影響英語母語者中文關係子句習得之研究

全文

(1)國⽴臺灣師範⼤學英語學系碩⼠論⽂ Master’s Thesis Graduate Institute of English National Taiwan Normal University. 影響英語母語者中⽂關係⼦句習得之研究 Factors Affecting English-Speaking Learners’ Acquisition of Chinese Relative Clauses. 指導教授: 陳純⾳博⼠ Advisor: Dr. Chun-yin Doris Chen 研究⽣: 杜慧銘 Student: Hui-ming Lois Du. 中華民國⼀百零⼋年⼀⽉. January, 2019.

(2) 摘要本論⽂旨在探討以英語為母語者對於中⽂關係⼦句的學習情況。本研究包括了三個實驗。實驗⼀主要研究⼆語學習者對於主語關係⼦句及賓語關係⼦句的產出與接受度，主要研究議題包括空位影響，母語及題型效應。實驗⼆旨在研究受試者對於空位關係⼦句及⾮空位關係⼦句的表現，研究議題包括指⽰量詞位序的影響，母語轉移及題型效應。實驗三深⼊探究了受試者對於指⽰量詞前置與指⽰量詞後置的賓語關係⼦句的理解，同時也探討了受試者的習得表現是否受到語境效應的影響。實驗⼀與實驗⼆都採⽤了句⼦⽣成測驗及可接受度測驗。實驗三採⽤了理解測驗，該測驗分為三部分，每部分包含不同的語境類型。本研究共招募了⼆⼗位在國⽴台灣師範⼤學國語中⼼學習中⽂的英語為母語的學⽣及⼆⼗位母語⼈⼠。⾸先，實驗⼀發現對於⼆語學習者來說，賓語關係⼦句⽐主語關係⼦句更容易產出及接受，且他們的習得表現受到空位的影響。再者，根據實驗⼆的結果可⾒，空位關係⼦句相較於⾮空位關係⼦句更容易產出及接受，指⽰量詞位序影響受試者的表現。實驗⼀及實驗⼆皆發現⼆語學習者在產出時受到母語影響，但題型效應並不顯著。實驗三發現，受試者對指⽰量詞前置的賓語關係⼦句與指⽰量詞後置的賓語關係⼦句的理解難度⼀致。指⽰量詞前置的賓語關係⼦句可能存在⾮限定與限定的理解。最後，研究發現受試者對不同位序的指⽰量詞之關係⼦句的理解受到語境效應影響。關鍵詞：中⽂關係⼦句、空位影響、母語影響、題型效應、語境效應、指⽰量詞位置、⼆語習得. i.

(3) ABSTRACT The current empirical research aimed at investigating English-speaking learners’ acquisition of Chinese relative clauses (RCs). Three studies were conducted. Study I examined the L2 learners’ production and preference of SRCs and ORCs, and addressed issues, such as gap condition, L1 transfer and task effect. Study II explored the subjects’ performance on gapped and gapless ORCs, and explored issues including DCL positions, L1 transfer and task effect. Study III further investigated the subjects’ interpretation of DCL-first and DCL-second ORCs and discussed contextual effect. Both Study I and Study II conducted two tasks: a sentence-making (SM) task and an acceptability judgement (AJ) task. Study III conducted an interpretation (IT) task which included three parts in different contexts (i.e., context-free, biasing and supporting). A total of twenty L2 learners and twenty Mandarin native speakers participated in our research. The overall results obtained from Study I revealed that ORCs were easier than SRCs for the L2 learners to produce and accept, and their performance was affected by the gap condition of RCs. In addition, it was found in Study II that gapped ORCs were easier than gapless ORCs, and the performance was influenced by DCL positions. Moreover, the L2 learners’ L1 knowledge was found to affect their production in Studies I and II, but the effect of tasks was not significant. In Study III, DCL-first ORCs were found as difficult as DCL-second ORCs for the L2 learners to interpret. The nonrestrictive and restrictive interpretations of DCL-first ORCs were found possible. Finally, the contextual effect indeed existed in the L2 learners’ interpretation of ORCs with DCL phrases in different positions. Keywords: Chinese Relative Clauses, gap influence, L1 influence, task effect, contextual effect, DCL positions, second language acquisition. ii.

(4) 謝辭歷經⼀年來撰寫論⽂的磨礪，終於到了提筆致謝的時刻，也意味著即將為兩年半的碩⼠⽣涯畫上完美休⽌符。光陰荏苒，回顧這條夢幻般的台灣求學之路，⼼中滿是不捨與感恩。借此謝辭，我想對我的師⾧、同學、好友及家⼈表⽰最誠摯的謝意。⾸先，我要感謝我的指導教授陳純⾳⽼師。在研究所的這段時間⾥，若⾮純⾳⽼師的⿎勵與⽀持，我絕對無法如此順利地⾛過我的求學之路。從論⽂的選題、構思、實驗、撰寫到最終定稿，純⾳⽼師都傾注了極⼤的⼼⾎。這⼀年來，⽼師總是不厭其煩地指點我，⼜在百忙之中抽空認真批改我的論⽂，字字句句把關。⽼師豐富的研究經驗、獨到且精闢的⾒解總能幫助我從研究的迷茫中豁然開朗。在純⾳⽼師的⾝上，我不僅學到了如何嚴謹地做研究、如何⽤清晰的邏輯思考及解決問題，還學到了如何謙遜地待⼈處世，如何積極樂觀地看待困境。此外，也要由衷感謝百忙之中抽空來當我的⼝試委員的范瑞玲教授及謝妙玲教授，謝謝她們從⼤綱⼝試到畢業⼝試所給予的幫助及寶貴建議，這都讓我能夠更順利地完成這篇論⽂。再者，我還要感謝研究所⾥曾經教導過我的所有師⾧，謝謝李臻儀⽼師、謝妙玲⽼師、吳曉虹⽼師、林蕙珊⽼師、張妙霞⽼師、颯楊⽼師以及甯俐馨⽼師，感謝⽼師們嚴格耐⼼的指導，讓我能扎扎實實地學好語⾔學知識，進⽽有⾜夠的能⼒完成這本論⽂。同時，我要謝謝所有在這兩年多時光⾥幫助過、關⼼過我的學⾧、學姐以及⼀直相伴我左右的同學們。特別要感謝李翰綸學⾧以及葉惠真學姐⼀直以來的貼⼼提點及⿎勵。還要感謝 Joyce(冠嫻)同學這⼀路與我並肩作戰，感謝她幫我⼀起做實驗，⼀起度過了最艱難的時光。並且我還要謝謝暖⼼可愛的 Eddie 蕭為允同學，願意傾聽我在寫論⽂期間遇到的困難，總是⽤他的幽默來化解我的不安。最後，我還要感謝我的⽗母及家⼈。在外求學⼗餘載，沒能陪伴在他們⾝邊⼀直⼼中有愧，但是他們仍然⼀如既往地信任、理解、⽀持我做每⼀件我想做的事。深知他們是我最堅實的後盾，才讓我能夠更勇敢地⾛⾃⼰的路，才能有今天的我。僅以此論⽂獻給所有幫助、關⼼以及愛我的⼈，以表我最誠摯的謝意及崇⾼的敬意。同時，也將此論⽂作為我的台灣求學之旅最重要、最珍貴的禮物獻給⾃⼰。感謝師⼤，感謝台灣。 iii.

(5) TABLE OF CONTENTS CHINESE ABSTRACT ................................................................................................i ENGLISH ABSTRACT .............................................................................................. ii CHINESE ACKNOWLEDGEMENTS.................................................................... iii TABLE OF CONTENTS ............................................................................................iv LIST OF TABLES AND FIGURES ........................................................................ vii LIST OF ABBREVIATIONS .....................................................................................ix. CHAPTER ONE INTRODUCTION ......................................................................... 1 1.1 Motivation ............................................................................................................ 1 1.2 Theoretical Background ....................................................................................... 4 1.2.1 L1 Influence .................................................................................................. 4 1.2.2 Task Effect ..................................................................................................... 5 1.2.3 Contextual Effect........................................................................................... 6 1.3 Research Questions .............................................................................................. 7 1.4 Significance of the Present Study ........................................................................ 8 1.5 Organization of the Thesis ................................................................................... 9 CHAPTER TWO LITERATURE REVIEW .......................................................... 10 2.1 Chinese RC Types .............................................................................................. 10 2.1.1 SRCs vs. ORCs ............................................................................................ 11 2.1.2 Gapped RCs vs. Gapless RCs ..................................................................... 13 2.1.3 DCL-First RCs vs. DCL-Second RCs ......................................................... 15 2.2 A Typological Comparison of Chinese and English RCs .................................. 19 2.3 Previous Empirical Studies of DCL Position in Chinese RCs ........................... 23 2.3.1 Xu (2009) .................................................................................................... 23 2.3.2 Li (2013)...................................................................................................... 27 2.3.3 Wu & Sheng (2014) ..................................................................................... 29 2.3.4 Xu (2014) .................................................................................................... 32 2.3.5 Summary of Empirical Studies .................................................................... 34 2.4 Summary of Chapter Two .................................................................................. 37 iv.

