t J. t. . . i \Ydfifiammu , :19. :8: I. ‘ , . . 5% ..V .L. ‘ 4:34. . I #51272? 1.. 31.1146. THE???» ‘J 070/53 59%?!223 This is to certify that the dissertation entitled On Chinese Temporal Clauses presented by Yu-fen Liou has been accepted towards fulfillment of the requirements for the Ph.D. degree in Lingistics Aid/Jr Major Professor's Signature 2/l$(03 Date MSU is an Affirmative Action/Equal Opportunity Institution PLACE IN RETURN BOX to remove this checkout from your record. TO AVOID FINES return on or before date due. MAY BE RECALLED with earlier due date if requested. DATE DUE DATE DUE DATE DUE MWMW 6/01 c:/CIRCIDateDuo.965-p. 15 ON CHINESE TEMPORAL CLAUSES By Yu-fen Liou A Dissertation Submitted to Michigan State University In partial fulfillment of the requirements for the degree of DOCTOR OF PHILOSOPHY Department of Linguistics and Germanic, Slavic, Asian and African Languages 2003 ABSTRACT ON CHINESE TEMPORAL CLAUSES By Yu-fen Liou This dissertation investigates three issues of temporal clauses in Mandarin Chinese: the syntax of de shihou ‘when’ clauses; the syntax of yi-qaz‘n/yi—hou ‘before/afier’ clauses, and the behavior of the preposition zai in post-subj ect positions when it takes a temporal clause. Temporal elements such as de-shihou ‘when’ and yi-qian/yi- hou ‘before/after’ have traditionally been analyzed as clause linking units or postpositions. In this thesis, I argue that shihou and qian/hou are in fact head nouns of noun phrases: de-shihou clauses are relative clauses while yi-qian/yi-hou ‘before/after’ clauses are noun complement clauses. Adopting Kayne's (1994) analysis of relative clauses, I argue that de is an enclitic determiner that takes CP as its complement and the adverbial element shihou raises to Spec CP. The remnant IP then raises to Spec DP to license de. In yi-qian/yi-hou clauses, I argue that qian and hou are temporal localizers which take a clausal complement. Since zhi is the classical form of de and zhi-qian/zhi-hou have the same fimction as yi—qian/yi-hou in temporal clauses, I argue that zhi and yi also head DPs with qian/hou as NP complements. The complement CP raises to Spec DP to license yi/zhi. The preposition zai usually occurs in a temporal clause in an adjunct position, and is optional, especially when the clause is sentence-initial. However omitting zai in a sentence like W0 *(zai) Lisi chusheng yi-qian zaojiu chuguo 1e ‘1 had gone abroad before Lisi was born’ where the temporal clause follows the main clause subject makes the sentence to be less acceptable. I argue that the omission of zai in sentence of this sort results in two DPs in a sequence, which leads to a garden path effect. When two DPs occur in the sentence initial position without zai, the two DPs can easily be interpreted as either a coordination or a topic-subj ect relationship. Inserting zai between the two NPs is preferred in order to avoid the garden path. Cepyright by Yu—fen Liou 2003 To My Parents TABLE OF CONTENTS CHPTER 1: INTRODUCTION 1.1 Introduction to de shihou and yi-qian/yi-hou clauses ......................................................... 2 1.2 Examining Chinese temporal clauses .......................................... 4 1.2.1 The status of de shihou clauses ................................... 5 1.2.2 The status of yi-qian/yi-hou clauses ............................... 7 1.2.3 The function of zai ................................................. 10 1.3 Problems to be investigated .................................................... 12 1.4 Organization ...................................................................... 13 CHAPTER 2: The Chinese DP 2.1 Introduction to de phrases ...................................................... 16 2.2 De as a complementizer ......................................................... 18 2.2.1 De as a head final complementizer .............................. 19 2.2.2 De as a head initial complementizer ............................. 24 2.3 De as a syntactic category ...................................................... 34 2.4 De as a determiner ............................................................... 37 2.4.1 Simpson’s analysis ................................................. 38 2.4.2 Lin, Murasugi and Saito’s analysis .............................. 47 2.5 Discussing the three analyses ................................................... 56 2.6 De phrase with a complement clause .......................................... 66 CHPTER3: De shihou clauses 3.1 Introduction ....................................................................... 69 3.2 Previous studies of de shihou clauses ......................................... 71 3.3 Problems with Gasde and Paul’s analysis .................................... 73 3.4 The position of demonstratives and quantifiers in Chinese relative clauses ...................................................... 81 3.5 Introduction to Chinese adjunct relative clauses ............................. 84 3.6 Movement effects in de shihou clauses ....................................... 89 3.7 The structure of de shihou clauses ............................................. 95 3.8 Conclusion ........................................................................ 97 CHAPTER 4: Yi—qian/yi-hou clauses 4.1 Introduction ....................................................................... 99 4.2 Previous analyses for the category yi-qian/yi-hou ........................... 102 4.2.1 Yi-qian/yi-hou as postpositions ................................... 103 4.2.2 McCawley’s analysis .............................................. 108 4.2.3 Problems with G & P’s and McCawley’s analyses ............ 112 4.3 Localizers ......................................................................... 1 18 4.3.1 Introduction ......................................................... 1 18 4.3.2 Localizers as postpositions ........................................ 125 vi 4. 3. 3 The noun analysis .................................................. 128 4. 3. 4 Analysis of the short form localizers as clitics .................. 131 4.3.5 Problems of Ernst’ s analysis ...................................... 133 4.4 The analysis ....................................................................... 141 4.4.1 Short form localizers are nouns that carry clitic properties. . .142 4.4.2 The structure of NP + Localizer .................................. 149 4.5 Temporal clauses and their structure .......................................... 159 4.6 Comparing yi-qian/yi-hou with before/after ................................. 165 4.7 Revising the structure of yi-qian/yi-hou clauses ............................. 175 4.8 Conclusion ........................................................................ 181 CHAPTER 5: The optionality of zai in temporal clauses 5.1 The problems ..................................................................... 184 5.2 Larson and Stroik ................................................................ 187 5.3 The optionality of zai in locative phrases .................................... 190 5.4 Zai in yi-qian/yi-hou phrases ................................................... 197 5.4.1 Zai is not syntactically required in temporal phrases .......... 198 5.4.2 Constructions starting with two noun phrases .................. 207 5.4.2.1 Coordination ............................................... 207 5.4.2.2 Topic-subject relationship ................................ 211 5.5 Explaining the problems ........................................................ 216 5.6 Zai in de shihou clauses ......................................................... 230 5.7 Conclusion ........................................................................ 233 CHAPTER 6: Conclusion .................................................................... 235 References ...................................................................................... 237 vii Chapter 1: Introduction Temporal clauses are clauses that contain time expressions. The following sentences are typical temporal clauses in English. (1) a. John arrived when Bill left. b. John arrived before Bill was fired. 0. John departed afier Bill left. The research of temporal clauses in English has focused mainly on their interpretations instead of their structures. One important study about English temporal clause structure dates back to Geis (1970). Geis (1970) argues that English temporal clauses are restrictive relative clauses in that sentences like (1 a, b, and c) are derived from the following sentences respectively. (2) a. John arrived at the time when Bill left. b. John arrived before the time at which Bill was fired. c. John departed after the time at which Bill left. That is, each sentence in (1) has a counterpart, which is shown in (2), that contains a noun phrase as the temporal clause’s antecedent. Afier Geis, basically, English temporal clauses have been analyzed similarly to relative clauses or clauses with movement of a null temporal operator within them. Temporal clauses in Chinese, however, have not been extensively studied. Some examples of Chinese temporal clauses are given in (3)1. (3) a. Wogzai) chufa de shihou kandao 1e taz. I ZAI depart DE time see LE him ‘I saw him when I departed.’ b. Wo (zai) huijia xi-gian/n'-hou dei zuo gongke. I ZAI return-home before/afier must do homework ‘I have to do homework before/ after going home.’ As we can see in (3), Chinese temporal clauses are different from their English counterparts. In what follows, I will introduce Chinese temporal clauses by comparing them with their English counterparts. I will first discuss the differences between de shihou ‘DE time’ clauses and when-clauses, then discuss the differences between yi- qian/yi-hou ‘before/after’ clauses and before/after clauses. 1.1 Introduction to de shihou and yi-qian/j’i-hou clauses Consider the following de shihou clauses. (4) a. Wo (zai) chufa de shihou kandao 1e ta. I ZAI depart DE time saw LE him ‘I saw him when I departed.’ ' Chinese adverbs such as, manners, locative adverbials and temporal adverbials usually occur after the subject and before the main verb phrase, as in (1) (Li & Thompson 1981, Ernst 1994). Some temporal bare noun phrases can also occur before the subject, as in (ii). (i) a. Wo mingtian dei shangxue. (ii). Mingtian wo dei shangxue. I tomorrow must go-to-school tomorrow I must go-to-school ‘I have to go to school tomorrow,’ ‘I have to go to school tomorrow.’ 2 Although zai is usually translated as ‘at’, in this study, I will gloss it as ZAI , since it does not have any meaning. I will discuss zai in section 1.2.3. b. Wo (zai) chufa de nei-ge shihou kandao 1e ta. 1 ZAI depart DE that-CL time saw LE him ‘I saw him at the time/moment when I departed.’ c. (Zai) chufa de shihou wo kandao le ta. ZAI depart DE time I see LE him ‘I saw him when I departed.’ An English when clause is usually argued to correspond to the de shihou clause in Chinese (Chao 1968, Li and Thompson 1981, Gasde and Paul 1996). De shihou occurs at the end of the temporal clause, opposite to the position of when in English. Furthermore, de shihou, unlike a single element, can be separated by a demonstrative, as shown in (4b) in that nei-ge ‘that-CL’ is between de and shihou. However, in English it is not possible for a single element when to be separated by a demonstrative’. Now consider Chinese before/after clauses, as shown in (5). (5) a. Wo (zai) huijia yi-gian/fl-hou dei zuo gongke. I ZAI retum-home before/after must do homework ‘I have/had to do homework before/ after going home.’ b. (Zai) huijia y_i-gian/yi-hou wo dei zuo gongke. ZAI retum-home before/after I must do homework ‘Before/afier going home, I have/had to do homework.’ 3 Another element functioning similar to zai is dang. Dang usually occurs with de shihou clauses in temporal clauses; however when a de shihou clause occurs with dang, it cannot occur in post-subject position as shown in the following: (i) a. (Dang) ta lai de shihou wo zheng-zai chifan DANG he come DE time I right-ZAI eat—meal ‘When he came I was eating.’ b. *ta . (Dang) lai de shihou wo bu zai jia he DANG come DE time I not ZAI home Intended to mean: ‘When he came I was not home.’ In (5), we can see that Chinese yi-qian/yi—hou clauses differ from their English counterparts in that before/after in English precedes their clause going home, but the Chinese yi-qian ‘before’ and yi-hou ‘afier’ follow their clauses huz'jia ‘return home’. Also, like de shihou clauses, they both optionally allow zai to start the temporal clause. Also the optionality of zai does not change the meaning of the sentence. However, English temporal clauses do not have this kind of optional element that does not indicate any meaning. 1.2. Examining Chinese temporal clauses Having discussed differences between the English and Chinese temporal clauses, it is very important to understand if de shihou and yi-qian/yi-hou are the same element as their English counterparts, when and before/after. That is, are de shihou and yi—qian/yi- hou single elements like when and before/after or are they different from their English counterparts? This question leads to the major goal of this study, namely to determine the syntactic status of de shihou, yi-qian, and yi-hou and investigate the syntactic structure of de shihou and yi-qian/yi-hou clauses. Moreover, since both de shihou and yi- qian/yi-hou clauses involve the occurrence of zai, it is also important to investigate how zai interacts with temporal clauses. In what follows, I will lay out some issues regarding the structure of de shihou and yi-qian/yi-hou clauses, as well as some issues about how zai interacts with temporal clauses. 1.2.1. The status of rte-shihou clauses In this section, I will briefly introduce the meaning of shihou as in de shihou, and then discuss de shihou’s status. Shihou in Chinese, depending on context, can mean ‘time’, ‘moment’, ‘during’, ‘while’, and ‘when’, as shown in the following examples: (6) 3. Ni shenme shihou lai de? you what time come DE ‘What time did you come?’ b. zai nei-ge shihou ZAI that-CL time ‘at that moment’ c. wanfan de shihou dinner DE time ‘during dinner’ (1. Ta zai chifan de shihou, wo zai kanshu ne. he ZAI eat-meal DE time I ZAI read-book NE ‘While he was eating, I was reading.’ e. Ta dao-le de shihou wo zheng zai chi-fan ne. he arrive-LE DE time I right ZAI eat-meal NE ‘When he arrived I was eating.’ In this thesis, I will focus on the combination of de and shihou since it has been treated as a single element that introduces a temporal clause, (Chao 1968, Li and Thompson 1981, Gasde and Paul 1996), and it is only when shihou goes with de, that it can introduce a temporal clause. De shihou traditionally is treated as a single element. Li and Thompson (1981) consider it a conjunction, while Gasde and Paul (1996) treat it as a postposition. Now let us look at when-clauses in Chinese again. (7) a. Ta lai de shihou wo zai chifan. he come DE time IZAI eat-meal ‘When he came I was eating.’ b. Ta lai de nei-ge shihou wo bu zai jia. he come DE that-CL time I not ZAI home ‘The moment he came I was not home.’ De shihou clauses especially those like (7b) look very much like relative clauses. A Chinese relative clause (RC) requires de to bridge the clause to its left and the head noun to its right. De is an element that does not have any semantic content. In (7), the de shihou clause, like an RC, also contains a clause, de, and shihou in which shihou is likely the head noun. A typical Chinese RC can be shown as in (8). (8) a. chuan hong yi-fu de ren wear red clothes DE person ‘the person who wears red clothes’ b. chuan hong yi-fii de na-ge ren wear red clothes DE that-CL person ‘that person who wears red clothes’ As (8a) shows, the relative clause consists of a clause, chuan hang yi-fu ‘wear read clothes’, de, and a head noun ren ‘person’. (8b) shows that between de and the head noun ren ‘person’ a demonstrative nei—ge ‘that-CL’ can be inserted. Further evidence that a de shihou clause is like a relative clause is that shihou can be modified by the demonstrative na-ge ‘that-CL’ as (70) shows. Moreover, a de shihou clause behaves similarly to other adjunct RCs in that adjunct RCs do not allow demonstratives to occur in pre-RC positions, as shown below: (9) a. *Na-ge ta lai de shihou wo bu zai jia That-CL he come DE time I not ZAI home ‘That time when he came I was not home.’ b. Ni xiu che de (na-ge) wanshang. you fix car DE that-CL evening ‘the evening when you fixed the car’ c. *Na—ge ni xiu che de wanshang. that-CL you fix car DE evening ‘that evening when you fixed the car’ (9a) is a de shihou clause that has its demonstrative appearing at the beginning of the clause. (9b) is an adjunct relative clause with a demonstrative, and moving the demonstrative to the beginning of the sentence, we get (9c). As we can see from (9a) and (9c), both de shihou clauses and the adjunct relative clauses do not allow pre-RC demonstratives. 1.2.2 The status of yi-qian/yi-hou clauses In this section, I will discuss other elements that can also correspond to before/after in Chinese, zhi-qian/zhi-hou and qian/hou. Furthermore, I discuss the syntactic status of yi-qian/yi-hou. In addition to yi-qian/yi-hou, other Chinese elements that correspond to before/after are zhi-qian/zhi-hou and qian/hou, as shown in the following sentences. (10) a. Wo (zai) huijia yi—gian/yi-hou dei zuo gongke. I ZAI return-home before/after must do homework ‘I have/had to do homework before/after going home.’ b. Wo (zai) huijia zhi—qianlzhi-hou dei zuo gongke. I ZAI return-home before/after must do homework ‘I have/had to do homework before /after going home.’ c. Wo (zai) huijia gian/hou dei zuo gongke. I ZAI return-home before/after must do homework ‘I have/had to do homework before/after going home.’ As we can see from (10), the meaning and fimction of yi-qian/yi-hou, zhi-qian/zhi-hou, and qian/hou are the same in temporal clauses. Nevertheless, the categorial status of yi- qian/yi-hou, zhi-qian/zhi-hou, and qian/hou is not clear. Let us focus on yi-qian/yi—hou for the time being and examine (5), repeated here as (11). (11) a. Wo (zai) huijia yi-gian/yi-hou dei zuo gongke. I ZAI retum-home before/after must do homework ‘I have/had to do homework before/ after going home.’ b. (Zai) huijia yi-gian lyi-hou wo dei zuo gongke. ZAI retum-home before/after I must do homework ‘fifore/ajter going home, I have/had to do homework.’ (11) shows that yi-qian and yi-hou correspond to English before and after respectively, and introduce embedded clauses as before and after do in English. However, yi-qian/yi- hou follow the clauses that they introduce instead of preceding the clauses as their English counterparts do. This implies that, instead of being prepositions, they may be postpositions as argued in Gasde & Paul (1996). However, yi-qian/yi-hou ‘before/afier’ phrases also behave like noun phrases. One argument made by McCawely (1992) is that these phrases can occur after ba as shown in (12)4. (12) Zhangsan ba wanfan vi-hou/vi-qiflkanzuo zuihao de xiuxi shijian. Zhangsan BA dinner after/before regard best DE rest time ‘He regards after/before dinner as the best rest time.’ (McCawely, 1992, 231) (12) shows that wanfan yi-qian/yi-hou ‘before/after dinner’ can occur after ba, and should be noun phrases, thus, yi-qian and yi-hou are very likely to be nouns. According to what I have discussed so far, the syntactic category of yi-qian and yi-hou is not certain; they may be postpositions or they may be nouns. However, if they are postpositions, why is it possible for them to occur after ba? Also, it is well known that zhi as in zhi-qian ‘before’ is de’s ( de as in de shihou) classical form (Xu 1996, Simpson 1999, 2000), and yi-qian/yi-hou, zhi-qian/zhi-hou, and qian/hou have ’ According to Li (2001), ha was a verb in classical Chinese which had the meaning of take, hold, and handle. It has undergone grammaticalization, and is not used with these meanings anymore. A ba construction usually has a construction like (i) and (ii), and the NP afier ha is the object of the verb. (i) NP+ba+NP+V+X (ii) NP+ba+NP+x+V According to Li, the X is anything that is non-null that can be an adverb or a PP. An example of this construction is as follows. (iii) Wo ba shuiguo chi-wan le I BA fruit eat-finish LE ‘1 at up the fruits.’ the same meaning and function as shown in (10); therefore, it is very likely that yi, zhi, and de are the same kind of element, and thus Chinese before/after clauses are very likely to be noun phrases. 1.2.3. The function of zai Zai is an element that occurs with both locative phrases as well as temporal phrases. According to Li and Thompson (1974), Li (1990), Ernst (1988), McCawley (1992), and Liu (1998), zai is a preposition. Zai does not have semantic content as the following sentences show. (13) a. Shu zai zhuozi shang book ZAI table top ‘The book is on the table.’ b. Shu zai xiangzi li book ZAI box inside ‘The book is in the box.’ c. Shu zai beizi qian-mian book ZAI cup front-side ‘The book is in front of the cup.’ From (13), we can see that zai does not contribute any meaning in any of the sentences. The meaning of on, in, front, seems to be from the localizers, i.e., words such as shang ‘top’ [1' ‘inside’, and qian-mian ‘front-side’ in (13), instead of from zai. Moreover, zai does not seem to be an obligatory element in temporal clauses as (4), (5) and (11) show. However, the following sentences show that it is sometimes obligatory. 10 (14) a. Zhangsan *(zari) Lisi chusheng vi-gian jiu chuguo qu 1e ZAI was-born before then go-abroad go LE Intended to mean: ‘Before Lisi was born, Zhangsan had gone abroad.’ b. (Zai) Lisi chusheng fl-gian, Zhangsan jiu chuguo qu le. was-bom before then go-abroad go LE ‘Before Lisi was born, Zhangsan had gone abroad.’ What is it that makes zai obligatory in (14a)? That is, why is zai obligatory when the matrix subject (Zhangsan) in (14a) is at the beginning of the sentence, but optional when the matrix subject is not at the beginning of the sentence, as shown in (14b)? Moreover, it is not always the case that when the matrix subject precedes the embedded clause, zai becomes obligatory, as we have seen in (4a) and (5a), repeated here as (15). (15) a. Wo (zai) chufa de shihou kandaole ta. I ZAI depart DE time saw LE him ‘I saw him when I dgparted.’ b. Wo (zai) huijia yi-gian/yi-hou dei zuo gongke. I ZAI retum—home before/after must do homework ‘I have to do homework before! after going home.’ Examine (14) and (15). We can see that the difference between them is that (15a) does not have a subject in the embedded clause, but (14a) has a subject in the embedded clause. This suggests that when the embedded subject and the matrix subject are in a sequence, zai is obligatory. However the following sentence shows that this is not the case. 11 ( 16) W3 (zai) gongke zuohao yihou jiu huijia. I at homework finish after then return-home ‘I will go home after I finish the homework.’ (16) has a subject, gongke ‘homework’, in the embedded clause, and it follows the matrix subject, but zai is not obligatory. Concluding from this section, zai does not contribute any crucial meaning to the sentence, and that may be the reason why it is optional. Since it is optional most of the time, its existence cannot be due to syntactic constraints. If this is the case, the presence of zai in (14a) should be for a different purpose. 1.3. Problems to be investigated Having discussed some of the basic data concerning Chinese temporal clauses, there are three main points to be investigated: the structure of de shihou clauses, the structure of yi-qian/yi-hou clauses, and the inconsistent optionality of zai . I will pursue the hypothesis that Chinese temporal clauses are all noun phrases. This can be illustrated by the following: first, de shihou clauses are very similar to - adjunct relative clauses in that they both involve the use of de and both have the same effect when encountering demonstratives. In Chinese before/after clauses, zhi in zhi- qian/zhi-hou is believed to be the classical form of de. Furthermore, phrases with yi- qian/yi-hou behave like noun phrases as McCawley argues. Moreover, the clauses introduced by de shihou and yi-qian/yi-hou precede them, which is similar to relative clauses. 12 Finally, the inconsistent optionality of zai seems to indicate that the optionality of zai is not due to syntactic constraints, but to something else. With this as a background, this study will be based on the framework of Principles and Parameters (Chomsky 1986b, 1995), and will pursue the following issues: 1 Is de shihou a single element? That is, can it be separated into smaller units? What is the structure of a de shihou clause? 2. Are yi-qian/yi-hou each a single element? That is, can they be separated into smaller units? What is the structure of a yi-qian/yi-hou clause? 3. What is the firnction of zai in Chinese temporal clauses? What are the constraints on its occurrence? 1.4. Organization In this study, I argue that Chinese temporal clauses, de shihou clauses and yi- qian/yi—hou clauses, are syntactically similar to clauses involving phrases with de (an element that introduces relative clauses and noun complement clauses). In addition, I discuss the optionality of zai in temporal clauses. I argue that the optionality of zai in temporal clauses in the post-subject position is not due to syntactic constraints. The rest of my dissertation is organized as follows: In Chapter 2, I review the literature regarding Chinese relative clauses. Since yi- qian/yi-hou clauses and de shihou clauses are similar to relative clauses, examining the structure of relative clauses is crucial. Two main analyses will be examined. One analysis takes de in a relative clause as a complementizer like the English that in English 13 relative clauses. The other analysis takes de as a determiner like the English the; however, de has lost its definiteness and cannot firnction like the English the. The de as a determiner analysis will be chosen partially because its classical form zhi was a demonstrative in classical Chinese. According to Simpson (2000), demonstratives often develop into determiners and it is very possible that de has lost its determiner properties and does not firrrction like a normal determiner. The other reason to take de as a determiner is that de does not function like a complementizer that connects clauses like that does. Finally, evidence from Lin, Murasugi, and Saito (2001) shows that de is a determiner since it firnctions similarly to English ‘3, that it can license noun phrase i deletion, such as this is John ’s. In Chapter 3, I examine the properties of de shihou clauses. Gasde and Paul (1996) argue that de shihou is a postposition and shihou is not a normal noun. However, I will show that Gasde and Paul’s argument cannot be right, and de shihou clauses behave like other time relative clauses as Ning (1993) describes. Chapter 4 deals with yi—qian/yi-hou clauses. Yi-qian/yi-hou can be substituted for by zhi-qian/zhi-hou, and qian/hou in temporal clauses. Qian/hou are also localizers that indicate locations of objects. Localizers have been studied more than temporal clauses. Therefore, it is necessary to examine localizers, and I will argue that qian/hou are temporal localizers that still possess nominal properties. Furthermore, this chapter argues that zhi and yi in temporal clauses are functional elements similar to de in relative clauses/noun complement clauses whose structural positions are the same. As a result, the construction of yi-qian/yi-hou clauses shows a similar structure to the relative clauses/noun complement clauses; the yi-qian/yi-hou clauses contain a clause, a 14 functional element de, which is yi or zhi, and a noun which is either qian or hou. Nevertheless, I will argue that they are in fact noun complement clauses since qian/hou do not undergo movement out of the temporal clause preceding them, assuming Simpson’s (2000) analysis of de phrases. In chapter 5, I discuss the syntactic behavior of zai in locative phrases and temporal phrases, specifically in yi—qian/yi-hou phrases. I argue that zai is necessary with locative phrases, but is optional with temporal phrases. Since zai is optional in preposed temporal clauses, the obligatory presence of zai in post-subj ect positions is to avoid a temporary parsing ambiguity caused by two noun phrases in a sequence. 15 Chapter Two: The Chinese DP As discussed in the previous chapter, Chinese temporal clauses are similar to noun phrases with de’. In this chapter, I discuss the structure of Chinese noun phrases with de (hence forth, de phrases), which are found in both relative clauses and complement clauses. In the literature, de has mainly been analyzed as either a complementizer or a determiner. In this chapter, I will argue in support of the determiner analysis put forth by Simpson (1999, 2000). The chapter is organized into six sections. In 2.1, I introduce Chinese de phrases and their relevant literature. In 2.2, I review the analyses that take de as a complementizer (Cheng 1986, Ning 1993, Xu 1997). In 2.3, I discuss Ning’s (1996) analysis of de as its own syntactic category. In 2.4, I review Simpson’s (1999, 2000) and Lin, Murasugi, and Saito’s (2001) determiner analyses. Section 2.5 is a discussion of the previous analyses and an overview of my reasons for supporting Simpson’s analysis. Finally, in 2.6 I focus on de phrases that contain a complement to the left of the head noun. 2.1 Introduction to de phrases The head noun of a Chinese noun phrase always occurs at the end of the phrase, as shown in (1). ' Although some linguists view this de the same as the adverbial de as in pao de kuai ‘run DE quick (run quickly)’, it is argued convincingly among many linguists that these DEs are different. (Huang 1982, Tang 1990) 16 "‘l (1) a. zhe-ben shu this-CL book ‘this book’ b. wo de shu c. wo de na-ben shu I DE book I DE that-CL book ‘my book(s)’ ‘the book which is mine’ d. piaoliang de hua e. wo mai de shu pretty DE flower I buy DE book ‘pretty flowers.’ ‘the book(s) that I bought’ ‘the flowers that are pretty f. ta mei chu-guo de shishi he not go-abroad DE fact ‘the fact that he did not go abroad’ (1a) shows that both the demonstrative and the classifier must precede the head noun. Examples (1b) to (1e) show that de is required for any phrase that relates to the head noun and has the structure of XP + de + NP. (lb) shows that de is like a genitive marker. Although (1c) has ahnost the same structure as (1b), from its English gloss, de seems to be like a relative clause marker. (1d) shows that de is like an adjective modifier marker. (1e) shows that de is like a relative clause marker, and (It) shows that de is like a complement marker of the head noun. (1b), (1c), (1d) and (1 f) do not show any gap in the XP position. (1c) shows that the XP contains a gap afier mai ‘buy’. In this chapter, I will focus on noun phrases with de instead of noun phrases with demonstratives, numbers, and classifiers as in (1a). Examples (1b) to (11) seem to show that de has different functions. Li and Thompson (1981) suggest that tie is a genitive marker, an associative marker, and a nominalizerz. However, most linguists who 2 According to Li and Thompson, when de occurs in between two nouns, the meaning of the whole phrase is determined by the association of the two nouns. l7 study Chinese agree that these des are the same de (Huang 1982, Ross 1982, Cheng 1986, Tang 1990, Ning 1996, Xu 1997, Simpson 1999, 2000) and the XP that precedes de may or may not contain a gap. That is, if the XP contains a gap, it is a relative clause, while if the XP does not contain a gap, the XP is likely a noun complement clause. There are three kinds of analyses of de with respect to its syntactic category. One takes de as a complementizer generated in C0 (Huang 1982, Cheng 1986, Ning 1993, Xu 1997), one takes de itself as a syntactic category that projects a de phrase (DeP) (N ing "' .l_'- 1996), and the third takes de as a determiner (Simpson 1999, 2000, Lin, Murasugi and Saito 2001). In the following sections, I will review these three analyses, which focus on de phrases containing gaps (i.e., relative clauses), and discuss their strengths and weaknesses. I will then give my reasons to support Simpson’s analysis. Finally, I will discuss the gapless CP that precedes de, i.e., noun complement clauses. In section 2.2, I will review complementizer analyses of Cheng (1986), Ning (1993) and Xu (1997). 2.2 De as a complementizer De phrases usually can be translated as relative clauses in English, as examples (1c), (1d) and (1e) above show. Even the English version of (l 0 contains a complementizer. Many de phrases do have the properties of relative clauses since the phrase preceding de contains a gap. The three analyses I will discuss in this section view all de phrases as having the same syntactic configuration as a relative clause. Because of this, de is usually taken to be a complementizer, such as the word that in English. Cheng (1986) and Ning (1993), following Huang (1982), argue that de is a complementizer, and 18 further claim that it is a head-final element, while Xu (1997), following Kayne (1994), argues that de is a head-initial complementizer and undergoes overt C0 to Do movement. In the following section, I will first review Cheng’s and Ning’s head-final complementizer approaches, and then discuss Xu’s head—initial complementizer approach. 2.2.1 De as a head-final complementizer Assuming X-bar theory and following Huang’s (1982) claim that Chinese nouns are head-final and de is a complementizer, Cheng (1986) argues that Chinese de is a head-final complementizer. Her analysis is as follows: (2) NP That is, de is a head-fmal complementizer that takes an XP to its left. The CP acts as the modifier of the head noun. Based on the structure in (2), she suggests that the de phrases in (3) have the structures in (4). 19 (3) a. Zhangsan de fangzi DE house b. kuaile de ren happy DE person ‘Zhangsan’s house’ ‘a happy person’ c. Ta dui wo de taidu hen hao. he to me DE attitude very good ‘His attitude towards me is very good.’ d. Ni xihuan de ren hen chengshi. you like DE person very honest ‘The person that you like is very honest.’ (4) a. NP 1). NP /\ /\ CP N’ CP N’ l | l l C’ fangzi ‘house’ C’ ren ‘person’ /\ /\ NP C AP C | l | l N’ de ‘DE’ A’ de ‘DE’ | | Zhangsan kuaile ‘happy’ c. NP (1. NP /\ /\ CP N’ CP 1\ ’ C’ taidu ‘attitude’ Op, C’ /\ /\ PP C [P C /\ I /\ N’ P’ de ‘DE’ NP 1’ I /\ /\ ta P NP I VP ‘he’ I l | ren dui N wo V’ 'person’ ctos l ‘1, /\~ wo ‘me’ V N l l xihuan ‘like’ e, de ‘DE’ 20 According to Cheng, there is no syntactic restriction on the complement that de selects. In (3a), it takes a noun phrase as its complement; in (3b), it takes an AP. In (3c), it takes a PP as its complement, and in (3d), it takes a clause. Among the structures in (4), the whole CP modifies the head noun. Cheng also emphasizes that de is a case marker, so Zhangsan in (3a) gets case from de, while the head of the whole phrase gets case from an external case marker. She concludes that Chinese not only has a head-final head noun, but it also has a head-final complementizer. Along the line of Huang and Cheng, Ning (1993) also takes de as the head of a CP and further elaborates the properties of de in a relative clause construction. He argues that there are three kinds of de phrases: possessive de, complement de, and relative clause de. All three are illustrated below in (5) respectively. (5) a. Zhangsan de shu DE book ‘Zhangsan’s book’ b. Ta mei lai de shishi he not come DE fact ‘the fact that he did not come’ c. Wo xihuan de shu I like DE book ‘the book that I like’ He argues that de in (5a) is a case marker. The de phrase in (5b) is a complement construction with de, which does not involve an operator movement. (5c) is a relative clause and in this case, de must select a clause with a gap as its complement. 21 Ning, following Rizzi (1990), assumes that there are four kinds of Co, as shown in (6), and each type attracts different kinds of elements to its Spec position. (6) a. [+wh, -pred] b. [+wh, +pred] c. [-wh, +pred] d [-wh, -pred] A C0 with a [+wh] feature attracts a wh-element to its Spec position for Spec-head agreement. A [-wh] feature in C0 does not allow a wh-element to occur in its Spec position. A C0 with a [+pred] feature heads a CP which contains a predicate. A {-pred] feature of C0 heads a CP which contains a proposition. Based on these assumptions about C0, Ning analyzes de in relative clauses as [-wh, + pred], and in noun complement clauses as [-wh, -pred]. Ning’s analysis for de in relative clauses is that de is a functional head which occupies the C0 position with the features [-wh, +pred]. Since the C0 position is [-wh, +pred], de does not require a wh-element to occur at its Spec position; however, an operator does occur in its Spec. According to Ning, ‘the semantic contribution of de as a functional head is to turn a proposition into a one-place predicate (p. 66).’ De requires the one-place predicate to be its complement which contains a variable gap, and thus an operator must be introduced to have an operator-variable construction, as shown in (7). (7) CP /\ Opt C’ /\ [P C0 A | [. . .ti. . .] de 22 (7) shows that de is generated in CO and selects an [P as its complement, which involves operator movement. A relative clause as in (8a) has the structure shown in (8b). (8) a. ta xihuan de shu he like DE book ‘the book he likes’ b NP /\ CP NP /\ 0P1 C’ /\ IP c0 A l [CP i [c1 [ta xihuan tij de]] shu, ‘he like’ ‘DE’ ‘book’ (8) shows that de heads a CP with a null operator in its specifier, and the operator is coindexed with the gap in the IP and the NP to form a chain at LF. As for a noun complement clause, (5b), repeated here as (9a), it has the structure shown in (9b). (9) a. Ta mei lai de shishi he not come DE fact ‘the fact that he did not come’ 23 b. NP /\ CP NP /\ IP C0 A I [ta mei lai], dei shishi, ‘he not come’ ‘DE’ ‘fact’ According to Ning, the C0 in a noun complement clause has [-wh, -pred] features, and thus CO in (9) selects a proposition; however, the C0 does not project a Spec position. Therefore, there is no operator variable construction involved, and the C0 is coindexed with the NP. In summary, Ning claims that de is a genitive case marker and also follows Huang (1982) and Cheng (1986) in assuming that de in relative clauses and noun complement clauses is in Co, which is a head-final element. He argues, supporting Rizzi (1990), that cross-linguistically there are the four kinds of C0 listed in (6). De in a relative clause construction has the features [-wh, +pred], and it selects an IP with a gap as its complement. Thus, it has a full projection and has a Spec position. However, de in a noun complement clause has the features [-wh, -pred], selects a proposition as its complement, and does not undergo full projection to have a Spec position. 2.2.2 Be as a head-initial complementizer Xu (1997), following Chomsky’s (1993) assumption that all languages possess the same set of functional categories, argues that de phrases are in fact CPs with de as their complementizer. 