論漢語右置移位現象

全文

(1)國立臺灣師範大學英語學系碩士論文 Master’s thesis Department of English National Taiwan Normal University. 論漢語右置移位現象 On Right Dislocation in Mandarin Chinese. 指導教授：吳曉虹教授 Advisor: Dr. Hsiao-hung Iris Wu 研究生：姜玉娟 Graduate: Lucy Yu-chuan Chiang. 中華民國 104 年 7 月 July 2015.

(2) 摘要. 本論文主要探討的是漢語右置移位現象。右置移位現象在其他語言中，像是德文、西文、法文、日文、韓文等，是十分常見的，過去已經有許多學者針對此現象做了有系統的句法分析。但是相對來說，漢語的右置移位現象的句法討論尚未有一套完整的分析討論，因此本論文旨在探究此句法結構在漢語中的行為表徵，並進一步地提出其對應的句法分析。首先，我針對右置移位以及另外兩個表面相似的結構，即句後補充及呼語結構，提出區辨的方式；繼而回顧了右置移位現象在文獻上曾提出的分析方式，分別是移位、基底衍生以及刪略。本文根據右置移位的定義以及過去學者所提之分析方式，逐一檢視了漢語在右置移位的句法表現，發現三個文獻的主要分析法都各有其優缺點。我進一步提出移位分析才能夠較適切地分析漢語右置移位結構。具體來說，漢語右置移位結構先經歷了主題化，然後再藉由殘餘成分移位至焦點的位置。這樣的分析法不僅能夠正確衍生出表面的語序，並且也能夠解釋右置移位片語的言談功能。本論文期望可以提供未來對於右置移位結構研究奠定一些基礎的語料及分析。關鍵字：右置移位、漢語、移位分析. i.

(3) Abstract. The aim of this thesis is to investigate the right dislocation construction in Mandarin Chinese. Right dislocation is widely observed in languages such as German, Spanish, French, Japanese, and Korean. However, there has not been enough attention paid to this construction in Mandarin Chinese and, accordingly, a systematic syntactic analysis is left desired. Thus, the present study aims to fill this gap. In the thesis, I first distinguish right dislocation from the other two superficially similar constructions, i.e. afterthought and vocative phrases. I compare the mainstream analytic options proposed for right dislocation in the literature, namely, movement analysis, base generation analysis, and ellipsis analysis; I further show that the three analyses all have their advantages and disadvantages in light of the Mandarin data. Nevertheless, I argue that among the three, the movement analysis can better capture the phenomenon observed in right dislocation of Mandarin Chinese. The movement analysis proposed in the present study consists of two operations, i.e., topicalization of the dislocated phrase followed by a remnant movement to the Spec, FocP position. This analysis not only generates the correct surface word order but also can differentiate the discourse function of the right dislocated phrases. This study wishes to provide the empirical and theoretical basis for the future study on the right dislocation in Mandarin Chinese. Key words: Right dislocation, Mandarin Chinese, Movement analysis ii.

(4) Acknowledgments. Un homme est bien fort quand il s’avoue sa faiblesse. --Honoré de Balzac. I never thought that I could finish my thesis within such a short time. The investigation on syntactic issues is never easy. This is a journey full of challenges; nevertheless, it is sweet as candy at the end of my graduate study. I have received helps from countless people when I study here. This thesis is not yet perfect, but with their help, I am stronger and have more confidence facing my thesis. They all deserve my appreciation from the bottom of my heart.. Firstly, I owe my advisor, Prof. Hsiao-hung Iris Wu, a debt of gratitude. Were it not for her guidance, I would not have taken the road to linguistics, especially syntax. I am honored to meet such an outstanding professor when I am here in NTNU. During the time of my thesis writing, I learned much more than I thought through the discussion with her and I was inspired when working with her. She is definitely the greatest supporter and mentor. Without her, I could never finish my thesis. I am greatly indebted to Prof. Gerardo Fernández-Salgueiro as well. I am blessed to have such an excellent professor to work with and I respect him a lot. He is always willing to discuss the syntax issues with me and spends time on giving me many useful and inspiring comments. If it were not him bringing out the iii.

(5) idea on right dislocation, I would never have the opportunity to investigate such an interesting topic. Also I would like to show my gratitude toward Prof. Chen-Shen Luther Liu. It is my pleasure to have such a distinguished syntax researcher included in my committee. He gave me abundant insightful comments on this challenging topic and helped me to make this thesis better and more complete.. In addition, I also want to thank the linguistics faculty members of the English Department of NTNU. I was benefit from their lectures and the linguistic training from their courses, Prof. Doris Chen, Prof. Miao-Ling Hsieh, Prof. Miao-Hsia Chang, Prof. Nissa Lin, Prof. Joy Wu, Prof. Jen-I Li, Prof. His-Yao Su, Prof. Lindsey Chen, and Prof. Jen-Ting. Also I want to thank the administration staffs, Mu-Han, Peggy, Hui-Chiao, Tzu-yi, and Hao-Ku.. Besides, I am grateful that I meet my classmates, Kevin, Ken, Lilian, Jan, Vicky, Doris, Bill, Ivy, Jennifer, John and Karen, and the lab members, Gracie, Lin, Helen, Vivi, Julia, Matt, Jeff and Ronald. They all showed great supports during the time in the graduate school. I especially want to thank Kevin Sung for his support. He is my greatest partner. We encouraged each other in doing the syntactic research and went through the hard time of thesis writing. Whenever I feel gloomy, he always pushes me to the bright side. It is my honor to have such a great partner on the road of the study in syntax. Also, I am thankful that iv.

(6) I have Ken Lin. He is one of the best friends, the best punch-line maker who always brings me laughter whenever I am stressful, and the best listener whenever I feel like I am in trouble. Even when he went to the Netherlands for exchange student, he was supportive as always just like he never left. I am delighted to know him in the graduate school. I want to thank Lilian Chen as well. She is the warmest person who listens to my complaints and gives me comfort whenever I meet her. I am happy to know Aries Cheng and Lin for they are the best school sisters, good listeners and supporters. They add abundant spices into my life. Gracie Liao is my role model in terms of conducting linguistic researches and her passion towards linguistics. Her courage of chasing her dream deeply enlightens me. Vivi Hsu is like my mentor. Whenever I have questions in mind, I can always find the answer after chatting with her. Thanks Andrew Syue for chatting with me and always making me laugh. I am pleased to have his support at the end of my thesis writing. My 1451 friends, Jocelyn, Soup, Hazelynn, Kate, I thank them for keeping encouraging me when I am pursuing my dreams. I would never forget the time when we study and live together. I appreciate for their mental support. My friend Amber Wang and Angus Hsu, although they are often occupied with their works, I could not thank them more for giving me strength when I need it. I treasure the time whenever we meet. Siōng-jû sī guá tha̍k tī gián-kiù-sóo ê sî-tsūn siōng hó ê pîng-iú. Tong guá bô huann-hí ê sî-tsūn, i long ē an-uì guá, puê guá khai-káng. Guá tsiok ài lán nn̄g lâng tsē tī-leh kheh-thiann, khuànn-tio̍h thang-guā ê iā-kíng, tsò-hué khai-káng, lim tsiú. Guá tsiok v.

(7) kám-siā guá ē-tàng sı̍ k-sāi lí, tsiok kám-siā guá ū tsı̍ t ê tsiah-nī hó ê pîng-iú. To-siā lí. Due to the space limits, I could not list all the people who have helped me, but I want to say, ‘Thank you! Merci beaucoup!’. Moreover, I sincerely appreciate the professors and participants of SICOGG 16 in Seoul because they gave me many insightful comments and suggestions on my preliminary analysis. It was an unforgettable experience and I enjoyed the time of discussing my topic with the participants in that conference. I am grateful that I had this great opportunity to share my work with others and know more about the studies related to right dislocation.. I would like to leave the last paragraph in Mandarin Chinese to show my gratitude to my beloved family members. 最後我要謝謝我的家人，我的爸爸、媽媽、弟弟、爺爺、奶奶，我愛你們。謝謝我的爸媽在我選擇走上研究所這條路的時候給了我無比的支持和鼓勵，也謝謝他們的照顧跟教育，讓我可以順利的完成研究所的學業。我也要謝謝從小到大一路吵吵鬧鬧但是感情只有更好的弟弟，每當我心情低落的時候，感謝有他可以一起聊天、一起散心，我真的很愛你。謝謝我的爺爺、奶奶，每次回家都一定會有很好吃的東西，不管有多累，只要回到家裡，就一定可以感受到家裡的溫暖。還有很多無法一一列出來的阿公、阿嬤、叔叔、阿姨，謝謝你們，因為有你們的照顧，才能有今天的我。. vi.