(6) CHAPTER THREE RESEARCH DESIGN ........................................................... 38 3.1 Study I ................................................................................................................ 38 3.1.1 Subjects ....................................................................................................... 38 3.1.2 Methods and Materials ............................................................................... 39 3.1.2.1 A Sentence-Making (SM) Task ............................................................ 40 3.1.2.2 An Acceptability Judgment (AJ) Task ................................................. 43 3.1.3 Procedure .................................................................................................... 44 3.2 Study II............................................................................................................... 45 3.2.1 Subjects ....................................................................................................... 46 3.2.2 Methods and Materials ............................................................................... 46 3.2.2.1 The Sentence-Making (SM) Task ........................................................ 46 3.2.2.2 The Acceptability Judgement (AJ) Task .............................................. 47 3.2.3 Procedure .................................................................................................... 48 3.3 Study III ............................................................................................................. 48 3.3.1 Subjects ....................................................................................................... 49 3.3.2 Methods and Materials ............................................................................... 49 3.3.3 Procedure .................................................................................................... 53 3.4 Summary of Chapter Three ................................................................................ 54 CHAPTER FOUR STUDY I ..................................................................................... 55 4.1 SRCs vs. ORCs .................................................................................................. 55 4.2 Gap Influence ..................................................................................................... 59 4.3 L1 Influence ....................................................................................................... 63 4.4 Task Effect ......................................................................................................... 65 4.4 Summary of Chapter Four ................................................................................. 67 CHAPTER FIVE STUDY II ..................................................................................... 68 5.1 Gapped ORCs vs. Gapless ORCs ...................................................................... 68 5.2 DCL Position influence ...................................................................................... 73 5.3 L1 Influence ....................................................................................................... 78 5.4 Task Effect ......................................................................................................... 81 5.5 Summary of Chapter Five .................................................................................. 81 CHAPTER SIX STUDY III ...................................................................................... 83 6.1 ORCs Interpretation ........................................................................................... 83 v.

(7) 6.2 DCL Phrase Ambiguity Resolution.................................................................... 87 6.3 Contextual Effect ............................................................................................... 92 6.4 Summary of Chapter Six.................................................................................... 94 CHAPTER SEVEN CONCLUSION ....................................................................... 95 7.1 Summary of Major Findings .............................................................................. 95 7.2 Pedagogical Implications ................................................................................... 97 7.3 Limitations of the Current Study and Suggestions for Future Research ........... 98 REFERENCES......................................................................................................... 100 APPENDIX A TEST ITEMS OF THE SENTENCE-MAKING TASK IN STUDY I .................................................................................... 105 APPENDIX B TEST ITEMS OF THE ACCEPTABILITY JUDGMENT TASK IN STUDY I .................................................................................... 106 APPENDIX C TEST ITEMS OF THE SENTENCE-MAKING TASK IN STUDY II................................................................................... 108 APPENDIX D TEST ITEMS OF THE ACCEPTABILITY JUDGMENT TASK IN STUDY II................................................................................... 109 APPENDIX E TEST ITEMS OF THE INTERPRETATION TASK IN STUDY III ................................................................................. 111 APPENDIX F THE CONSENT FORM ................................................................. 119 APPENDIX G THE ANSWER SHEET OF THE STUDY .................................. 120. vi.

(8) LIST OF TABLES AND FIGURES Table 2-1 A Typological Comparison of Chinese and English RCs ................ 22 Table 2-2 A Summary of Four Empirical Studies ............................................ 35 Table 3-1 A Summary of the Subjects.............................................................. 39 Table 3-2 A summary of the Test Items for the SM Task of Study I................ 42 Table 3-3 An Example of the SM Task of Study I ........................................... 43 Table 3-4 A Summary of the Test Items for the AJ Task of Study I................. 43 Table 3-5 An Example of the AJ Task of Study I............................................. 44 Table 3-6 A summary of the Test Items for the SM Task of Study II .............. 46 Table 3-7 An Example of the SM Task of Study II .......................................... 47 Table 3-8 A Summary of the Test Items for the AJ Task of Study II ............... 48 Table 3-9 A Summary of the Test Items for the IT Task of Study III .............. 49 Table 3-10 An Example of a Test Item of Part I in Study III ........................... 50 Table 3-11 An Example of a Test Item of Part II in Study III .......................... 51 Table 3-12 An Example of a Test Item of Part III in Study III ........................ 52 Table 4-1 Subjects’ Mean Scores on SRCs and ORCs in the SM Task ........... 56 Table 4-2 P-values for the within-Type Differences between the Subjects’ Performance on the SM Task ........................................................... 56 Table 4-3 Subjects’ Mean Scores on SRCs and ORCs in the AJ Task ............. 57 Table 4-4 P-values for the within-Type Differences between the Subjects’ Performance on the AJ Task............................................................. 57 Table 4-5 Subjects’ Overall Performance on SRCs and ORCs with Gapped and Gapless Structures in the SM Task .................................................. 60 Table 4-6 Subjects’ Mean Scores on SRCs and ORCs with Gapped and Gapless Structures in the AJ Task .................................................... 61 Table 4-7 Percentages of Different Error Types Found in the SM Task .......... 63 Table 5-1 Subjects' Mean Scores on Gapped and Gapless ORCs in the SM Task ................................................................................. 68 Table 5-2 P-values for the within-Type Differences between the L2 and NS Groups in the SM Task..................................................................... 69 Table 5-3 Subjects’ Mean Scores on Gapped and Gapless ORCs in the AJ Task ................................................................................... 69 vii.

(9) Table 5-4 P-values for the within-Type Differences between the L2 and NS Groups in the AJ Task ...................................................................... 70 Table 5-5 Subjects’ Performance on Gapped and Gapless RCs with DCL Structures in the SM Task ................................................................ 73 Table 5-6 Subjects’ Mean Scores on Gapped and Gapless RCs with DCL Structures in the SM Task ................................................................ 76 Table 5-7 The Percentages of Different Error Types in the SM Task .............. 78 Table 6-1 Subjects’ Correct Interpretation of the DCL-first and DCL-second ORCs .......................................................................... 83 Table 6-2 P-values for the Within-Type Differences between the L2 and NS Groups........................................................................................ 84 Table 6-3 Subjects’ Correct Interpretation of DCL-first and DCL-second RCs in Biasing and Supporting Contexts ................................................ 85 Table 6-4 P-values for the Within-Type Differences between the Three Context Types .................................................................................. 86 Table 6-5 Subjects’ Overall Association Preference for the Noun in the DCL-first ORCs ............................................................................... 87 Table 6-6 P-values for Differences among the Three Association Options ..... 88 Table 6-7 Subjects’ Association Preference in Different Contexts .................. 89 Table 6-8 P-values for Differences among the Three Association Options in Different Contexts........................................................................ 90 Figure 3-1 The Summary of the Procedures of Study I ................................... 45 Figure 3-2 The Summary of the Overall Procedures of the Current Study ..... 54. viii.

(10) LIST OF ABBREVIATIONS L1. first language. NL. native language. L2. second language. RC. relative clause. DCL. demonstrative classifier. SRC. subject relative clause. ORC. object relative clause. SDT. Structural Distance Theory. CWO. Canonical Word Order Hypothesis. LDT. Linear Distance Theory. FGD. Filler Gap Domain Theory. Cl. classifier. ClP. classifier phrase. ix.

