping the table

(

)

11

George is dancing at the local club

(

)

12

Henry is feeling cold

217

(

)

13

Her eyes are blinking

(

)

14

It is raining

(

)

15

Jack’s feet are hurting

(

)

16

Jessica is having black hair

(

)

17

Jim is disagreeing with us

(

)

18

Kole is shooting the tiger

(

)

19

Marie is wanting another cake

(

)

20

Mary is appearing on the stage

(

)

21

Mary is believing the truth of his story

(

)

22

Mary is hearing the story

(

)

23

Mary is owning two cars

(

)

24

Michael is being a Canadian

(

)

25

Mike is reading a new book

(

)

26

Money is mattering much to Bob

(

)

27

Nick is banging the door

(

)

28

The water is boiling

(

)

29

Nick is stopping the car

(

)

30

Peter is realizing his fault

(

)

31

The branches are shaking against the wall

(

)

32

Rosy is cutting cloth with scissors

(

)

33

Sam’s English is improving

(

)

34

The lightening is flashing across the sky

(

)

35

Smith is drowning in the swimming pool

(

)

36

The food is smelling bad

(

)

37

Stephan is being a tall boy

(

)

38

Ted’s back is aching

(

)

39

The ball is knocking Mike on the head

(

)

40

The box is containing candies

(

)

41

Rebecca is dying of cancer

(

)

42

The engine is running well

(

)

43

The house is exploding in the fire

(

)

44

She is intending to visit New York city

(

)

45

The sun is shining

(

)

46

Vivian is sitting in the armchair 218

(

)

47

Nick is flying to London

(

)

48

The weather is getting warmer

(

)

49

This computer is belonging to Sam

(

)

50

Tim is reaching the top of the mountain

(

)

51

Tony is jumping in the backyard

(

)

52

The water in the lake is freezing

219

Appendix II Aspect Sentence Interpretation Task (ASIT) 同学们！你们将参与一个阅读测验。你们将通过电脑屏幕阅读几篇短文。每篇 短文有两道陈述题，一道在文中，另一道在文后。请根据你的理解，判断这两 道陈述题是否符合短文内容。如果符合，请在答题纸相应的空格上添上（T）， 反之，则填上（F）。 请看示例 Passage 0 Mary had a little lamb. She took the lamb to the riverside every afternoon. 0.1

True (T) or False (F)

( ) Mary took her lamb to the riverside in the morning. One afternoon, however, her lamb refused to go to the riverside. Mary wondered what happened to the lamb. 0.2.

True (T) or False (F)

( ) Mary had no idea what went wrong with her lamb. Now the test starts. 现在开始测试。

Passage One Cathy was Bob’s long standing girl friend. Bob liked Cathy a lot but was finding himself attracted to some other women. He was intimate (亲昵) with Rose at a party when Cathy wasn’t looking. Bob danced with Rose to his favorite song. 1.1

True (T) or False (F) :

( ) Bob and Rose danced to the end of Bob’s favorite song. He told Rose how attractive she was. Before he left the party, he asked Rose out on a date （约会）. Cathy found out about Bob and Rose from a friend.

220

She told him that their relationship was over. (78 words) 1.2

True (T) or False (F) :

( ) Bob loved Cathy more than any other girl. This is the end of Passage One. Now please get ready to read Passage Two. Passage Two Davy liked to go boating on weekends. He loved the feeling of lying in the boat and doing nothing at all. One Sunday, he went into his boat as usual. A few kids were playing football by the riverside and Davy was looking into the sky. 2.1 (

True (T)or False (F) ) The boys were playing football in the water.

Suddenly the football flew straight at him and struck him in the head. The boat toppled (翻). Poor Davy was drowning (drown, 淹). The boys were too scared. They ran away in all directions immediately. (79 words) 2.2 (

True (T) or False (F) ) Davy lost his life in the accident.

This is the end of Passage Two. Now please get ready to read Passage Three . Passage Three Ted’s brother was a hunter. On his tenth birthday, Ted asked his brother to take him to hunt. His brother promised to take him along on Ted’s next birthday. The big day came. 3.1 (

True (T) or False (F) ) Ted went hunting with his brother on his tenth birthday.

Ted and his brother were lurking (潜伏) in the grass when a tiger appeared thirty yards away. Ted was shooting the tiger. He acted like a real hunter. His brother believed that Ted would be the best hunter in the local area. (79 words)

221

3.2 (

True (T) or False (F) ) The tiger was killed by Ted .