(8) Table of Contents. Chinese Abstract………………………………………………………………………………i English Abstract……………………………………………………………………………...ii Acknowledgments..………………………………………………………………………….iii Table of Contents…………………………………………………………………………....vii List of Tables………………………………………………………………………………..viii Chapter 1 Introduction………………………………………………………………………1 Chapter 2 Literature Review………………………………………………………………...5 2.1 Dislocation…………………………………………………………………………...5 2.1.1 Left dislocation and right dislocation…………………………………………6 2.1.2 Right dislocation and afterthought…………………………………………..10 2.1.3 Distinguishing RD and AT in Mandarin Chinese……………………………18 2.1.4 Distinguishing RD and vocative phrases in Mandarin Chinese……………..23 2.2 Previous analyses of right dislocation………………………………………………32 2.2.1 Movement analysis…………………………………………………………..33 2.2.2 Base generation analysis…………………………………………………….35 2.2.3 Ellipsis analysis……………………………………………………………...38 Chapter 3 Right dislocation in Mandarin Chinese………………………………………..43 3.1 Facts of right dislocation in Mandarin Chinese……………………………………..43 3.2 Analysis……………………………………………………………………………..63 3.2.1 Movement approach…………………………………………………………63 3.2.2 Base generation approach……………………………………………………65 3.2.3 Ellipsis approach…………………………………………………………….68 3.2.4 Discussion………………………...…………………………………………75 3.3 Proposal…………………………………………………………………………….77 Chapter 4 Conclusion……………………………………………………………………….83 References…………………………………………………………………………………...85. vii.

(9) List of Tables. Table 1 the comparison between right dislocation and afterthought………………………...18. viii.

(10) Chapter one Introduction. Dislocations have received a great amount of attention in syntactic studies. Among them, right dislocation (henceforth, RD) gains more and more attention in recent decades in generative syntax. RD is widely discussed in a number of languages under the generative approach, such as Italian, French, German, Dutch, Spanish, Catalan (Cecchetto 1999; Ott & de Vries 2012, 2014, in press; López 2009, and among others), Japanese (Tanaka 2001 and others), Korean (Chung 2009; Lee 2009, Ko 2014 and among others) and many other languages. The examples of RD of these different languages are provided in (1). As seen in Italian, German, French and Spanish, the case-marked resumptive pronoun such as lo, ihm and il ‘him’ are required in this construction. Importantly, only resumptive pronoun is allowed. It cannot be the proper noun. In contrast, Korean represents a different pattern that the proper noun, Yuni, does not have to be replaced by a resumptive pronoun.. (1). a. Italian (Cecchetto 1999) Io lo I. odio, Gianni.. him hate. Gianni. ‘I hate him, Gianni.’ 1.

(11) b. German (Ott & de Vries 2012) Ich habe ihm I. geholfen, dem. have him-Dat helped. Peter.. the-Dat Peter. ‘I helped him, Peter’ c. French Il. est gentil,. He is. Jean.. nice. Jean. ‘He is nice, Jean.’ d. Spanish Yo lo I. odio, a. him hate. to. Juan. Juan. ‘I hate him, Juan.’ e. Korean (Chung 2009) Cheli-ka. Yuni-lul. manna-ess-e, Yuni-lul.. Cheli-Nom. Yuni-Acc meet-Pst-De, Yuni-Acc. ‘Cheli saw Yuni.’. As for the example of RD in Mandarin Chinese, both declarative sentence and interrogative sentence in (2) show that the resumptive pronoun, ta ‘he’, is optional and that the proper noun 2.

(12) Zhangsan cannot repeat twice, which is superficially different from the languages mentioned above. Also, uniquely, the sentence final particles ‘le’ and ‘ma’ are required in the RD in Mandarin Chinese. I will show that the appearance of the sentence final particles signals the end of the clause, which is similar to the prosodic differences at the end of the clauses exclusively in RD in Romance languages. Thus, in some cases, sentences final particles are absent because the fact that there has already been an element that signals the end of the sentence.. (2) a. Context : Students are playing hide-and-seek. Zhangsan’s teammates want to make sure whether Lisi is found or not. Thus, a person replies with this sentence. (Tai)/*Zhangsani. zhaodao Lisi le. he. find. Zhangsan. Zhangsani.. Lisi SFP Zhangsan. ‘He found Lisi, Zhangsan.’. b. Context : Students are gossiping in class. Most of them think that Zhangsan likes Lisi, but one student doubts that. Thus, the student asks this question. (Tai)/*Zhangsani he. Zhangsan. xihuan Lisi like. Lisi. ma Zhangsani? SFP Zhangsan. ‘Does he like Lisi, Zhangsan?’ 3.

(13) Unlike the studies of RD in other languages, the literature of RD in Mandarin Chinese mainly focuses on the function or discourse perspective of this construction (Guo 1999, Luke 2012, and so on). It has not yet been fully discussed under the generative approach. Thus, the aim of this thesis is to investigate the phenomenon of RD in Mandarin Chinese under the minimalist framework (Chomsky 1995) since little attention has been paid to this construction in the literature (though see Cheung (1997, 2005, 2009) and Law (2003) on RD in Cantonese). The research question in the present study is the following. What is the syntactic structure of RD construction in Mandarin Chinese? With the syntactic structure, it can be testified which approach proposed in previous scholarship, namely, ellipsis, base generation or movement (or none of them), can better analyze the RD in Mandarin Chinese. The thesis is organized as follows. Section 2 is the literature review of dislocation and the previous analyses on RD. Section 2.1 firstly introduces two types of dislocation that are obviously different from the surface, i.e., left dislocation and right dislocation. Furthermore, I differentiate some seemingly similar structures from RD, i.e., afterthought and vocative phrases, because these constructions all involve a right-dislocated phrase on the surface. However, the syntactic properties of these constructions are different. Then in Section 2.2, the previous analyses of RD, i.e., movement, ellipsis and base generation, which are the bases of the study, are reviewed in this section. Section 3 provides the data and the analysis and my proposal of RD in Mandarin Chinese. Section 4 concludes the paper. 4.

(14) Chapter 2 Literature Review. In this section, the properties and the definition of RD as well as some related dislocation constructions are introduced. In section 2.1, the previous research on RD, afterthought, vocative phrases and left dislocation are summarized. From the previous literature, I differentiate RD from the other constructions, and provide the criteria to define RD in Mandarin Chinese. Then the data that this study targets at, RD in Mandarin Chinese, are presented following the definition of RD and the discussion of the literature. Section 2.2 is the review of the three main analyses of RD, namely, movement approach, base generation approach and ellipsis approach.. 2.1 Dislocation Dislocation can be divided into left dislocation and right dislocation. Left dislocation (henceforth, LD), as the name indicates, is a construction that has a dislocated XP in the leftmost position of a sentence. On the other hand, RD is the opposite in that a dislocated XP is in the rightmost position of a sentence. However, there are also some other constructions, including afterthought and vocative phrases, that apparently have a dislocated XP on the right of the sentence. In the following part, I will discuss the differences among these 5.