(11) CHAPTER ONE INTRODUCTION 1.1 Motivation Attributing to innate language abilities, human beings can successfully acquire their first languages (L1s) or native languages (NLs) (Chomsky 1965). However, the path through second language (L2) acquisition is not as smooth or even as that in L1 acquisition (Lightbown & Spada 2013). As pointed out by Selinker (1972), one of the factors affecting L2 leaners in mastering their target language results from an interlanguage system, in which L2 learners’ errors are likely to result from the influence of their L1. These interlingual errors can be further classified into types: phonological, morphological, semantic, syntactic, and so on (Keshavarz 1994). Syntactic errors, in particular, can be predicted by comparing L1 and L2 in typology (Selinker 1972). To demonstrate this point, many L2 researchers focused on Chinese relative clauses (RCs) in acquisition (e.g., Wu & Sheng 2014, Zhao 2015, Du 2016) and pointed out that L1 influence could explain why L2 learners made certain syntactic errors. For example, Du (2016) examined the L2 acquisition of Chinese RCs by English-speaking learners and found that one of their common errors was wrong word order, as shown in (1a). 1.

(12) (1) a *Bi pen. [Wangwu. gang. mai. de]. huaidiao. le.. Wangwu. just. buy. DE. break. Asp. gang. mai. de]. bi. huaidiao. just. buy. DE. pen. break. b. [Wangwu Wangwu. le. Asp. ‘The pen that Wangwu just bought was broken.’ (Du 2016:12) Different from the correct sentence as in (1b), (1a) shows that the head noun bi ‘pen’ was incorrectly placed before the RC by the L2 learners. This wrong order can be attributed to the head-initial word order of English (Wu & Sheng 2014). In the process of learning RCs, L2 learners have common difficulty in production and interpretation. In production, one of the difficulties is word order as mentioned above, which has been evidenced in many studies (see Xu 2009, Wu & Sheng 2014, Zhao 2015). Besides, another common difficulty is in interpretation. The research focus on interpretation of RCs mainly falls into L1 acquisition (see Hu et al. 2016, Du 2017). In particular, the difficulty can be more obviously observed when a demonstrative classifier (DCL) phrase occurs with a head noun since this phrase may result in ambiguity of an RC, as (2) shows, where the DCL can be used to modify the first noun huoche ‘truck’ or the head noun jiaotache ‘bike’.. 2.

(13) (2) a. [Na liang that CL. huoche zhuangdao truck. knock. de]. jiaotahce hai. DE bike. tingzai lubian.. still park. roadside. ‘The bike that the truck knocked still parks by the roadside.’ b. Na liang that CL. [huoche zhuangdao truck. knock. de] DE. jiaotache hai tingzai. lubian.. bike. roadside. still park. In Du (2017), she investigated Chinese native children’s production and interpretation of RCs. When her child subjects were asked which vehicle was modified by the DCL phrase naliang, most of them tended to associate naliang with the adjacent noun huoche ‘truck’ rather than the long-distance noun, also the head noun jiaotache ‘bike’. This shows that children tend to associate a DCL phrase with an adjacent noun immediately but they may have difficulty in associating a DCL phrase with a head noun since the distance between the DCL phrase and the head noun is longer. This interpretation difficulty in L1 acquisition can be further examined in L2 acquisition. As mentioned above, we have found it is challenging for L1 and L2 learners to produce and interpret Chinese RCs. These difficulties in acquiring RCs are mostly found in the studies of RCs with gaps (i.e., gapped RCs) (Xu 2009, Wu & Sheng 2014, Du 2016, 2017). However, Chinese also has gapless RCs which do not involve gaps (see Huang, James, Audrey Li and Yafei Li (henceforth HLL) 2009, Xu 2009), as illustrated in (3). 3.

(14) (3) Zhe jiu this. exactly. shi [[ta. changge. de]. shengyin].. is. sing.song. DE. voice. he. ‘This is his singing voice.’ (HLL 2009:234) In (3), ta ‘he’ is the subject of the RC and the verb changge ‘sing a song’ is intransitive (Yip & Don 2004), which means there is no gap required in the RC. In this case, the head noun shengying ‘voice’ is related to the entire RC instead of a subpart of it. To examine whether L2 learners will encounter any difficulty in the production and interpretation of gapped RCs and gapless RCs, we were motivated to conduct an empirical study on L2 acquisition of Chinese RCs by English-speaking learners. 1.2 Theoretical Background 1.2.1 L1 Influence Chomsky (2000) states that errors are inevitable in language learning. Corder (1976) points out that L2 learners’ errors, which are termed interlingual errors and intralingual errors by Richards (1971), are significant for researchers to investigate L2 learners’ learning process. Interlingual errors result from negative influence of L2 learners’ native language while intralingual errors are errors attributed to partial exposure to target language (Richards 1971). These errors are signs that L2 learners test the hypothesis they form in the process of mastering the grammar or rules of target. 4.

(15) language. To master target languages, L2 learners tend to develop an interlanguage system and L1 influence is regarded as one of five significant processes influencing this system: language transfer, strategies of L2 communication, transfer of training, strategies of L2 acquisition and overgeneralization of target languages (Selinker 1972). Many studies have been investigated the effect of L1 on different perspectives of L2 learning (Kubota 1998, Felser et al. 2003, Papadopoulou & Clahsen 2003, Chan 2004, Wen et al. 2015, Yan & Mattews 2017). For instance, Yan & Matthews (2017) found that the most frequent error type made by bilingual children is postnominal RCs, which is considered as proof for transfer from the participants’ L1 (i.e., English). However, L1 effect on the acquisition of RCs was not found in Wen et al. (2015), which studied on the acquisition of English RCs by Chinese-speaking L2 learners. Based on the opposite findings obtained in these studies, the current study investigated whether L1 knowledge takes effect on L2 learners’ interpretation and production of gapped and gapless Chinese RCs. 1.2.2 Task Effect Based on previous studies of language acquisition (e.g. Chapman & Miller 1975, Clark 1993, Hendriks & Koster 2010), L2 learners have been found to perform 5.

(16) differently on an interpretation task and on a production task. Hendriks & Koster (2010) found that their L2 learners’ production of a particular linguistic form was different from their interpretation of this form. Though the participants performed well on the interpretation task, they did not do well on the production task. According to some researchers, L2 learners’ comprehension is inferior to production (e.g. Chapman and Miller 1975, Chapman 1978); however, some have a different finding (e.g. Rice 1980, Unsworth 2007). Thus, in the current study, we investigated whether L2 learners’ interpretation and production of Chinese RCs show a similar tendency. 1.2.3 Contextual Effect Few studies have investigated whether L2 learners’ interpretation abilities are influenced by discourse-level cues such as extra-sentential information (e.g. Dekydtspotter & Outcalt 2005, Hopp 2009, Pan & Felser 2011). Pan et al. (2015) found that referential context could affect L2 learners’ interpretation of ambiguous RCs in English. The learners’ initial analyses of RCs were strongly influenced by two types of contexts: supporting and biasing. Inspired by this effect on the acquisition of English RCs, the researcher would like to investigate whether English-speaking L2 learners will interpret Chinese RCs differently in supporting and biasing contexts. 6.

(17) 1.3 Research Questions Based on the previous discussion of difficulties in acquisition of RCs, some research questions were addressed to examine English-speaking learners’ interpretation and production of Chinese RCs and factors influencing their performance. Research questions were separately addressed in three studies as follows: Study I 1.. In what way do English-speaking L2 learners of Chinese produce and accept Chinese SRCs and ORCs?. 2.. Does gap condition affect the English-speaking learners’ performance on SRCs and ORCs?. 3.. Is the English-speaking learners’ performance influenced by their native language in their acquisition of SRCs and ORCs?. 4.. Do English-speaking learners perform differently in different tasks on Chinese SRCs and ORCs?. Study II 5.. In what way do English-speaking L2 learners of Chinese produce and accept gapped ORCs and gapless ORCs in Chinese?. 6.. Do DCL positions influence the English-speaking learners’ performance on gapped and gapless ORCs in Chinese?. 7.. Is the English-speaking learners’ performance affected by their native language in their acquisition of gapped and gapless ORCs?. 8.. Do English-speaking learners perform differently in different tasks on gapped and gapless ORCs?. Study III 1.. In what way do English-speaking L2 learners of Chinese interpret DCL-first and DCL-second ORCs in Chinese? 7.

