台灣兒童複雜句的理解與表達之研究

全文

(1)Doctoral Dissertation Graduate Institute of English National Taiwan Normal University. Chinese Children’s Comprehension and Production of Complex Sentences. Advisor: Dr. Chun-Yin Doris Chen. Li-Chin Jennifer Yang. 中華民國一百零二年一月 January, 2013.

(2) 108 3. 4. 5. 6. 7. Huang (1982) X0. X’. 7. 6. Gasser & Smith, 1998; Mintz Gleitman, 2002; Naigles & Gelman, 1995; Smith & Sera, 1992; Waxman & Booth, 2001. Pollard &. Sage, 1987. Keenan & MacWhinney, 1987 Mcdaniel et al., 2010. Mcdaniel et al., 2010; Zeevat, 2001 NVN-schema hypothesis hypothesis. Bever, 1970. Hawkins, 1994. Locality Theory. Early Immediate Constituent Syntactic Prediction. Gibson, 1998 i.

(3) Huang (1982). X’. ii.

(4) ABSTRACT This dissertation aims at investigating (a) whether directionality is a factor influencing the acquisition of complex sentences in Mandarin, including complement clauses (CCs), relative clauses (RCs) and adverbial clauses (ACs), (b) how the constructions of the three complex sentence types affect the acquisition order of the complex sentences, (c) how the comprehension and production differ in acquiring complex sentences, and (d) how Optimality Theory (OT) can account for the acquisition of complex sentences. In pursuit of answering to these research questions, three experiments concerning complex sentences with left- and right-branching CCs, RCs and ACs were conducted. Each experiments included two tasks, that is, the comprehension and production tasks, in which subjects of the study orally responded to the utterances or elicited questions matched with pictures broadcast by the computer. There were 108 subjects equally divided into five child groups (aged 3, 4, 5, 6 and 7 years old) and one adult group (undergraduates), participating in each experiment. The results showed that for each age group in complex sentences with CCs right-branching structures are easier than left-branching ones, while in complex sentences with RCs and those with ACs the preference of directionality reversed. This finding is compatible with Huang’s (1982) view in relation to the unmarked branching direction in Mandarin, in which the preference of directionality should trace back to X0 and X’ layers according to the syntactic category of the head. In addition, it is found that although 7-year-olds were not able to master the three complex sentence types, the fact that CC complex sentences were acquired better than those of RC or AC complex sentence type before the age of 6 was quite clear, a finding that is in agreement with the priority of arguments over modifiers in child acquisition (Gasser & Smith, 1998; Mintz Gleitman, 2002; Naigles & Gelman, 1995; Smith & Sera, 1992; Waxman & Booth, 2001), the associated semantics between the arguments and their heads (Pollard & Sage, 1987) as well as the language economy. Furthermore, our results also revealed that generally speaking, the comprehension task bears less difficulty than the production task; however, when tackling the iii.

(5) unmarked or acquired constructions, Mandarin speakers tend to comprehend and produce equally well, sometimes even produce better than comprehend. This can be explained by the fact that production is dependent on comprehension in acquisition (Keenan & MacWhinney, 1987), the claim that the demand for working memory resource in production is higher than in comprehension (Mcdaniel et al., 2010) as well as the hypothesis that for a language user who has mastered the language, production may be easier than comprehension because he can control what he is going to say but not what he is going to hear (Mcdaniel et al., 2010; Zeevat, 2001). Moreover, the results obtained from the comprehension task and from the production task respectively demonstrated the validity of the NVN-schema hypothesis (Bever, 1970) and the Early Immediate Constituent hypothesis (Hawkins, 1994) in comprehension and that of the Syntactic Prediction Locality Theory (Gibson, 1998) in production. Finally, gathered the results from the three experiments, by adopting the existing OT constraints and relevant theories, it is suggested that OT might be a feasible theoretical framework to account for the acquisition of complex sentences although some deficiencies still await further modifications. To conclude, the results refuted the branching direction theory claimed by Bever (1970), Dryer (1992) and Goodluck (1991) but supported the X’-structure of Chinese proposed by Huang (1982). Our findings also proved that the attributes of RCs and ACs were similar since they were acquired in the same way. Additionally, OT accounts, which allow the possibilities of more than one structure in a language, may be a more practical theoretical framework.. Keywords: branching direction, complex sentences, language acquisition, sentence comprehension and production, Optimality Theory. iv.

(6) ACKNOWLEDGMENTS I would like to express my gratitude to all those who gave me the possibility to complete this dissertation. Without their assistance and support, this dissertation could have never been possible. My deepest gratitude goes to my dissertation advisor, Prof. Chun-yin Doris Chen, who was as perfect a dissertation director as can be. Her guidance and critical comments during the period in which the research for this dissertation was carried out as well as her personal advice, comfort, care and thoughtfulness have made the long road of writing the dissertation become joyful. Her breadth of linguistic knowledge and her endless enthusiasm in teaching have also inspired me to work hard as a researcher and to love my job as a teacher in the future. She is more than just a supervisor to me. I would also like to express my very great appreciation to my committee members, Professors Feng-fu Tsao, Yu-fang Wang, Yuh-fang Chang and Yi-ching Su, for the valuable advice. The detailed and useful suggestions given by each member have shown that all of the professors had carefully read through the draft and were more than willing to help me to refine my dissertation even though they might at the same time struggle for writing NSC projects since the oral defense was held one month before the deadline or might be busy dealing with administrative work. The invaluable advice given by my advisor and the committee members helps me improve both my understanding and my exposition of the issues discussed in this dissertation and is essential to the accomplishment of this dissertation. I would also offer my special thanks to Prof. Jen Ting and Prof. Chien-jer Charles Lin. Taking Prof. Ting’s seminars on Chinese syntax has not only consolidated my knowledge in Chinese syntax but also enriched my ability of writing papers. Additionally, in Prof. Lin’s seminar on sentence processing I profited and was inspired a lot in children’s sentence comprehension and production. Because of their instructions in the courses and extra-curriculum guidance, I may be able to investigate how Chinese children cope with complex sentences in Mandarin. The experimental study could not have been finished without a number of people, including one of my friends as well as my colleagues and students at MDU. I v.