This is the end of Passage Three. Now please get ready to read Passage Four. Passage Four It was a cold winter night. John was watching TV with his family when someone was knocking at the window. John opened the window. It was a stranger. He explained that he was extremely thirsty after a long journey and he wanted something to drink. John handed him a cup of water. The water was freezing as it was extremely cold outside. 4.1 (

True （T）or False （F） ) The water turned into ice when John handed it out.

John invited him in but the stranger declined. He thanked John and went on his journey. (79 words) 4.2 (

True （T）or False （F） ) John was unwilling to give any water to a stranger.

This is the end of Passage Four. Now please get ready to read Passage Five. Passage Five John, an 18-year-old boy, dreamed of being a pilot. He loved the feeling of flying a plane. But he couldn’t be a pilot until he was 20 years old. He couldn’t wait to be a pilot after his 20th birthday. The big day of the pilot recruitment (招募)approached. 5.1 (

True (T) or False (F) ) The day of the pilot recruitment arrived at last.

But John had his left leg broken in a car accident on his trip to pilot recruitment centre. John’s dream of being a pilot remained a dream. 5.2 (

True (T) or False (F) ) John’s dream of being a pilot came true on his 20th birthday.

222

This is the end of Passage Five. Now please get ready to read Passage Six.

Passage Six Lucy was Tom’s girlfriend. One day they went to the cinema. After they took their seats, Lucy told Tony that she felt thirsty. She said she would like to have some ice-cream. Tony was out for the ice-cream . However, the ice-cream store was a few blocks away. （block, 街区） 6.1 (

True (T) or False (F) ) The ice-cream store was close to the cinema.

Tony hurried back to the cinema. The ice-cream melted. Tony was all in sweat but Lucy’s heart was melted. She was moved by what Tony did for her. （79 words） 6.2 (

True (T) or False (F): ) All the ice-cream melted away when Tony was back.

This is the end of Passage Six. Now please get ready to read Passage Seven. Passage Seven Ever since he was a kid, Gabe was afraid of heights. He decided to get rid of it once and for all. His plan was to spend the night on the top of the local fire tower. He took his friends along with him that night. Gabe was reaching the top of the tower. 7.1

True (T) or False (F)

( ) Gabe got to the top of the tower that night. He was never more scared. His friends tried to calm him down. Gabe tried his best to listen to them. The next weekend he rode the rollercoaster for the first time. (79 words) 7.2

True (T) or False (F)

223

( ) Gabe spent the night on the tower with his friends. This is the end of Passage Ten. Now please get ready to read Passage Eight. Passage Eight Tom was walking along the street. He saw a young artist playing the violin. The young artist was earning money from the passers-by. Tom decided to give a little money to the young man. But he thought his large note (大票子) was too big for the young man. 8.1

True (T) or False (F)

( ) The young man had a large note in his pockets. He found some coins (硬币) in his pockets. He really enjoyed the music. He even thought it was better than the concert music. (78 words) 8.2

True (T) or False (F) :

( )Tom managed to take some coins from his pockets. This is the end of Passage Twelve. Now please get ready to read Passage Nine. Passage Nine Mrs. Smith had a cat. One evening, however, her cat was missing. She got a letter instead. It said that her cat would be returned unless $ 1,000 was paid. Mrs. Smith was not sure whether she should report to the police. Finally, She was drawing some money from her bank account. The policemen thought that she was crazy. 9.1 True (T) or False (F) : ( ) The police encouraged her to pay the money to the kidnappers. But Mrs. Smith said the cat was dear to her. She could not imagine a life without the cat around. 9.

2 True (T) or False (F):

( ) Mrs. Smith took out $1,000 from her bank account. This is the end of Passage Thirteen. Now please get ready to read Passage Ten.

224

Passage Ten Wendy and Joan were best friends in college. It had been ten years since they last saw each other. One day, Wendy decided to visit Joan in New York. 10.1

True (T) or False (F) :

( ) Joan invited Wendy to visit her home in New York City. Wendy drove from Montana towards New York. She called Joan on her way, telling that she was coming to see her. But she discovered that Joan was not the same person. Joan seemed unwilling to spend any time with her. Wendy felt very disappointed. 10.2 True (T) or False (F) : ( ) Wendy completed his trip from Montana to New York. This is the end of Passage fifteen. Now please get ready to read Passage Eleven. Passage Eleven Carl was the best athlete （运动员）when he was at school. Unfortunately, Carl’s physical condition was going down after he left the school. One of his friends bet him $10 that he couldn’t even do five pushups (俯卧撑). He had to accept the challenge (挑战). He began to do the pushups . He won the bet. 9.1 (

True (T) or False (F) ) Carl managed to finish the five push-ups.