(18) 2.. How do English-speaking L2 learners associate DCL phrases in DCL-first ORCs?. 3.. Do contextual differences influence the interpretation of DCL-first and DCLsecond ORCs and ambiguity resolution by English-speaking learners?. 1.4 Significance of the Present Study The theoretical issues of Chinese RCs and DCL phrases have been investigated in several studies (e.g., Zhang 2007, HLL 2009, Yin 2009). Although there has been much research on these issues in syntax, few studies have explored them in acquisition (e.g. Xu 2009, Wu & Sheng 2014). Despite the existing research has examined the acquisition of Chinese RCs and DCL positions (Hsiao & Gibson 2003, Lin & Bever 2010, Gibson & Wu 2013, Yip & Methews 2007, Hsu e al. 2009), only gapped RCs have been investigated. Gapless RCs have rarely been examined in acquisition. In addition, Wu & Sheng (2014) mentioned that L1 transfer could explain L2 learners’ errors in production; however, they did not further discuss whether their errors can be treated as evidence for L1 influence. Moreover, although Xu (2014) has already conducted both interpretation and production tasks on Chinese RCs, there was no comparison of results obtained from two types of tasks. Furthermore, these L2 studies (e.g. Xu 2009, Wu & Sheng 2014) on RCs did not examine the effect of context on L2 learners’ interpretation. Therefore, to fill the gaps (e.g. RC types, L1 influence, task 8.

(19) effect, contextual effect) found in literature, the researcher would like to provide a better picture of English-speaking learners’ interpretation and production of Chinese RCs. 1.5 Organization of the Thesis This thesis is organized as follows: In Chapter Two, theoretical issues of Chinese RC types are reviewed and the processing ease of different RC types is discussed. In addition, Chinese and English RCs are compared in typology. Chapter Three introduces the research designs of three studies. Moreover, the results and discussions of three studies are provided in Chapter Four, Chapter Five and Chapter Six, respectively. Finally, Chapter Seven makes a brief conclusion of this thesis, including the major findings of the present study and suggestions for future research.. 9.

(20) CHAPTER TWO LITERATURE REVIEW Some theoretical and empirical studies of Chinese relative clauses (RCs) are reviewed in this chapter. Section 2.1 discusses the types of Chinese RCs and their processing ease, and Section 2.2 presents a typological comparison of Chinese and English RCs. Section 2.3 reviews some empirical studies of DCL positions in Chinese RCs. Finally, a brief summary of this chapter is provided in Section 2.4. 2.1 Chinese RC Types This section discusses the types and processing of RCs in Chinese. In typology, Chinese RCs are considered as a pre-nominal modifier (e.g. Ding et al. 1961, Zhu 1982). According to Xu’s classification (2009), RCs are firstly divided into two types without involving DCL positions: subject RCs (SRCs) and object RCs (ORCs). Each type can be further classified into gapped and gapless. Moreover, by taking into consideration DCL positions, another type of RCs is DCL-RCs which contain DCL-first RCs and DCL-second RCs. The following sections will introduce these RC types: SRCs vs. ORCs, gapped RCs vs. gapless RCs, and DCL-first RCs vs. DCL-second RCs.. 10.

(21) 2.1.1 SRCs vs. ORCs First of all, RCs in Chinese can be classified into SRCs and ORCs, as shown in (1) (both adopted from Xu (2009, p.27)). (1a) is an SRC where a gap is generated in the subject position of the clause, and (1b) is an ORC with a gap in the object position. (1) a. [. xihuan like. Xiaoming. de. ren]. Xiaoming. DE person. jiao. Zhenni.. named. Jenny. ‘The person who likes Xiaoming is named Jenny.’ b. [Xiaoming Xiaoming. xihuan. de. ren]. like. DE person. jiao named. Zhenni. Jenny. ‘The person that Xiaoming likes is named Jenny.’ Two theories have been proposed to predict the processing asymmetry of SRCs and ORCs. One theory supports that SRCs are easier than ORCs to process and the other theory keeps a reverse position. Firstly, the Structural Distance Theory (SDT)1 proposed by O’Grady (1997) states that the number of XP categories is likely to measure complexity of an RC. Based on the SDT, (1a) and (1b) can be compared in structures as shown in (2), respectively. (2) a. [CP [NP [S ei [VP xihuan [NPo Xiaoming]] de] reni] [VP jiao zhenni]] b. [CP [NP [S [NPs Xiaoming] [VP xihuan ej] de] renj] [VP jiao zhenni]] In (2a), only the S category intervenes between the ei and the head noun reni while. 1. The Structural Distance Theory (SDT) is proposed by O’Grady (1997), and it is claimed that the complexity of an RC by the amount of ‘XPs categories between a gap and the element with which it is associated’ (1997:136). 11.

(22) in the ORC (2b), a VP and an S category intervene between the ej and the filler renj. It can be found that the ej in the ORC is in the more embedded position under the VP than the ei in the SRC, which is outside the VP. Thus, an ORC has more XP categories between the gap and the element where it is associated, which means that an ORC in Chinese is more complicated in structure. Thus, the SDT supports that SRCs are easier than ORCs to process. On the other hand, the Canonical Word Order Hypothesis (CWO)2 proposed by Bever (1970) and Slobin & Bever (1982) illustrates that if the word order of a sentence is similar to the order of canonical sentences of a language, this sentence is likely to be easier to process. A word order of the SRC (1a) is presented in (3a) while a word order of the ORC (1b) is in (3b). (3) a. [[. V N de] N]. b. [[NS V. de] NO]. According to the CWO, the word order of the ORC is more similar to the canonical order in Chinese [N-V-N]. Thus, ORCs are predicted to be easier than SRCs in processing; in other words, a sequence of processing ease for these two types of RCs is. 2. The Canonical Word Hypothesis (CWO) is proposed by Bever (1970) and Slobin & Bever (1982) that the processing difficulty is determined by whether the word order of sentences is unmarked in that language. Sentences can be easily processed if they have similar word orders to the canonical order of the language. 12.

(23) as follows: ORCs > SRCs. To sum up, a sequence of RC processing ease predicted by the SDT from easy to difficult is as follows: SRCs > ORCs, while the CWO indicates the sequence should be: ORCs > SRCs. 2.1.2 Gapped RCs vs. Gapless RCs Lin (2010:413) identifies RCs as ‘the clauses that are embedded inside noun phrases, whereby one nominal argument in the clause is co-referenced with the head of the higher noun phrase.’ Chinese RCs can also be grouped into gapped RCs and gapless RCs. In a gapped RC, the embedded nominal argument is left empty forming a gap and it can be a subject or an object, resulting in subject-extracted and object-extracted gapped RCs (see in (4a) and (4b), the same as the examples in (1)). Gapped RCs in Chinese are usually considered as regular RCs (Zhang 2007, Tsai 2008). (4) a. [. xihuan like. Xiaoming. de. ren]. Xiaoming. DE person. jiao. Zhenni.. named. Jenny. ‘The person who likes Xiaoming is named Jenny.’ b. [Xiaoming Xiaoming. xihuan. de. ren]. like. DE person. jiao named. Zhenni. Jenny. ‘The person that Xiaoming likes is named Jenny.’ Another type of RCs distinct from regular RCs is gapless as seen in (5), which refers to the RCs without a gap or a resumptive pronoun (HLL 2009) and this type can 13.

(24) be further grouped into SRC gapless (5a) and ORC gapless (5b). Zhang (2007) states that the head noun of a gapless RC must be a relational noun and its licensor is the clause. (5) a. [Yifu clothes. hen. zang. de. ren]. jiao. Yuehan.. very. dirty. DE. person. named. John. ‘The person whose clothes are very dirty is called John’ b. [[Ta he. changge. de]. shengyin]. sing. DE sound. hen. haoting.. very. good. ‘The sound of his singing is very good.’ In (5a), yifu ‘cloth’ shows the feature of the head noun ren ‘person’ and it is the subject of the RC. In (5b), the argument structure of chang ‘sing’ has been fulfilled by ge ‘song’, resulting in the fact that there is no gap in the RC. The head noun shengyin ‘sound’ is a predicate and the clause is the subject. According to Ning (1993), the head noun shengyin associates with a covert verb [V E] meaning of dedao ‘obtain’ and a structure of the RC in (5b) is proposed as (6). Thus, (5b) is regarded as an object RC. (6) [Opi [Ta changge he sing.song. [VP [V E] ti] de]]. [shengyin]. DE. sound. According to the Linear Distance Theory (LDT)3 proposed by Gibson (1998), the linear distance between a gap and a head noun of an RC can be used to measure. 3. The Linear Distance Theory (LDT) (also called Dependency Locality Theory (DLT)) is a psycholinguistic model by Gibson (1998, 2000) that the integration cost of an RC is measured by its linear distance. 14.