(7) wish to acknowledge the help provided by the wonderful students, Yu-ling Wang, Yo Lin, Jessie Lin, Vani Chen, Yaya Jiang, Mei-mei Chen, Wen-yi Yang and Polly Huang, for collecting the data from 8 o’clock in the morning and never being late. Assistance provided by Yu-ling Wang and Yo Lin, who also helped transcribe the data, was greatly appreciated. Special thanks should also go to my student, Johan Chao, who is now a graduate student of Linguistics at Providence University, for being the rater of the data. My colleagues in MDU, particularly my colleagues, Sara Gong, Mark Tsai, Alan Lin and Roger Wang, should also deserve my words of gratitude for their help and encouragement. In addition, I am particularly grateful to my friend, who is my fellow graduate at NTNU and my ex-colleague at MDU, Yu-da Stephen Lai, for leading me to study for the doctoral degree, helping me recruit subjects and research assistants, giving me suggestions on my dissertation, lessening the burden of my departmental work, and sharing my happiness and sadness in both NTNU and MDU school life. Without his all-out efforts, I would have never received the diploma. Thanks also go to my students, Yo Lin, Johan Chao, Stanley Lin, Fiona Ho and Emma Chen, for their assistance with everything they have done for me whenever I was in need.. Lunnie. I dedicate this dissertation to all of you.. vi.

(8) TABLE OF CONTENTS CHAPTER ONE INTRODUCTION ……………………………………………… 1 1.1. Motivation ……………………………………………………………. 1. 1.2. Optimality Theory ……………………………………………………. 9 1.2.1 OT Framework ………………………………………………. 11 1.2.2 Language Acquisition in OT ………………………………… 13 1.2.3 Sentence Comprehension and Production in OT ……………. 15 1.2.3.1 Comprehension …………………………………….. 17 1.2.3.2 Production ………………………………………….. 19. 1.3. Linguistic Properties of Complex Sentences & Their Impacts on Comprehension and Production ……………………………………… 20. 1.4. Research Questions ………………………………………………….. 23. 1.5. Significance of the Present Study ……………………………………. 24. 1.6. Organization of the Dissertation …………………………………….. 24. CHAPTER TWO COMPLEMENT CLAUSES …………………………………. 26 2.1. Linguistic Property of Mandarin Complement Clauses……………… 27. 2.2. Acquisition of Complement Clauses ………………………………… 32. 2.3. Research Design……………………………………………………… 37 2.3.1 Participants …………………………………………………… 37 2.3.2 Materials……………………………………………………… 38 2.3.2.1 Comprehension……………………………………… 38 2.3.2.2 Production…………………………………………… 39 2.3.3 Procedure……………………………………………………… 42 2.3.3.1 Pilot Study…………………………………………… 43 2.3.3.2 Formal Study………………………………………… 45. "##! !.

(9) 2.3.4 Data Analysis ………………………………………………… 45 2.3.4.1 Scoring Procedure ………………………………….. 45 2.3.4.2 Statistical Analysis …………………………………. 48 2.4. Results and Discussion ………………………………………………. 49 2.4.1 The Branching Direction Effects …………………………….. 50 2.4.2 The CC Type Effects ………………………………………… 52 2.4.3 The Task Effects ……………………………………………... 55 2.4.4 The Age Effects ……………………………………………… 61 2.4.5 The Error Patterns …………………………………………… 64 2.4.6 The Interactions among Factors ……………………………… 66. 2.5. Conclusion …………………………………………………………… 70. CHAPTER THREE RELATIVE CLAUSES …………………………………….. 72 3.1. Linguistic Property of Mandarin Relative Clauses ………………….. 73. 3.2. Acquisition of Mandarin Relative Clauses …………………………... 79. 3.3. Research Design……………………………………………………… 85 3.3.1 Participants …………………………………………………… 86 3.3.2 Materials……………………………………………………… 86 3.3.2.1 Comprehension……………………………………… 86 3.3.2.2 Production…………………………………………… 88 3.3.3 Procedure……………………………………………………… 90 3.3.3.1 Pilot Study …………………………………………... 91 3.3.3.2 Formal Study ………………………………………... 92 3.3.4 Data Analysis ………………………………………………… 92 3.3.4.1 Scoring Procedure …………………………………... 92 3.3.4.2 Statistical Analysis ………………………………….. 95. "###! !.

(10) 3.4. Results and Discussion ………………………………………………. 95 3.4.1 The Branching Direction Effects …………………………….. 96 3.4.2 The RC Type Effects ………………………………………… 98 3.4.3 The Task Effects …………………………………………… 104 3.4.4 The Animacy Effects ……………………………………….. 114 3.4.5 The Age Effects …………………………………………….. 116 3.4.6 The Error Patterns ………………………………………….. 118 3.4.7 The Interactions among Factors ……………………………. 121. 3.5. Conclusion ………………………………………………………….. 125. CHAPTER FOUR ADVERBIAL CLAUSES ………………………………… 127 4.1. Linguistic Property of Mandarin Adverbial Clauses ………………. 128. 4.2. Acquisition of Adverbial Clauses ………………………………….. 133. 4.3. Research Design …………………………………………………… 135 4.3.1 Participants ………………………………………………… 136 4.3.2 Materials …………………………………………………… 136 4.3.2.1 Comprehension …………………………………… 136 4.3.2.2 Production ………………………………………… 137 4.3.3 Procedure …………………………………………………… 141 4.3.3.1 Pilot Study ………………………………………… 142 4.3.3.2 Formal Study ……………………………………… 143 4.3.4 Data Analysis ……………………………………………… 143 4.3.4.1 Scoring Procedure ………………………………… 143 4.3.4.2 Statistical Analysis………………………………… 145. 4.4. Results and Discussion……………………………………………… 146 4.4.1 The Branching Direction Effects …………………………… 146. "#! !.

(11) 4.4.2 The AC Type Effects ……………………………………….. 149 4.4.3 The Task Effects …………………………………………… 151 4.4.4 The Age Effects …………………………………………….. 160 4.4.5 The Error Patterns…………………………………………… 162 4.4.6 The Interactions among Factors ……………………………. 164 4.5. Conclusion ………………………………………………………….. 169. CHAPTER FIVE A COMPARISON OF COMPLEX SENTENCES ………….. 171 5.1. The Directionality Effects ………………………………………….. 172 5.1.1 The Overall Findings ……………………………………….. 172 5.1.2 Discussion ………………………………………………….. 175. 5.2. The Construction Effects …………………………………………… 177 5.2.1 The Overall Findings ……………………………………….. 177 5.2.2 Discussion ………………………………………………….. 179. 5.3. The Task Effects …………………………………………………… 181 5.3.1 Results and Discussion Concerning the Tasks……………… 181 5.3.1.1 The Overall Findings ……………………………… 181 5.3.1.2 Discussion ………………………………………… 184 5.3.2 Results and Discussion Concerning Comprehension and Production…………………………………………………… 186 5.3.2.1 The Overall Findings Concerning Comprehension... 188 5.3.2.2 Discussion Concerning Comprehension ………….. 195 5.3.2.3 The Overall Findings Concerning Production ……. 195 5.3.2.4 Discussion Concerning Production ……………….. 199. 5.4. An OT Account for Complex Sentence Acquisition ……………… 199 5.4.1 OT Constraints for the Current Study ……………………… 200. "!!.