24 He suggests that des should not be considered as different markers, such as genitive markers, relative clauses markers, or markers for adjectives. He emphasizes that these des should all be considered the same. Based on Kayne’s (1994) Linear Correspondence Axiom and his relative clauses analysis, Xu proposes that all de phrases can be analyzed as relative clauses and that de is a head-initial element generated in Comp. Before turning to Xu’s analysis, we shall look at Kayne’s analysis of a relative clause. Kayne (1994) proposes the Linear Correspondence Axiom (LCA) which enforces a strict relationship between linear order and hierarchical structure: for any two non- terminal nodes X and Y, if X asymmetrically c-commands Y, then the terminals dominated by X must precede the terminals dominated by Y. Asymmetric c-command is defined as follows. (10) X asymmetrically c-commands Y iff X c-commands Y and Y does not c-command X. (p. 4) Under the LCA, word order is strictly determined by the hierarchy of phrase structure. A head must asymmetrically c—command its complement. Therefore, the head’s terminal must precede the complement’s terminals as shown in (11). 25 (11) VP V 9 /\ V0 NP l l S66 NO I John In (11), the head V0 asymmetrically c-commands N0, and Vo’s terminal see precedes NO’s terminal John. Since the hierarchical structure of a phrase determines the linear order of the terminals, one consequence of the LCA is that right-adjunction is disallowed. The reason is that an XP right-adjoined to a YP will asymmetrically c-command all of YP’s terminals, which means that the XP’s terminals will precede the YP’s terminals, and thus result in an incorrect word order. This can be illustrated in (12)3 . (12) YP 3 According to Kayne, the restriction for c-command is as follows: ‘X c-commands Y iff X and Y are categories and X excludes Y and every category that dominates X dominates Y’. ( Kayne 1994, 16) When an )0) adjoins to a YP, the two YPs are segments of one category, namely YP. The two YPs are not in a c-command relation since c-command is restricted to categories rather than segments. In (12), XP asymmetrically c-commands all categories that are dominated by the lower YP, and thus, following LCA, XP’s terminals should precede YP’s terminals. 26 A consequence of the LCA is that the standard analysis of relative clauses cannot be correct. The standard analysis analyzes a relative clause, as shown in (13), as an adjunction to the head noun (Chomsky 1977, Safir 1986, Browning 1991) as in (13b). (13) a. the claim that he made b. [01) the [Np [Np claim] [cp Opi that he made ti]]] If the relative clause is an adjunction to the head noun, as the standard analysis claims, it will c-command the head noun and result in an incorrect word order. (13b) shows this discrepancy, as we can see from its tree structure in (14). (14) DP /\ D, /\ D0 NP I /\ the NP CP | A N’ that he made | NO I claim The structure in (14) has the CP adjoined to the head NP. The lower NP and the higher NP are two segments of the same category. Since segments cannot enter into c-command relations, the CP in (14) asymmetrically c-commands the N0, and thus the CP’s terminals 27 should precede the head noun, which would derive the word order the that he made claim, which is not a grammatical English relative clause. However, Kayne (1994) adopts a head-raising analysis (Vergnaud 1974) which can solve the problem that the standard analysis encounters. The head-raising analysis emphasizes that the head noun is derived through a direct movement from the relative clause. A relative clause, like (13a) the claim that he made, has its head noun moved from the relative clause, as shown in (15a) with its tree structure shown in (15b). (15) a. [DP the [cp[Np claim], [C that he made ti] b. DP /\ D, /\ D0 CF I /\ the NPi C , I /\ N’ c0 [P | l A N0 that he made ti claim (15) shows that the NP moves to the Spec of CP and it asymmetrically c-commands CO and IP, but not the other way around, since C’ is a segment which does not involve a c- command relation. Following the LCA, it has the correct word order, the claim that he made. Another consequence of the LCA is that the head direction parameter does not exist. Since right adjunction is not allowed, an XP can only have the following structure: 28 That is, the specifier position is the site for adjunction. The head must asymmetrically c- command and precede its complement, and thus the universal order of the elements of a phrase are specifier, head, and complement. The head directionality parameter does not exist and the word order of each language is derived by movement with respect to the order of the specifier, head and complement sequence. Thus, an SOV order has a derivation as shown in (17). (l 7) IP DP 1’ (subject) I0 FP DP, F’ (object) F0 VP That is, the complement of the head V0 will make a leftward movement to the Spec of an FP (a functional phrase, such as an Agr Phrase), resulting in SOV word order, which is the correct word order for languages like Japanese. 29 Since head directionality does not exist, the underlying structure of final head relative clauses must be the same as the structure of head-initial relative clauses. For a language like Amharic, which has a definite article between its relative clause and its head noun, (i.e., relative clause + definite article + noun), Kayne proposes the structure shown in (18). (18) DP > (18) shows that head-final relative clauses have the same structure as head-initial relative clauses. However, in order to get the head-final relative clause, the IP originally containing the head noun moves to Spec of DP, after the head noun raises to Spec of CP, and thus the head noun appears at the end of the relative clause. Following Kayne’s proposal, according to Xu, a Chinese CP will have the same structure as other languages, as shown in (19). (19) CP /\ Spec C ’ /\ C0 IP 30 Since Chinese has noun final relative clauses, the Chinese relative clause has a similar structure to (18). According to Xu, a Chinese relative clause such as (20a), will have a structure as in (20b). (20) a. wo mai de shu I buy DE book ‘the book I bought’ b DP /\ IPj D’ A /\ wo mai tk D0 CP ‘1 1m)" I /\ deg NPk C, ‘DE’ A |\ shu c0 1P,- ‘book’ | That is, the head noun moves to Spec of CP. De starts out in C0 and moves to DO. This allows IP to make a legal movement to Spec of DP. According to Xu, de is not a determiner since de + NP does not form a constituent, as shown in (21). (21) *de shu DE book Therefore, the most likely position to generate de in is C0. After de moves to Do, the distance from the lower IP to [Spec, CP] equals to the distance to [Spec, DP] under the assumptions of Minimalist Program (Chomsky 1993). Thus, [Spec, DP] is a possible target for IP to move to. Xu argues that C0 to DO movement is important for two reasons. 31 First, by moving to Do, it can check the feature of CP. Second, according to him, “assuming that IP features have to be checked in the [Spec, DP] position and that these features (due to selection) have to be checked by C0 instead of D0, substituting the empty D0 with a lexical C0 such as de will make this checking possible” (1997, 92). Based on Xu’s analysis, the sentences in (22) will have the structures in (23). (22) a. Zhangsan de shu DE book ‘Zhangsan’s book’ b. zai zhuo shang de shu ZAI table t0p DE book ‘the book which is on the table top.’ 0. haokan de dianying good-looking DE movie ‘the movies that are good to watch’ d. Lisi shang guo de daxue top GUO DE university ‘the university that Lisi attended’ (23) a. [Dphp Zhangsan Pred. I] ]k [D0 131% ICP [Np shu ]i [ C0 ej [1p e]k ]]4 b. [mm t, zai zhuo shang]1< [1)0 DE,- [CF [Np shu], [c0 e,- [Ip em] 0 [DP [11> t, h30kan ]k [D0 D131 [CP [NP dianying], [CO CI] [1P elklll d. [DP [1p L181 shang-guo ti ]k [D0 DEj [Cp [NP daxue]g [C0 ej [1p e]k]] As a result of his analysis for relative clauses, Xu further applies this structure to yi-qian/yi-hou and zhi-qian/zhi-hou ‘before/after’ clauses. He suggests that yi/zhi should ’ Xu, following Hashimoto (1971), analyzes (22a) Zhangsan de shu ‘Zhangsan’s book’ as deriving from Zhangsan you shu ‘Zhangsan has books’. You ‘has’ is deleted and therefore, (23a) shows 3 Fred in the structure. 32 be analyzed as de, and qian/hou ‘before/after’ should be analyzed as head nouns that are modified by temporal clauses. A temporal clause such as in (24a) would have the structure in (24b). (24) a. wo likai zhongguo yi-qian/yi-hou. I leave China before/after ‘before/after I left China...’ b. [DP [W0 lrkai ZhOI'lggUOL' [Dp yli [cp [Np qian/hou] [cp ti IPj]]]]]- . . That is, qian/hou are head nouns that are base-generated in Spec of CP, and yi is in the same position as de. Xu also emphasizes the fact that (24) can occur after the preposition zai, suggesting that the temporal clause is a DP. This is shown as (25). (25) zai wo likai Zhongguo yi-qian/yi-hou ZAI I leave China before/after ‘before/after I lefi China’ In summary, following the Minimalist framework, and Kayne’s analysis of head- f"1171211 relative clauses, Xu argues that de is in C0. The head noun moves to Spec of CP and d e moves from C0 to D0. The head movement extends the domain of C0, and this allows IP to move to Spec of DP to derive the correct word order. 33 2.3 De as a syntactic category Ning (1996) proposes that de and Cl(assifier) are functional heads in Chinese that project their own phrases: de projects to a DeP (de phrase), and CL projects to a ClP (classifier phrase). Ning suggests that de selects a clause to the left as its complement. This complement must be an AgrsP that contains a gap. The gap for Ning is created by movement to Spec of DeP for feature checking, as (27) shows, just as wh-phrases move for checking [+wh] features in English5 . Semantically, a DeP is like a predicate formed by A abstraction and any bare DeP expresses a set of entities that shares the property that is defined in the pre-de clause. Examples of bare DePs are shown in (26), and the proposed structure of a bare DeP is shown in (27). (26) a. ta xihuan de he like DE ‘a set of entities having the property of x such that he likes x’ b. zhuotian huilai de yesterday come DE ‘a set of entities that came yesterday.’ (2 7) DeP /\ DE’ NP /I AgrsP DE A l [ ti ...] de Inc, ‘3 s Ning labels the moved element [ne] for Null Entity, but it is unclear what significance is to be attached to this label. 34 With this bare de structure, the head noun adjoins to DeP, and the structure is as follows: (23) DeP l\ DeP NP /\ DE’ NP //1 AgrsP DE A I [ ti ] de Inc, e I That is, the head does not extract directly from the relative clause, but it adjoins to the DeP. Like Xu, Ning analyses all de phrases as relative clauses. Sentences such as (29), which appear not to have a gap, are analyzed as having a gap, as in (30). (29) a Zhangsan de shu DE book ‘Zhangsan’s book’ b. zai MIT de na-ge xuesheng ZAI DE that-CL student ‘the student at MIT’ c. hen qimiao de gushi very wonderful DE story ‘a very wonderful story’ (30) a. [NP[ e Zhangsan] de [NP shu]] b. [[e zai MIT] de [nage xuesheng]] 35 c. [[e hen qimaio] de [gushi]] (29a) is analyzed such that Zhangsan de is a predicate, as shown in (31). (31) Zhe shu shi [Zhangsan de]. this book is DE ‘This book is Zhangsan’s.’ Since Zhangsan de can occur after the copula shi ‘be’, it is a predicate and able to be selected by de and merges with de to form [Zhangsan de ] de ], a DeP. After this merge, then one de is deleted; later, the NP shu adjoins to the DeP, and (29a) is derived. (29b) zai MIT is also taken as a predicate and it has a subject gap. Evidence for this comes from the following comparison. (32) a. *l e cong MIT| de xuesheng from DE student ‘the student(s) who is/(are) from MIT/comes fi'om MIT.’ b. *na-ge xuesheng cong MIT that-CL student from ‘That student is from MIT/comes from MIT.’ (33) a. [ e zai MIT] de na—ge xuesheng ZAI DE that-CL student ‘that students who is at MIT’ b. Na—ge xuesheng zai MIT. that-CL student ZAI ‘That student is at MIT.’ 36 (33b) shows that the subject na—ge xuesheng ‘that student’ takes zai MIT ‘at MIT’ as a predicate, but (32b) na-ge xuesheng ‘that student’ does not take cong MIT ‘from MIT’ as a predicate. Therefore, leaving out a subject in (33a) to make a relative clause with zai MIT is fine, but not in (32a) because it is ungrammatical before the subject is relativized". The same analysis applies to (300) that adding a subject to hen qimiao ‘very wonderful’ is a grammatical construction. To sum up, Ning suggests that de is a head-final functional category that selects a clause with a gap created by movement to Spec DeP. If a de phrase contains a head noun, the head noun adjoins to DeP to the right. 2.4. De as a determiner In this section, I will discuss two analyses that take de as a determiner. One is Simpson (1999, 2000) who follows Kayne’s noun final relative clauses analysis and argues that de is an enclitic determiner. The other is Lin, Murasugi, and Saito (2001) (henceforth Lin et al.) who argue that de is a determiner that licenses a DP Spec position, and the relative clause is base-generated in the Spec of DP. I will review Simpson’s argument first then go on to review Lin et al. ‘ 6 'Ihe correct sentences for (32a) and (32b) can be expressed in (i) and (ii) respectively. (i) cong MIT lai de ren (ii) Na-ge xuesheng cong MIT lai from come DE person that-CL student from come ‘The person who comes from MIT.’ ‘He comes from MIT’ That is, the difference between zai and cong is that zai phrases can be predicates, but cong phrases are pure Drepositional phrases. Since (32a) lacks a predicate, it is ungrarrnnatical before the relativization takes place. 37 2.4.1 Simpson’s analysis Simpson (1999, 2000), following Kayne’s analysis for relative clauses, argues that de is an enclitic determiner that takes CP as a complement. Since de + NP does not form a constituent, as Xu (1997) argues, such an analysis does not seem to be straightforward. However, Simpson examines determiners cross-linguistically and finds that there are similarities between de and determiners in other languages. He also looks at the historical derivation of how determiners changed among languages and concludes that de is an enclitic determiner. There are three difficulties in claiming that de is a determiner. First, de can occur more than once in a DP, as shown in (34), and it is not common for a DP to have more than one determiner as the head, especially in English. (34) wo dc; Chomsky xie Q nei-ben shu I DE write DE that-CL book ‘my book which is written by Chomsky’ Simpson argues that we cannot use our understanding of West European languages to judge this phenomenon. He gives examples from Giusti (1997) to show that Hebrew, as in (35a), Albanian, as in (35b), and Greek, as in (35c), all have phrases that contain more than one determiner. (35) a. ha—yit h_a-gadol the-house the-bi g ‘the big house’ 38 b. dial-i_ i-mire c. afto _t_g oreo t9 vivlio boy-the the-good this the good the book ‘the good boy’ ‘this good book’ As shown in (35), the determiners in these languages are able to occur more than once within a DP. Therefore, a determiner that occurs more than once should not be something that is forbidden. The second problem with de not being a suitable determiner is that it does not have any contribution to mark definiteness of a noun phrase. Simpson draws attention to Longobardi’s (1994) argument that in some languages, the definite determiners are simply expletives which are place-holders, and do not necessarily mark definiteness. That is, the occurrence of the definite determiner is not contributing to the definiteness of the whole noun phrases. One of examples from Italian is shown in (36). (36) a. il mio Gianni the my Gianni b.G1ann1. mio ti o. *mio Gianni According to Longobardi, a definite determiner can be a placeholder. The definiteness of the DP in (36b) is established by overt movement from N0 to D0. Otherwise, a definite article has to fill the deterrniner’s position as (36a) shows. Otherwise, failure to insert an eXpletive in D0 or moving the name to D0 will end up ungrammatical as in (36c). Also, in French, definite determiners, when used in instances of inalienable possession, do not have defrnite interpretation, but an indefinite interpretation as shown in (37). 39 (37) Olga a le bras enfle. Olga has the arm swollen ‘Olga has a swollen arm.’ Also English definite article may be used without marking definiteness when it is used as a book or story title, as shown in (38). (38) The Crow and the Fox Therefore, the fact that de does not have the function of definiteness is not a crucial argument in support of the conclusion that de is not a determiner. A third possible reason to reject de as a determiner is that de can co-occur with demonstratives. As shown in (39), the determiner and demonstrative should not co-occur in English since the demonstrative is traditionally considered to be in D0 position. (39) a *the that book b. *that the book However, examples from Spanish and other languages show that it is not impossible for the definite article to co-occur with a demonstrative, as shown in (40). (40) el hombre este the man this ‘this man’ 40 Furthermore, evidence from Grosu (1988) shows that demonstratives in Romanian are not generated in D0, but are phrases, i.e., XPs. This explains why N-movement over a demonstrative is not barred, as in (41 a), but AP movement over a demonstrative is barred as shown in (MY. (41) a. baiati-uo acesta frumos ti boy-the this nice ‘this nice boy’ b. * frumosi-ul acesta ti baiat nice-the this boy If demonstratives are not in Do, it is not surprising to see a determiner and a demonstrative present in the same phrase, as with de and other determiners. The data above demonstrate that, although de behaves differently from English ‘the’, this should not be taken as reason to exclude de as a determiner. Simpson further argues that de is an enclitic determiner in Chinese by applying Grosu’s (1988) argument supporting an enclitic determiner analysis for the Romanian definite determiner. According to Grosu, the Romanian definite determiner is an enclitic determiner and needs phonological support to exist. Therefore, the determiner triggers some other element to move to precede it, as shown in (42). (42) a. potreti-ul t, unei fete portrait-the a.Gen. girl ‘the portrait of a girl’ \ 7 Rizzi (1990) shows that movements should obey relativized minirnality. A category can only be blocked by the same kind of category. Therefore, an XP would be blocked by an XP, but not an X0. 41 b. un portet al unei fete a portrait of-the one girl ‘a portrait of the girl’ (42a) shows that id ‘the’ triggers potret ‘portrait’ to move to its left for phonological supports. (42b) shows that the indefinite determiner, un ‘a’, does not trigger any element to its left. Simpson argues that this situation is just like de phrases. De always needs some element to precede it for phonological support. Furthermore, he follows Szabolcsi’s (1994) analysis of Hungarian determiners and assumes that de in Chinese is similar to the Hungarian determiner. According to Szabolcsi, Hungarian D0 is only a position for the determiner to fill in, and the determiner does not have definite interpretation. Szabolcsi argues that the Hungarian determiner is just a subordinator; the definiteness is assumed to come from other elements. For example, demonstratives appear lower than the determiner. Along the line of Szabolcsi, Simpson takes de as an underspecified determiner that does not contribute a definite interpretation. However, a demonstrative can co-occur with the determiner and assign definiteness to the whole phrase, as shown in (43). (43) ta mai de nei-ben shu he buy DE that-CL book ‘that book that he bought’ Assuming that de is a determiner, Simpson takes the fimction of de “to introduce a predicative restriction on some nominal” (1999, 9). The reason that de is required is ‘ 8 It is unclear what Simpson means by “phonological support”; however, it can be assumed that a Do requires its specifier position to be filled with an XP. 42 that the presence of a modifier for a head noun depends on the presence of a determiner. (44) shows this fact in Vietnamese. (44) a. Trong nha-hat kia co muoi-bay ghe. in cinema this are 17 chair ‘There are 17 chairs in this cinema.’ b. Trong nha-hat kia co muoi-bay *(cai) ghe lam bang cay tot. in cinema this are 17 CL chair make of wood good ‘There are 17 chairs make of good wood in this cinema.’ Classifiers in Vietnamese have determiner-like properties, according to Daley (1995). (44) shows that Vietnamese does not need a classifier for counting. However if a modifier phrase occurs, the classifier is obligatory. English also shows a similar situation. Although it is not obligatory for English modifiers to co-occur with the definite determiner the, sentences in (45) show that there is a connection between the determiner and the noun modifier. (45) a. *the every whim a’. the every whim of Margaret Thatcher b. that sweater of John’s (is expensive.) b’. * the sweater of John’s (is expensive.) b’ ’. the sweater of John’s that you showed me last night (is expensive.) All these examples support the idea that the determiner has a strong bond with the modifier of a head noun. 43 Therefore, according to Simpson, the reason that (46) is ungrammatical is because de in (46a) needs a modifier’s phonological support to its left. (46b) shows that movement of the head noun to the left of de does not satisfy the structural requirements of de. (46) a. *de ren b. *reni de t.- DE person person DE Intended to mean: ‘the person’ Intended to mean: ‘the person’ That is, the connection between the modifier and the determiner is a cross-linguistic phenomena. De, in the same sense, is necessary when a modifier is needed to modify the head noun. Therefore, in order for a modifier to precede the head noun, de is necessary. Simpson also looks through the development of determiners cross-linguistically in other languages that have determiners behaving quite parallel to de. He argues that de is indeed an enclitic determiner. Deterrniners frequently develop from demonstratives and then gradually lose their definiteness. For example, French 1e and la ‘the’ are derived from Latin demonstratives il-le and il-la. Greenberg (1978) also claims that quite a few Afiican languages have determiners that developed fi'om demonstratives, and gradually lost their definiteness. Hence, a new set of demonstratives will be used for specifying the definiteness of a noun phrase. Chinese de is thought to develop from the classical form zhi, which firnctions like de in introducing modifiers, but also functions as a demonstrative (Pulleyblank 1995). Simpson believes that this is good evidence that tie is an enclitic DO element. Moreover, there is also evidence from Buginese (47) that its definite determiner shows a similar 44 phenomenon like de that introduces a modifier and noun relationship, as in (47a), and it uses the definite determiner to form relative clauses, as shown in (47b). (47) a. iaro buku-e malotonng-e lima-e those book-the black-the five-the ‘those five black books’ (Nishiyarna 1997) b. buku lima-e uvlii-e iaro book five-the 1SG.buy-the that ‘those five books which I bought’ In summary, de is a semantically bleached determiner that introduces a predicate/modifier phrase before the head noun. Since it needs phonological support, the predicate/modifier it introduces has to move to its left. This situation is similar to the Romanian data in (42). Now turning to Simpson’s analysis of a Chinese de phrase, Simpson proposes the structure in (48b) as an analysis of (48a). (48) a. wo zuotian mai de shu I yesterday buy DE book ‘the book that I bought yesterday.’ b. . DP /| [P D’ A l\ [wo zuotian mai ti]m D CP ‘ I yesterday buy’ | /\ A de NP C’ ‘DE’ A l\ shui C [P ‘b00k’ A T l tli lm 45 (48) shows that shu ‘book’ moves from the IP to the Spec of CP, and the IP moves to the Spec of DP. In addition to relative clauses like (48), de phrases such as genitive phrases and adjective phrases are also interpreted as relative clauses under Simpson’s analysis. The examples in (49) are analyzed as in (50). (49) a. wo de shu I DE book ‘my book’ b. In de huaping green DE vase ‘green vase’l ‘the vase that is green’ (50) a. [DP [1P W0 10 [VP 6 ti ]]k (16 [CP Shui tk] I b- [DP [1P ti 1ij lm [D de [CP huaping, tm ll] (50b) is not problematic since it is widely known that Chinese adjectives can be predicated without be like in English. In (50a), Simpson follows Kayne’s analysis, analyzing wo shu ‘I book’ meaning ‘I have (a) book’ as an IP headed by a null verb. The element e in (50a) is equivalent to English have. Simpson argues that the copula shi ‘be’ can be dropped in certain environments, for instance, wo (shi) zhongguo ren, ta (shi) yingguo ren. ‘I am a Chinese person, he is an English person’. Therefore, it can be taken as a default verb in sentences which have two nouns without verbs. However, it is Uncommon to have the copula followed by a noun in a de phrase and link with the head noun, as shown in (51). 46 (51) *shi zhongguo de ren is China DE person ‘*people who are China’ Therefore, Simpson suggests that if this default interpretation of the copula does not apply to de phrases, then when de connects a noun phrase and a head noun, the null verb possession reading can occur, as (50a) shows. In summary, Simpson argues that we cannot apply the English use of definite determiners to refirte the determiner analysis of de. Examples of determiners in other languages that behave differently from English the have very similar syntactic properties to de. For instance, some of these determiners introduce modifiers, and some need phonological support, and some have both characteristics, like de. Also when looking into history, determiners that developed from demonstratives are usually semantically bleached. Since de is believed to be developed from zhi, and zhi is also a demonstrative, it is likely that de is a determiner. 2.4.2 Lin, Murasugi and Saito’s analysis Lin, Murasugi and Saito (Lin et al.) (2001), based on Murasugi and Saito’s (M &S) (1990) study of Japanese NP deletion, claim that Chinese de is a determiner. M & S (1990) argue that only under the DP hypothesis can a noun phrase deletion be explained in a principled way. It is said that the deletion of a noun phrase occurs because it Undergoes N’ deletion (J ackendoff 1971), as shown in (52), and the same situation applies to Japanese, as shown in (53). 47 (52) [Np Linclon’s [N. portrait]] didn’t please me as much as [Np Wilson’s [N’ e]]. (53) [NpHaruki no [Nve]]-wa [NP Masako no [N1 e]] yori kirei da. NO picture-Top NO than beautiful be ‘Haruki’s picture is more beautiful than Masako’s.’ Deleted N’s must be preceded by a genitive phrase, and like VP deletion, the deleted constituent needs an antecedent. As (52) and (53) show, the deleted noun phrase has an antecedent, which is portrait in (52) and e ‘picture’ in (53), and they both have a stranded genitive phrase which is Wilson ’s for English and Masako no in Japanese. M & S argue against an N’ deletion approach on the following grounds. First, if as J ackendoff argues, N’ deletion and VP deletion share the same basic properties, it misses a structural parallel with the VP deletion that N’ is not a maximal projection. Under the assumption that all empty categories need to be licensed, VP deletion is licensed by Infl, while N’ deletion does not have a head similar to Infl to license the deleted noun phrase. Moreover, by examining (54), we find that the N’ deletion analysis also encounters empirical problems. (54) a. This [book] is John’s [e]. a’. Kono [hon]-wa John no [e] da. this book-Top John NO be ‘This book is John’s. ’ b. *That [bad attitude toward research] is John’s [e]. b’. *Sono [yokunai kenkyuu-ni taisuru taido ]-wa John no [e] da. that bad research-Dat toward attitude-Top NO be ‘That bad attitude toward research is J ohn’s.’ c. John’s attitude toward research is good, but Mary’s is bad. 48 c’ . [John no [kenkyuu-ni taido]]-wa ii ga, [Mary no [ e]]-wa yokunai. NO research-Dat attitude-Top is-good though, NO -Top bad ‘J ohn’s attitude toward research is good, but Mary ’3 is bad.’ (54a) shows that the deleted noun phrase refers back to book. The same meaning in Japanese as in (54a’) shows that the deleted noun phrase after no has an antecedent hon ‘book’ and is grammatical. (540) also shows that the gap refers back to attitude toward research and the same meaning in Japanese as in (54c’) shows that the deleted noun phrase after the second no can refer back to the noun phrase, kenkyuu-ni taido ‘research attitude’. However, the gap in (54b) cannot refer back to bad attitude toward research, and the gap in (54b’) cannot refer back to yokunai kenkyuu-ni taisuru taido ‘bad attitude toward research’. Given N’ deletion, it is puzzling why the gaps in (54b) and (54b’) cannot refer back to their antecedents. M & S (1990) suggest that a DP hypothesis not only provides a parallel structure with the VP deletion, but it can also explain the situation in (54). With a DP hypothesis, the deleted noun phrase is an NP, a maximal projection just as in VP deletion. Also, NP deletion will have D0 to license the deleted NP just as I0 licenses the deleted VP, as illustrated in (55). (55) a. [1121 [r [1] [VP left] because [ 1}: John [1* [1 dld] [vp e]]] b. [op John [D’ [D ’s] [Np t, [N- reliance on his advisor]]] is more problematic than [DP Mary [D' [D ’5] [NP ti ell] Also, the situation for English in (54) can be well explained, as shown in (56). 49 (56) a. [DP This [[NP bOOk] is [DP John’s [NP e]]. b. *[m That [Np bad attitude toward research]] is [DP Johnj [D ’s [Np tj e]]]. c. [DpJohn’si [Npti attitude toward research]is good, but [13pr [9’5 [tj e]]is bad. Given the DP hypothesis, (56a) has ‘s in D0. John is the possessor of book, and the possessor is based-generated in Spec of DP. In (56b) and (56c), John and Mary are not the possessors of attitude, so they are not base-generated in Spec of DP. However, they start out in the NP to receive the 0-role of experiencer from attitude toward research and then move to Spec of DP to get case from Do. In (56b), the deleted NP has John move to Spec of DP. The deleted NP contains a trace, and therefore cannot refer back to the higher NP bad attitude toward research since there is no trace in the higher NP, and thus the result is ungrammatical. In (560), both the higher NP and the deleted NP contain a trace, and therefore, the deleted NP can refer back to the higher NP. A similar analysis can be applied to the Japanese data in (54), as shown in (57). ( 5 7) a. [pp Kono [Np hon]-wa [DpJohn no [Np e]] da. this book-Top John NO be ‘This book is John’s.’ b. *[Dp Sono [NP yokunai kenkyuu-ni taisuru taido ]-wa [DPJOhnj no [tj e]] da. that bad research-Dat toward attitude-Top NO be ‘That bad attitude toward research is John’s.’ 0. [pp John, no [Npti kenkyuu-ni taido]]-wa ii ga, NO research-Dat attitude-Top is-good though [DPMa-I'Yj no [Np tj e]]-wa yokunai. NO -Top bad ‘J ohn’s attitude toward research is good, but Mary ’3 is bad.’ 50 Under the DP hypothesis, (57b) is ruled out for the same reason as (56b) -- a deleted NP with a trace cannot refer back to an NP that does not contain a trace. However, Japanese no cannot be equal to English ‘3. For M & S, no is inserted whenever two noun phrases or a PP and a noun phrase are adjacent to each other. Therefore, a phrase like Kyooto no hikai ‘Kyooto’s destruction’ can have either structure in (5 8). (58) a. [pp [Np Kyooto-no hakai]] destruction ‘Kyooto’s destruction’ b. [pp Kyooto-no, [Np t, haikai]] destruction ‘Kyooto’s destruction’ (M & S 1990, 296) That is, Kyooto can stay as the complement of hakai ‘destruction’ and then insert no, as (5 8a) shows; or it can move to Spec of DP and then insert no, as (5 8b) shows. Nevertheless, when NP deletion occurs, the movement to Spec of DP is forced. Following their analysis of the NP deletion and no phrases, Murasugi and Saito filrther examine the reason why a sentence parallel to (57c) is ungrammatical, as (59) shows, (5 9) a. *Saikin-wa hare-no hi-ga ame-no yorimo ooi. recently-Top clear-NO day-Nom rain-NO than plentiful ‘Recently, sunny days are more plentiful than rainy ones.’ b. *Saikin-wa [pp lnarei —no [Np t, hi]]-ga [pp amej-no [Np tj e]] yorimo ooi. recently-Top clear-NO day-Nom rain-NO than plentiful 51 According to M & S, NP deletion requires the phrase preceding no to move to Spec of DP, as (58b) shows. Hence, (59a) has its structure of (59b) that hare ‘clear’ and ame ‘rain’ move to Spec of their DPs. However, adjunct NP movement is known to be an illegitimate movement’. In (59), both hare and ame are adjuncts moving to Spec of DP, and thus, the illegitimate adjunct movement causes (59) to be ungrammatical. If (59) is analyzed under an N’ deletion analysis, it would be difficult to explain why (59) is ungrammatical, since under N’ deletion, no movement would occur. Lin et al. (2001) suggest that there are many similarities between Japanese no and Chinese de. Based on M & S’s (1990) NP deletion analysis that was discussed above, Lin et al. also suggest that the same analysis can apply to Chinese noun phrase deletion, as shown in (60). (60) has a structure parallel to (56) and (57). (60) a. [pp Zhe-liang [Np che] shi [pp John de [Np e]]. this-CL car is DE ‘This car is John’s.’ b. *[Zhe-zhong [Np taidu] shi [J ohn, de [Np t, e]. this-CL attitude is DE “This attitude is John’s.’ c. [DP John de [NP t, taidu]] bi [1)? Mary, de [Np t, hao]]. DE attitude compare DE good ‘J ohn’s attitude is better than Mary’s.’ \ 9 bggcording to Murasugi and Saito (1990), the blocking of adjunct movement can be illustrated as in (i) El) ( , ‘Two pounds, are weighted t, by this book. (p. 297) Sub 3), two pounds does not get any O-role fiom weight, and therefore cannot undergo NP movement to the J ect position. 52 However unlike the Japanese no phrases that do not allow NP deletion when an adjunct precedes no, as (59) shows, Chinese de phrases allow NP deletion when an adjunct precedes de, as shown in (61). (61) W0 zuotian kandao de nuhao bi ni zuotian kando de piaoliang. I yesterday saw DE girl compare you yesterday saw DE beautiful ‘The girl I saw yesterday is more beautiful than the one you saw yesterday.’ (61) shows that although the relative clause is an adjunct, Chinese de still allows the NP deletion. Lin et al. conclude that Japanese no and de, although similar in many aspects, are different in NP deletion. In Japanese, no does not play a crucial role in NP deletion. It is inserted when the NP moves to Spec of DP, as in (5 8b). In Chinese, de is crucial in that it licenses an NP deleted position although a relative clause precedes it. Therefore, in Chinese, de must be a determiner occurring in D0 that can license its Spec position. A relative clause is an adjunct and does not have a O-role. It cannot undergo NP movement according to M & S (1990), as shown in (59b), thus, the relative clause which is preceding de must be base-generated in Spec of DP. In addition to relative clauses, they a180 claim that all thematic expression phrases that relate to the head noun are also base- generated in the Spec of DP that is projected by de, instead of starting below with the head noun and moving to Spec of DP. This is so, as illustrated by the fact that a relative clause can occur between a theta marked noun phrase and the head noun, as shown in (62)- S3 (62) Luoma de (&ci) m_anzu zao-chengde huimian. Rome DE (that-time) barbarian cause DE destruction ‘The destruction of Rome that the barbarians caused (that time).’ According to Murasugi and Saito’s (1990) analysis of NP deletion, Luoma ‘Rome’ has the patient 0-role from huimian ‘destruction’ and is supposed to start out as the complement of huimian ‘destruction’ and move to the highest Spec of DP as in (62). However, Lin et al. observe that this movement would cross the relative clause and thus violate Relativized Minirnality, as stated in Rizzi (1990). Therefore, the sentence should be ungrammatical. But, (62) is grammatical. Thus, there should not be movement from the head noun to the highest Spec of DP position. The explanation for (62) is that de can license a relative clause that is base-generated in Spec of DP, and it does not matter if a thematic expression phrase relates to the head noun. According to Lin et al., (63) shows that this explanation is valid since (62) can also be expressed as (63) in that the relative clause can exchange positions with Luoma ‘Rorne’ without changing the meaning of the whole phrase. (63) (gr-oi) mgnzu zao-chengde Luoma de huimian. (that-time) barbarian cause DE Rome DE destruction ‘The destruction of Rome that the barbarians caused (that time).’ Therefore, (62) has the structure shown in (64). 54 (64) DP /\ Luoma D’ ‘Rrome’ /\ 1|)° DP /\ de nai-ci D’ ‘DE’ ‘that time’ /\ manzu zao-cheng D0 NP ‘barbarians caused’ I A de huimian ‘destruction’ In summary, Murasugi and Saito (1990) claim that under DP hypothesis, noun phrase deletion can be explained in more principled way. NP deletion has D0 to license the gap just as VP deletion has I0 to license the gap. They conclude that, in order to have a successful NP deletion, the element preceding the genitive elements (i.e., ‘s in English and no in Japanese) must be a possessor which is base-generated in Spec of DP, or it has to bear a thematic role and undergo movement to Spec of DP. However, this movement does not apply to adjunct genitive phrases in Japanese since adjuncts do not have a thematic role. The movement of the adjunct is barred and thus adjunct genitive phrases do not license NP deletion. Chinese de phrases also show NP deletion, and adjunct + de can even allow NP deletion, which is not allowed in Japanese. Since Japanese no is an inSel‘ted item and does not stay in Do, this leads Lin et al. to conclude that de must be baSe~generated in D0, and therefore de is a determiner that can license NP deletion. They also observe that thematic expressions relating to the noun following de can be intelWiened by relative clauses. Exchanging the thematic expressions and the relative clauses does not change the meaning of the whole phrase. This further leads them to ConCIUde that de as a determiner can license the Spec of DP position and thus any 55 thematic expression relating to the head noun does not need to be base-generated close to the head noun but can instead be base-generated in Spec of DP. 2.5 Discussing the three analyses In this section, I will examine the three analyses with the goal of choosing one on which to base my subsequent analysis of Chinese temporal clauses. I will first examine the head-final complementizer (Cheng 1986 and Ning 1993) and the head-final de (N ing 1996) analyses. Then I will examine Xu’s head-initial complementizer analysis and show that Xu’s analysis is a better choice than Cheng and Ning’s analyses. Next, I will examine Lin et al. (2001). I will show that it is difficult to interpret the relationship between the head and the relative clause if relative clauses are base-generated in Spec of DPs. Finally, I will examine Xu’s complementizer analysis and Simpson’s enclitic determiner analysis together since their analyses are very similar. I will ultimately choose Simpson’s analysis on which to base my analysis of temporal clauses. Basically, Cheng (1986) and Ning (1993) both argue for de as a head-final complementizer and take de as a case assigner. The difference is that Cheng does not ConStrain the category of phrases occurring before de, while Ning insists that only clauses that contain gaps can precede de. However, Cheng and Ning’s analysis of de as a case mar ker is problematic. Tang (1990) has argued convincingly that de is not a case marker, as Shown in (65). (65 ) Zhangsan (de) meimei de haizi DE sister DE children ‘The children of Zhangsan’s sister’ 56 Although de occurs between two nouns here, the first de is optional. If de is a case marker, it is not clear why the first de is optional. Also, de is usually preceded by a clause, as in a relative clause and a noun complement clause. Clauses do not need case (Stowell 1981), but de is obligatory. The obligatory presence of de is apparently irrelevant to case marking. Furthermore, Cheng’s analysis that de unrestrictively selects an XP as its complement is problematic, since it is simply not true that de unrestrictively selects an XP to be its complement, as (66) illustrates clearly. (66) a. *cong Michigan de ren from DE people ‘people who are from Michigan’ b. *xuesheng dai yanjing de ren students wear glasses DE people ‘*people who students wear glasses’ (663) shows that it is not true, as Cheng claims, that de can take any PP as its complement. (66b) shows that de does not take any clause as its complement. Thus, Cheng’s claim is too strong. Actually, none of the analyses that attempt to prove that de is a complementizer are Satisfactory, including Cheng (1986) and Ning (1993). De may mark a subordinate Clause, but it is not sufficient to claim that de is a complementizer. Szabolcsi (1994) Shows that a definite determiner can be a subordinator too. If de is a complementizer, it Won Id be odd that cross-linguistically de may be the only complementizer that marks the gem tive relation between two nouns. And, it would be odd that de may be the only 57 complementizer that cannot mediate a matrix clause and an embedded clause as a complementizer usually does, as shown in (67). (67) Ta shuo (*de) ta bu zhidao he say DE he not know ‘He said that he doesn’t know.’ In addition, the Chinese lexical complementizers, ruguo/yaosh/jiarui ‘if’, precede their complement clauses. Wu (2000), following Hwang’s (1998) observation that Chinese languages grammaticalize say, also suggests that de cannot be a complementizer. According to Hwang, Mandarin Chinese, Taiwanese and Cantonese say is undergoing grarmnticalization into a complementizer in certain environments, such as after the main verb think and tell, and before complements, as shown in (68)"). (68) a. Ta xiang shuo ta bu lai le. he think say he not come LE ‘He thinks that he is no longer coming.’ b.Ta gaosu wo shuo ta bu lai 1e. he tell me say he not come LE ‘He told me that he is no longer coming.’ (Wu 2000, 104) Therefore, for dc to be a complementizer is very unlikely, since Chinese does possess he"=1C1—initial complementizers. Generally speaking, Cheng (1986) and Ning (1993) take de as a head-final complementizer that can assign case. Cheng emphasizes de’s lack of restriction in \ 10 _y ls expressed as shuo, gang, and gang 1n Mandarrn Chrnese, Ta1wanese and Cantonese respectrvely. 58 selecting a complement to its left, while Ning argues that de must select a gapped-clause. However, as (65) shows, de does not assign case and (66) shows that de cannot take any category to its left. Ning (1996) takes de as its own syntactic category. This is too language specific and appears to be too ad hoc. Furthermore, neither analysis has a good story to support why de is a complementizer. Finally, their head-final analysis poses a language typology problem since, in a language like Chinese, verbs, prepositions and the lexical complementizers ruguo/jiaru/yaoshi ‘if’ are head-initial. Therefore, for Chinese to have a head-final noun phrase and a head-final functional category de seems odd. Now let us examine Xu and Simpson’s analyses since their analyses are very similar in that they both apply Kayne’s proposal for noun final relative clauses. The difference is that Xu takes de as a complementizer, while Simpson takes de as a determiner. These are illustrated below in (69), respectively: (69) a DP b DP /\ /\ [P D’ IP D’ l /\ | /\ [ tk ]j DO CF I: I}, ]j D, CP DE, NPk C’ DE NPk C’ |\ /\ I C,° 1P,- C0 1PJ (69a) is Xu’s analysis. He argues that de is not a determiner because it does not allow any noun to follow it to form a phrase. He argues that de is generated in C0 and moves to D0 to expand de’s domain. This allows the lower [P to move to Spec of DP. Among the 59 complementizer analyses, Xu’s analysis seems to be the best choice. His head-initial analysis follows the language typology. He does not have an ad hoc analysis that takes de as a specific syntactic category, as Ning (1996) does. His analysis is able to account for all the de phrases without claiming that some de phrases have de as a case marker, and some do not. His analysis seems to be more explanatory than either Cheng’s or Ning’s. (69b) is Simpson’s analysis. He argues that de is a determiner based on cross- linguistic support and de’s development from zhi, which is a demonstrative in old Chinese. According to him, the IP in (69b) moves for the phonological support of de, and this phenomenon can be found in Romanian and Buginese. Simpson’s determiner analysis also follows the language typology system by being head-initial. In addition, there is a wide range of cross-linguistic support for de as a determiner that introduces adjectives and clauses to modify nouns. For instance, Romanian and Buginese definite determiners show very striking parallels with Chinese de in that they can introduce adjectives for modifying nouns. Also, these determiners attract movement of other elements to their left for phonological support, and de also requires this kind of movement. Moreover, Buginese’s definite determiner is similar to de in that they both can reoccur in phrases and form relative clauses. Now let us examine (69) again. Both structures follows Kayne’s (1994) noun final relative clause analysis, regardless of the overt C0 to D0 movement in (69a). Before we go further in comparing these two analyses, we should be aware of two problems with Kayne’s analysis. One is that the movement of IP to Spec of DP violates the minimality requirement put forth by Chomsky (1993). The other is that this movement is a remnant 60 movement that contains an empty category that is not governed by its antecedent, the moved NP which is in the Spec of CP. With respect to the first problem, Kayne suggests that the problem can be solved by abstracting the incorporation of CO to Do. As to the second problem, Kayne does not discuss the issue directly, instead referring to similar remnant movements existing in German, in which the IP movement is a legitimate movement. It is well known that German allows remnant topicalization and that the remnant contains a trace that is not bound by its antecedent. However, it is not the case that any remnant movement is allowed, as (70) shows. (70) a. [NP Ein Buch t, ]j hat [pp fiber die Liebe], niemand t, gelesen. a book has about the love no-one read ‘No one read a book about love.’ (Mtlller 1996, 362) b. *daB[Np ein Bueh t, ]j niemand [pp fiber die Liebe], t, gelesen hat. that a book no-one about the love read has ‘that no one read a book about love (Miiller 1996, 363) To explain (70), Miiller (1996) proposed that remnant movements are legitimate only when the remnant moves to a different category from the category type that the moved item moves to. That is, when an item moves to Spec of XP, its remnant cannot move to another Spec of XP. The only legitimate site for the remnant to move to is a position other than an XP. In (70a), the PP fiber die Liebe adjoins to IP, and its remnant, the NP Ein Buch t,, moves to Spec of TP, which is a different category from the site that the PP moves to and thus is a legitimate movement. In (70b), the PP fiber die Liebe adjoins to IP, while its remnant ein Bach t, adjoins to Spec of the higher IP, and ends up to be ungrammatical. Along the lines of Mitller, Kayne’s IP movement is legitimate since it 61 moves to Spec of DP, while the original moved item, the NP, moves to Spec of CP, which is a different site from the Spec of DP that contains the NP. Now let us go back to Xu and Simpson’s analyses in (69). Xu rejects generating de in D0 because de+NP is an impossible constituent in Chinese. Simpson argues for de being determiner because it developed fi'om zhi, which is a demonstrative in old Chinese. If we set aside the impossibility of de being a complementizer as I discussed above, then both analyses are very strong. Their analyses are better than Cheng and Ning’s analyses in that their analyses both follow the language typology that de is a head-initial functional category, and they both treat de in a general way that de is analyzed as a single item. However, I will choose Simpson’s analysis over Xu’s for two reasons. First, as Kayne states, the minimality requirement can be solved through covert C0 to D0 movement. Therefore, although Xu’s analysis of de as a complementizer moving to D0 seems to solve the minimality requirement, if de is a determiner generated in D0, and a covert C0 to D0 movement takes place, then the overt C0 to D0 movement seems to be unnecessary. Second, de seems to act more like a determiner than a complementizer. In addition to the fact that de does not function like a complementizer, as discussed above, evidence from Lin et al. (2001) shows that de is a determiner since Chinese de phrases show NP deletion even when a relative clause precedes de. A higher DP must exist in the de phrase as the highest maximal projection in order to allow the NP deletion. De must be a determiner that projects the DP; otherwise, it would not allow NP deletion when an adjunct precedes de, as Murasugi and Saito (1990) argue. In the following, I examine Lin et a1 (2001). We have seen that Simpson argues for dc to be a determiner since it is developed from the classical demonstrative zhi. Lin et al.’s analysis further gives empirical 62 evidence to support the claim that de is a determiner that can license a deleted NP. However, Lin et al.’s analysis of the phrases preceding de being base- generated in Spec of DP needs to be further examined. Lin et al. basically claim that because de is able to license the Spec of DP, phrases preceding de can be base-generated in the Spec of DP. Evidence from (62) and (63) that complement phrases (a phrase that receive its thematic role from the head noun of the DP) and adjunct phrases (a relative clause) can be switched without changing meaning further leads them to confirm that whether or not the phrase preceding de has any relationship with the head noun they are all base-generated in the Spec of DP that is projected by de. Although some phrases preceding de show no order preference, this does not hold for adjuncts and complements that are preceding de at the same time. Simpson (1999) cites a Tsao (1997), shown here in (71). It shows that adjunct and complement order in de phrases is significant. (71) a. [wo zuotian tingdao] de [Deng Xiao-ping shishi] de xiaoxi I yesterday hear DE Deng Xiao-ping die DE news ‘the news that Deng Xiao-ping had died which I heard yesterday’ b. *[Deng Xiao-ping shishi] de [wo zuotian tingdao] de xiaoxi Deng Xiao-ping die DE I yesterday hear DE news ‘the news that Deng Xiao-ping had die which I heard yesterday’ (71) shows that the complement Deng Xiaoping shishi ‘Deng Xiao-ping had died’ must occur right before the head noun xiaoxi ‘news’ , while the adjunct wo zuotian tingdao ‘I heard yesterday’, which is a relative clause, is far away from the head noun. If phrases preceding de are base-generated in Spec of DP and make no meaning change when the 63 order is switched, it will be difficult to explain why it is impossible for the two phrases preceding dc to exchange their position. Simpson (1999) shows that (71) can be explained by movement along the lines of Kaynes’s analysis of noun final relative clauses as the derivation shown in (72). (72) a. [pp dc [NP xiaoxi [ Deng Xiao-ping shishi]] DE news die b. [pp[ Deng Xiao-ping shishi], [p dc [NP xiaoxi t, ] A l c. wo zuotian tingdao [pp DengXiaoping shishi de xiaoxi] Iyesterday hear die DE news (1. [pp dC[Cp [pp Deng Xiao-ping shishi d2 xiaoxi], [,p we zuotian tingdao ti ]]] e. [pp [,p wo zuotian tingdao t, ]k [pde[cp [pp Deng Xiao-ping shishi de xiaoxi], tk]] (72a) and (72b) shows that Deng Xiao-ping shishi de xiaoxi ‘the news that Deng Xiao- ping had died’ is a complement clause and that xiaoxi ‘news’ takes Deng Xiao-ping shishi ‘Deng Xiao-ping had died’ as its complement. Then de in D0 requires an IP to fill in its Spec position. Therefore, Deng Xiao-ping shishi moves to precede de resulting in (72b). (72c) shows that the DP, Deng Xiao-ping shishi de xiaoxi, is the object of tingdao ‘hear’. Then (72¢) undergoes relativization, so in (72d) the head noun Deng Xiao-ping shishi de xiaoxi moves out of the IP. Then again the higher de requires the IR, wo zuotian tingdao ‘I heard yesterday’, to fill its Spec of DP. However, the reverse order as in (71b) would not be possible. First, the relative clause has to form wo zuotian tingdao de xiaoxi ‘the news I heard yesterday’, but then there is no way to connect xiaoxi ‘news’ with Deng Xiao-ping shishi ‘Deng Xiao-ping had died’ without violating cyclicity. And there is even no way to connect Deng Xiao- ping shishi with wo zuotian tingdao de xiaoxi since Deng Xiao-ping shishi is not an unsaturated clause. Further, following Kayne’s analysis, since the IP movement to Spec of DP is a remnant movement to a different target than the first moved item, there is no movement violation as stated in Lin et al. To sum up, in this section I have reviewed de phrases with relative clauses. De has been analyzed as a complementizer, a syntactic category, and a determiner. In the complementizer analyses, both Cheng and Ning’s analyses do not follow language typology, and their claim that de is a case marker is problematic. Ning’s analysis of de as a syntactic category also does not follow language typology and taking de as a syntactic category is too language specific. Among these analyses, only Xu, Simpson and Lin et al.’s analyses follow the language typology and have a general analysis for de. They both take the de phrases as DPs, and have competitive analyses. In order to choose a proper analysis, I compared Xu and Simpson’s analyses, and take Simpson’s analysis because a covert movement of C0 to D0 is less costly, and de does not behave like a normal complementizer. Also, evidence from Lin et al. shows that de as a determiner can license the NP deletion. Furthermore, both Xu and Simpson agree that de developed fiom zai which was a demonstrative. A determiner analysis therefore seems to be the best. Lin et al. also argue for de as a determiner, however they argue that there is no movement involved in generating phrases preceding de. These phrases are based-generated in the Spec of DP since the phrases preceding de do not show order difference. However, as shown in (71), order does matter for clausal adjuncts and complement. Simpson’s analysis, on the other hand based on Kayne (1994), is able to account for the order of the 65 clausal adjuncts and complements. Therefore, I take Simpson’s analysis as the basis for my consequent analysis of temporal clauses. As a consequence of adopting this analysis, I also adopt the idea of demonstratives being generated in a lower position of a DP, as shown in (43). 2.6. De phrase with a noun complement clause Having discussed de phrases that contain a relative clause that is an adjunct to the head noun, now we will examine de phrases that contain a complement of the head noun preceding de, as shown in (73). (73) ta mei lai de shishi he not come DE fact ‘the fact that he did not come’ As shown in the English gloss, this kind of noun phrase does not show any movement within the clause. That is, the structure of this kind of noun phrase, although it has a very similar syntactic configuration to a relative clause, has a different relationship between the head noun and the clause. The head noun does not move out from the clause, but instead, the head noun takes the clause as its complement. The structure of the fact that he did not come is shown as (74). 66 (74) DP /\ D0 NP I /\ the N’ /\ N0 CP | A fact that he did not come Similarly, the structure of (73) in Chinese, under Simpson’s (1999) analysis appears in (75). (75) DP /\ CP, D’ A /\ ta mei lai D0 NP ‘he not come’ I /\ DE NO CP 1 | 81118111 I; ‘fact’ (75) shows that the CP is the complement of the head noun. Since the determiner requires phonological support, the CP moves to the Sec of DP. In the following chapters, I will take the syntactic structure that I have discussed here and use as the basis for an investigation of the structure of temporal clauses. I will investigate the de shihou clause first and argue that its structure is a relative clause. Then 67 I will examine yi-qian/yi-hou clauses and argue that these clauses have a structure similar to that shown in (7 5). 68 Chapter 3 De shihou clauses In this chapter, I will argue that the Chinese when-clauses, de shihou ‘DE time’ clauses, are relative clauses. I will demonstrate that de shihou ‘DE time’ is not a particle equal to the English when (Chao 1968), not a linking element that links two sentences (Li and Thompson 1981) and not a postposition heading a postpositional phrase (Gasde and Paul 1996). Instead, shihou ‘time’ is the head noun that follows de in a relative clause. 3.1 Introduction De shihou ‘DE time’ has been treated as a single element that selects for a clause (Chao 1968, Li and Thompson 1981, Gasde and Paul 1996). Chao (1986) treats it as a compound particle which can be equated with when. Li and Thompson (1981) consider it a sentential linking element, but they never give a clear explanation of what syntactic role de shihou plays. Gasde and Paul (1996) also treat it as a single element and argue that it is a postposition. Despite these analyses, it is clear that de shihou clauses have many properties in common with relative clauses. First, as we have seen in Chapter 2 the combination of clause + de + NP is a typical Chinese relative clause construction, and de shihou clauses have the same order, as shown in (1 a). Furthermore, shihou can be modified by a demonstrative as in (lb). (1) a. Ta lai de shihou wo zai chifan. he come DE time I ZAI eat-meal ‘When he came I was eating.’ 69 b. Ta lai de nei-ge s_hihou wo bu zai j ia. he come DE that-CL time I not ZAI home ‘The moment he came I was not home.’ (1a) shows that de shihou clauses are like other de phrases in that they take a clause preceding de. (1b) shows that the demonstrative is generated to the right of de. These two properties of de shihou clauses are like some relative clauses that we have discussed in Chapter 2. However, unlike most relative clauses, de shihou clauses do not allow the demonstrative to occur in a pre-relative position as shown in (2). (2) *Nei-ge ta lai de shihou wo bu zai jia. that-CL he come DE time I not ZAI home Gasde and Paul argue that the impossibility of the demonstrative occurring in the pre—RC position shows that de shihou clauses are not relative clauses. However, Ning (1993) shows that time adjuncts in de phrases are relative clauses. He does not discuss de shihou clauses, but de shihou clauses do contain time adjuncts. In this chapter, I will review Ning’s (1993) account of adjunct relative clauses and show that de shihou clauses are in fact relative clauses. This chapter is organized as follows. In 3.2, I will review Gasde and Paul (1996) and in 3.3 and 3.4 I raise some problems concerning their analysis. In 3.5, I review Ning (1993) and show the characteristics of adjunct relative clauses. I will also discuss issues regarding the position of demonstratives and quantifiers in Chinese relative clauses and further confirm that de shihou clauses are relative clauses. Finally, in 3.6, I will show that de shihou clauses show long distance dependencies, following Geis (1970) and Larson (1990). 