(15) constructions.. 2.1.1 Left dislocation and right dislocation To begin with, the dislocation structure mainly has two types which are distinct from the surface, that is, LD and RD. The two constructions are illustrated in (3). The main difference between the two lies in the landing site of the dislocated XP which originates from the host clause and has a resumptive pronoun to fill in the gap after dislocation. In (3a), the dislocated XP is on the left in LD while in (3b) the dislocated XP is on the right in RD.. (3) a. left dislocation: [[dislocated XP] [host clause]] b. right dislocation: [[host clause] [dislocated XP]]. The studies of LD can date back quite early and the body of literature is considerably large (Rivero 1980, Ziv 1994, Anagnostopoulou et al. 1997, Rizzi 1997, Cecchetto 1999, Lambrecht 2001, Alexiadou 2006, and many others). In the literature, LD1 is described as a construction that has an XP occupying a position in the leftmost of the clause and preceding the subject of the clause2. The examples of LD in different languages are shown in (4), 1. In a broad sense, LD may include topicalization since both constructions have a left dislocated XP on the surface. However, some may argue that these two constructions are not the same in terms of coreferentiality. In the present study, the difference between LD and topicalization is not within my focus; for the detailed studies on the distinction, readers can refer to Rivero (1980), Lambrecht (2001), and among others. 2 LD can be further divided into contrastive left dislocation (CLD), clitic left dislocation (CLLD) and hanging topic left dislocation (HTLD), whose differences has been greatly discussed in the literature (Riemsdijk and 6.

(16) mainly taken from Lambrecht (2001:1051). As can be seen, the left dislocated NP, the Romans in different languages is in the leftmost position of the clause. In the host clause, there is a resumptive pronoun, they, coindexing with the dislocated NP.. (4) Left dislocation in different languages (Lambrecht 2001:1051) a. French [Les Romains]i, ilsi sont fous. the Romans. they are crazy. b. German [Die Römer]i, diei spinnen. the Romans they are.crazy c. Italian [I. Romani]i,. The Romans. son-oi. pazzi. be-3.PL crazy. d. English [The Romans]i, theyi are crazy.. Zwarts 1974, Haaften et al. 1983, Cinque 1990, Alexiadou 2006, and among others). One significant difference between CLLD and HTLD is the agreement between the resumptive pronoun and the dislocated phrase. The resumptive pronoun and the dislocated phrase are coreferred in CLLD but not in HTLD. Also, the dislocated phrase can undergo reconstruction in CLLD while reconstruction is not possible in HTLD. In the present study, thus, my focus will be on CLLD only since it is more related to the construction that I am going to investigate in Mandarin Chinese, i.e., a construction that has co-indexation and that can undergo reconstruction. Detailed discussions of different types of LD can be found in Alexiadou (2006) and the papers cited therein. 7.

(17) On the contrary, the amount of the studies of RD is less and the discussion of RD appears later than that of LD. In RD, it is obvious to observe that the dislocated XPs are in the rightmost position of the clause. The examples of RD taken from Lambrecht (2001) are shown in (5). The dislocated NP, these Romans, is at the end of the clause in French, German, Italian, and English respectively3. Similar to LD, there is a co-indexed resumptive pronoun, they, in the host clause in the four languages.. (5) Right dislocation in different languages (Lambrecht 2001:1051) a. French Ilsi sont fous, [ces. Romains]i.. they are crazy these Romans b. German Diei spinnen, [diese Römer]i. they are.crazy these Romans c. Italian Son-oi. pazzi, [questi Romani]i.. be-3.PL crazy these Romans 3. It is observed that the choice of the determiners of the NP between the LD and RD is different. In LD, the definite articles are preferred and sound more natural than the demonstrative ones (Lambrecht 2001), while in RD, the demonstratives are acceptable. Such a difference in Romance and Germanic languages is not within my focus in the present study although the difference may be related to some phenomenon observed in RD in Mandarin Chinese, as I will discuss later in section 3. It seems that specificity plays an important role in RD. Also it is important to note that Ott and de Vries (2014) observe this fact and use it as a piece of evidence to support their ellipsis analysis. The detailed analysis can refer to their paper. 8.

(18) d. English Theyi are crazy, [these Romans]i.. In the early research on dislocation, many view RD as a mirror of LD since the only difference between the two seems to be the position of the dislocated phrases. Also, the movement to the left is expected in the syntactic analysis while the movement to the right is more special. It would be more reasonable to propose that LD is the basic construction and RD is derived from LD, sharing a unified analysis with LD. Nevertheless, Cecchetto (1999) argues that RD is not the same as LD by showing the discrepancies from the following four main effects, anticonstruction, ECP, the right roof constraint, and Aux-to-Comp constructions4. López (2009) also provides evidence to argue that clitic left dislocation (CLLD) is not exactly the same as clitic right dislocation (CLRD). Moreover, the examples in Spanish in (6) illustrate that LD and RD behave a little different. The preposition a ‘to’ is obligatorily required in RD in (6a). On the contrary, in (6b), the preposition can be omitted in LD. This piece of evidence may indicate that the operation between LD and RD may not be the same.. 4. There are some debates on the arguments and examples proposed by Cecchetto (1999). For instance, Frascarelli (2004) points out that one of the discrepancies used to argue against the difference between CLLD and CLRD does not totally support Cecchetto’s proposal. More discussion on this issue can refer to Frascarelli (2004) and the papers cited therein. 9.

(19) (6) LD and RD in Spanish a. Right dislocation Yo lo I. odio, *(a). him hate. to. Juan Juan. ‘I hate him, Juan’ b. Left dislocation (A) Juan, yo. lo. to. him hate. Juan. I. odio.. ‘Juan, I hate him.’. Due to these discrepancies and facts, it can be clearly seen that RD is indeed syntactically different from LD and that the analysis for LD is not totally suitable for RD. Hence, RD should have its own analysis.. 2.1.2 Right dislocation and afterthoughts Before I proceed to discussing the focus of this study, it is necessary to differentiate RD and afterthought (henceforth, AT) since they both seem to appear in the sentence-final position on the surface5. A great amount of literature has discussed the differences between 5. However, in the latest version of Ott and de Vries’s (2014) analysis, they propose that their ellipsis approach can account for both AT and RD. That is, even though AT and RD have different restrictions and the syntactic structures may be different, Ott and de Vries aim to provide a unified analysis for this two constructions that 10.

(20) the two from the perspective of their functions (Ziv 1994, Averintseva-Klisch 2008, Luke 2012, and many others). According to Averintseva-Klisch (2008), the function of RD is to identify the discourse topic while the function of AT is mainly about the discourse repair strategy. Crucially, AT is more subordinating while RD is more independent in terms of their relationships with the host clause. Precisely, the relationship of the right-dislocated NP in RD is closer to the host clause than that in AT. In particular, Ziv (1994) and Averintseva-Klisch (2008) provide tests to distinguish these two constructions, and the properties can be summarized as the followings. Generally speaking, the differences come from two aspects, prosody and syntax. In terms of prosodic features, RD does not have a pause or special tone preceding the right-dislocated NP. In contrast, ATs have a distinct pause and special tone preceding the dislocated NP and thus the dislocated NP is often stressed. Averintseva-Klisch (2008) further points out that optional additions like I mean is not allowed in RD while they can appear in AT, as shown in (7).. (7) English a. Right dislocation Context: John is chatting with Mary and talking about Mary’s brother. Joh wants to tell Mary that he met Mary’s brother the other day. Thus, he utters this sentence.. both include a right-dislocated element. 11.

(21) I met him two weeks ago, (*I mean), your brother.. b. Afterthought Context: John is chatting with Mary and he wants to tell Mary that he met Mary’s brother the other day. However, when he uttered the sentence, he realized that Mary did not know the person that John was talking about. Thus, he adds ‘I mean’ for clarification. I met him two weeks ago, (I mean), your brother.. As for the distinct syntactic features, it can be evidenced from three aspects. The first one is the restriction of the position of the right-dislocated NP. The right-dislocated NP is always in the clause-final position in RD while the position of the dislocated NP is freer in ATs, as illustrated in (8). As shown here, your brother, in AT can be moved to the sentence internal position, while in RD, the dislocated NP must stay in the sentence final position. The difference between the RD and AT data lies in the discourse function. In the example of RD in (8a), the right-dislocated phrase is the topic, so it usually appears at the edge of the sentence. For the second sentence in (8a) to the grammatical, the adverbial phrase, two weeks ago, is AT, meaning that the original sentence for RD is only ‘I met him, your brother’. As for AT in (8b), the discourse function is repair and clarification or adds more information; thus, 12.