(12) 5.4.2 An OT Account……………………………………………… 205 5.4.3 Summary …………………………………………………… 209 5.5. Conclusion …………………………………………………………. 209. CHAPTER SIX CONCLUSION ………………………………………………… 211 6.1. Summary of the Results ……………………………………………. 211. 6.2. Theoretical Implications ……………………………………………. 214. 6.3. Limitations of the Study and Suggestions for Future Research …… 217. REFERENCES …………………………………………………………………… 219 Appendix I ……………………………………………………………………… 227 Appendix II ……………………………………………………………………… 231 Appendix III ……………………………………………………………………… 232 Appendix IV ……………………………………………………………………… 236 Appendix V ……………………………………………………………………… 238 Appendix VI ……………………………………………………………………… 243 Appendix VII …………………………………………………………………….. 247 Appendix VIII …………………………………………………………………… 248 Appendix IX ……………………………………………………………………… 253 Appendix X ……………………………………………………………………… 255 Appendix XI ……………………………………………………………………… 260 Appendix XII …………………………………………………………………….. 264 Appendix XIII …………………………………………………………………… 265 Appendix XIV …………………………………………………………………… 270 Appendix XV……………………………………………………………………… 272. "#! !.

(13) LIST OF TABLES Table 2.1 The Distribution of the 12 Test Items for the CC Comprehension Task ……………………………………………………………….. 39 Table 2.2 Examples of Responses Scored as 0.5 and 0 Point for CC Production … 47 Table 2.3 The Mean Percentages & the Frequency Counts (in parenthesis) of the HalfCorrect CC Responses by CC Types ……………………………… 65 Table 2.4 The Mean Percentages & the Frequency Counts (in parenthesis) of the Incorrect CC Responses of the WH-Question CC Type by Error Category …………………………………………………………… 65 Table 2.5 The Mean Percentages & the Frequency Counts (in parenthesis) of the Incorrect CC Responses of the A-not-A-Question CC Type by Error Category …………………………………………………………… 66 Table 2.6 ANOVA Results of Interactions among Factors in the CC Study ………. 67 Table 3.1 The Distribution of the 16 Test Items for the RC Comprehension Task ……………………………………………………………….. 87 Table 3.2 Examples of Responses Scored as 0.5 and 0 Point for RC Production … 94 Table 3.3 The Mean Percentages & the Frequency Counts (in parenthesis) of the Incorrect RC Responses of the SRC Type by Error Category …… 120 Table 3.4 The Mean Percentages & the Frequency Counts (in parenthesis) of the Incorrect RC Responses of the ORC Type by Error Category …... 120 Table 3.5 ANOVA Results of Interactions among Factors in the RC Study ………121 Table 4.1 The Distribution of the 16 Test Items for the AC Comprehension Task ……………………………………………………………… 137 Table 4.2 Examples of Responses Scored as 0.5 and 0 Point for AC Production … 144 Table 4.3 The Mean Percentages & the Frequency Counts (in parenthesis) of the Incorrect AC Responses of the Temporal AC Type by Error Category …………………………………………………………. 163 Table 4.4 The Mean Percentages & the Frequency Counts (in parenthesis) of the Incorrect AC Responses of the Conditional AC Type by Error Category ………………………………………………………….. 163 "##! !.

(14) Table 4.5 The Mean Percentages & the Frequency Counts (in parenthesis) of the Incorrect AC Responses of the Concessive AC Type by Error Category …………………………………………………………. 164 Table 4.6 The Mean Percentages & the Frequency Counts (in parenthesis) of the Incorrect AC Responses of the Causal AC Type by Error Category………………………………………………………….. 164 Table 4.7 The Mean Percentages & the Frequency Counts (in parenthesis) of the Incorrect AC Responses of the AC Type by Wrong AC Type ….. 164 Table 4.8 ANOVA Results of Interactions among Factors in the AC Study …………………………………………………………….. 165 Table 5.1 The Acquisition of the Three Complex Sentence Types of Diverse Branching Directions ……………………………………………. 205. "###! !.

(15) LIST OF FIGURES Figure 2.1 The Mean Scores and Standard Deviations of CCs in each Branching Direction by Age Group ……………………………………………50 Figure 2.2 The Mean Scores and Standard Deviations of CCs of each CC Type by Age Group ………………………………………………………… 53 Figure 2.3 The Mean Scores and Standard Deviations of the Left-branching CC Comprehension and Production by Age Group …………………… 56 Figure 2.4 The Mean Scores and Standard Deviations of the Right-branching CC Comprehension and Production by Age Group …………………… 56 Figure 2.5 The Mean Scores and Standard Deviations of the Statement CC Comprehension and Production by Age Group …………………… 58 Figure 2.6 The Mean Scores and Standard Deviations of the WH-question CC Comprehension and Production by Age Group …………………… 59 Figure 2.7 The Mean Scores and Standard Deviations of the A-not-A-question CC Comprehension and Production by Age Group …………………… 59 Figure 2.8 The Mean Scores and Standard Deviations of CCs of each Task by Age Group ……………………………………………………………… 63 Figure 2.9 The Mean Scores and Standard Deviations of CC Types in Comprehension by Age Group ………………………………………………………69 Figure 2.10 The Mean Scores and Standard Deviations of CC Types in Production by Age Group ………………………………………………………… 69 Figure 3.1 The Mean Scores and Standard Deviations of RCs in each Branching Direction by Age Group ……………………………………………97 Figure 3.2 The Mean Scores and Standard Deviations of RCs of each RC Type by Age Group ………………………………………………………… 99 Figure 3.3 The Mean Scores and Standard Deviations of RCs of each RC Complex Sentence Type by Age Group …………………………………… 101 Figure 3.4 The Mean Scores and Standard Deviations of the Left-branching RC Comprehension and Production by Age Group …………………. 105. "#$! !.