He was absolutely exhausted (精疲力竭) by how hard it was. Two weeks later, he joined a health club. (77 words) 9.2 (

True (T) or False (F) ) Five pushups was too hard for Tom even though he used to be a good athlete.

This is the end of Passage Eleven. Now please get ready to read Passage Twelve. Passage Twelve Stephanie and Kate had been best friends.

225

They had not seen each other since Kate’s family moved to another city. Finally, Stephanie decided it was time to visit Kate. She was on her way to the airport when she got a flat tire (车胎瘪了). Stephanie was changing the flat tire. 11.1

True (T) or False (F) :

( ) Stephanie managed to change the flat tire. She was worried about being late for her flight. She was afraid that she would have to catch a later flight. She was extremely disappointed. 11.2

True (T) or False (F) :

( ) Stephanie and Kate were no longer good friends since Kate moved to another city.

This is the end of the test. Thank you for participation.

226

Appendix III Movie Clip Retell Task 同学们： 你们将参加一次有关外语学习的测验。在这项测验中，你们将观看英国电影 《憨豆先生》的几个片断。看完每个片断后，我们将呈现该片断中出现过的四 幅图片，这些图片会帮助你回忆片断中的内容。每看完一个片断，你们要根据 电影片断和图片的内容，用所给动词的适当形式填空。 所有的答案必须填写在答题纸上。 谢谢你的参与和合作。

CLIP 1

Perfume

56”03

Mr. Bean got a new charge card （付款卡）and he decided to use it to buy some daily necessities. But when he went into the supermarket （超市）, a strong odor （气味） __(1.1)__ (bother, 让…难受，干扰) him. Some women were spraying (喷 洒) perfume （香水） on their wrists （手腕）to test its quality （质量）. The odor of the perfume was so strong that Mr. Bean could hardly breathe. He crawled on the ground to avoid the bad smell in the air. When he __(1.2)__(stand) up, he nearly knocked down (撞倒) another woman who __(1.3)__ (hurry) towards the cosmetic counter. Clip 2

Toothbrush

44”09

Mr. Bean wanted to buy a toothbrush （牙刷）. Various types of toothbrushes were displayed on the shelves. Mr. Bean showed great interest in one of them. He took it out and tore open its packaging （外包装）. Then he __(2.1)__(brush) his teeth with it before he decided to buy it. But he __(2.2)__(prefer) to have a brand new （崭新）

227

one with new packaging. He __(2.3)__(look) around . When he found there was not anyone nearby, he replaced it with a new toothbrush. Clip 3

Telephone

1’04’’07

Mr. Bean wanted to buy a new telephone. When he __(3.1)__(reach) the telephone section of the supermarket, a woman __(3.2)__ (make) a call at a desk. Mr. Bean chose one telephone on the shelf to see whether it worked well. As the telephone was not connected to the network（联网）, they were not working for the moment. Mr. Bean, however, was not aware of this. When he was hearing the ringing of those telephones on the shelf, he __(3.3)__(wonder) what went wrong with them. Finally, he __(3.4)__(pick) up the office telephone that the woman had used. He was pleased to find that this one worked properly. Mr. Bean decided to buy this one and put it into his shopping basket. Clip 4

Knife

32”03

When Mr. Bean was waiting for his meal in a restaurant, he became very impatient （不耐烦）. He took the knife on his table and thrust it into the air. The woman sitting next to his table was startled. She __(4.1)__(shoot) Mr. Bean a disproving （不 赞同） look to stop his ill manner. Mr. Bean __(4.2)__(appear) very unhappy. He __(4.3)__(stop) playing with the knife reluctantly. Instead, he __(4.4)__ (tap) on the glasses and cups instead when the woman was drinking. He __(4.5)__ (make) such a great noise that most of the dinner guests at the restaurant could not go on with their dinner. They __(4.6)__ (express) their annoyance (不满，恼火) silently but Mr. Bean did not seem to care.