(22) the dislocated phrase can appear at different positions and it does not have to be in the edge. The position is related to the discourse context not directly to the syntactic structure. (8) English a. Right dislocation Context(i) and (ii): John is chatting with Mary and talking about Mary’s brother. Joh wants to tell Mary that he met Mary’s brother two weeks ago. Thus, he utters this sentence. (i). I met him two weeks ago, your brother.. (ii) *I met him, your brother, two weeks ago.. b. Afterthought Context (i): John is chatting with Mary and he wants to tell Mary that he met Mary’s brother two weeks ago. However, when he uttered the sentence, he realized that Mary did not know the person that John was talking about. Thus, he adds ‘your brother’ for specification. (i) I met him two weeks ago, your brother. Context (ii): John is chatting with Mary and talking about Mary’s brother. Joh wants to tell Mary that he met Mary’s brother. When he uttered the sentence, he suddenly wants to specify the time that he met Mary’s brother. Thus, he adds ‘two weeks ago’ in 13.

(23) the end and after a pause of saying ‘your brother’. (ii) I met him, your brother, two weeks ago.. Similar distributional distinction is also observed in German. The examples in (9), taken from Averintseva-Klisch (2008), exhibit the same properties. That is, the dislocated NP, der Taifun ‘the typhoon’, is always in the rightmost position in (9a), while in (9b), the NP den Peter ‘Peter’ in AT can appear in the clause internal position. Moreover, in these two examples, it can be seen that ich meine ‘I mean’ is not allowed in RD but is acceptable in AT.. (9) German (Averintseva-Klisch 2008: 220-221) a. Right dislocation Da. war eri, (* ich meine) der Taifuni.. there. was it. I mean the typhoon. ‘There it was, the typhoon.’ b. Afterthought Ich habe. ihn, [ich meine den Peter],. gestern. nur. mit. Mühe. I. him I. yesterday. only. with. effort. have. mean the Peter. wiedererkannt. recognized. ‘I hardly recognized him yesterday, I mean Peter.’ 14.

(24) Second, in RD, the right-dislocated NP must be coreferential and morphologically agrees with the NP in the host clause. However, ATs do not necessarily require coreferentiality and agreement, as shown in (10) and (11). The examples in English in (10) illustrate that correferenciality is required in RD as indicated by the coindexation in (10a). On the contrary, AT does not have this restriction. It also can be evidenced that the optional addition ‘I mean’ can appear either on the right or on the left of the dislocated phrase, as in (10b). The sequence of the dislocated phrase in AT and the optional addition is flexible since they both function as discourse repair.. (10) English (Ziv 1994) a. Right dislocation Hei/*j is there, Jimi. b. Afterthought Context: Mary is chatting with Alice, and she wants to tell Alice that she met Bill yesterday. However, when Mary utters the sentence, she mistakenly said John at the very beginning. Thus, she has a self-correction by adding ‘I mean’. (i) I met John yesterday, Bill, I mean. (ii) I met John yesterday, I mean, Bill.. 15.

(25) Likewise, German in (11a) also shows that the right-dislocated NP, den Mann ‘that man’, should have the same case and gender agreeing with its resumptive pronoun, ihn ‘him’, in RD. However, the example of AT in (11b) shows that the case agreement between the resumptive pronoun and the NP is not obligatorily needed.. (11) German (Averintseva-Klisch 2008: 220) a. Right dislocation Ich kann. ihn. nur bedauern,. den Mann/. I. him-Acc.Masc. only regret. the man-Acc.Masc. can. *der. Mann.. *the. man-Nom.Masc. ‘I feel sorry to him, that man.’ b. Afterthought Esi. ist ein bisschen. gleichförmig, deine Melodiei.. It-Neut. is. homogeneous your. a bit. melody-Fem. ‘It is a little homogeneous, your music.’. Island sensitivity is the third syntactic difference between RD and AT. Island constraints are observed in RD but are absent in AT, as illustrated in (12). The German example in (12a) 16.

(26) shows, the right-dislocated NP, dem Meyer ‘the Meyer’, cannot move out from the complex NP island, das Gerücht, dass ihm die Mafia geholfen hat ‘the rumor that the mafia has assisted him.’ In contrast, the AT in (12b), is perfectly fine with the NP, dem Meyer ‘the Meyer’, dislocated in the right. This indicates that RD is sensitive to islands while AT does not have this restriction.. (12) German (Averintseva-Klisch 2008: 223) a. Right dislocation *Ich habe I. have. schon oft [NP das Gerücht, dass. ihmi die Mafia. already often. him the mafia. the rumor. geholfen. hat]. gehört, dem Meyeri.. helped. has. heard the Meyer. that. ‘I have already often heard the rumor, that the mafia has assisted him, this Meyer’ b. Afterthought Ich habe. schon. oft [NP das Gerücht, dass. ihmi die Mafia. I. already. often. him the mafia. have. the rumor. that. geholfen. hat]. gehört, ich meine dem Meyeri.. helped. has. heard I. mean the Meyer. ‘I have already often heard the rumor, that the mafia has assisted him, this Meyer’. 17.

(27) In sum, there are five crucial distinctive differences between RD and AT, namely, pause, optional addition, fixed position, correferentiality and agreement as well as island sensitivity. Table 1 briefly summarizes the discussion in this section. These distinctive criteria not only reflect the fundamental differences between the function of AT and RD but also suggest that the seemingly similar surface structures do not have the same underlying syntactic structure. In other words, even though they both have a right-dislocated phrase in the sentence, RD exhibits more connectivities or dependency relations to the host clause and more restrictions while AT does not.. Table 1 The comparison between right dislocation and afterthought. Prosody Syntax. Right dislocation. Afterthought. Pause. No. Yes. optional addition ‘I mean’. No. Yes. fixed position. Yes. No. coreferentiality & agreement. Yes. No. island sensitivity. Yes. No. 2.1.3 Distinguishing RD and AT in Mandarin Chinese After introducing the properties and showing the differences among LD, RD, and AT, I present the examples of RD in Mandarin Chinese based on the definition discussed in section 2.1.2. In this paper, the notation that distinguishes RD and AT in Mandarin Chinese is represented by the absence of the comma. RD sentences do not have a comma meaning that 18.

(28) there is no pause while a comma in AT sentences stands for a pause. As discussed in the previous section, there are five criteria to distinguish RD and AT, namely, pause, optional addition, fixed position, coreferentiality and agreement and island sensitivity. In this section, I only focus on pause, optional addition and fixed position. I will also show some data of island sensitivity that differentiate RD and AT but leaving the detailed discussion of island sensitivity in RD to section 3. To begin with, the RD construction in Mandarin Chinese consists of a host clause and a dislocated XP as formulized in (13a). Two typical types of RD in Mandarin Chinese are presented in (13b) & (13c), in which Zhangsan is the right-dislocated noun phrase. The part where Zhangsan moves out and contains the sentence final particle, ‘le’ and ‘ma’, and the resumptive pronoun ‘ta’ is the host clause.. (13) a. [[host clause] [XP]] b. Context : Zhangsan looks upset. His mother worries about him, so she asks Zhangsan’s classmate ‘What happened to Zhangsan?’ And Zhangsan’s classmate replies. (Tai) he. kandao. Lisi le. Zhangsani.. see. Lisi SFP Zhangsan. ‘He saw Lisi, Zhangsan.’. 19.