(25) integration cost. The longer the linear distance is, the more head-dependent integration will get involved. Gapped RCs with gaps require integration cost for integrating gaps and fillers. However, compared with gapped RCs, gapless RCs do not cost integration as much as gapped RCs do since they do not involve gaps. Thus, gapless RCs are likely to be predicted as the easier type in processing by the LDT. The processing ease of gapped and gapless RCs is predicted as follows: gapless RCs > gapped RCs. 2.1.3 DCL-First RCs vs. DCL-Second RCs In Chinese, a demonstrative classifier (DCL) phrase is able to appear in different positions of RCs. The current study will mainly discuss two possible positions of DCL phrases and the processing of these two positions. In Chinese RC constructions, a basic component of a nominal phrase can occur with a numeral or a demonstrative, which needs a classifier to specify the unit. HLL (2009) summarizes the order of these components as in (7), and the position of an RC is indicated by numerals I-III as in (8). (7) Demonstrative+ Number+ Classifier+ Noun zhe/na this/that. yi/san one/three. zhi/bei mao/shui cl/cup. cat/water. ‘this/that (one) cat; these/those three cups of water’. 15.

(26) (8). Demonstrative+ I. Number+ Classifier+ II. Noun III (HLL 2009:2014). According to HLL (2009), Position II is not natural enough for an RC (see Lu 1998), and Positions I and III are respectively considered as ‘restrictive’ and ‘nonrestrictive’ (see Chao 1968, Hashimoto 1971, Huang 1982). They show that RCs in Chinese are likely to be put after demonstratives (DCL-first RCs), or before demonstratives (DCL-second RCs), as shown in (9). (9) a. na ge. chuan. hong. that CL. wear. red. yifu. de tongxue. clothes. DE student. (DCL-first RC). ‘the student who wears red’ b.. chuan. hong. yifu. de. na. ge. tongxue. wear. red. clothes. DE the CL. student. (DCL-second RC). ‘the student who wears red’ (Zhang 2007:46) The question as to whether Chinese has non-restrictive RCs (see e.g. Chao 1968, Huang 1982, Del Gobbo 2003) is discussed in terms of the different interpretation of RCs. Chao (1968) claims that in a DCL-first structure, an RC is non-restrictive, while in a DCL-second structure, it is restrictive. DCL-first RCs are considered to be ‘descriptive’ while DCL-second RCs to be restrictive (Chao 1968). Huang (1982) also supports this claim and develops the ideas of scope where RCs are in the scope of Dem. 16.

(27) (demonstrative) in the first structure and Dem is included in the scope of RCs in the second structure. On the contrary, according to the ideas of Zhang (2001a, b) and Del Gobbo (2003), Chinese only exhibits restrictive RCs. Based on Huang’s analysis (1982), Del Gobbo (2003) states that in DCL-first RCs, the demonstratives are used deictically, while they are used anaphorically in DCL-second RCs. Zhang (2007) summarizes from Huang and Chao’s analyses that the distinction between DCL-first and DCL-second RCs is not the same as that between restrictive and non-restrictive RCs in English (see Zhang 2001a, b; Del Gobbo 2003), arguing that there are no equivalent non-restrictive Chinese RCs as those in English. Regarding DCL positions and interpretations of Chinese RCs, the processing ease of RCs with different DCL positions can be predicted by the Filler Gap Domain (FGD) Theory 4 proposed by Hawkins (2004), the filler-gap domain is likely to yield the processing difficulty in RCs. When an NP projection is assumed to exist in an RC structure, it can be found that a DCL-second structure has two more nodes than a DCL-. 4. The Filler Gap Domain (FGD) Theory proposed by Hawkins (2004) is used to test the processing ease of RCs by the filler-gap domain (FGD) that ‘the smallest set of terminal and non-terminal nodes dominated by the mother of a filler and on a connected path that must be accessed for gap identification and processing’ (2004: 175). 17.

(28) first structure does, involving additional nodes: the head Classifier (Cl) and Classifier Phrase (ClP). Based on the FGD Theory, it shows that DCL-second RCs have more complex FGDs than DCL-first RCs do. Thus, DCL-second RCs are predicted to be more difficult than DCL-first RCs in processing. A sequence in terms of processing ease is predicted as follows: DCL-first RCs > DCL-second RCs. By comparing with SRCs and ORCs with different DCL positions, it is worth mentioning that ambiguity occurs while processing DCL-first ORCs. The ambiguity results from the fact that the preceding DCL phrase in a DCL-first ORC can modify the head noun or the adjacent noun if the classifier is a general classifier (such as ge, liang, wei, etc.), as illustrated in (10). (10) a. Na liang that CL. huoche truck. zhuangdao. de. knock.down. jiaotache tingzai. DE bike. park. lubian. roadside. ‘That bike which was knocked down by the truck parked at the roadside.’ b. Na ge that CL. xiaonanhai changge de. yangzi. hen. ke’ai.. little.boy. look. very. cute. sing. DE. ‘The look of the little boy while he sings is very cute.’ Sentence (10a) is a gapped ORC while (10b) a gapless ORC. The classifiers in each sentence match the head noun and the subject of the ORC, which means that the DCL phrases in both the gapped and gapless ORCs can modify the head noun and the subject. 18.

(29) of the ORC. In (10a), naliang is likely to modify both huoche ‘truck’ and jiaotache ‘bike’ while nage in (10b) can modify both xiaonanhai ‘little boy’ and yangzi ‘look’. Thus, ambiguity appears in both sentences. To sum up, two positions of DCL phrases, DCL-first and DCL-second, are discussed in this section. In DCL-first RCs, RCs modify [(Demonstrative) + Number+ Classifier+ Noun] while in DCL-second RCs, RCs only modify [Noun]. The demonstratives in these two constructions have different functions: a deictic expression (in DCL-first RCs), or an anaphoric expression (in DCL-second RCs). Based on the FGD Theory, the sequence of processing ease is predicted as follows: DCL-first RCs > DCL-second RCs. Considering this prediction, we can propose that ambiguity of DCLfirst ORCs is unlikely to influence the interpretation of RCs. 2.2 A Typological Comparison of Chinese and English RCs This section compares Chinese and English RCs in terms of branching directions, RC types, and DCL phrases. Firstly, Chinese and English have different branching directions. Chinese has a left branching direction, which shows a head-final pattern in RCs (Greenberg 1996, Dryer 1992). However, although English is also a VO language like Chinese, it is right19.

(30) branching with a head-initial feature, as shown in (11) and (12), respectively. (11) a. [Zhangsan Zhangsan. mai de] buy. qiche hen. DE car. gui.. very. expensive. ‘the car that Zhangsan bought is very expensive.’ b. *Qiche [Zhangsan car. Zhangsan. mai buy. de] DE. hen. gui.. very. expensive. (12) *a. [John bought ] that the car is very expensive. b. The car that [John bought. ] is very expensive.. we can see that it is grammatical that an RC precedes a head noun in Chinese, as shown in (11a). However, this order is not grammatical in English, as illustrated in (12a). On the contrary, English requires a head noun preceding an RC, as in (12b). In Chinese, a head noun is not permitted to precede an RC, as (10b) shows. Secondly, both two languages have distinct types of RCs. Chinese exhibits gapped and gapless RCs (see Xu 2014, are mentioned in 2.1.1). A gap can be either in an RC or not, as in (13). Sentence (13a) is a gapped RC while (13b) is gapless. In comparison of Chinese, English only has gapped RCs. As in (14a), a gap must be generated in an RC. However, the sentence without a gap is ungrammatical in English, as presented in (14b), showing that English does not exhibit gapless RCs. (13) a. [. chuan. zhe. zang. yifu. de. ren]. wear. Asp. dirty. clothes. DE person. jiao named. ‘The person that wears dirty clothes is named Jenny.’ b. [Yifu. hen. zang. de. ren] 20. jiao. Zhenni.. Zhenni. Jenny.

(31) clothes. very. dirty. DE. person. named. Jenny. ‘The person whose clothes are very dirty is called Jenny.’ (14) a. [The person that. wears dirty clothes] is called Jenny.. b. *The person [that her clothes are very dirty] is called Jenny. Lastly, Chinese has a unique feature of DCL phrases in which ambiguity will occur in DCL-first ORCs (as mentioned in Section 2.1.3), as (15) shows. In contrast, ORCs only have one interpretation in English, as (16) shows. (15) Na zhi that CL. laohu. zhui. de. xiaoyang. pao. de hen. kuai.. tiger. chase. DE. sheep. run. DE very. fast. ‘The sheep that a tiger chasing ran very fast.’ Meaning 1: A sheep ran very fast and that tiger was chasing it. Meaning 2: That sheep ran very fast and a tiger was chasing it. (16) The sheep that a tiger was chasing ran very fast. Meaning: That sheep ran very fast and a tiger was chasing it. Sentence (15) has two interpretations where nazhi can associate with laohu ‘tiger’ or xiaoyang ‘sheep’. In other words, the DCL phrase may modify an adjacent noun or a head noun. However, in (16), the is used to modify only the head noun sheep, not tiger. That is to say, it is impossible for English speakers to get the interpretation where an article adjacent to a head noun modifies a noun in the clause. Table 2-1 summarizes the comparison of Chinese and English RCs in typology.. 21.