(16) Figure 3.5 The Mean Scores and Standard Deviations of the Right-branching RC Comprehension and Production by Age Group ………………… 105 Figure 3.6 The Mean Scores and Standard Deviations of the SRC Comprehension and Production by Age Group ………………………………………. 107 Figure 3.7 The Mean Scores and Standard Deviations of the ORC Comprehension and Production by Age Group ………………………………………. 108 Figure 3.8 The Mean Scores and Standard Deviations of the SS Comprehension and Production by Age Group ………………………………………. 110 Figure 3.9 The Mean Scores and Standard Deviations of the SO Comprehension and Production by Age Group ………………………………………. 111 Figure 3.10 The Mean Scores and Standard Deviations of the OS Comprehension and Production by Age Group ………………………………………. 111 Figure 3.11 The Mean Scores and Standard Deviations of the OO Comprehension and Production by Age Group ………………………………………. 112 Figure 3.12 The Mean Scores and Standard Deviations of RCs of each Animacy Pair by Age Group …………………………………………………… 115 Figure 3.13 The Mean Scores and Standard Deviations of RCs of each Task by Age Group ……………………………………………………………. 117 Figure 3.14 The Mean Scores and Standard Deviations of RC Types in Comprehension by Age Group ………………………………….. 123 Figure 3.15 The Mean Scores and Standard Deviations of RC Types in Production by Age Group ………………………………………………………. 124 Figure 4.1 The Mean Scores and Standard Deviations of ACs in each Branching Direction by Age Group ………………………………………… 147 Figure 4.2 The Mean Scores and Standard Deviations of ACs of each AC Type by Age Group ………………………………………………………. 150 Figure 4.3 The Mean Scores and Standard Deviations of the Left-branching AC Comprehension and Production by Age Group …………………. 152 Figure 4.4 The Mean Scores and Standard Deviations of the Right-branching AC Comprehension and Production by Age Group …………………. 153 "#! !.

(17) Figure 4.5 The Mean Scores and Standard Deviations of the Temporal AC Comprehension and Production by Age Group …………………. 155 Figure 4.6 The Mean Scores and Standard Deviations of the Conditional AC Comprehension and Production by Age Group …………………. 156 Figure 4.7 The Mean Scores and Standard Deviations of the Concessive AC Comprehension and Production by Age Group …………………. 156 Figure 4.8 The Mean Scores and Standard Deviations of the Causal AC Comprehension and Production by Age Group …………………. 157 Figure 4.9 The Mean Scores and Standard Deviations of ACs of each Task by Age Group ……………………………………………………………. 161 Figure 4.10 The Mean Scores and Standard Deviations of Branching Directions in Temporal ACs by Age Group …………………………………… 168 Figure 4.11 The Mean Scores and Standard Deviations of Branching Directions in Conditional ACs by Age Group …………………………………. 168 Figure 4.12 The Mean Scores and Standard Deviations of Branching Directions in Concessive ACs by Age Group …………………………………. 169 Figure 4.13 The Mean Scores and Standard Deviations of Branching Directions in Causal ACs by Age Group ………………………………………. 169 Figure 5.1 The Mean Scores and Standard Deviations of Branching Directions by Age Group ……………………………………………………………. 173 Figure 5.2 The Mean Scores and Standard Deviations of CCs, RCs and ACs in each Branching Direction by Age Group ……………………………… 174 Figure 5.3 The Mean Scores and Standard Deviations of the Three Complex Sentence Types by Age Group …………………………………………….. 178 Figure 5.4 The Mean Scores and Standard Deviations of CC Comprehension and Production by Age Group ……………………………………….. 182 Figure 5.5 The Mean Scores and Standard Deviations of RC Comprehension and Production by Age Group ……………………………………….. 182 Figure 5.6 The Mean Scores and Standard Deviations of AC Comprehension and Production by Age Group ……………………………………….. 183 "#$! !.

(18) Figure 5.7 The Mean Scores and Standard Deviations of the CC, RC and AC Comprehension and Production in Diverse Branching Directions by Age Group ……………………………………………………….. 184 Figure 5.8 The Mean Scores and Standard Deviations of CCs in each Branching Direction in Comprehension by Age Group …………………….. 191 Figure 5.9 The Mean Scores and Standard Deviations of SO and OO RCs in Comprehension by Age Group ………………………………….. 192 Figure 5.10 The Mean Scores and Standard Deviations of SS and OS RCs in Comprehension by Age Group ……………………………………193 Figure 5.11 The Mean Scores and Standard Deviations of ACs in each Branching Direction in Comprehension by Age Group ……………………... 194 Figure 5.12 The Mean Scores and Standard Deviations of CCs in each Branching Direction in Production by Age Group ………………………….. 196 Figure 5.13 The Mean Scores and Standard Deviations of RCs in each Branching Direction in Production by Age Group ………………………….. 197 Figure 5.14 The Mean Scores and Standard Deviations of ACs in each Branching Direction in Production by Age Group ………………………….. 198. "#$$! !.

(19) !.

(20) CHAPTER ONE INTRODUCTION. 1.1. Motivation Research on the performance of Mandarin complex sentences has primarily. conducted to investigate how adults comprehend or produce the complex structures based on experiments, corpus or discourse in terms of the sentence types. Of the three types of complex sentences, the structure with relative clauses (RCs) has drawn much attention. All the studies on this clause type are of experiment, aiming to explore how adults perceive RCs as well as how they deal with the processing difficulties brought by the garden path when an RC modifies the object of a main clause, yet none of the studies examined the production (Chen et al., 2008; Hsiao & Gibson, 2003; Lin & Bever, 2006). Unlike the comparative abundance of literature allied to adults’ performance of Mandarin RCs, there was only one study on complement clauses (CCs), which was discourse corpus-based, hence, the production-focused (Huang, 2003). Likewise, research on adverbial clauses (ACs) has been done to tackle with the production, either via corpus (Wong, 2005) or discourse collected from different occasions (Wang, 1996, 1999, 2002, 2006). Regarding acquisition, the performance of Mandarin complex sentences has even hardly been delved, except for the complex structure with RCs. Within the studies on RCs, considerable concern has arisen over both comprehension and production (Chang, 1984; Cheng, 1995; Chiu, 1996; Hsu, 2006; Su, 2004, 2006). Although these experimental studies showed inconsistent results, they presented insightful discussions. Nevertheless, it appears that no research findings are yet. 1.