(29) c. (Tai) he. xihuan. Lisi ma Zhangsani?. like. Lisi SFP Zhangsan. ‘Does he like Lisi, Zhangsan?’. The examples in (14) show the prosody facts that distinguish RD and AT, namely, pause and optional additions. It can be seen in (14) that there are a pause and an optional addition, wo shi shuo ‘I mean’, in AT, while pause and optional addition in RD are disallowed. The reason why RD does not allow wo shi shuo ‘I mean’ is that the pause does not exist between the right-dislocated phrase and the host clause. On the contrary, the pause in AT suggests that the following phrase is a correction or additional information. Thus, adding ‘I mean’ further confirms that the pause does exist in AT and the right-dislocated phrase does not have any dependency relations to the element in the host clause. Additionally, adding ‘I mean’ may make the sentence sound better. That is because the appearance of ‘I mean’ suggests that there is another piece of information which corresponds to the function of AT, discourse repair strategy, and the discourse information structure. On the other hand, RD does not need the optional addition to introduce the dislocated phrase since RD is a complete sentence. The low acceptability of adding ‘I mean’ in the RD example may contribute to the fact that Mandarin Chinese does not prefer something appearing or left on the right edge of the sentence. In Mandarin Chinese, it is more restricted in terms of leaving a constituent at the 20.

(30) end of the sentence comparing with other languages. (cf. Cheng 2015). (14) a. Right dislocation Context: Zhangsan’s mother cannot find Zhangsan in the house so she asks her husband about the place that Zhangsan may go to. Tai zai xuexiao ma (*wo shi shuo). Zhangsani?. he in. Zhangsan. school. SFP. I shi say. ‘Is he at school, Zhangsan?’ b. Afterthought Context: Zhangsan and Lisi are brothers. Their mother cannot find Zhangsan, so she wants to check with her husband whether Zhangsan is at school or not. However, when she asks the question, she says the wrong name, Lisi. Thus, she corrects herself immediately at the end of the utterance. Lisi zai. xuexiao ma,. (wo shi. Lisi in. school SFP. I shi. shuo), Zhangsan? say. Zhangsan. ‘Is Lisi at school, I mean, Zhangsan?’. Examples in (15) illustrate the differences between RD and AT in terms of the fixed position. The right-dislocated NP, Zhangsan, is at the rightmost position in RD. In contrast, 21.

(31) Zhangsan in AT does not have to be in the right most position; instead, it can be in the clause internal position.. (15) a. Right dislocation Lisi kan-guo. tai. ba. Zhangsani.. Lisi see-Asp. he. SFP Zhangsan. ‘Lisi has seen him, Zhangsan.’ b. Afterthought Lisi gei ta,. Zhangsan, wo. shi shuo,. yi-ben shu.. Lisi give he. Zhangsan. shi say,. one-CL book. I. ‘Lisi gives him, Zhangsan, I mean, a book.’. Lastly, in example (16), it can be seen that island effects are observed in RD but not AT. In (16a), the NP Lisi cannot be moved out from the wh-island to the right; otherwise, it violates the island effect. On the contrary, in (16b), the NP Lisi can move out from the island causing no ungrammaticality. In the AT example, adding a pause or optional additions ‘I mean’ can save the sentence from ungrammatical.. 22.

(32) (16) a. Wh-island in right dislocation *Zhangsan xiang. zhidao tai. xihuan. shenme a. Zhangsan want. know he. like. what. Lisii?. SFP Lisi. ‘Intended: Zhangsan wants to know what does he like, Lisi?’ b. Wh-island in afterthought Zhangsan xiang. zhidao tai. xihuan. shenme a,. wo shi shuo , Lisii?. Zhangsan want. know he. like. what. I. SFP. shi say. Lisi. ‘Zhangsan wants to know what does he like, I mean, Lisi?’. In this section, I have presented the data in Mandarin Chinese are all follow the criteria proposed in section 2.1.2. It is seen that RD and AT are indeed two different construction in Mandarin Chinese and need to have separate analysis. In the next section, I will further distinguish RD from another construction that also looks similar with RD on the surface.. 2.1.4 Distinguishing RD and vocative phrases in Mandarin Chinese In addition to AT, there is another set of sentences that is similar to RD and that needs to be addressed, namely, vocative phrases. On the surface, vocative phrase may be situated in the right-most position sharing the same property of RD, as in (17).. 23.

(33) (17) qiguai. e. ni!. strange. SFP you. ‘You are strange, you!’. However, to take a closer examination at the host clause, it is seen that the host clause is not a complete sentence but rather an adjectival predicate, qiguai ‘strange’6. Some may argue that (17) still can be counted as RD. I will show how the sentence in (17) involves a vocative phrase based on Moro (2003). Moro provides some diagnostics for vocative phrases. First, the position of the vocative phrase is usually at the very beginning or at the very end of the sentence. Second, the function of the vocative phrase is, in general, to attract someone’s attention. The vocative phrase can be a pronoun or an entity. Lastly, the vocative phrase is not within the theta-grid of the main predicate. I will present data in Mandarin Chinese according to the criteria respectively. To begin with, consider the sentences in (18) which show the property of the position of vocative phrases. Compared the examples in (18) with those in (19), it is seen that both vocative phrases and dislocated phrases in RD can be situated in the sentence final position,. 6. Note that one-syllable adjectives cannot fit in this construction, e.g., gao ‘tall’, ai ‘short’, unless the degree modifier hen ‘very’ is added before the adjectives. The reason why one-syllable adjectives cannot be independently used may be a phonological problem. It may also be affected by the syntactic property as well. Nevertheless, I will not further discuss the reasons behind the differences among the adjectives. 24.

(34) namely, ni ‘you’ and Zhangsan.7 From the examples in (18), the difference at this stage is not yet clear and thus some may consider RD and vocative phrases to be the same. However, let us reexamine the sentences in (18) and (19). It can be seen that the resumptive pronoun behaves differently between vocative phrase and RD. One is that the resumptive pronoun cannot appear. The overt resumptive pronoun causes ungrammaticality in vocative phrases. In contrast, the resumptive pronoun in RD is optional. The overt resumptive pronoun does not cause any ungrammaticality. Moreover, some consider (18b) to be less satisfactory than (18a). Such difference is not observed in RD. From the facts discussed suggests that vocative phrase and RD may not be the same underlyingly.. (18) Vocative phrases a. (*Ni) you. qiguai. e. ni!. strange. SFP you. ‘You are strange, you!’. 7. The vocative phrase can also appears at the very beginning, which makes it look similar to left dislocation or topicalization. The examples are illustrated in (i). It is important to note that without pause between the vocative phrases, ni ‘you’ and Zhangsan, and the adjectival predicate, the sentence is ungrammatical. With the pause, the dislocated phrases serve as vocative phrases and have the function of calling for attention. (i) a. Ni, you. qiguai. e.. strange. SFP. ‘You, you are strange!’ b. Zhangsan, Zhangsan. qiguai. e.. strange. SFP. ‘Zhangsan, you are strange!’ 25.

(35) b. ?(*Tai) he. qiguai. e. Zhangsani!. strange. SFP Zhangsan. ‘Zhangsan is strage, Zhangsan!’. (19) Right dislocation a. (Ni) you. hen qiguai. e. ni!. very strange. SFP you. ‘You are very strange, you!’ b. (Tai) he. hen qiguai. e. Zhangsani!. very strange. SFP Zhangsan. ‘Zhangsan is very strange, Zhangsan!’. Then, taking the second criterion, the function, into consideration, a contrast is also shown between RD and vocative phrases. As shown that (18b) is not totally perfect, but for those who consider (18b) to be acceptable, they interpret Zhangsan as ni ‘you’. To be precise, when this sentence is uttered, the speaker is actually pointing at Zhangsan, so Zhangsan is involved in the conversation and needs to be present. Thus, in this case, Zhangsan, in fact, is a substitution for the second person pronoun, ni ‘you’. This piece of interpretation corresponds to the function of vocative phrases, i.e. attracting someone’s attention. When the 26.