(32) Table 2-1 A Typological Comparison of Chinese and English RCs Languages. Word Order. RC Type. DCL phrases. head-. head-. gapped. gapless. modifying. modifying. final. initial. RCs. RCs. head. adjacent. word. word. nouns. nouns. order. order. Chinese. +. -. +. +. +. +. English. -. +. +. -. +. -. According to Larsen-Freeman and Long (1991:53), “Where two languages were similar, positive transfer would occur; where they were different, negative transfer, or interference, would result.” Foley and Flynn (2013) also emphasize that similarities between L1 and L2 will facilitate L2 acquisition while L1-L2 differences can impede L2 learners’ learning. Thus, following their claims, attributing to the positive influence of L1, we can predict that gapped RCs are easier than gapless RCs and that L2 learners may only get one interpretation where a DCL phrase modifies a head noun. On the other hand, due to the negative influence of L1, English-speaking learners are predicted to place a head noun preceding an RC, and to have difficulty acquiring gapless RCs. Additionally, it is assumed that L2 learners will fail to interpret a DCL phrase as a modifier of a noun adjacent to it in a DCL-ORC structure. The current study aims to investigate whether these predictions can be.. 22.

(33) 2.3 Previous Empirical Studies of DCL Position in Chinese RCs This section reviews studies on the L2 acquisition of DCL positions in Chinese RCs from the aspects of processing and production. Four studies are reviewed, namely Xu (2009), Li (2013), Wu & Sheng (2014), Xu (2014). 2.3.1 Xu (2009) Xu (2009) investigated the L2 learners’ processing and production of different types of RCs with two studies (Study I and Study II). Study I was a self-spaced word order judgment task, and thirty-two Englishspeaking subjects participated in it ranging from eighteen to thirty-six years old. This study aimed to test the L2 processing ease of the following four types of RCs: (17) a. DCL+SRC+N (DCL-first) b. SRC+DCL+N (DCL-second) c. DCL+ORC+N (DCL-first) d. ORC+DCL+N (DCL-second) According to Del Gobbo’s classification (2003), Xu addressed two sequences of RCs, including DCL-first ((17a) and (17c)) and DCL-second RCs ((17b) and (17d)). In each sequence, Xu further classified RCs into SRCs ((17a) and (17b)) and ORCs ((17c) and (17d)). A total of one hundred and twenty sentences were presented in the task, including forty-eight grammatical sentences, forty-eight ungrammatical sentences and. 23.

(34) twenty-four fillers. From Study I, Xu concluded that in both DCL-first structures and DCL-second structures, SRCs ((17a) and (17b)) were generally easier than ORCs ((17c) and (17d)). This result was identical to the prediction made by the SDT, rather than the LDT. It was found that the L2 learners were able to read SRCs faster than ORCs; however, they still had a slightly high error rate of SRC-DCL sentences. That is to say, DCL positions were likely to influence the L2 learners’ interpretation of different RC types. In addition to the examination of RC types, Xu further investigated the effect of DCL positions. It was found that DCL-first RCs ((17a) and (17c)) were easier than DCL-second ones ((17b) and (17d)) on the whole, in that the L2 learners had shorter response time (RT) in judging DCL-first RCs and also had a lower error rate of this RC type. This finding was also consistent with the prediction made by the SDT. The processing ease of the four types of RCs from easy to difficult was: (17a) > (17c) > (17b) > (17d). In addition, Xu conducted a written sentence completion task in Study II for examining whether these findings could be supported in a production task. Forty English-speaking subjects and twenty-eight Chinese native speakers were recruited to. 24.

(35) complete the task, which contained forty sentences (sixteen test items and twenty-four fillers) in total. The items consisted of eight in DCL-RC-N structures and eight in RCDCL-N structures and aimed to test the L2 learners’ production preference for SRCs and ORCs by asking the participants to complete sentences with given verbs, as in (18a). The participants were required to fill in blanks to complete sentences with either SRCs or ORCs and a possible answer as in (18b). (18) a.. de na ge nansheng zai Beijing. daxue. DE that CL boy. University study (fall.in.love.with). ‘The boy that b. Wo aishang I. at Beijing. dushu. (aishang). studies at Beijing University. (fall in love)’ de na ge nansheng zai Beijing. fall.in.love.with DE that CL boy. daxue. dushu.. at Beijing University study. ‘That boy that I fell in love with studies at Beijing University.’ (Xu 2009:217) In Study II, the production of RCs in different types was measured by the accuracy rate. Xu (2009) observed that the accuracy rate of DCL-first sentences (80%) was roughly similar to that of DCL-second ones (84%), which did not show which type of RCs was easier for the L2 learners in production. In DCL-first structures, DCL-SRC sentences were proved to be easier for the L2 learners who produced more grammatical sentences with SRCs than those with ORCs. However, as for DCL-second structures, the subjects’ accuracy rates of SRCs and ORCs were nearly equal. By comparing the 25.

(36) L2 learners and native speakers, Xu found that the L2 learners’ production was nativelike. In summary, the first finding of Study I was that DCL-SRC structures were easier than DCL-ORC structures. Secondly, SRC-DCL structures were also easier than ORCDCL sentences. Lastly, DCL-first structures were easier than DCL-second structures. However, the results of Study II were only identical to the first result of Study I, but they did not support the other two findings of Study I. Although Xu (2009) designed both a processing task and a production task, there are some inadequacies. First of all, even though the research questions of Studies I and II are the same for investigating the interpretation and production of different types of RCs, the participants of these two studies are different and are not in the same size either. Study I has thirty-two subjects excluding Chinese native speakers as a control group while Study II has sixty-eight subjects including Chinese native speakers. Secondly, the materials used in these two tasks in Xu’s study do not include gapless RCs and contextual effect, which fail to provide a whole picture of L2 acquisition of Chinese RCs. Thus, the current study will further discuss these issues.. 26.

(37) 2.3.2 Li (2013) Li (2013) investigated the distribution of classifiers in Chinese RCs in the HSK corpus of Huo (2008), who collected writing samples from exams of L2 learners from 1992 to 2005. He explored a total of 308 RCs made by English-speaking learners from Australia, Canada, America and Britain, and analyzed them through several aspects: the position of head nouns in main clauses, the types of RCs, classifiers or nonclassifiers, the position of RCs and classifiers, the collocation of the classifiers, the animacy of head nouns, and the types of verbs in clauses. It was found that his L2 learners produced more RCs without classifiers (80.5%) than those with classifiers (19.5%). In addition, among the RCs with classifiers, the classifiers with numerals were more frequently produced than those with demonstratives (χ2(1, 60) = 32.267, p < .001). Especially, the L2 learners preferred to produce RC sentences with classifiers and yi ‘one’. Secondly, regardless of RC types, sentences with classifiers in pre-RC positions were more frequently produced than those with classifiers in post-RC positions (χ2(1, 58) = 27.59, p < .001). On the other hand, when RC types were taken into consideration, the subjects produced classifiers with SRCs more frequently than with ORCs (χ2(1, 60) = 27.

(38) 17.067, p < .001), which was identical to the results obtained from L1 studies (Pu 2008, Wu 2009, Ming & Chen 2010). As for SRCs, it was found that the L2 learners preferred to place classifiers in pre-RC positions (χ2(1, 46) = 31.391, p < .001). With regard to ORCs, there was no obvious asymmetry of sentences with different classifier positions. Li claimed that classifier-ORC sentences were not easier than ORC-classifier sentences since semantic conflicts might occur while classifiers in pre-RC positions cannot modify subjects of ORCs, as (19) shows. The numeral-classifier yiju ‘one sentence’ cannot modify bieren ‘other people’. (19) Dang when. ni neng. xiechu. you can. write.out one CL other.people. de] juzi. [bieren. de shihou, na zhong. DE sentence DE time wukebini. yi ju. neng. mingbai. can. understand. manzugan. that CL. sense.of.satisfaction. shi SHI. de.. beyond.comparison. DE. ‘When you can write a sentence that other people can understand, the sense of satisfaction you get from it is beyond description.’ (Li 2013:171) In brief, Li (2013)5 found that the L2 learners tended to produce sentences with classifiers in pre-RC positions than those with classifiers in post-RC positions.. 5. Li (2013) also found that the animacy affected the L2 learners’ production. The L2 learners preferred to put classifiers before RCs no matter the RCs modify animate or specific inanimate head nouns. Moreover, in the RCs with classifiers in pre-RC position, the head nouns being modified are mainly abstract inanimate nouns, which slightly showed the animacy effects in the L2 learners’ production of DCL positions in RCs. 28.