70 3.2. Previous studies of de shihou clauses Traditionally, de shihou has been treated as a single element, like when, to introduce a clause to form an adverbial phrase. Chao (1968) claims that the clause before de shihou is an adjectival clause that modifies the noun shihou ‘time,’ and shihou is fi‘equently pronounced in a neutral tone. So de and shihou form a compound particle that can be equated with when. Li and Thompson (henceforth L & T) (1981) describe de shihou as a sentential linking element. Basically, whether de shihou is a particle or a sentential linking element, Chao and L & T both treat de shihou as a single element without any further syntactic explanation; the only discussion of de shihou that provides a syntactic explanation is Gasde and Paul (1996) (henceforth G & P). G & P argue that de shihou is a postposition, and that shihou is not a noun, and therefore not the head of a relative clause. Their reasons for treating de shihou as a postposition and not as a relative clause head noun rely on the following arguments. First, shihou does not take an adj ectival modifier like a regular noun. Second, extraction out of a de shihou clause is possible. Third, de shihou clauses do not allow demonstratives to precede their clauses. In the following, I will summarize these three arguments. G & P argue that shihou ‘time’ is not a noun because it does not take an adjective modifier as in (3a), while a regular noun like shijian ‘time’ does take an adjective modifier, as in (3b). (3) a. *hen duan de shihou very short DE time 71 b. hen duan de shijian very short DE time ‘a very short time.’ According to G & P, shihou is part of the de shihou postposition that takes an IP or a DP as a complement, but that does not take adjectives as complements. (3b) shows that shijian ‘time’ is a noun that can take adjectival phrases as its modifiers, as other nouns do, but shihou in (3a), not being a noun, cannot be modified by an adjectival phrase. Therefore, G & P conclude that shihou is not a noun. The second argument relies on the following sentences. (4) a. Zhei-jian zang fl'fu, ni xi t, de shihou kending yao hua bushao liqi. this-CL dirty clothing you wash DE time certainly must spend much energy ‘This dirty garment, when you wash it you will certainly have to make a lot of effort.’ b. *Xigua,, nongrnin cai-shou t, de jijie hen kuai jiu hui guoqu 1e. watermelon peasant pick-harvest DE season very fast then will pass LE ‘Watermelons, the season when the peasants can harvest them will soon be 9 over. (4a) shows that topicalizing an object out of the temporal clause does not make the sentence ungrammatical. However, (4b) shows that a relative clause does not allow topicalization. That is, if the temporal clause is a relative clause, then the topicalization should cause an ungrammatical result due to the Complex NP constraint. Furthermore, according to G & P, a normal RC would allow a demonstrative to either follow the relative clause, as in (5a,) or precede the relative clause, as in (5b). 72 (5) a. Dai yanjing de nei-ge xuesheng jiao Zhangsan. wear glasses DE that-CL student call ‘That student, who wears glasses, is called Zhangsan.’ b. Nei-ge dai yanjing de xuesheng jiao Zhangsan. that—CL wear glasses DE student call ‘That student that wears glasses is called Zangsan.’ (6) a. Ta lai de nei-ge shihou wo bu zai jia. he come DE that-CL time I not at home ‘When he came, I was not home.’ b. *Nei-ge ta lai de shihou wo bu zai jia. that-CL he come DE time I not at home ‘When he came, I was not home.’ However, a de shihou clause only allows the demonstrative to follow it as in (6a), but not to precede the clause as in (6b). Therefore, the de shihou clause is not a relative construction. 3.3 Problems with Gasde and Paul’s analysis Although at first G & P’s (1996) arguments seem to work, they need to be further examined. In this section and the following one, I question their arguments. First, the fact that shihou can not be modified by an adjectival phrase does not show that it is not a noun. There are nouns that cannot be modified by an adjectival phrase. (7) is an example of this fact. (7) a. Wo bu zhidao ta de guxiang. I not know he DE direction Literately means: ‘I don’t know his direction.’ (I don’t know where he is going.) 73 b. * hen yuan/jin/chang/duan/zhengque de guxiang very far /near/long/short/correct DE direction (7a) shows that quxiang ‘direction’ is a noun because it follows de. However, (7b) shows that it cannot be modified by an adj ectival phrase. If we follow G & P’s analysis of shihou, then quxiang in (7) should not be a noun either. Furthermore, data from English temporal relative clauses also shows a similar pattern to de shihou clauses. (8) a. I have studied for a long time. b. The time that/when John cm I was studying. c. * The longishort time that/when John came I was studying. Note that the time is translated as shijian, as in (3b), in Chinese only when it is not the head of the when-clause, while the time is translated as shihou only when it is the head of a when-clause. When the time in (8c) means shihou, it has the same property as shihou in that it cannot be modified by an adjective as in (3 a); however (8b), without an adjective modifier, is a relative clause. Therefore, a relative clause can have an un-modifiable noun as its head. Now turning to the extraction effect for de shihou clauses, we will see that it is not obvious that (4a) is sufficient to show the real extraction from the adverbial clause. Two independent properties of Chinese, gapless topics and null objects, make (4a) less straightforward than it might appear. (4) is repeated here as (9) for convenience. 74 (9) a. Zhei-jian zang fl'fu, ni xi t, de shihou kending yao hua bushao liqi. this-CL dirty clothing you wash DE time certainly must spend much energy ‘This dirty garment, when you wash it you will certainly have to make a lot of effort.’ b. *Xigua,, nongrnin cai-shou t, de jijie hen kuai jiu hui guoqu-1e. watermelon peasant pick-harvest DE season very fast then will pass-LE ‘Watermelons, the season when the peasants can harvest them will soon be over.’ It is well known that Chinese verbs sometimes take null objects (Huang 1984a). This property also correlates with having gapless topics, as topic is shown in (10). (10) Nei-chang huo, xingkui xiaofang-dui lai de kuai. that—CL fire fortunately fire-brigade come DE quickly ‘That fire, fortunately the fire brigade came quickly.’ (L & T 1981, 96) Thus, having a topic in a sentence does not mean that the topic is derived by movement. Furthermore, even if there is a gap in the sentence, the topic may not have moved from the gap site. Instead, the topic is base-generated in the topic position and associates with a resumptive pronoun or a null object derived by operator movement as shown in (11) (Huang 1984a, 1989). (11) a. XP, [...pronoun, ] b. XP, [Op,[ ....... e, ]] Consider the examples in (12). 75 (12) Speaker A: Ni kanjian John le ma? You see LE MA ‘Did you see John?’ Speaker B: a. John, wo kanjian ta 1e. I saw him LE ‘John, I saw him.’ b. Wo kanjian le e. I saw LE ‘1 saw him.’ c. John, wo kanjian e 1e. I saw LE ‘John, I saw him.’ (12) shows that speaker B can answer speaker A’s question with a topic without a gap as shown in (12a). Speaker B can even utter (12b) with a null object without an overt antecedent. Speaker B can also utter (12c) with a topic and a null object. That is, to have a topic does not mean to have a gap, and the converse is also true. With these characteristics as a background, (9a) can be rephrased in the following way: (13) Speaker A: Wo dei qu xi zhe-jian yifu. I have to go wash this-CL clothes ‘I want to go and wash this clothing.’ Speaker B: Ni xi e de shihou kending yao hua bushao liqi. you wash DE time certainly must spend much energy ‘when you wash it you will certainly have to make a lot of effort.’ As shown in (13), speaker B can utter the sentence with a null object without an overt antecedent, which is like (12b) and has the same meaning as (9a). (9a) has zhejian yifu ‘this clothing’ as a topic with a null object, which is like (12c), and this shows that (9a) '76 could be a combination of a topic with a null object instead of an extraction of the topic out of the embedded clause. However, it is necessary to examine why (9b) does not allow t0picalization. Let us examine topicalization similar to (9b) as shown in (14). ( 14) a. Norrmin cai-shou (zhexie) xigga de shihou kending hui hen mang. peasant pick-harvest (these) waterrnelons DE time certainly will very busy ‘The time when the peasants harvest waterrnelons will be very busy.’ b. Zhexie xigga,, nonmin cai-shou t, de shihou kending hui hen mang. these watermelon peasant pick-harvest DE time certainly will very busy ‘These watennelons, the time when the peasants harvest them will be very busy.’ c. *Xiggg, nonmin cai-shou t, de shihou kending hui hen mang. watermelon peasant pick-harvest DE time certainly will very busy ‘Watermelons, the time when the peasants harvest them will be very busy.’ (14a) is a de shihou clause without topicalization, while (14b) and (14c) are de shihou clauses with topicalization. However, only a definite NP can be topicalized, as in (14b), but not a bare NP, as in (140). If G & P are right about extraction in de shihou clauses, it would be predicted that (14c) should be acceptable, but it is not. Examining (14b) and (14c), it seems that the difference between the two is that (14b) involves a definite NP topicalization, while (14c) involves a bare NP topicalization. This predicts that a definite NP extraction in a sentence similar to (9b) will be good, and this is the case as shown by the acceptability of (15). (15) Na-xie nongzuowu,, nongrnin cai-shou t, de j ij ie hen kuai jiu hui guoqu-1e. That-ME crop peasant pick-harvest DE season very fast then will pass-LE ‘Those crops, the season when the peasants can harvest them will soon be over.’ 77 If this is the case, we should further examine other cases of topicalization. Now consider (16). (16) a. Renren ai chi (zhexie) xigua. Everyone love eat (these) watermelon ‘Everyone loves to eat watermelons.’ b. Egg renren ai chi. watermelon everyone love eat ‘Watermelons, everyone loves to eat (them).’ c. Zhexie xiggg, renren ai chi. these watermelon everyone love eat ‘These waterrnelons, everyone loves to eat (them).’ (16) shows that either a bare NP, as in (16b), or a definite NP, as in (16c), can be topicalized. However, the following topicalizations show an asymmetry between a bare NP and a definite NP. (17) a. Wo jian-guo (neixie) huai shangren. I see-GUO those bad business-men ‘I have seen (those) bad business men.’ b. Nexie hua shangren, wo jian-guo. Those bad business-men I see-Guo ‘Those bad business men,, I have seen (them,).’ c. Nexie hua shangren,, wo jian-guo tamen,. Those bad business-men I see-Guo them ‘These bad business men,, I have seen them,.’ d. Huai shangren, wo jian-guo. bad business-men I see-Guo ‘Bad business men,, I have seen (them,).’ 78 e. *Huai shangren,, wo jian-guo tarnen, Bad business-men I see-GUO them ‘Bad business men,, I have seen (them,).’ (17) shows that both bare NPs and definite NPs can be topicalized, but only definite NPs can associate with a resumptive pronoun as shown in (17c) and (17e). This data implies that bare NP topics cannot be base-generated, but instead must move. Thus, while a definite NP topic can be represented as (18a) or (18b), a bare NP topic can only be represented as (18c). (18) a. XP, [...pronoun, ] b. XP, [Op,[ ....... e, ]] C. XP, [........e, ....] Accordingly, in (l 7) the definite topic can either be associated with a resumptive pronoun or a null operator chain, while a bare NP topic must have moved directly to topic position. Now consider the following topicalizations in relative clauses: (19) a. Xi zhe-jian fl'fir de ren kengding hen xinku. wash this-CL clothes DE person certainly very hard-working ‘The person who washes this clothing must have worked very hard.’ a’ Zhe-jian fl'fu, xi de ren kengding hen xinku. this-CL clothes wash DE person certainly very hard-working b. Mai Zhe-dong fanggi de ren kengding hen you qian. buy this-CL house DE person certainly very have money ‘The person who bought this house must be very rich.’ b’ Zhe dong fangzi, mai de ren kengding hen you qian. This-CL house buy DE person certainly very have money 79 (20) a. Xi @ de ren kengding hen xinku. wash clothes DE person certainly very hard-working ‘The person who washes this clothing must have worked very hard.’ a’ *Yifir, xi de ren kengding hen xinku. clothes wash DE person certainly very hard-working b. Mai f ggi de ren kengding hen you qian. buy house DE person certainly very have money ‘The person who bought this house must be very rich.’ b’ *Fangzi, mai de ren kengding hen you qian. house buy DE person certainly very have money (19) shows that a normal relative clause also allows definite NP topics, but (20) shows that a relative clause does not allow bare NP topics. Following the Complex NP constraint, it seems that (20a’) and (20b’) involve movement, but (19a’) and (19b’) do not. If we follow this argument for the definite NP topic and the bare NP topic in (l 7), (l9a’), (l9b’) and (20a’), (20b’) should contain a trace in the object position of the RC. However, (19a ’) and (19b’) are not ungrammatical, because they involve null operator movement and the null operator does not move out of the relative clause, according to (18). On the other hand, (20a’) and (20b’) are ungrammatical because bare NPs do not undergo null operator movement, and must move directly out of the RC to the topic position. If this analysis is on the right track, the reason why (14b) is fine but (14c) is not can be explained if de shihou clauses are relative clauses. (14b) has a definite NP topicalization similar to (19), which undergoes operator movement, while (14c) has a bare NP topicalization similar to (20), and is a real extraction. The topic in (14b) does not involve direct movement out of the de shihou clause; therefore, (14b) is allowed, but 80 (14c) is not allowed. The parallels between the data in (14) and the data in (19) and (20) seem to suggest that de shihou clauses obey the Complex NP constraint and are thus relative clauses. Gasde and Paul’s analysis of de shihou as a postposition does not seem to be convincing. As I have shown, the inability of shihou to take adjectives is a property of shihou. Also, some usages of English time cannot take adjectives either, as shown in (8c). Furthermore, as shown in data from (13) to (20), the topicalization test is not a good test to claim that de shihou clauses are not relative clauses since definite NP topicalization does not involve movement. In the next section, I will discuss G & P’s third argument, which relates to the position of demonstratives in a relative clause. 3.4 The position of demonstratives and quantifiers in Chinese relative clauses G & P argue that de shihou clauses are not relative clauses because they do not allow a demonstrative to occur in the pre-RC position, as shown in (6), repeated here as (21). (21) a. Ta lai de nei-ge shihou wo bu zai jia. he come DE that-CL time I not at home ‘When he came, I was not home.’ b. *Nei-ge ta lai de shihou wo bu zai jia. that-CL he come DE time I not at home In this section, I show that the position of a demonstrative is also not a good test for defining a structure of relative clause. Consider the following sentences: 81 (22) a. Mary jianfei zhongyu jian-diao ta yao de nei-ge zhongliang. lose-weight finally lose-away she want DE that-CL weight ‘Mary finally lost the weight that she wants.’ b. ??Mary jianfei zhongyu jian-diao nei-ge ta yao de zhongliang. lose-weight finally lose-away that-CL she want DE weight ‘Mary finally lost the weight that she wants.’ c. Magjian-diao de xuduo zhongliang you zeng huilai le. lose De much weight again add back LE ‘Much of the weight that Mary lost came back again.’ (1. ?Xuduo My ]‘im-dique zhongliang you zeng huilai 1e. Much lose DE weight again add back LE (22a) contains a relative clause since it contains an unsaturated clause and follows the typical relative clause patterns (XP + de + NP). (22b) shows that a pre-RC demonstrative is not allowed, but we cannot say that (22b) does not contain an RC. Also, (22c) and (22d) show a similar situation. (22c) allows the quantifier to occur after the RC, while (22d) is less acceptable with the quantifier preceding the RC. Now consider the de shihou clauses again. (23) a. Ta lai de shihou wo bu zai jia. he come DE time I not at home ‘I was not home when he came.’ b. Ta lai de neige shihou wo bu zai jia. he come DE that-CL time I not at home ‘I was not home at that time that he came.’ c. *Neige ta lai de shihou wo bu zai jia. that—CL he come DE time I not at home (24) a. Ta lai de xuduo shihou wo dou bu zai jia. he come DE many time I all not at home ‘Many times, when he came I was not home.’ 82 b. *Xuduo ta lai de shihou wo dou bu zai jia. many he come DE time I all not at home ‘Many times when he came I was not home.’ De shihou clauses behave quite similarly to de zhongliang ‘weight’ clauses in that they are both sensitive to a demonstrative being in the pre-RC position, as in (22b) and (230). When they interact with quantifiers, the de shihou clauses do not allow quantifiers to precede the RC (24b), and de zhongliang clauses, though not as dramatically as de shihou clauses, are still sensitive to the quantifier preceding it (22d). Nevertheless, de zhongliang clauses do allow quantifiers to follow the RC, as in (22c) and (24a). This interaction with quantifiers is opposite to what happens in a normal RC with its subject relativized. A subj ect-relativized RC does not allow the quantifier to follow its RC, but it is fine with an object-relativized RC (Tang 1976, Huang 1983, Hou and Kitagawa 1987), shown as follows: (25) a. Xuduo dai yanjing de xuesheng hen yonggong. many wear glasses DE students very hardworking ‘Many students who wear glasses are hard working.’ (subject-relativized RC) b.* Dai yanjing de xuduo xuesheng hen yonggong. wear glasses DE many students very hardworking ‘Many students who wear glasses are hard working.’ (subj ect-relativized RC) c. Xuduo wo renshi de xuesheng dou hen yonggong. Many I know DE students all very hardworking ‘Many students I know are hard working.’ (object-relativized RC) (1. W0 renshi de xuduo xuesheng dou hen yonggong. I lmow DE many students all very hardworking ‘Many students I know are hard working.’ (object-relativized RC) (Tang 1976, 252-253) 83 (25) shows that only an object-relativized RC allows a quantifier to either precede or follow the RC, while a subject-relativized RC does not allow a quantifier to follow the RC. However, we cannot say that a subj ect-relativized RC is not an RC just because a quantifier cannot occur after the RC. If we pay attention to the meanings of shihou ‘time’ and zhongliang ‘weight’, we can find that they both are semantically non-individual denoting nouns. It is possible that this difference might explain why they behave differently from the individual denoting nouns. If they behave differently from the individual denoting nouns, it does not mean that they are unable to be the head of an RC. Therefore, G & P’s data do not show that de shihou clauses are not relative clauses. In the next section, I shall discuss Chinese adjunct relative clauses, and I will show that other adjunct relative clauses also behave similarly to de shihou clauses. This further confirms that de shihou clauses are in fact adjunct relative clauses. 