(36) proper noun is used in the vocative phrase, it carries a stronger deictic sense. It can be evidenced from the examples in (20) and (21). In the sentences of vocative phrase in (20), changing the third person resumptive pronoun to the second person resumptive pronoun for the vocative phrase makes the sentence sound better. The second person resumptive pronoun usually serves as specifying the hearer in the conversation. Thus, this distinction is in cope with the function of the vocative phrase, calling for someone’s attention.. (20) Vocative phrase a. *Tai he. qiguai. e. Zhangsani!. strange. SFP Zhangsan. ‘Zhangsan is strange, Zhangsan!’ b. Nii you. qiguai. e. Zhangsani!. strange. SFP Zhangsan. ‘You are strange, Zhangsan!’. Moreover, this split distinction is not found in RD, as illustrated in (21). In the case of RD, two kinds of resumptive pronouns are both grammatical, which is different from vocative phrases. It should be noted that in (21a), the second person pronoun, ni ‘you’, is co-indexed with the proper noun, Zhangsan, since it is restricted to the context of a conversation. Thus, 27.

(37) in this case, this sentence is RD not AT based on my definition (i.e. coreferentiality).. (21) Right dislocation a. Nii you. hen qiguai. e. Zhangsani!. very strange. SFP Zhangsan. ‘Zhangsan is very strange, Zhangsan!’ b. Tai he. hen qiguai. e. Zhangsani!. very strange. SFP Zhangsan. ‘Zhangsan is very strange, Zhangsan!’. Lastly, the crucial criterion based on Moro (2003) is the theta-role; that is, the vocative phrase does not get a thematic role from the predicate in the main clause. As in the data presented above, the crucial element that sets RD and vocative phrase apart has not yet been discussed, that is, the degree modifier hen ‘very’8. From (18) to (21), it is shown that the sentence with hen modifying the adjectives can have a co-indexed resumptive pronoun. On the contrary, the sentence without the degree modifier hen could hardly have a co-indexed resumptive pronoun. Recall the definition of RD, it says that a co-indexed resumptive pronoun is obligatory in RD and RD can undergo reconstruction, suggesting that the 8. In the present study, I do not discuss the morpheme, hen ‘very’ and the syntactic reason why this kind of sentences is acceptable once hen is added. Liu (2010) suggests that hen is a morpheme indicating comparison. The properties of hen and its analysis can refer to Liu (2010) and the papers cited therein. 28.

(38) right-dislocated phrase has a strong connection with the host clause. Thus, it can be inferred that the right-dislocated phrase in RD does receive a theta-role from the main predicate. From the above-mentioned examples, the co-indexed reumsptive pronouns are supported. I provide more data regarding reconstruction in the following part. Based on this definition, I compare the following two sets of sentences, as in (22) and (23). The data in (22) show that the sentences with the degree modifier hen can undergo reconstruction, which are actually RD. On the other hand, the data in (23) are sentences without hen and they cannot undergo reconstruction, suggesting that hen plays a defining role in differentiating RD and vocative phrases and that the vocative phrases do not receive theta-roles from the main predicate. It is noteworthy that for some speakers, they consider the set of sentences in (23) grammatical because they think that there is a covert ‘hen’ preceding the adjectives. This interpretation supports the idea that the degree modifier ‘hen’ usually appears with the adjectival predicates.. (22) Right dislocation a. Ni. hen qiguai. you very strange. e! SFP. ‘You are very strange!’. 29.

(39) b. Zhangsan Zhangsan. hen. qiguai. e!. very. strange. SFP. ‘Zhangsan is very strange!’. (23) Vocative phrase a. *Ni you. qiguai. e.. strange. SFP. ‘Intended : You are strange.’ b. *Zhangsan qiguai Zhangsan strange. e. SFP. ‘Intended: Zhangsan is strange.’. In addition, there is another pivotal set of sentences regarding this criterion. Consider the examples in (24). It is said that the dislocated phrase in RD receives the theta-role from the main predicated while the vocative phrase does not receive any theta-role from the main predicate. From (24), the examples show that the dislocated phrase can be omitted in vocative phrases but not in RD. In other words, the adjectival predicate in (24b) can be independent without any covert or overt NP. Nevertheless, the adjectival predicate in (24a) shows that an overt NP is needed. If the NP is covert in (24a), it must be clearly mentioned in the context. 30.

(40) This can be evidence that without the degree modifier ‘hen’ the dislocated NP cannot receive the theta-role from the main predicate and that with hen the dependency relation of the dislocated phrase and the host clause is closer.. (24) a. Right dislocation hen. qiguai. e. very. strange. SFP. *(ni)! you. ‘You are very strange, you!’ b. Vocative phrases qiguai. e. (ni)!. strange. SFP you. ‘You are strange, you!’. To have a short summary, based on these three criteria, RD can be distinguished from vocative phrases. The vocative phrases are not within the theta-grid of the predicate in the main clause and do not have any dependency relations with the main clause. Also their function is attracting attention and their position is at the edge of the sentences. On the contrary, right-dislocated phrases in RD do receive a theta-role from the main predicate and the connectivity with the main clause is observed. As for the function of RD, it is not 31.

(41) restricted to attracting someone’s attention but rather identifying the discourse topic. The position of RD, as discussed in the previous section, is always in the sentence final position. Thus, again, the seemingly similar surface structure does not have the same syntactic structure. In this section 2.1.3 and 2.1.4, I gave a brief introduction on the data of RD in Mandarin Chinese. So far, the criteria used to distinguish RD and AT in other languages are also attested in Mandarin Chinese. Besides, the criteria distinguishing RD and vocative phrases are evidenced by the data in Mandarin Chinese. Later in section 3, more discussion will be provided on the data and facts of RD in Mandarin Chinese. To sum up, in section 2.1, I discussed the different constructions in dislocations from LD to RD and showed that they are indeed different. From RD, I further distinguished it from AT with five criteria and from vocative phrases with three criteria. In the next section, the syntactic analyses of RD in the literature will be introduced.. 2.2 Previous analyses of right dislocation In this study I focus on three mainstream analyses proposed for RD9, namely, movement analysis, base generation analysis and ellipsis analysis. For the movement analysis, it assumes that RD is monoclausal. As for the base generation approach, it is proposed that the 9. In addition to the movement analysis, the ellipsis analysis and the base generation analysis, there are other possible analyses of RD, such as, rightward movement (Ross 1967), base generation with clitic doubling (Kayne 1994) and so on. 32.

(42) right-dislocated phrase is base-generated in the very high position. While for the ellipsis approach, it suggests that RD is derived from the biclausal structure. I will further discuss these three analyses in the following. In section 2.2.1, I begin with a summary of the movement analysis for RD, and then followed by the alternative analysis, base generation analysis and ellipsis approach, in section 2.2.2 and section 2.2.3, respectively.. 2.2.1 Movement analysis As discussed in the previous section, it is seen that RD is sensitive to islands in contrast to AT. This property is described as clause-external properties, suggesting that movement is involved in the derivation of this construction. It is, thus, reasonable to propose that RD is the result of movement since islands block movement. The typical analysis, as López (2009) summarized, is that the right-dislocated phrase undergoes a leftward movement and then followed by a remnant movement of the vP (Cecchetto 1999, Frascarelli 2000, Villalba 2000, Belletti 2004), as illustrated in (25).. (25) Steps for movement analysis for RD (López 2009: 102) (i) Jo els I. he. them have. llegit els llibres read. the books. 33.

(43) (ii) Jo [els llibres]i els he llegit ti. (object raising). (iii) Jo [els he llegit ti]k [els llibres]i tk. (VP raising). The right-dislocated object, els llibres ‘the books’, firstly undergoes raising. Then the whole vP, including the trace, moves to a higher position than the right-dislocated phrase. This analysis seems to be simple and plausible, but it may be challenged by some problems. First, in the discussion of RD in Korean, Chung (2009) mentions that Lee (2007) points out that this approach is criticized because, in some cases, it seems to violate Proper Binding Condition (PBC) (Fiengo 1977). Namely, the trace of the XP should not be c-commanded by the XP. The formal formulation of PBC is illustrated in (26).. (26) Proper Binding Condition (PBC) (Fiengo 1977) In surface structure Sα, if [e]NPn is not properly bound by […]NPn, then Sα is not grammatical.. Based on the definition of PBC, it suggests that a trace should be c-commanded. It is noteworthy that this violation is not problematic if we adopt Kayne’s (1994) assumption on remnant movement, namely, the remnant movement is a way to have the correct surface word order. Second, it does not specify the time when the resumptive pronoun appears and how it 34.