Clip 5

Mr. Bean’s Dinner

228

1’47”26

When Mr. Bean was waiting for his dinner to come, the head waiter (领班) came up to him. He helped Mr. Bean to lay the napkin （餐巾）across his lap (膝上). The waiter’s manner __(5.1)__(amuse，让…感到可笑) Mr. Bean a great deal. He imitated it several times before he, without warning (出其不意地), threw the napkin to the neighboring table. A woman __(5.2)__(have) dinner with her friends at that table. When she __(5.3)__(find) out who was so mischievous （淘气，恶作剧）， Mr. Bean pretented that it had nothing to do with him. Just then the head waiter came and served Mr. Bean’s dinner. He paid for the dinner and the headwaiter uncovered （揭开盖子）the dish. It didn’t seem like a delicious dish. As Mr. Bean was tasting his horrible food, the restaurant owner and the headwaiter asked whether he was satisfied with it. Mr. Bean __(5.4)__ (nod) with a smile as if he were enjoying the food.

Clip 6

Parking （泊车，停车）

1’39”21

It was the school’s Open Day. Mr. Bean was one of the adults in the community who__(6.1)__ (come) to take the various courses the school offered on this day. When he drove to the school, a group of soldiers __(6.2)__(drill) on the playground. Mr. Bean drove all around the playground. He was finding a place to park his car when his car almost knocked down (撞倒) an officer, who__(6.3)__ (give) orders to the soldiers. At last, Mr. Bean __(6.4)__ (drive) to the parking zone, only to find that there was no more room for his own car. When Mr. Bean saw another car that looked exactly like his own, he had an idea. He __(6.5)__(push) that car out of the parking zone and parked his own car in its place. Mr. Bean appeared very satisfied with the way he solved the problem.

229

Clip 7

Drilling

1’15”00

Mr. Bean was amused at the way a group of soldiers were acting, They were following each instruction of an officer and making different gestures （姿势，动 作）. He came near to the soldiers when the officer left for a while and the group of soldiers were waiting for him. Mr. Bean was surprised to find that the soldiers suddenly __ (7.1)__ at attention (立正) when he __(7.2)__ his throat （清嗓子）. He hid himself behind a car and uttered some more sounds. But it seemed that the soldiers could understand his meaningless sounds and acted accordingly (相应地). When the soldiers, following Mr. Bean’s last instruction, was standing on one foot and __(7.3)__ (wave) their hands in the air, the officer came back, wondering what his soldiers__(7.11)__(do).

Clip 8

Calligraphy （书法）

1’12”23

Then Mr. Bean went to another room where more art works were displayed. Some people __(8.1)__ (finish) their projects (展品). An old man was looking at his stamp collection through a reading glass when Mr. Bean was going by his desk. Mr. Bean yawned （打哈欠） and knocked his head upon the stamps. When he raised his head from the collection, a stamp was sticking (贴) on his forehead（额头） and it finally fell to the ground but the old man were searching for it everywhere on the desk. Then Mr. Bean went to trouble another old man. He kept turning on and off the desk lamp (台灯) when the old man __(8.2)__(concentrate) on his handwriting practice. After Mr. Bean left, the man went back to his work . As he was writing the fine letters on the paper with the greatest care, Mr. Bean __(8.7)__(utter) a sudden cry from behind.

230

The poor old man was startled（惊吓）and his careful handwriting work was spoiled （破坏）.

Clip 9

At the Art Room

1’50”19

When Mr. Bean went to the art room, a lady __(9.1)__ (show) a group of adult learners how to draw fruit. The lady led Mr. Bean to an empty seat and told him to draw whatever he saw on the platform. Then she went away to see the work done by other learners. Mr. Bean was drawing a banana when the lady came back. She added some strokes to Mr. Bean’s drawing. These additions made Mr. Bean feel so uneasy (不自在) that he tore the drawing into pieces and threw it away. Then he __(9.2)__(draw) another banana. He was concentrating on his work so much that he did not realize the other students had started a new session （阶段，课段）.

Clip 10

At the Chemistry Lab

1’54”18

When Mr. Bean went into the chemistry lab, a set of chemistry equipment was arranged on the table. Mr. Bean was very curious about the experiment. He poured some black liquid into a cup and the colorless liquid in it __(10.1)__(turn) into red color. This made Mr. Bean feel more interested in the experiment. He wondered at the magic he was creating. Then he poured the black liquid into another larger cup which contained some white powder (粉末). As he __ (10.2)__ (see) what chemical reaction would take place, the liquids in the cups, bottles and tubes was turning blue. Mr. Bean was shocked to see that white gas (气体) was pouring out and the whole set of chemical equipment __(10.3)__ (shake). Just then a boy entered and watched what was going on with great interest. Mr. Bean ran out of the chemistry lab. He

231

__(10.4)__ (stand) outside the lab for a while. As he __(10.5)__ (decide) whether to leave or stay, there was a loud explosion （爆炸）in the lab. Some heavy blue smoke were pouring out when Mr. Bean peeped through the door to see what had happened to the boy inside.