(21) available concerning the acquisition of Mandarin CCs and ACs, let alone the comparison of the development of the complex sentences. As a matter of fact, there exists literature on the issue of how children develop the three complex sentence structures; however, the literature, which was corpus-based, has been concerned about English acquisition and has emerged only in a slowly and scattered way (Diessel, 2004; Limber, 1973). As Chomsky (1959, 1965) has proposed, children adopt language acquisition device that is common to all human languages to make sense of the utterances heard around them. Therefore, it is expected that children are able to detect the characteristics shared by the three types of complex sentences and acquire them in a similar way during acquisition. Since each type of complex sentences consists of a subordinate clause and a main clause, the impacts brought by where the subordinate clause positions to its main clause may be similar. This relative position between the two clauses is defined by the branching direction. When a complex sentence comprises a subordinate clause positioning to the left of a main clause, it is called a left-branching sentence, as in (1a), (2a) and (3a). On the contrary, a complex sentence with a subordinate clause positioning to the right of a main clause, it is a rightbranching sentence, as in (1b), (2b) and (3b). The impacts of the branching direction on acquiring complex sentences could be predicted by the proposal that the consistency of the branching direction between complex structures and the nature basic syntax of the language causes less processing difficulties (Bever, 1970; Goodluck, 1991). Mandarin, according to Huang (1982) and Lust & Chien (1984), is essentially considered as a left-branching language in that its major recursive structures, e.g., RCs and ACs, located to the left of their head or matrix sentence are unmarked. An apparent example is that an RC (e.g., nage nusheng chi de) is placed 2.

(22) before its head (e.g., pingguo), as shown in (1). The unmarked branching direction suggests that despite the complex sentence types, the left-branching complex sentences should be easier to process than the right-branching complex sentences. (1) a. Left-branching RC [[RC Nage nusheng chi de] pingguo] dadao xiaogou. that girl. eat DE apple. hit. puppy. ‘The apple which that girl ate hit the puppy.’ b. Right-branching RC Nage nusheng xihuan [[RC pingguo dadao de] xiaogou]. that girl. like. apple. hit. DE puppy. ‘That girl likes the puppy which the apple hit.’ (2) a. Left-branching CC [CC Milaoshu. chuan qunzi] bu heshi.. Mickey Mouse wear skirt. not appropriate. ‘It is not appropriate for Mickey Mouse to wear a skirt.’ b. Right-branching CC Wo kanjian [CC Milaoshu I. see. chuan qunzi].. Mickey Mouse wear. skirt. ‘I saw that Mickey Mouse wore a skirt.’ (3) a. Left-branching AC [AC Ruguo Milaoshu if. kan. dianshi. dehua], ta jiu hui chi baomihua.. Mickey Mouse watch television in case he then will eat popcorn. ‘If Mickey Mouse watches television, then he will eat popcorn.’. 3.

(23) b. Right-branching AC Milaoshu. hui chi baomihua, [AC ruguo ta kan. Mickey Mouse will eat popcorn. if. dianshi. dehua].. he watch television in case. ‘Mickey Mouse will eat popcorn if he watches television.’ In the existing research on Mandarin RC acquisition, the effect of the branching direction on comprehension and production has been explored. To be specific, it has been shown that an RC in the left-branching position of a main clause is easier than that in the right-branching, that is, center-embedded, position of a main clause in comprehension (Chang, 1984; Cheng, 1995; Chiu, 1996), a finding which accords with the proposal that complex structures located in a right-branching position in a left-branching language would cause difficulties, and vice versa (Bever, 1970; Dryer, 1992; Goodluck, 1991). Nevertheless, in production, no branching direction effect has been found (Hsu, 2006). The left- and right-branching structures also exist in the other two complex sentence types, CCs and ACs; however, as has been outlined earlier, little is known about how children acquire the two complex sentence types. This motivates us to wonder (i) if the branching direction is also a factor that influences children’s acquisition of CCs and ACs. Despite the similarity that each complex sentence type bears both left- and right-branching structures, the linguistic properties of CCs, RCs and ACs are not identical. CCs, which are syntactically embedded clauses, function as arguments of the matrix clause predicates. Nevertheless, RCs, which are also syntactically embedded, function as modifiers of their heads in subject or object positions of matrix clauses. Like RCs, ACs, which are adjuncts, function as modifiers of an associated main clause or a verb phrase. These distinct attributes may give rise to diverse acquisition order of the three complex sentence types, an issue that seems not to 4.

(24) receive much attention. Accordingly, (ii) how the three complex sentence types in Mandarin are acquired in order becomes one of our interests. In addition to the branching directions and the constructions of the complex sentences, the asymmetry between the comprehension and production may provide a view of how well the complex sentences are acquired. As MacDonald (2000) has claimed, the processes of comprehension and production differ; therefore, the performance of the two different tasks may not be the same. Since the non-linguistic planning for content as well as the integration between the message and the syntactic structure in production (Bock and Levelt, 1994) cause much more difficulties for young children than the comprehension process does (Mcdaniel et al., 2010), the production may lag behind the comprehension. However, for adults or children who have acquired a construction, the production may reflect their competence; in other words, they may produce equally well as they comprehend the construction or they may produce even better than they comprehend. This invokes our interests of investigating (iii) how the comprehension and production of complex sentences differ in acquisition. In relation to the comprehension and production, how the complex sentences are acquired differently in the two tasks is another issue worth exploring. Take comprehension of the complex sentences as an example. The same word sequences of the first few words may lead to similar difficulties in acquisition in spite of the complex sentence types. As proposed by Bever (1970), in an SVO language, any Noun-Verb-Noun (NVN) sequence within a potential internal unit in the surface structure corresponding to actor-action-object (SVO) is easier to comprehend. Grounded on this claim, owing to the fact that the unmarked word order in Mandarin is SVO, the first NVN sequence which matches the SVO word order should result in 5.