(44) agrees with its antecedent. This analysis does not discuss the connectivity. Precisely, this analysis cannot accommodate the fact such as case agreement, preposition stranding, reconstruction, and so on. I have argued that case agreement is another important properties to distinguish RD from AT besides island effects; however, the movement approach does not specify how the right-dislocated NP agrees with its resumptive pronoun. This is the major problem that movement analysis encounters.. 2.2.2 Base generation analysis In order to solve the problems that the movement analysis left, Zwart (2001) proposes a slightly revised version of movement analysis based on the observation of RD in Dutch, that is, the base generation analysis. He argues that the right-dislocated phrases are based-generated in a very high specifier position and the word order on the surface results from the leftward movement of the main clause, as illustrated in (27)10.. (27) (i) [dislocated XP [main clause]] (ii) [[main clause]i [dislocated XP ti. 10. ]]. I revise the formula from the original one in Zwart(2001) by changing the backgrounding into dislocated XP for the consistency in this study. It should also be noted that the RD in Dutch and German, they include backgrounding and afterthought (Zwart 2001, Averintseva-Klisch 2008, Ott and de Vries 2014). Backgrounding is more equivalent to the RD that is discussed in this study and hence the examples presented here are, in fact, backgrounding in these langauges. More discussion can refer to Zwart (2001), Averintseva-Klisch (2008), Ott and de Vries (2014) and papers cited therein. 35.

(45) If the main clause has movement, it is prior to the merge with the dislocated phrase. Thus, the violation of the island effects is only affected in the main clause and the ungrammaticality of the main clause will make it not able to merge with the dislocated phrase. The advantages of this approach are that it escapes from the problem of PBC violation by assuming the right-dislocated phrases are base-generated and that there is a weak pronoun in the main clause associating with the dislocated XP. Moreover, this analysis, as Zwart states, can capture the scope interaction of adverbs and numerals. When the adverb is right dislocated, the sentence can have two interpretation, exemplified with the Dutch sentence in (28). One interpretation is that the numeral, twee keer ‘two times’, scopes over the right-dislocated adverb, vermoedelijk ‘probably’, which corresponds to the order of the surface structure. The other interpretation is that the right-dislocated adverb scopes over the numeral, which is different from the surface order.. (28) Dutch (Zwart 2001) Hij heeft. twee. keer. he. two. times claimed that. has. beweerd dat. hij overleden. was, vermoedelijk. he. was probably. ‘He claimed twice that he had probably died’ ‘He probably claimed twice that he had died’. 36. deceased.

(46) Following Zwart’s assumption that the right-dislocated phrase is originally high above the main clause, one can explain the reason why the sentence has the reverse interpretation. In other words, the reverse interpretation comes from the dislocated phrase base-generated in the specifier position that can take scope over the main clause. After the movement of the main clause, the interpretation that corresponds to the surface order can be generated. Even though Zwart’s analysis is appealing, there are some details left to be explained. Importantly, he leaves many problems regarding island effects in this approach. He only solves the problem of connectivity sine he states that there is an association between the dislocated phrase and the pronoun in the main clause. However, this approach does not provide an explanation on island sensitivity. If the dislocated phrase is base-generated and the moved element is the main clause, then it is not clear why some extracted elements cause ungrammaticality due to the violation of islands. That is, as this approach suggests that the ungrammaticality of the main clause will block it from moving to the left. A problem appears in terms of associating the dislocated phrase and the resumptive pronoun in the main clause, namely, the timing for the association. Moreover, even though he states that the leftward movement works in the main clause, there does not seem to be a motivation that will trigger the movement within the main clause. If there is any reason to trigger the leftward movement, he has to explain the positon that the moved element in the main clause goes to. Also, as I will argue in section 3, this approach encounters some problems in Mandarin Chinese. 37.

(47) Particularly, Zwart’s assumption about the right-dislocated phrase situating in the highest position may be challenged by the sentence final particles which are analyzed in the highest position in Mandarin Chinese. In order to solve the problems that movement analysis and base generation analysis face, I will turn to another analysis in section 2.2.3.. 2.2.3 Ellipsis analysis In section 2.2.1, I present that the movement analysis for RD is proposed due to the clause-external properties that RD exhibits. However, as discussed in section 2.1, in addition to island sensitivity, RD also shows coreferentiality and agreement, which are categorized as clause-internal properties. These properties show some connectivity between the host clause and the right-dislocated phrase, posing some threats to the movement analysis. The basic idea of this alternative analysis for RD, ellipsis approach, is proposed in Ott and de Vries (2012, 2014, in press). The proposed formula is shown in (29).. (29) [CP1 ….correlate....] [CP2 dXPi [...ti....]]. (29) illustrates their analysis assuming that RD is a bi-clausal construction; that is, the right-dislocated phrases are inside a clause (CP2) which is different from the clause containing the resumptive pronoun (CP1). They argue that the connectivity exhibited in RD 38.

(48) favors the ellipsis approach because they have some evidence (e.g. short answer) to show that similar behaviors are not exclusive for RD but also for other constructions. Precisely, they build on Merchant’s (2004) idea of the analysis of ellipsis; that is, given two semantically equivalent coordinated clauses, the second clause can undergo ellipsis. The step-by-step derivation is illustrated in (30).. (30) Dutch (Ott & de Vries 2012) Joop heeft ze Joop has. gezien, die. al. them already. seen. nieuw. those new. tablet-pc’s. tablet-pc’s. ‘Joop saw them already, those new tablet PCs.’. (i). [Joop heeft. ze. Joop has. them already. [Joop heeft die Joop has. al. nieuwe. those new. gezien]: seen tablet-pc’s al. gezian] fronting. tablet-pc’s already. seen. (ii) [Joop heeft ze al gezien]: [die nieuwe tablet-pc’s heeft Joop al t gezian] PF-deletion (iii) [Joop heeft ze al gezien]: [die nieuwe tablet-pc’s heeft Joop al t gezian]. 39.

(49) I follow the representation used in Ott and de Vries. The coordination is marked by a colon because it is not overt on the surface and this marking is only for the illustration of the underlyingly representation. In the first step of the derivation (30i), there are two semantic equivalent clauses coordinated. One clause is with the resumptive pronoun, ze ‘them’, and the other is with the DP, die nieuwe tablet-pc’s ‘those new tablet PCs’. At this stage, the resumptive pronoun and the DP have already co-indexed and agreed. In step two (30ii), the DP is fronted by leftward A-bar movement in the second clause. The last step is PF-deletion (30iii), except the fronted DP, the rest of the elements in the second clause are deleted at PF. Ott and de Vries (2014) further specify that their approach is based on parallelism and clausal ellipsis. And they claim that this approach can capture the two conflicting characteristics of RD, namely, independence and connectivity. They use the term independence to illustrate that the right-dislocated phrase is not part of the host clause, which they view it as a piece of evidence of favoring the bi-clausal ellipsis approach. As for connectivity which, as they claim, strongly supports the ellipsis approach, this term describes a fact that there are some dependencies between the right-dislocated phrase and its resumptive pronoun. This approach can capture the fact of connectivity by claiming that the two clauses are semantically equivalent, which movement approach fails to account for. Moreover, this approach can predict the island sensitivity/locality effects of RD. In their proposal, Ott and de Vries assume that CP1 and CP2 are anaphoric juxtaposition and thus the second clause can 40.