232

Appendix IV Descriptive Statistic Data for the Figures

A. Means of Progressive and Perfective Marking for Five Situation Types in MCRT (Figure 11)

Level Low Intermediate Advanced Native

Aspect PERF PROG PERF PROG PERF PROG PERF PROG

STA Mean SD 2.633 .57 .817 .44 2.850 .60 .550 .42 2.633 .66 .633 .43 2.312 .83 .844 .60

ACT Mean SD 1.350 .54 2.033 .75 2.050 .72 1.750 .42 1.983 .52 1.950 .51 2.469 .13 1.531 .13

ACC Mean SD 2.567 .68 1.250 .65 2.267 .60 1.567 .57 2.133 .41 1.817 .40 2.000 .00 2.000 .00

ACH Mean SD 3.133 .45 .650 .42 3.033 .63 .683 .43 2.883 .57 .917 .49 2.344 .24 1.656 .24

SMF Mean SD 2.950 .71 .800 .61 2.800 .95 1.033 .82 2.800 .55 1.183 .53 2.531 .43 1.469 .43

B. Probability of Interpreting Aspect Sentence as Denoting Completed Events by all Participant groups in ASIT (Figure 12)

Low Intermediate Advanced Native

ACH Mean .800 .700 .708 .625

ACC Mean .529 .595 .621 .609

SD .19 .19 .20 .27

SD .17 .19 .23 .10

C. Mean Scores for Acceptability of Progressive Marking in Five Situation Types by L2 Participants and the Native Controls (Figure 13) Level Low Intermediate Advanced Native

STA Mean SD -.592 .56 -.953 .57 -.773 .58 .023 .38

ACT Mean SD 1.058 .63 1.300 .37 1.650 .33 1.891 .11

ACC Mean SD .931 .39 1.008 .38 1.367 .30 1.889 .11

233

ACH Mean SD -.254 .58 .012 .53 .079 .67 1.375 .56

SMF Mean SD .146 .54 .433 .49 1.017 .49 1.142 .51

D. Mean Scores for Progressive Aspect Marking by All Participant Groups on Each Situation Type in MCRT (Figure 14)

Level Low Intermediate Advanced Native

STA Mean SD .817 .38 .550 .42 .633 .43 .844 .30

ACT Mean SD 2.033 .65 1.750 .72 1.950 .51 1.531 .13

ACC Mean SD 1.250 .65 1.567 .57 1.817 .40 2.000 .00

ACH Mean SD .650 .42 .683 .43 .916 .49 1.656 .24

SMF Mean SD .800 .61 1.033 .82 1.183 .53 1.469 .43

E. Mean Scores for Perfective Aspect Marking by All Participant Groups on Each Situation Type in MCRT (Figure 15)

Level Low Intermediate Advanced Native

STA Mean SD 2.633 .57 2.850 .60 2.633 .65 2.312 .83

ACT Mean SD 1.350 .54 2.050 .71 1.983 .52 2.469 .13

ACC Mean SD 2.567 .68 2.267 .60 2.133 .41 2.000 .00

ACH Mean SD 3.133 .45 3.033 .63 2.883 .57 2.344 .24

SMF Mean SD 2.950 .71 2.800 .95 2.800 .55 2.531 .43

F. Interpretation of Critical Aspect Sentences by Four Participant Groups (Figure 16)

PROG-ACH Mean .800 .650 .467 .375

Low Intermediate Advanced Native

PERF-ACC Mean .525 .600 .717 .937

SD .31 .32 .72 .39

SD .26 .30 .27 .11

G. Acceptability of Aspect Marking on Five Subclasses of Stative Verbs by Four Participant Groups (Figure 17)

Level Low Intermediate Advanced Native

Intellectual Mean -.783 -1.05 -1.04 .933

SD .72 .85 .66 .47

Relational Mean -.592 -.683 -.608 -1.25

SD .88 .80 .84 .65

Emotional Mean -.500 -1.02 -.658 .583

234

SD .79 .79 .81 .56

Property Mean -.792 -1.16 .-1.38 -1.633

SD .73 .61 .62 .44

Bodily Sensation Mean SD -.292 .70 -.858 .64 -.175 .87 1.483 .32