(25) less processing burden in comprehension, regardless of the types of complex sentences. As shown in (4), (5) and (6), repeated from (1), (2) and (3), the first NVN sequences formed in the complex sentences with CCs and RCs1 may or may not match the SVO word order, while those formed in the complex sentences with ACs always match. The preference of SVO word order would then suggest that complex sentences like (4a) and (5a) be easier than complex sentences like (4b) and (5b). However, complex sentences like (6a) and (6b) are equally easy due to the same word sequence.2 (4) a. Left-branching RC [[RC Nage nusheng chi de] pingguo] dadao xiaogou. that girl. eat DE apple. hit. puppy. N. V. N. V. N. S. V. O. ‘The apple which that girl ate hit the puppy.’. 1. The RCs used in (1) and (4) are all object-extracted RCs. Only the object-extracted RCs in both leftbranching and right-branching positions will result in the same word sequence for the first three words, namely, NVN, but different word order, namely, SVO and SVS. There is the other type of RCs, that is, the subject-extracted RCs, yet this RC type does not give rise to the NVN sequence, and hence is not considered in the present chapter. For more details about the acquisition of the RC types, please see the discussion in Chapter Three. 2 Notice that the RC marker de and the subordinators in ACs may serve as a cue in the comprehension. As claimed by Hsiao and Gibson (2003), the RC marker de, like the other lexical items, helps a listener to predict the incomplete head-dependencies that require his storage resources. Lin and Bever (2006) also proposed that when comprehending a relative clause, a listener would not construct the filler-gap relations of the relative clause until the RC marker de and the head nouns are reached. In addition, Hawkins (1994) has argued that the emergence of the subordinator of an AC tells a listener that the sentence they are dealing with is a complex sentence. Even though both the RC marker de and the subordinators in ACs indeed play important roles in the comprehension, these words, compared to the nouns and verbs, are less prominent for children since children may pay more attention to the content words rather than the function words. Accordingly, this dissertation mainly takes the NVN-schema hypothesis into consideration and leaves the detailed discussions of the RC marker de and the subordinators in ACs for future research. 6.

(26) b. Right-branching RC Nage nusheng xihuan [[RC pingguo dadao de] xiaogou]. that girl. like. apple. hit. N. V. N. V. S. V. S. DE puppy N. ‘That girl likes the puppy which the apple hit.’ (5) a. Left-branching CC [CC Milaoshu. chuan qunzi] bu heshi.. Mickey Mouse wear skirt N. V. N. S. V. O. not appropriate V. ‘It is not appropriate for Mickey Mouse to wear a skirt.’ b. Right-branching CC Wo kanjian [CC Milaoshu I. see. N. V. N. S. V. S. chuan qunzi].. Mickey Mouse wear. skirt. V. N. ‘I saw that Mickey Mouse wore a skirt.’ (6) a. Left-branching AC [AC Ruguo Milaoshu if. kan. dianshi. dehua], ta jiu hui chi baomihua.. Mickey Mouse watch television in case he then will eat popcorn. N. V. N. S. V. O. N. V. N. ‘If Mickey Mouse watches television, then he will eat popcorn.’. 7.

(27) b. Right-branching AC Milaoshu. hui chi baomihua, [AC ruguo ta kan. Mickey Mouse will eat popcorn N. V. N. S. V. O. if. dianshi. dehua].. he watch television in case N V. N. ‘Mickey Mouse will eat popcorn if he watches television.’ Previous studies relevant to RCs have shown that there indeed exists the preference of basic word order during comprehension (Su, 2006). When a hearer encounters a NVN sequence that does not match the SVO basic word order as in (4b), he will first interpret the subject of the RC as the object of the main clause and need a reanalysis when he finds out that he has wrongly interpreted the sentence, which means more processing efforts and hence processing difficulties. This has inspired us to suspect (iv) if the comprehension and production of complex sentences will be affected by some factors, e.g., the NVN sequence, other than the branching direction. Apart from the dearth of studies on acquisition of Mandarin CCs and ACs, the development of acquiring Mandarin complex sentences have not yet been explained via a theory in which the universality, continuity and learnability have been taken into consideration. It is well known that children are equipped with an innate template for language, that is, Universal Grammar (UG). The considerable similarity in the way grammatical patterns are acquired across different children and languages suggests that there must be a set of innate principles and adjustable parameters that are universal to all human languages. Additionally, the continuity, which means that every child grammar is a possible human grammar, is a core concept in acquisition (Goodluck, 2010). Nevertheless, the child’s grammar may deviate from that of the language he is acquiring. By correcting the errors that have been made, the child will 8.

(28) ultimately acquire the language. In other words, there must be a developmental process that links the child grammar and the adult grammar. This process entails that grammars are learnable. These three concepts, namely, universality, continuity and learnability, have been so intrinsic in language acquisition that we wonder (v) if the theory, Optimality Theory (OT), which regards these concepts as essential properties, can account for the acquisition of complex sentences. In view of little research on the acquisition of Mandarin complex sentences and the need of a theory that can account for the performance of the complex sentences, the purpose of the current dissertation is to investigate how Mandarin-speaking children of different age comprehend and produce the complex sentences, including CCs, RCs and ACs, as well as to make certain if the grammar at each stage can be captured by OT, which sees the continuity, learnability and universality as essential properties.. 1.2. Optimality Theory According to Müller & Sternefeld (2000), theories of syntax differ with respect. to how they determine the wellformedness or illformedness of a sentence. The wellknown Chomsky’s (1981, 1986) Principles-and-Parameters Theory (P&P), in which grammatical principles are universal and inviolable, can check a given syntactic object without considering the properties of other syntactic objects. In other words, the parameters determine whether a principle is active or not at individual levels of syntactic representation in individual languages. To avoid the universal uniformity, P&P would merely ‘turn off’ the parameters. On the contrary, the OT for syntax, which appears to be discussed first by Smolensky (1996) and Grimshaw (1997), is treated as a competition-based approach, taking other relevant syntactic objects into 9.

(29) consideration. It allows the constraints to be violated and the grammaticality of the sentence either violating a constraint or not to be determined by the higher-ranked constraint. Unlike P&P, the effects of the violated constraints still exist in the language though they are just dominated. In OT, a constraint cannot hold in one language and simply disappear in another. To put that differently, the same OT constraints exist in all languages, yet the language-specific facts can be illustrated by the diverse rankings of the OT constraints. As a matter of fact, OT was initiated by Prince and Smolensky (1993) to account for phonological phenomena. Following their theory attending to the sound formation, some researchers set off in search of the possibility of applying OT to the sentence formation (Broadwell, 2002; Broekhuis & Dekkers, 2000; de Hoop & Lamers, 2006; Grimshaw, 1997; Hendriks et al., 2005; Lee, 2001; Legendre, 2001; Legendre et al., 1998; Morimoto, 2001; Müller & Sternefeld, 2000; Smolensky, 1996; Stevenson & Smolensky, 2005; Zeevat, 2001). Although these OT syntax studies made efforts to illustrate certain grammatical phenomena, such as the wh-phrase movements, the anaphor resolution, the relative pronoun selections, etc., they inherit the OT framework for phonology proposed by Prince and Smolensky (1993), aiming at explaining language phenomena with diverse rankings of the universal constraints. The universality claimed in OT seems to be consistent with the notion of a core grammar from language typology. Apart from the language typology, language acquisition also advocates the conception that there should be a core grammar (Kager, 1999). To be specific, language acquisition contributes the hypothesis of the innateness of UG in that children acquire their first language in noticeably similar ways in the course of the language development. In addition, the continuity and learnability are also shared by language acquisition (Goodluck, 2010; Lightfoot, 1999) 10.