(39) Additionally, classifiers were more inclined to be put by the L2 learners before SRCs rather than ORCs. However, Li only adopted the exam data from a corpus, and he did not include an interpretation task in his study. In addition, he focused on classifiers in production but failed to examine the effect of the positions of demonstratives on the interpretation of RCs. Also, although the materials involved gapped SRCs and ORCs, gapless RCs were not included. 2.3.3 Wu & Sheng (2014) Wu and Sheng (2014) used a word-based sentence production paradigm to investigate how the L2 learners produced Chinese RCs online and how they sequenced DCL phrases and RCs. The participants were twenty-three overseas advanced Englishspeaking learners of Chinese from different universities in Shanghai and Beijing. They were required to pass a pre-test on the interpretation of forty Chinese phrases selected from Levels 5 to 6 of the HSK. And the formal test included a total of sixty sentences: twenty-four items were RCs (SRCs and ORCs) with DCL phrases modifying subjects and thirty-six fillers. Each sentence was segmented into four constituents: demonstrative classifiers, RCs, head nouns, predicates of main clauses. A paradigm was made by the Perception Research Systems to present stimuli to 29.

(40) the participants, who were asked to read four segments of sentences in wrong order presented on the computer screen and to produce complete sentences by combining four parts in correct order. Their production was recorded into audio files. Based on the data, firstly, it was found that ORCs were easier than SRCs for the L2 learners to produce in that more ORCs in head-final order were produced than SRCs in head-final order (73.11% vs. 56.27%). Additionally, the error rate of ORCs was lower than that of SRCs (4.92% vs. 11.03%). These results were not identical to the findings of L1 and L2 acquisition studies (Pu 2008, Wu 2011, Sheng & Wu 2013, Tang 2007, Sheng & Wu 2013, Li 2013). Although the L2 learners made more head-final sentences, they still produced some sentences where head nouns were put before RCs in both SRC and ORC structures. This finding indicated that the L2 learners were influenced by their L1 knowledge. Besides the investigation of RC types, they explored DCL positions and found that it was easier for the L2 learners to make RCs with DCL-first structures than those with DCL-second structures (χ2(1, N=383) = 233.42, p < .001). In SRCs in head-final structures, DCL phrases tended to be placed before SRCs (χ2(1, N=148) = 64.89, p < .001), which. 30.

(41) corresponded to the results from previous studies (e.g. Pu 2008, Wu 2011, etc.). In another structure, i.e. ORCs in head-final structures, the L2 learners also produced sentences with DCL phrases in pre-RC position more frequently (χ2(1, N=193) = 13.48, p < .001), which was not identical to the findings in previous studies. In terms of the findings mentioned above, DCL positions were likely to affect the L2 learners’ production; thus, they examined the production of SRCs and ORCs in the structures with different DCL positions. In DCL-first structures, SRCs (49.80%) in head-final order produced by the L2 learners were approximately as many as ORCs (50.20%) in head-final order. However, in DCL-second structures, the L2 learners produced more ORCs in head-final order than SRCs in head-final order (χ2(1, N=96) = 22.04, p < .001). This finding was just opposite to the results of Xu (2009), but it was identical to the results of Pu (2008) and Wu (2011) in their L1 studies. To sum up, ORCs were easier for the L2 learners to produce, which was different from the findings of Xu (2009) and Li (2013). Both in SRCs and ORCs, the L2 learners produced more DCL-first RCs than DCL-second ones. However, Wu and Sheng only compared the results in this L2 study with those obtained from other L1 studies and it. 31.

(42) was hard to say whether the L2 learners’ performance was native-like or not, since the production task in their study was different from the tasks in previous L1 studies. Additionally, they did not examine whether the results obtained in the production task could be supported by the results of an interpretation task. Lastly, RC types investigated in this study only involved gapped RCs but not gapless RCs. 2.3.4 Xu (2014) Xu (2014) examined the processing of Chinese RCs by forty-nine Englishspeaking adult learners. A self-paced word-order judgment task was conducted, including a total of one hundred and eight items: forty-eight sentences in correct order, forty-eight sentences in incorrect order and twenty-four fillers. The test items included four types, as shown in (20a-d). (20) a. DCL+ SRC+N (DCL-SRC) b. SRC+DCL+N (SRC-DCL) c. DCL+ ORC+N (DCL-ORC) d. ORC+ DCL+ N (ORC-DCL) The participants were asked to judge whether the given sentences were in correct sequence or not by the DMDX software (Forster and Forster 2003) on computers and they had twelve seconds to judge each item. Through comparing the L2 learners’ response time (RT) in the study, Xu (2014) 32.

(43) found that DCL-first RCs were generally easier than DCL-second RCs in that the participants read DCL-first RCs faster than DCL-second RCs. They also made fewer errors in DCL-first RCs, which was identical to the prediction made by the FGD Theory6. Additionally, regarding RC types, Xu found that the participants processed SRCs better than ORCs, which was supported by the shorter RT of SRCs. In DCL-first RCs, SRCs were easier than ORCs to process in that the error rate of DCL-SRCs was lower. But there was no evident asymmetry of processing ease of SRCs and ORCs in DCLsecond structures. These findings showed that RC types might influence the L2 learners’ interpretation. Thus, Xu further investigated the interpretation of DCL-first and DCLsecond sentences with different RC types. In SRCs, DCL-first structures were the easiest, which was consistent with the findings of L1 corpus research (e.g. Ming & Chen 2010). This finding was supported by the lowest error rate of DCL-SRC sentences.. The Filler Gap Domain (FGD) Theory proposed by Hawkins (2004) is used to test the processing ease of RCs by the filler-gap domain (FGD) that ‘the smallest set of terminal and non-terminal nodes dominated by the mother of a filler and on a connected path that must be accessed for gap identification and processing’ (2004: 175). 6. 33.

(44) On the other hand, the data did not show a distinct asymmetry of processing ease of ORCs with different DCL positions. The RT of DCL-ORC sentences was roughly as long as that of ORC-DCL structures. To summarize, Xu (2014) adopted an online task to test processing ease of four types of Chinese RCs. DCL-first structures were testified as the easier type. The overall results also showed that it was easier for the L2 learners to process SRCs, especially in DCL-first structures. However, Xu did not explore whether the findings of the interpretation task could also apply to a production task. He only investigated gapped SRCs and ORCs and did not examine gapless RCs. Also, contextual effect was not explored in the task. Thus, more studies should be conducted to fill these gaps. 2.3.5 Summary of Empirical Studies This section presents a summary of four empirical studies including their major findings and inadequacies, as shown in Table 2-2.. 34.

(45) Table 2-2 A Summary of Four Empirical Studies Studies Xu. Major Findings. Inadequacies and Limitations. 1. Position of DCL: DCL-first 1. Participants: not in same size;. (2009). easier than DCL-second. no. 2. Type of RC Produced: SRC. control. group. in. processing task. before ORC in either DCL-first 2. Materials: gapless types of or DCL-second structures. RCs, contextual effect not examined. Li. 1.. (2013). Position. of. classifiers: 1.. classifier-first. easier. than. classifier-second 2.. Sheng (2014). 2. 3.. 2.. 1.. (2014) 4.. control. Materials:. demonstratives. Type of RC Produced: SRC. and gapless types of RCs not. before ORC. examined Task: only a corpus study. Position of DCL: DCL-first 1.. Participants: no control. easier than DCL-second. group. Type of RC Produced: ORC 2.. Material: gapless types of. before SRC. RCs not examined. Factors Affecting Production:. 3.. L1 knowledge Xu. no. group. 3. Wu & 1.. Participants:. Task: only one production task. Position of DCL: DCL-first. 1.. Participants: no control. easier than DCL-second. group. Type of RC Preferred: SRC 2.. Materials: gapless types of. before. RCs, contextual effect not. ORC. in. DCL-first. structures, but not in DCLsecond structures. examined 4.. Task: only one processing task. Some linguists (Xu 2009, Xu 2014) examined the processing ease of different RCs with different DCL positions by L2 learners, while others (Xu 2009, Li 2013, Wu & Sheng 2014) investigated L2 learners’ preference for DCL positions in production. 35.