3.5 Introduction to Chinese adjunct relative clauses In this section, I will review Ning (1993) and Li (1999), who argue that Chinese adjunct relative clauses do in fact show movement. Furthermore, I will show that those adjunct relative clauses that Ning argues have operator movement behave like de shihou clauses in that they do not allow pre-RC demonstratives. Ning (1993) argues that a relative clause must contain a gap in its clause, i.e., a relative clause is an unsaturated expression. Consider a sentence like (26). (26) Ta zuotian zai wu li yong bi xie xin. he yesterday ZAI room inside with pen write letter ‘Yesterday he wrote a letter with a pen in the house.’ 84 (26) has the following structure and semantic roles: (27) [ta [zuotian] [zai wu 1i] [yong bi] xie xin] he yesterday ZAI room inside with a pen write letter agent time place instrument patient With any gap fi'om any semantic role in (27), a relative clause can be formed, as shown in (28). (28) a. [e [zuotian] [zai wu li] [yong bi] xie xin de ren] yesterday ZAI room inside with a pen write letter DE person ‘the man who wrote the letter with a pen in the room yesterday’ b. [ta [e] [zai wu li] [yong bi] xie xin de nei tian] he ZAI room inside with a pen write letter DE that day ‘the day when he wrote the letter with the pen in the room’ c. . [ta [zuotian] [e] [yong bi] xie xin dc difang] he yesterday with a pen write letter DE place ‘the place where he wrote the letter with the pen yesterday’ (1. . [ta [zuotian] [zai wu li ] [e] xie xin de na-zhi bi] he yesterday ZAI room inside write letter DE that-CL pen ‘the pen with which he wrote the letter in the room yesterday’ e. . [ta [zuotian] [zai wu li] [yong bi] xie e de xin] he yesterday ZAI room inside with a pen write DE letter ‘the letter he wrote with the pen in the room yesterday’ However, if no such gap occurs, a de phrase is ungrammatical, as shown in (29). (29) I"ta xihuan juzi de shuiguo he like orange DE fruit ”the fruit that he likes orange’ 85 According to (28), relativization is possible from both argument positions and adjunct positions which denote time, place or manner. Ning further shows that there are four adjunct relative clauses that show null operator movement. These correspond to English single word wh-adjunct operators (as opposed to operators like in which, for whom etc.,) as shown in (30). (30) a. Locative adjunct: ta xiu che de cheku he fix car DE garage ‘the garage where he fixed his car’ b. Time adjunct: ta xiu che de na-ge wanshang he fix car DE that-CL evening ‘the evening 31% he fixed the car’ 0. Manner adjunct: ta xiu che de fangfa he fix car DE way ‘?the way ho_whe fixed the car’ (1. Reason adjunct: ta xiu che de yuanyin he fix car DE reason ‘the reason why he fixed the car’ As the gloss in (30) shows, the English single word wh-operator relative clauses have Chinese counterparts. However, the English pied-piping relative clauses do not have Chinese counterparts, as shown in (31). (31) a. *wo tiaowu de guniang I dance DE girl Intended to mean: ‘the girl with whom I danced’ 86 b. *wo cha xie de xiansheng I polish shoe DE gentleman Intended to mean: ‘the gentleman for whom I polished his shoes’ The English relative clauses in the gloss of (31) involve a preposition with the wh- operator; however, their Chinese counterparts do not have the same mechanism to form pied-piped relative clauses. With this asymmetry between single word wh-operator relative clauses and pied-piped relative clauses, Ning concludes that the relative clauses that can correspond to their English counterparts must use the same strategy as English for movement]. That is, relative clauses like (30) all use operator movement. Li (1999) further illustrates this fact. She shows that Chinese adjunct relative clauses obey island constraints. (32) shows that long distance movement is possible for adjunct relative clauses and (33) shows that the long distance dependency for these adjunct relative clauses must obey island constraints. (33a) shows a Complex NP constraint violation and (33b) shows an adjunct island constraint violation. (32) a. zhe jiu shi [[ women juede [ ta yinggai qu t, nian shu ] de] difang, this exactly is we feel he should go study DE place ‘this is the place where we feel he should go study.’ b. zhe jiu shi [[ta renwei [ nimen t, yingai likai ] de] yuanyin,] this exactly is he think you should leave DE reason ‘this is the reason why he thinks you should leave.’ ' According to Ning, although the Chinese does not have pied-piped wh-operator movement, Chinese uses a resumptive strategy to form relative clauses in (31). Therefore, (31) is frne with this strategy. (1) W0 gen ta tiaowu de guliang. I with her dance DE girl ‘The girl with whom I danced.’ (ii) Wo wei ta cha xie de xiansheng. I for him polish shoe DE gentleman ‘the gentleman for whom I polish his shoes.’ 87 (33) a *zhe jiu shi [[[[ ta xihuan [ t, nian guo shu ] de] ren] de] difang, this exactly is he like read ASP book DE person DE place ‘this is the place where he likes the person that has studied (there)’ b. *zhe jiu shi [[[ruguo ta t, shengqi] ni hui bu gaoxing ] de ]yuanyin, this exactly is if he angry you will not happy DE reason ‘this is the reason(x) that you will not be happy if he gets angry (because of x) (Li 1999, 9) 9 Having enough evidence to show that Chinese adjunct clauses do show island effects, now we should turn to their interaction with demonstratives. Let us examine the following sentences: (34) a. Ta xiu che de nei-ge difang hen zang. he fix car DE that-CL place very dirty ‘That garage where he fixed his car is very dirty.’ a.’ ??Na-ge ta xiu che de difang hen zang. that-CL he fix car DE place very dirty ‘That place where he fixed his car is very dirty.’ b. Ta xiu che de na-ge wanshang, women dou bu zai jia. he fix car DE that-CL evening we all not at home ‘That evening when he fixed the car, we were not home.’ b’.??Na-ge ta xiu che de wanshang, women dou bu zai jia. that-CL he fix car DE evening we all not at home ‘That evening when he fixed the car, we were not home.’ c. Ta xiu che de na-ge fangfa hen hao. he fix car DE that-CL way very good ‘?That way how he fixed the car is good.’ c’. *Na-ge ta xiu che de fangfa hen hao. that-CL he fix car DE way very good ‘That way how he fixed the car is good.’ (1. Ta xiu che de na-ge yuanyin hen qiguai. he fix car DE that-CL reason very strange ‘That reason why he fixed the car is strange.’ 88 d’.*Na—ge ta xiu che de yuanyin hen qiguai. that-CL he fix car DE reason very strange ‘That reason m he fixed the car is strange.’ Not surprisingly, these relative clauses all have the same restriction on pre-RC demonstratives. The head nouns in adjunct relative clauses are also not individual denoting nouns. This strongly supports the idea that the restriction on pro-RC demonstratives and quantifiers is related to the semantics of the head noun. Therefore, the fact that a relative clause does not allow pro-RC demonstratives cannot be used to argue that it is not a relative clause. 3.6. Movement effects in de shihou clauses As I have mentioned briefly in Chapter 1, Geis (1970) argues that when-clauses and clauses following before/after are relative clauses. He believes that the sentences in (35) are derived from (36). (35) a. John anived when Bill left. b. John anived before Bill was fired. c. John departed after Bill left. (36) a. John arrived at the time when Bill left. b. John anived before the time at which Bill was fired. c. John departed after the time gt which Bill left. 89 That is, Geis believes that when-clauses and clauses after before/after are adjuncts of deleted head nouns, i.e., when- clauses and the clauses after before/after are relative clauses. To argue that the sentences in (35) are relative clauses, he gives evidence to show that when-clauses and clauses after before/after observe the Complex NP effect, can be substituted for by pronominal items, and show time adverbial movement. In what follows, I show how Geis argues to support that when-clauses are relative clauses; in Chapter 4 I will discuss his arguments regarding clauses following before/after. In addition to Geis (1970), I will also show Larson’s (1990) analysis for when-clauses. Finally, I will show that de shihou clauses, like when-clauses, also observe the Complex NP constraint, can be substituted for by items that are similar to pronominals, and involve time adverbial movement. The first piece of evidence that Geis gives to support the idea that when-clauses are noun phrases is the fact that when-clauses seem to observe the Complex NP constraint. If when-clauses are noun phrases, they would not allow any element to move out of the when-clauses. The following sentences illustrate this assumption: (3 7) a. Mary began to cry when Harry kissed Lelita. b. *Which woman did Mary begin to cry when Harry kissed? c. *The woman who Mary began to cry when Harry kissed is named Lelita. (3 8) a. Mary began to cry 2_1t the moment when Harry kissed Lelita. b. *Which woman did Mary begin to cry at the moment when Harry kissed? c. *The woman who Mary began to cry at the moment when Harry kissed is named Lelita. (p. 76) 9O (3 73) does not have any element that moves out of the when-clause. (3 7b) is ungrammatical because it involves a wh-movement out of the when-clause, while (37c) relativizes an object out of the when-clause, and it is ungrammatical because the relativization crosses the when-clause. Furthermore, (3 7) behaves like (3 8), which contains at the time as the when-clause’s antecedent. Therefore, Geis argues, when- clauses must be analyzed like the relative clause in (38)2. Further evidence to support the idea that when-clauses are relative clauses is the fact that when-clauses can be pronominalized like a regular noun phrase, as shown in (39). (39) John left when Harry left and George left thir, too. (p. 74) (39) shows that then can substitute for when Harry left. Therefore, when Harry left is very likely a noun phrase. Finally, Geis shows that “when is derived from some time adverbial that moves to clause-initial position” (p. 77), which is like adverbial movement in the relative clause counterparts. This can be illustrated in (40) and (41). (40) a. John arrived when Harry told Mary that she should leave. b. John arrived when Harry told Mary about his desire that she should leave. 2 It should be noted that this argument is not very strong, since movement out of adjuncts would be blocked by the CED (Condition on Extraction Domain) (Huang 1982). 91 (41) a. John arrived at the moment when Harry told Mary that she should leave. b. John arrived at the moment when Harry told Mary about his desire that she should leave. (p. 77-78) (40a) is ambiguous in that when can be interpreted as modifying either the verb told or the verb leave. That is, it can mean that John anived at the time that Harry told Mary the time that she had to leave, and it can also mean that John arrived at the time that Mary was told to leave. (40b) is not ambiguous like (40a) because there is only one reading. (40b) means that John arrived at the time when Harry showed his desire to Mary regarding her leaving time, but it does not mean that John arrived at the time that Mary was told to leave. The sentence in (40b) does not show two readings because the time adverbial in the most embedded clause is blocked by the complex NP his desire that she should leave. This suggests that there must be an adverbial movement from the most embedded clause, and the complex NP blocks the reading from the most embedded clause. The same phenomenon occurs in relative clause counterparts of (40) shown in (41). (41a) has two readings like (40a), while (41b), which contains a complex NP, only has one reading. This is further evidence to support the idea that when-clauses are relative clauses. In summary, Geis’ arguments to support the idea that when-clauses are relative clauses are the following: when-clauses can be the antecedent of the temporal pronoun then. When-clauses are subject to the Complex NP constraint, which is similar to when clauses that contain antecedents in (3 8); thus when-clauses very possibly have then or at some time as their antecedent. Furthermore, he suggests that when-clauses are derived by 92 moving a time adverbial to the initial position of when-clauses, just as in other relative clauses. Following Geis, Larson (1990) discusses the two-reading effect, as shown in (40), and proposes an operator movement to account for it, as shown in (42). (42) a. I saw Mary in New York [CP] when, [she claimed [cpz she would arrive] t, H. b. I saw Mary in New York [CP] when, [she claimed [(3132 she would arrive t,] H. In (42a), the trace is the adjunct of the verb claim and has the reading of seeing Mary at the time when she made the claim. In (42b), the trace is the adjunct of the verb arrive and has the reading of seeing Mary at the time of anival she had earlier specified. According to Larson(1985), in order for the traces to be interpreted, the traces must have case; since when bears Oblique case, the two traces in (42) are case-marked, and therefore, it is possible for the two readings to be accessible’. Whether the movement is a time adverbial or an null operator, both Geis and Larson show that when-clauses involve movement". Now, following Geis, I will show that de shihou clauses also show similar effects to those of when-clauses. Let us consider the following data in Chinese. (43) a. Mary zai John fin Sue de gliihop ku-le. ZAI kiss DE time cry-LE ‘Mary cried when John kissed Sue.’ 3 Larson (1985) discusses how bare NPs and adjunct relative clauses can occur in adverbial position without any apparent casemarking, since in the Government and Binding framework, NPs must be case marked. He proposes that bare NPs and wh-operators of adjunct relative clauses bear Oblique case; hence they are able to occur in adverbial position without apparent case. ‘ Larson’s analysis does not treat temporal clauses as relative clauses. 93 b. *Mary zai John gin t, de shihou ku-le de ren, shi Sue,. ZAI John kiss DE time cry-LE DE person is Sue ‘*The person who Mary cried when John kissed is Sue. (43a) contains a de shihou clause. (43b), which contains (433), relativizes the object Sue out of the de shihou clause and has an ungrammatical result. It is accepted that Chinese relative clauses observe island effects, although its question clauses do not. If de shihou clauses are relative clauses, it follows that de shihou clauses give rise to the Complex NP effect, as (43b) shows. Turning to pronominalization, Chinese does not seem to have a pronominal form corresponding to then; however, we can still test the substitutability of temporal phrases by using the DP na-ge shihou ‘that time’; consider (44). (44) a. Zhangsan zai [san dianl, likai, Lisi ye zai [na-ge shihou], likai. ZAI three o’clock leave also ZAI that-CL time leave ‘Zhangsan left at three, and Lisi left then, too.’ b. Zhangsan zai [Lisi zou le de shihou], daoda, ZAI walk LE DE time arrive Wangwu ye zai [na—ge shihou], daoda. also ZAI that-CL time arrive ‘Zhangsan arrived when Lisi left, and Wangwu arrived then, too.’ In (443), na-ge shihou ‘that time’ refers back to the temporal noun phrase san dian ‘three o’clock’. Na—ge shihou ‘that time’ in (44b) refers back to Lisi zou le de shihou ‘when Lisi left’. Thus, Chinese de shihou clauses can be complex NPs. Now let us examine if de shihou clauses also show two readings like their English counterparts. 94 (45) a. John zai Harry gaosu Mary ta yinggai likai de na-ge shihou daoda. ZAI told she should leave DE that-CL time arrive ‘John arrived at that time when Harry told Mary that she should leave.’ b. John zai Harry, gaosu Mary, youguan ta, yao tar, likai de ginggiu ZAI told about he want her leave DE desire de na-ge shihou daoda. DE that-CL time arrive ‘John arrived at that time when Harry told Mary about his desire for her to leave.’ (45a) has the reading that the time that John arrived can be either the time that Harry told Mary something or the time that Mary should leave; however, (45b), which contains a complex NP, shows that the lower clause reading is not available. That is, the complex NP inside the de shihou clause blocks the operator from moving out of the lower clause, and thus the lower clause reading is disallowed. This chapter has argued that de shihou clauses behave much like relative clauses. In what follows, I will show the structure of de shihou clauses. 3.7. The structure of de shihou clauses In Chapter 2, I stated that Simpson’s (1999, 2000) analysis for the structure of Chinese relative clauses will be used in my analyses of temporal clauses. According to Simpson, a Chinese relative clause as in (46a) has the structure shown in (46b). (46) a. wo zuotian mai de shu I yesterday buy DE book ‘the book I bought yesterday’ 95 b. DP ’,,,,/”1 IP D’ 22: r\~\\\\\\ [wo zuotian mai t,]m D CP ‘1 yesterday buy’ | /\ de NP C’ 13E’ 4:22;. [“‘\\\\r shu, C IP ‘book’ ‘::::::sr T [ tlilm l I have argued that de shihou clauses are relative clauses. Following the structure for relative clauses given in (46b), the de shihou clause in (1), repeated here as (47a), has the structure shown in (47b). (47) a. Ta lai de shihog wo zai chifan. he come DE time I ZAI eat-meal ‘When he came I was eating.’ b. DP .//I IP D ’ 22: [“‘\\\\r [ta t, lai]m D CP ‘he come’ I /\ a de NP C’ 135’ 4:122r l“‘r~\\\ shihou, C IP ‘time’ A 96 That is, shihou moves out of the IP to the Spec of CP, and then the remnant IP moves to the Spec of DP to fulfill Spec head agreement with Do. 3.8. Conclusion In this chapter, I have analyzed de shihou clauses. I argue against G & P’s postpositional analysis to show that shihou is a noun and that de shihou clauses are relative clauses. G & P’s arguments supporting a postposition analysis are based on the following beliefs: first, shihou cannot be modified by adjectives; second, de shihou does not show the Complex NP effect; third, de shihou clauses do not allow demonstratives to precede the clauses before de shihou. However, I argue that it is not sufficient to claim that an element that cannot be modified is not a noun. I gave evidence to show that a noun like quxiang ‘direction’ cannot be modified but is a noun, as in (7). Moreover, English time cannot be modified when it is used as a head of a when-clause, as in (80). I also show that topicalization cannot be used as a method to test movement effects in Chinese, since Chinese definite NP topics are base-generated in topic positions, and only bare NP topics involve real extraction. The NP topics that G & P use to support their claim are definite NP topics, and thus they do not show the Complex NP effect. In addition, G & P argue that regular Chinese relative clauses allow both pre- relative and post-relative demonstratives; however, demonstratives cannot occur before de shihou clauses, and thus de shihou clauses are not relative clauses. Nevertheless, I show that the distribution of demonstratives cannot be used as a way to judge whether or not a phrase is a relative clause. Since quantifiers occur in pre-RC or post-RC positions depending on the linguistic context, the position of quantifiers in relative clauses cannot 97 be used to judge whether a clause is a relative clause. Moreover, I reviewed Ning’s (1993) analysis of Chinese adjunct relative clauses and found that adjunct relative clauses behave the same way as de shihou clauses in that they do not allow demonstratives to occur before them. Finally, following Geis’ analysis for when-clauses, I examine whether de shihou clauses can be substituted for by the DP na-ge shihou ‘that time’, as well as if de shihou clauses observe island effects. Example (44) shows that de shihou clauses can be substituted for by na-ge shihou ‘that time’, and (43) and (45) show that de shihou clauses do involve movement just as regular relative clauses do; this further gives evidence to support the idea that de shihou clauses are relative clauses. 