(30) and OT (Kager, 1999; Smolensky, 1996). Due to the universality, continuity and learnability involved in OT and language acquisition as essential elements, OT is considered feasible for explaining the child language acquisition (Fikkert & Hoop, 2009; Hendriks et al., 2005; Hendriks & Rij, 2011; Smolensky, 1996). In the following sections, we first introduce the basic concepts of OT framework proposed by Prince and Smolensky (1993) and how the OT framework functions for syntax. After that, we illustrate how the OT concepts are applied in acquisition. Finally, we review the existing studies on the comprehension and production in OT framework in order to sketch how OT works in the sentence comprehension and production. 1.2.1. OT Framework Based on the notion that UG consists largely of a set of constraints on well-. formedness, Prince and Smolensky (1993) have claimed that individual grammars are constructed out of the conflicting rather than consistent constraints. These universal but conflicting constraints compete with each other, and hence, every possible output of a grammar will necessarily violate at least one constraint. The violation, however, does not directly result in ungrammaticality. Rather, what determines the optimal output of a grammar is the least costly violation of the constraints. In the framework, all the output candidates, the surface form, are generated from the input, the underlying form, by the Generator, and are evaluated by the Evaluator, that is, the set of ranked constraints. Output candidates that violate the higher-ranking constraints are fatal, while the output candidate that violates the least and the lower-ranking constraints is selected. There is no cumulativity in OT framework. The amount of violation of lower-ranking constraints will never override the violation of a single higher-ranking constraint. Along with different rankings of the constraints, the most 11.

(31) harmonic output candidate will differ. Therefore, although all languages have the same set of universal constraints, with the language-specific rankings, the selected output forms will vary. The constraints can be divided into two types: markedness and faithfulness constraints. Markedness constraints evaluate output candidates and aim at avoiding marked structures, while faithfulness constraints evaluate the input-output mapping and aim at minimizing the difference between input and output form. The least marked outputs necessitate the most complex derivations to generate, whereas the most faithful outputs to inputs require the least complex derivations. Accordingly, markedness and faithfulness are usually in conflict (Kager, 1999). Since the OT framework was then adopted for OT syntax (Legendre, 2001; Müller & Sternefeld, 2000), we extracted an English example from Legendre (2001) to illuminate how the OT framework addressed above works in OT syntax. In English there are expletive elements like it, while in Italian no counterpart can be found, as in (7). Two constraints, capturing the core ideas of the EPP and the Principle of Full Interpretation (Chomsky, 1991), proposed by Grimshaw (1997) are given in (8). The constraint SUBJECT will be violated by any clauses without a subject in the canonical position, whereas the constraint FULLINT will be violated by any lexical items without an interpretation. As shown in (9), the English example, according to the underlying form, the Generator will generate some output candidates, shown vertically in random order in the tableau. The constraints are presented from the higher-ranked to the lower-ranked horizontally from left to right. When a candidate violates a constraint, the cell is marked by the violation mark ‘*.’ The optimal output is indicated by the index ‘!,’ that is, (9a), which does not violate the higher-ranked constraint SUBJECT. On the contrary, its competitor (9b) violates the higher-ranked constraint; therefore, it 12.

(32) is fatal, indicated by the accompanying exclamation mark ‘!.’ Note that although (9a) violates constraint FULLINT, it still wins out because the violation content is no longer relevant, presented with the shading cells. The Italian example, as in (10), can be explained by the same OT account though the result is reversed. (7) a. English: It rained. b. Italian: Piove.. (Legendre, 2001: 4 (3)). (8) a. SUBJECT: The highest A-specifier in an extended projection must be filled. b. FULLINT: Lexical items must contribute to the interpretation of a structure. (Legendre, 2001: 5 (4)) (9) English (Input: rain V [present]) SUBJ ! a. It rained b. Rained *!. FULLINT * (Legendre, 2001: 6 Tableau T1.2). (10) Italian (Input: piovere V [present]) FULLINT a. EXPL piove *! ! b. Piove. SUBJ * (Legendre, 2001: 6 Tableau T1.1). As mentioned earlier, the OT framework has been adopted to account for how children acquire grammars. In the following section, we then present the language acquisition in OT. 1.2.2. Language Acquisition in OT UG, by hypothesis, is innate rather than acquired, and hence, the study of first. language acquisition may offer a window onto properties of UG. When first language acquisition is under discussion, two central concepts should be borne in mind: continuity and learnability. As pointed out by Goodluck (2010), continuity means that every child grammar is a possible human grammar though it is not necessarily the 13.

(33) grammar of the ambient language. In other words, there must be a developmental process which links the child grammar and the adult grammar. The process entails the learnability of grammars. OT, regarding continuity as an essential property (Kagger, 1999; Smolensky, 1996), assumes that UG defines a set of universal and violable constraints and the ranking of the universal constraints is learnable. To put it in another way, a language learner needs to extract the information from the outputs of the target language and figure out the ranking of the constraints which predicts the outputs. The learnability of constraint rankings and the continuity as the essential properties of OT make the theory more on a par with the theory of language acquisition. Smolensky (1996) even claimed that there is only one child grammar, and it is an OT grammar. In respect of the innate universal constraints, there are two different viewpoints. One is that all the universal constraints are born with a baby though they are not in any ranking (Kager, 1999). The other is that constraints emerge in the course of development, and faithfulness constraints are absent rather than dominated (Ellison, 2000). No matter which viewpoint depicts the facts, markedness constraints must outrank faithfulness constraints because an initial grammar with lower-ranking faithfulness provides child language production its highly unfaithful character, leading to less grammatical complexity (Fikkert & de Hoop, 2009; Smolensky, 1996). In order to learn a target adult language, children, assumed to have a mentallyobjective constraint hierarchy replete with universal constraints, are required to determine the relative rankings of the constraints. In the course of development, there will be a partial ordering of constraints or particular child-specific markedness constraints, which explain not only the existence of particular child forms, but also the frequency with which these forms appear (Legendre et al., 2002). When children are 14.