(46) All the researchers found that DCL-first structures were easier to process (Xu 2009, Li 2013, Wu & Sheng 2014, Xu 2014). SRCs were testified as the easier type in the L2 learners’ processing and production (Xu 2009, Li 2013, Xu 2014). However, in other studies, the results showed a preference for ORCs in the L2 learners’ production (e.g. Wu & Sheng 2014). Therefore, the effect of RC types was still controversial. All these studies had some inadequacies. Firstly, the RCs investigated in these studies were gapped (Xu 2009, Li 2013, Wu & Sheng 2014, Xu 2014), and gapless RCs were not discussed. Moreover, all researchers did not further investigate the effect of L1 transfer on L2 acquisition of RCs. Also, Li (2013), Wu and Sheng (2014) and Xu (2014) only conducted a single task in their studies, either an interpretation task or a production task, which could not provide a whole picture of L2 acquisition of RCs. Lastly, Xu (2009, 2014) examined the processing of RCs, but she did not test whether context could affect L2 learners’ interpretation of RCs. To fill in the gaps mentioned above, the current study attempts to examine the English-speaking learners’ interpretation and production of Chinese RCs by providing a thorough picture of their L2 acquisition of RCs in Chinese.. 36.

(47) 2.4 Summary of Chapter Two In this chapter, Chinese RC types and a typological comparison of Chinese and English RCs were discussed and four empirical studies on the L2 learners’ processing and production of Chinese RCs and DCL phrases were reviewed. Based on theoretical and empirical studies done before, the current empirical study will explore the effect of possible factors such as types of RCs, DCL phrases, L1 knowledge, task types, context to see if they will affect English-speaking learners’ interpretation and production of Chinese RCs. The research design of this study will be introduced in Chapter Three.. 37.

(48) CHAPTER THREE RESEARCH DESIGN In this chapter, the detailed design of the current research is presented. To investigate the L2 learners’ performance on different RC types (SRCs and ORCs), we conducted Study I. According to Wu & Sheng’s findings (2014) discussed in Chapter Two, ORCs are easier for L2 learners to produce than SRCs. In order to examine our subjects’ production and interpretation on ORCs with or without gaps (gapped or gapless ORCs), we also conducted Study II. To further explore whether context will affect L2 learners’ interpretation when DCL phrases occur in different positions (DCLfirst and DCL-second ORCs), we conducted Study III. The research design of each study includes the subjects, the methods and materials, and the procedure. Section 3.1 describes Study I, Study II is illustrated in Section 3.2, and Study III is presented in Section 3.3. Finally, Section 3.4 summarizes this chapter. 3.1 Study I Study I focuses on examining L2 learners’ production and acceptability of SRCs and ORCs. 3.1.1 Subjects A total of 40 subjects were recruited and grouped into an English group and a. 38.

(49) Chinese group. The English group included 20 high-intermediate English-speaking learners of Chinese who were using Books 3 to 5 of A Course in Contemporary Chinese as their textbooks at Mandarin Training Center (MTC) of National Taiwan Normal University (NTNU). These English-speaking subjects took a 10-hour language class and 5-hour supplementary class a week in a regular program at the MTC. The Chinese group consisted of 20 Chinese native speakers who were the students majoring in or minoring in Mandarin Chinese at NTNU. This group was used to examine whether the performance of the English group was native-like. A summary of the subjects is illustrated as follows. Table 3-1 A Summary of the Subjects Group. Language Proficiency. Mean Age. Number. English. High Intermediate Level. 24.3. 20. Chinese. Native. 22.7. 20. Total. 40. 3.1.2 Methods and Materials Qualitative and quantitative approaches have been used in many studies on second language acquisition (e.g. Xu 2009, Walliman 2011, Williams & Menard-Warwick 2014). In a qualitative approach, the development of a subject’s performance is. 39.

(50) generally investigated by collecting the subject’s speech over a period of time (LarsenFreeman & Long 1991). This approach is able to provide a deeper understanding of the process of L2 learners’ acquisition. However, a qualitative approach is ungeneralizable and it also takes much time. In contrast, a quantitative approach can be used to test an assumption by objective tasks and statistical analyses (Larsen-Freeman & Long 1991). Researchers collect data from a larger group of subjects at only one point in time and the results obtained by this approach are generalizable. Hence, to examine L2 learners’ interpretation and production of Chinese RCs, the previous studies (Wu & Sheng 2013, Xu 2014) have used quantitative approaches, such as production tasks, judgment tasks, since they are more generalizable and outcome-oriented. Thus, the current study followed them to adopt a quantitative approach. However, the previous researchers only conducted a single task in their experiments. In order to avoid the bias resulted from using a single task, we conducted two tasks in Study I: a sentence-making (SM) task and an acceptability judgement (AJ) task. 3.1.2.1 A Sentence-Making (SM) Task For investigating the production of Chinese RCs by L2 learners, written sentence completion tasks and online word-based sentence production tasks have been used in 40.

(51) previous studies (Xu 2009, Wu & Sheng 2014). However, the sentence completion tasks employed by Xu (2009) and Wu & Sheng (2014) mainly investigated L2 learners’ preference of SRCs or ORCs, not their performance on gapped and gapless RCs. Therefore, in order to investigate the production of different RC types, Study I adopted an SM task in which the test items were further grouped into SRCs and ORCs based on whether a subject or an object was extracted. Thus, a total of 4 subtypes were included as follows: gapped SRCs, gapped ORCs, gapless SRCs and gapless ORCs, where each subtype consisted of two sentences. To focus on factors’ (such as gap condition, task types, etc.) affecting the L2 learners’ performance on SRCs and ORCs, we mainly designed the test items with RCs modifying the subject NP, i.e., the S pattern, in the present study. A summary of the test items for the SM task is presented in Table 3-2.. 41.

(52) Table 3-2 A Summary of the Test Items for the SM Task of Study I Major. Examples. Number Questions. Types SRCs. 1. Da le Zhangsan jingchaju. taopao. de. huairen. cong. 4. le.. Qs 2,3,5,8. ‘The bad guy who hurt Zhangsan escaped from police office.’ 2. Chengji nian. henhao. de. xuesheng. dou. qu. yanjiusuo le.. ‘The students with good grades all go to graduate schools.’ ORCs. 1. Zhangsan. jieshao. de. laoshi. xihuan. zhonghua.. 4. Qs 1,4,6,7. ‘The teacher that Zhangsan introduced likes planting flowers.’ 2. Wangwu. shuohua. de shengyin. henda.. ‘The speaking voice of Wangwu is very loud.’ Total. 8. In the SM task, each test item was separated into three constituents: a clause, a head noun, a VP. All parts of a sentence were numbered in random order. The subjects were asked to unscramble test items with the numbers according to their syntactic intuition. An example of the SM task of Study I is illustrated in Table 3-3. For the complete version of the SM task, please refer to Appendix A.. 42.

(53) Table 3-3 An Example of the SM Task1 of Study I Sentence Hen keai / Lisi zhaogu de ①. Answer. / xiaonvsheng. ②. ②③①. ③. 3.1.2.2 An Acceptability Judgment (AJ) Task Previous empirical studies (Xu 2009, 2014) have conducted grammaticality judgement (GJ) tasks to test L2 learners’ syntactic intuition about Chinese RCs. However, they did not conduct an AJ task to examine L2 learners’ preference for Chinese RCs. In Study I, the AJ task was conducted for investigating our L2 learners’ acceptability of different types of RCs. A total of four subtypes of RCs were included, and each of them consisted of two sentences. A summary of the test items for the AJ task is illustrated in Table 3-4. Table 3-4 A Summary of the Test Items for the AJ Task of Study I Types SRC. Example Zhichi hen. Wangwu. de. pengyoumen. dou. Number. Questions. 4. Qs 2,3,4,5. 4. Qs 1,6,7,8. congming.. ‘The friends who support Wangwu are all very smart.’ ORC. Xuesheng. kaoshi. de. chengji. feichang. hao. ‘The grades of students are very good.’ Total. 8. 1. The test sentences in the SM task were presented in written form with traditional Chinese characters without pinyin or English translations. 43.