(50) undergo ellipsis. It seems that ellipsis approach seems to be more plausible than the movement approach and the base generation approach since the problems that the two analyses encountered can be solved under the ellipsis approach. However, this approach is not flawless. Firstly, regarding the clause coordination or parallelism, it is not clear at which level that the two clauses coordinate or merge. It is possible that the second clause is coordinated after fronting if the fronting does not change the semantic meaning. It is also possible that RD is not a bi-clausal construction and does not need coordination at all. Moreover, referring back to the fundamental question, it is not specified how to illustrate this approach with the tree diagram; namely, what is the syntactic structure as a whole and what is the assumption of the operation for the clausal operation. Also, as Ott and de Vries (2014) claim, the two parallel sentences do not cause any syntactic problem in their approach since at PF the coordinated sentence will undergo deletion. Nevertheless, it is still doubtful whether the two almost identical sentences can be generated in the syntactic system without the PF deletion. The reason or motivation of making PF deletion obligatory needs to be explained. Another problem is that Ott and de Vries (2014) aim to provide a unified analysis for both RD and AT. As discussed in the previous section, it is seen that RD and AT do not have the same underlying representation and the differences between the two cannot be explained if a unified analysis is provided. In section 2.2, I review three approaches for RD, movement, base generation, and 41.

(51) ellipsis approach. It is seen that there are pros and cons in these three analyses respectively. In the next section, I will apply these three analyses to examine whether they can accommodate the RD in Mandarin Chinese and also to see whether the Mandarin Chinese data can shed some light on the previous analyses.. 42.

(52) Chapter 3 Right dislocations in Mandarin Chinese. 3.1 Facts of right dislocation in Mandarin Chinese In the previous section, I have discussed some characteristics of RD and one of them is that the NP appears in the sentence final position. The position that the NP originates from may be occupied by a resumptive pronoun. I also present some examples of RD in Mandarin Chinese in section 2.1.3 based on the criteria in section 2.1.2. In this section, I will elaborate more on the Chinese data. To begin with, Mandarin Chinese, in general, does not have overt case/agreement system, unlike Romance languages, Germanic languages, Japanese and Korean. In (31), the typical examples of RD in Mandarin Chinese also follow this general fact. In (31), all the NPs, Zhangsan, ta, Lisi, do not have an explicit case/agreement marker.. (31) Right dislocation with different types of SFP a. (Tai) kandao he see. Lisi *(le) Lisi. Zhangsani.. SFP Zhangsan. ‘He saw Lisi, Zhangsan.’ b. (Tai) kandao. Lisi. le. a. Zhangsani!. he see. Lisi. Asp SFP Zhangsan. ‘He saw Lisi, Zhangsan!’ 43.

(53) c. (Tai) kandao he see. Lisi le. ma Zhangsani?. Lisi Asp SFP Zhangsan. ‘Did he see Lisi, Zhangsan?’. Then, from the data, some properties of RD in Mandarin Chinese can be observed11. First, in general, it does not necessarily need an overt resumptive pronoun, which is often seen and always required in Romance and Germanic languages.12 In Mandarin Chinese, the pronoun ta can be omitted.13 It should be noted that there are some cases that a resumptive pronoun is prohibited, such as the dislocated NP is non-human and inanimate (Xu 2006). However, it cannot be denied the possibility of having a pro in the position of the resumptive pronoun. Second, as shown in (31), sentence final particles (henceforth, SFP) are obligatorily. 11. The example (31c) here is a yes-no question. The two facts also can apply to wh-questions; namely, resumptive pronoun is not obligatorily needed and a sentence final particle is required. The example (ii) below illustrates these properties. (ii) (Tai) mai-le shenme *(a) Zhangsani? He buy-ASP what SFP Zhangsan ‘What did he buy, Zhangsan?’. 12. There is a set of sentence that cannot be included in RD in the present study. In (iii), the sentence with wangqiu ‘tennis’ at the end of the sentence cannot have a resumptive pronoun. It seems that it is a counter example of my first observation. However, there are some interfering factors in this sentence. First, the time adverbial liangnian ‘two years’ can appear at both side of the NP, wangqiu. It is not clear that, in this case, it is the NP moves or just the adverbial adjoins at different levels. Second, the pronoun in Mandarin Chinese often refers to people. It is seldom used to refer to inanimate objects or animals. This may be a reason why it cannot have resumptive pronoun in this sentence. (iii) Zhangsan da-guo (*tai) lianglian wangqiui. Zhangsan hit-Asp it two year tennis ‘Zhangsan has played for two years, tennis.’. 13. Whether the pronoun in RD in Mandarin Chinese is a resumptive pronoun or not may need further investigation. Some native speakers’ judgment suggests that they treat the pronoun ta as other referent in the discourse. They do not associate the pronoun with the right-dislocated NP. This question is more related to the discourse function which is not the main focus in the present study. 44.

(54) required in RD (Cheung 2009), whereas Romance and Germanic languages, Japanese and Korean do not exhibit such a property.14 It should also be noted that the SFP not only is required in a question as in (31c) but also appears in a declarative sentence (31a) or an exclamative sentence (31b). Besides, one may doubt whether ‘le’ is a SFP from (31a), but it can be evidenced from (31b) where a true SFP ‘ma’ can occur with the ‘le’ that marks for aspect. It is just a fact that two kinds of ‘le’, one for aspect and the other is for SFP, cannot co-occur on the surface due to some phonological constraints or other reasons in Mandarin Chinese15. No matter what the real reason is, it is not within my discussion in the present research. In addition to the facts shown in (31), now let us consider some other facts of RD in Mandarin Chinese. As revealed from the examples in (32) and (33), the right-dislocated NP. 14. Consider the sentence in (iv). Some may think that the sentence poses a challenge to my observation. Nevertheless, there are some questions in this sentence. First, this sentence should have a pause between the time adverbial liangnian ‘two year’ and the NP. According to my definition, whenever there is a pause, it is AT not RD. Second, in order to make the sentence better and avoid the pause, adding a SFP ‘le’ is a possible solution. Thus, my argument still holds under such circumstances. (iv) a. Wo jiao-guo tai liangnian, Zhangsani. I teach-Asp him two year Zhangsan ‘I have taught him for two years, Zhangsan.’ b. Wo jiao-guo tai liangnian le Zhangsani. I teach-Asp him two year SFP Zhangsan ‘I have taught him for two years, Zhangsan.’. 15. A set of sentences in (v) can illustrate the contrast. In (a), two different functions ‘le’ that are adjacent to each other make the whole sentence ungrammatical. If the aspect ‘le’ is moved near to the verb as in (b), the sentence is improved though not perfect. (v) a. *Zhangsan kandao Lisi le le. Zhangsan see Lisi Asp SFP ‘Intended: Zhangsan saw Lisi.’ b. ?Zhangsan kandao le Lisi le. Zhangsan see Asp Lisi SFP ‘Intended: Zhangsan saw Lisi.’ 45.

(55) can originate either from the subject or object position in the host clause. When the resumptive pronoun appears in the subject position, it usually can be omitted. However, when it is in the object position, the omission of the pronoun may vary due to different types of verbs or its antecedent, i.e., co-indexed non-human dislocated phrase.16 As in (33), it is seen that the omission of the resumptive pronoun in object position is affected by different types of verbs. For psych verbs, xihuan ‘like’, in (33a), the resumptive pronoun can be omitted while in (33b), the resumptive pronoun cannot be omitted with the verb kandao ‘see’. This may suggests that the relation of the dislocated object is closer to verbs. Thus, the internal structure of the verb may influence whether the resumptive pronoun has to appear or not.. (32) Right dislocated subject a. (Tai) xihuan he. like. Lisi *(ma). Zhangsani?. Lisi SFP Zhangsan. ‘Does he like Lisi, Zhangsan?’. 16. It may be the fact that the third person singular pronoun in Mandarin Chinese, regardless of its gender, all sound the same, namely, ta ‘he/she/it’. Only the writing system differentiates the gender and human/ non-human of the pronoun. Thus, when speakers utter the sentence, it is more common to refer the pronoun to a human being and disregard its gender. It is ungrammatical or unacceptable to link the pronoun with the non-human nouns, as in (vi) and (vii). (vi) Zhangsan xihuan tai ma *maoi/?na zhi maoi? Zhangsan like it SFP cat that CL cat ‘Does Zhangsan like it, cats/that cat?’ (vii) Zhangsan xihuan tai ma *mieni/ *na wan mieni? Zhangsan like he SFP noodle that CL noodle ‘Does Zhangsan like it, noodles/ that bowl of noodle?’ 46.