(34) faced with the fact that an output form cannot be predicted by the rankings of constraints in their grammar, and there is ‘positive evidence’ in the form of a constraint violation in the optimal output, the constraint reranking occurs. Reranking always involves the ‘demotion’ of a constraint below another constraint so that the output form will be more harmonic. To demote a constraint means to descend the ranking of the constraint from its original ranking level. For example, when the constraint C4 is demoted from a set of constraints {C1, C2, C3… Cn}, the ranking will be {C1… C3, C5… Cn} » {C4}. Alternatively, in the case of the emergence of new constraints during development, later-developed constraints, which are perhaps similar to the faithfulness constraints, are inserted.3 In pursuit of adopting OT to explain children’s comprehension and production of complex sentences, we now turn to the next section concerned with how sentence performance is handled in OT. 1.2.3. Sentence Comprehension and Production in OT The two types of sentence performance, namely, sentence comprehension and. sentence production, contain different input and output forms in OT, hence, should be accounted for from different angels (Fikkert & de Hoop, 2009; Mattausch, 2004; Smolensky, 1996; Wilson, 2001; Zeevat, 2001). In sentence comprehension, the fixed form is the surface form, that is, an utterance, and the competitions are meanings. This is the perspective of a hearer, and hence, the sentence comprehension should be illustrated by OT semantics. By contrast, in sentence production, the fixed form is the. 3. There exist different perspectives towards how constraint reranking is accomplished. Some researchers believed that learning can result in an instantaneous change in the ranking of constraints, and hence, lower-ranking constraints will override high-ranking ones. Some researchers, however, claimed that learning can be more gradual. In this case, the relevant markedness and faithfulness constraints move closer towards each other, but they do not necessarily give rise to a reranking of the constraints. Other researchers refuted the existence of reranking. They proposed that constraints are floating at a particular stage, leading to variation. 15.

(35) input, which is a meaning and is composed of the basic components of the intended utterance, such as lexical items, predicate-argument structure, and tense and aspect information. The competitions are the underlying forms, which are syntactic forms. This is the perspective of a speaker, and hence, the sentence production ought to be explained by OT syntax. In both OT semantics and OT syntax, since the input and the output are either the meaning and the form or the form and the meaning, faithfulness constraints evaluate the relation of association between the input and output rather than of identity of the two (Mattausch, 2004). In addition, since markedness constraints are output oriented, and the output differs in comprehension and production, constraints may bring different influences in OT semantics and OT syntax (Hendriks & Rij, 2009). This led linguists to wonder if OT semantics and OT syntax share the same constraints. Some researchers proposed that the grammar only differ in the faithfulness constraints (Pater, 2004); some contended that there is an asymmetric grammar for comprehension and production (Boersma, 1998; Hendriks et al., 2005). Nevertheless, these claims are less convincing because different sets of constraints for different performance do not seem to take economy into account. As a matter of fact, it would be more feasible if both comprehension and production contain the same set of constraints and the same order of these constraints. What differs between comprehension and production is that some constraints will only be activated in one direction of optimization and not in the other (Smolensky, 1996; Fikkert & de Hoop, 2009). OT semantics and OT syntax cannot be independent of each other since both theories discuss the relation between the form and the meaning (Zeevat, 2001). Wilson (2001) suggested that there is a model of the syntax-semantics interface that 16.

(36) merges comprehension and production. The idea of conflating the interpretive and expressive optimization was raised by Blutner (2000), who has pioneered a first version of bidirectional OT. A sentence is optimal only when both its structure is the optimal one for a particular interpretation and its interpretation is the optimal one for its structure (Ackema & de Hoop, 2006). The bidirectional OT forces the hearer or the speaker to simultaneously take both his and the interlocutor’s perspectives into consideration. This is essential in adult communication, whereas children under 6-7 years old have not acquired yet (Chapman & Miller, 1975). As noted by Fikkert & de Hoop (2009), children may interpret both unmarked and marked forms as the unmarked regardless of the reason why a speaker who could have used unmarked forms generate the marked forms. Therefore, not until children learn to optimize from meaning to form and from form to meaning simultaneously have they acquired an intact adult language. 1.2.3.1 Comprehension Unlike OT syntax, which elucidates a speaker’s two opposing principles of generating a sentence, that is, to mark every semantically relevant property of the input by some syntactic feature for a hearer’s convenience and to use minimal effort for the speaker’s economy, it seems that the same conflict does not exist in OT semantics because if an interpreter wants to minimize his efforts, he may probably run into the situation of not finding the speaker’s intention. In other words, the hearer attempts to make the most of the signal he receives rather than considering economy. Accordingly, if there are conflicting constraints in OT semantics, they must derive from constraints OT syntax (Blutner, 2000; Zeevat, 2001). The central concept of OT semantics is that semantic interpretation is on the basis of OT syntax. Blutner (2000) contended that the proper treatment of OT in the 17.

(37) interpretation has to take both OT semantics and OT syntax simultaneously. He used anaphor resolution, as shown in (11), as an evidence to illustrate the existence of syntactic representation in OT semantics. When interpreting (11b), a hearer knows the coreferential reading is impossible because there is a blocking effect caused by the form (11a), which obeys a weak constraint saying ‘bound NPs are marked reflexive.’ Knowing the failure of coreferential reading of (11b), the hearer must have the syntactic representation of the more cheaply generated sentence (11a) in mind. The interpretation from a semantic presentation to the syntactic representation and from the syntactic representation to the semantic representation is known as bidirection of optimization. (11) a. Johni washes himselfi. b. *Johni washes himi. (Blutner, 2000: 194 (31, b)) Since OT semantics is established on the basis of OT syntax, most constraints for the interpretation are from the production. There is just a small body of universal constraints in OT semantics that assist interpreters in deciding between different readings allowed by OT syntax (Zeevat, 2001). Compared to the constraints in OT syntax, which are assumed to be hard, immune to contextual variation, the constraints in OT semantics are rather soft constraints, which are subject to context effects (Keller, 2000). Therefore, except for the constraints derived from OT syntax, OT semantics involve constraints relevant to pragmatics. Only does the syntactic, semantic and pragmatic information integrate in a system of ranked constraints will the set of constraints correctly derive the optimal interpretations (Chapman & Miller, 1975; Hendriks & de Hoop, 2001). According to Zeevat (2001), the optimal interpretation at the sentence level is selected by two steps. Firstly, the OT syntax system determines the set {Content : 18.