98 Chapter Four: Yi-qian/Yi-hou clauses 4.1 Introduction In the previous chapter, we have seen that, in Chinese, modifiers and complements in noun phrases appear to the left of the head noun. I also showed that de shihou ‘time’ clauses are relative clauses where the modifier of the head noun is to the left of shihou ‘time’. Chinese before/after clauses are very much like the de shihou clauses in that their clauses are obligatorily at the beginning of the whole temporal clause. Geis (1970) treats clauses that are introduced by before/after as relative clauses. If Chinese before/after clauses are similar to de -shihou clauses and the English clauses introduced by before/after, they are very likely to be noun phrases. However, the syntactic category of yi-qian ‘before’ and yi-hou ‘after’ in Chinese temporal clauses has been an issue because they have been treated as both postpositions (Gasde and Paul 1996) and nouns (McCawely 1992). If they are treated as postpositions, the yi-qian/yi—hou clauses will be analyzed as postpositional phrases. If they are treated as nouns, the yi— qian/yi-hou clauses will be analyzed as noun phrases. In this chapter I will further investigate this issue, and will show that a noun phrase analysis for yi-qian/yi-hou clauses is a better choice. As discussed in chapter 1, the function of yi-qian/yi-hou in a temporal clause can also be performed by zhicqian/zhi-hou and qian/hou ‘before/after’. When qian/hou denote time, they indicate ‘before/after’; however, they are also location-denoting words that mean ‘front/back’. Chao (1968) categorizes words such as qian/hou ‘front/back’, 99 shang ‘on, top’, and li ‘in’ as localizers. Li and Thompson (henceforth L & T, 1981) also coined the term locative particles, and these particles denote the location of nouns that they attach to (in this thesis I will refer to them as localizers). These localizers usually co-occur with zai, according to L & T, as shown in (1). (l) a. Shu Q1; zhuozi shang. book ZAI table top ‘The book is on top of the table.’ b. Ta z_ai tushuguan h kanshu. he ZAI library inside study ‘He is studying in the library.’ c. Ta zai shixiang gigg wanshua. he ZAI statue front play ‘He is playing in front of the statue.’ d. Ta gi wuzi hou changge. he ZAI house back sing ‘He is singing at the back of the house.’ The localizers above are called short form localizers; there are also localizers called long form localizers (Ernst 1988, Liu 1998)‘. Long form localizers are the combination of the short forms plus words like bian ‘top/side’, mian ‘side’, and tau ‘end/side’z. Bian mian, and tau can be used interchangeably if they are used to mean ‘side’, as shown in (2). ' In Li and Thompson (1981), the short forms are called monosyllabic locative particles, and the long forms are called disyllabie locative particles. 2 Ernst (1988) considers these words to be suffixes; however since they can follow demonstrative as in zhe mian ‘this side’ and can also follow classifiers and/or as numbers as shown in the following phrases, I consider them words. (i) liang-ge bian/mian ii. liang bian/mian/tou two-CL side two side lOO (2) (3) The meanings of (1) and (3) are the same and they do not show any difference between the short forms and the long forms. While long forms are analyzed as nouns, the syntactic category of short forms is controversial. They have been treated as postpositions (Ernst 1988), nouns (Li 1990), and clitics (Liu 1998). These analyses of short forms all have drawbacks. Since qian/hou are also short form localizers, I will investigate the structure of these locative phrases and argue for a parallel structure between locative phrases and temporal phrases with qian/hou, yi-qian/yi-hou and zhi- qian/zhi-hou. Since locative phrases and temporal phrases seem to have a very close relation, and they have been investigated in more depth than temporal phrases have, in Shu zai zhuozi shang-bian/ shang-mian/ shang-tou. book ZAI table top-side / top-side/ top-side ‘The book is on the table.’ Now turn to compare (1) and (3). a. Shu gai zhuozi shang-bian. book ZAI table top side ‘The book is on top of the table.’ b. Ta z_ai tushuguan li-tou kanshu. he ZAI library inside study ‘He is studying in the library.’ c. Ta pg shixiang gian-bian wanshua. he ZAI statue front-side play ‘He is playing in front of the statue.’ (1. Ta gai wu-zi hou-mian changge. he ZAI house back-side sing ‘He is singing at the back of the house.’ 101 this chapter, I will discuss the locative phrases first. I will propose that short forms are nouns which have become clitics, and with their noun properties, short form localizers take DPs as their complements. I will then show the structural parallels between locative phrases and temporal phrases and apply the structure of locative phrases to the yi-qian/yi- hou clauses. This chapter is organized as follows. In 4.2, I review two analyses of yi-qian/yi- hou and discuss their problems. In 4.3, I discuss the behavior of localizers, the literature regarding their categorization, and the problems of their analyses. In 4.4, I propose that these localizers are in fact nouns that have clitic properties and take DPs as complements. In 4.5, I show the parallel between locative and temporal clauses and show that temporal clauses are noun phrases. Based on the parallel, I provide a structure for yi-qian/yi-hou phrases will be provided. Furthermore, I compare yi-qian/yi-hou phrases with before/after phrases in 4.6, and then in 4.7, I distinguish the differences between yi- qian/yi-qian clauses from de shihou clauses. 4.2 Previous analyses for the category of yi-qian/yi-hou There are two ways of treating yi-qian ‘before’ and yi-hou ‘after’: either as postpositions or as nouns. Gasde & Paul (1996) put yi—qian/yi-hou in the category of postpositions, but McCawley (1992) categorizes yi—qian/yi-hou as nouns. In this section, I will show how these linguists define the categorial status of these words, and some problems with their analyses. Finally, I will show that the noun analysis is a better choice to pursue. 102 4.2.1 Yi-qian/yi-hou as postpositions Gasde and Paul (1996) (henceforth G & P) argue that zai, yi-qian ‘before’, and yi- hou ‘after’ are adpositions: zai is a preposition and yi-qian/yi-hou are postpositions. There is no dispute regarding zai as an adposition, but yi-qian/yi-hou’s status as adpositions is debatable because G & P also consider yi-qian/yi-hou to have a nominal background. Nevertheless, they argue that yi-qian/yi-hou are not nouns since yi-qian/yi- hou clauses do not observe the Complex NP effect. They also show that yi-qian/yi-hou are adpositional lexical heads that are potential governors which can bar traces from being properly governed by their antecedents. In this section, I will review their analyses of yi-qian/yi-hou as postpositions. According to G & P, the structure of the temporal adjuncts in (4) is (5). (4) a. Wo h_uivi iiesfl vi-hog jin cheng. I meeting finish after go downtown ‘After the meeting is over, I will go downtown.’ b. (Zai) huiyi iieshu fl-hou wo jiu jin cheng. ZAI meeting finish after I then go downtown ‘After the meeting is over, I will go downtown.’ (5) PP /\ Spec P’ /\ P0 PostpP (zai) /\ ‘ZAI’ IP Postpo huiyi jieshu yi-hou ‘meeting finish’ ‘after 3 103 They argue that the yi-qian/yi-hou phrases are not noun phrases, but postpositional phrases (PostpP), of which yi-qian/yi-hou are the heads. The Postp0 only projects one bar level, i.e., there is no Spec of PostpP. (5) shows that yi-hou ‘after’ is a final head which projects the postpositional phrase. According to G & P, the possibility of extracting things out of temporal clauses supports the PostpP analysis rather than the noun analysis. The evidence for their claim comes from the following sentence. (6) Zhei-zuo fangzi ni mai yi-qian yinggai qu zhao yi-ge gongzhengren. this-CL house you buy before should go seek one-CL notary public ‘This house, before you buy it, you should consult a notary public.’ According to G & P, (6) has the structure in (7). (7) [Zhei-zuo fangi], [postppD’li mai tl, ] yi-qian] yinggai qu zhao yi-ge this-CL house you buy before should go seek one-CL gongzhengren. notary public ‘This house, before you buy it, you should consult a notary public.’ In (6), zhei-zuofangzi ‘this-CL house’ is topicalized. Following the structure in (7), if the temporal clause, ni mai t yi-qian ‘before you buy’, is an NP, topicalizing zhei-zuofangzi ‘this-CL house’ out of the temporal clause should not be possible due to the Complex NP constraint. However, since (6) is grammatical, the temporal clause must not be a noun phrase. 104 They further argue that the Postp0 only projects one bar level. According to them, the following sentences demonstrate this fact. When zai is present, the topicalization of an object can only be to the position which is to the left of but not lower than zai. (8) a. *_Z__a1_; zhei-zuo fangzi ni mai yi-qian yinggai qu zhao yi-ge gongzhengren. ZAI this-CL house you buy before should go seek one-CL notary public b. Zhei-zuo fangzi gai ni mai yi-qian yinggai qu zhao yi-ge gongzhengren. this-CL house ZAI you buy before should go seek one-CL notary public ‘This house, before you buy it, you should consult a notary public.’ The phenomenon in (8) is analyzed as follows: v (9) a. 1"In: .Zli [PostpP [DP zhei-zuo fangzi l1 [Postp’ [1P Ili mg ti] yi-qian ]]]yinggai ZAI this-CL house you buy before should qu zhao yi-ge gongzhengren. go seek l-CL notary public b. [pp [pp Zhei-zuo fan i ], [p' _z_ai_ [postpphp ni mailt,] yi-qian ]]] yinggai qu zhao this-CL house ZAI you buy before should go seek yi-ge gongzhengren. one-CL notary public ‘This house, before you buy it, you should consult a notary public.’ According to G & P, extracting the object zhei-zuofangzi ‘this house’ in (9a) to the right of zai is blocked because there is no Spec of PostpP. Extracting zhei—zuofangzi ‘this house’ in (9b) to the left of zai is acceptable because Spec of PP, which is headed by zai, 105 offers a place to which zhei-zuofangzi ‘this house’ can move. Therefore, the Postp0 only projects one bar level. Another argument that G & P give for postpositional analysis is that yi-qian/yi- hou can block an antecedent from governing its trace. The following sentence is used to support their postposition analysis. (10) *Huiyi zai jieshu yi-qian/yi-hou wo jin cheng. meeting ZAI finish before/ after I enter town ‘Before/after the meeting was finished I entered the town.’ The structure of (10) is as (l 1) according to G & P. (1 1) PP /\ Spec P’ | /\ huiyi P0 PostpP ‘meeting’ 2m /\ ‘ZAI’ IP Postpo A | t, jieshu yi-hou ‘ finish’ ‘after’ \/ /\ (11) shows that extracting the subject huiyi ‘meeting’ is barred. In order to fit their postposition analysis, G & P follow Chomsky’s (1986a) ECP (empty category principle) and assume that a functional head does not intervene between an antecedent governor and 106 its trace, but a lexical head does’. G & P analyze (11) as similar to a that-trace effect in that the head of PostpP, which is yi-hou, is a possible governor that bars the antecedent from governing the trace in the subject position. An example of a that-trace effect is shown in ( 12). ( 12) a. *Who, do you think that t, left? b. Who, do you think that John saw t,? In (12), that is a potential governor for its extracted subject, which bars who from antecedent governing its trace, while the object extraction in (12b) is legal because the trace is lexically governed by saw. G & P claim that (11) is ungrammatical because yi- hou is a potential governor of the trace in t jieshu ‘t finish’. According to Chomsky (1986a), only a lexical category can 0-govem a phrase’. Since yi-hou blocks the trace fi'om being governed by its antecedent huiyi ‘meeting’, it must be that yi-hou is a postposition that L-marks the [P t jieshu ‘t finish’ and therefore, it is a lexical head. However this situation will not apply to object extraction as in (6), where the trace is properly governed by the verb mai ‘buy’. In summary, G & P argue that yi-qian/yi-hou are postpositions because clauses that contain them do not show Complex NP effects. The postpositional phrase has only one bar level because extraction to the Spec of postpositional phrase is impossible as in 3 The ECP in Chomsky ( 1986a) is defined as follows: a nonpronorninal empty category must be properly governed. a properly governs [3 iff a O-governs or antecedent-govems [3 (p. 17). " a L-marks (Lexically-marks) [3 iff a is a lexical category. 107 (11); instead it is only possible to extract out of the postpositional phrase to the Spec of the prepositional phrase headed by zai. G & P also assume that only lexical categories block an antecedent from governing its trace. With their data in (11) showing that yi-hou blocks an antecedent from governing its trace, they conclude that yi—qian/yi-hou are lexical heads and therefore must be postpositions. 4.2.2. McCawley’s analysis McCawley (1992) argues that yi-qian/yi-hou are nouns since Chinese yi-qian/yi- hou ‘before/after’ are similar to localizers, and localizers behave like nounss. Chinese localizers are a group of words that indicate locations, and they usually occur after nouns. For example, in the Chinese phrase chuang xia ‘under the bed,’ chuang means bed, and xia means under, firnctioning as a localizer. McCawley first argues against the idea that yi-qian/yi-hou are adpositions. According to McCawley, an adposition in Chinese has the following characteristics: (13) i. Objects of an adposition cannot undergo extraction. ii. Objects of an adposition can serve as the antecedent of dou ‘all’. Based on the criteria in (13), (14) yi-qian ‘before’ appears to be an adposition. (14) a. *Nei-chang giusai,, ta e,- yi-gian mai le pijiu. that-CL ball-game he before buy LE beer ‘that ball-game, he bought beer before’ 5 McCawley analyzes yi-qian/yi-hou phrases with NP arguments rather than clausal arguments, i.e., he only discusses NP +yi-qian/yi-hou phrases instead of Clause +yi-qian/yi-hou phrases. 108 b. *ta e, y_i-gian mai 1e pijiu de nei-Chang qiusai, he before buy LE beer DE that-CL ball-game ‘that game that he bought beer before’ c. Ta nei-san-chang giusai . fi-gian 9193; mai 1e pijiu. he that-three—CL ball-game before fl buy LE beer ‘He bought beer before all three of those games.’ If yi-qian is an adposition, (14a) shows that the object of yi-qian, nei-chang qiusai ‘that ball game’, cannot undergo extraction, and (14b) shows that the object of yi-qian can’t form a relative clause while (14c) shows that the object of yi-qian can serve as the antecedent of dou ‘all’. However, with regard to the use of dou, McCawley shows that nouns do not in fact differ from adpositions. For example, a nominal adjunct can also be the antecedent of dou as in (15). Here the adjunct sange ren ‘three people’ is the antecedent of dou, which is similar to (14c) in that nei-san-chang qiusai ‘the three ball games’, the object of yi-qian, is the antecedent of dou. (15) Ta §_an-ge ren de s_h_u dou kan—guo. he three-CL people DE book all see-GUO ‘He has read the books of all three persons.’ (‘all three persons’ not ‘all the books of the three persons’) The following sentences further illustrate the fact that nouns and adpositions behave alike when they occur with dou. 109 (16) a. Zai J iulong—chezhan bi zai qita chezhan dou shufu. ZAI Kowloon-station than ZAI other station all comfortable ‘It is more comfortable in Kowloon Station than in all other railway stations.’ b. Ta de shu bi qita zuozhe de shu dou you yisi. he DE book than other author DE book all have interesting ‘His books are more interesting than all other authors’ books.’ In (16a), bi ‘than’ is a preposition which takes the prepositional phrase zai qita chezhan ‘at other stations’ as its complement. In (16b), bi takes the noun phrase qita zuozhe de shu ‘other authors’ books’ as its complement. Dou ‘all’ in (163) has the adpositional phrase zai qita chezhan ‘at other stations’ as its antecedent, while dou in (16b) has the noun phrase qita zuozhe de shu ‘other authors’ books’ as its antecedent. That is, dou takes either an adpositional phrase or a noun phrase as its antecedent. McCawley firrther shows that an NP combined with a localizer as the object of an adposition can be the antecedent of dou as in (17), but the fact is that no matter whether the NP + localizer is an NP or PP, it always can be the antecedent of dou. That is, if localizers are adpositions, (17) is analogous to (163) because nei sanjia puzi-li ‘inside the three stores’ is the adpositional complement of the preposition zai. If localizers are nouns, (17) is analogous to (16b) because nei sanjia puzi-li ‘inside the three stores’ is a noun phrase complement of the preposition zai. (17) Zhangsan zai nei san-jia puzi-li dou mai guo dongxi. (Ii is the localizer) ZAI that three-CL store-inside all buy GUO things ‘Zhangsan has bought things in all three of those stores.’ 110 (l 7) shows that the NP + localizer combination of nei san jia puzi-li ‘inside the three stores’ behaves like an adpositional phrase as well as a noun phrase; however, McCawley argues that localizers are nouns because the NP + localizer combination can occur either after ba or precede bei, locations where only noun phrases can occur6. (18) a. Haizi-men ba (*zai) jia-1i nong-de hen hao-kan. Child-plural BA ZAI home-inside make-DE very good-looking ‘The children made inside the house pretty.’ b. Jia-li bei haizi-men nong-de luanqibazao. home-inside BEI child-plural make-DE messy ‘The inside of the house was made a mess by the children.’ (18a) shows that ba cannot take the prepositional phrase zai jia-1i ‘at home-inside’ as a complement, but can take jia-Ii ‘home-inside’ as its complement. McCawley argues that yi-qian ‘before’ and yi-hou ‘after’ behave very similarly to the localizers in that they look like adpositions as shown in (14), but in fact they are nouns, since the noun phrase containing them can occur after ba and precede bei as shown in (19). (19) a. Ta ba wanfan yi-hou kanzuo zuihao de xiuxi shijian. he BA dinner after regard best DE rest time ‘He regards after dinner as the best rest time.’ 6 The bei construction and the ba construction are shown in (ia) and (ib) respectively, leaving out the irrelevant details with respect to this thesis. (I) a. NP] + bet +(NP2)+ VP b.NP,+ba+NP2+VP The bei construction is considered to be a passive sentence (Li and Thompson 1981, Li 1990, Shi 1997, and others) in which the NP, is the theme and NP; is the agent. The ba construction is usually considered to be an alternative way of expressing an active sentence (see Li 2001 for various analyses in the literature); the NP, is usually the agent and NP; is the theme. Examples for the two constructions are shown as follows: (ii) Ta bei (Zhangsan) da-le (iii) ta ba pingguo chi-1e he BEI beat-LE he BA apple eat-LE ‘He was beaten by (Zhangsan)/someone. ’ ‘He ate the apple’ 111 b. Wanfan yi-hou bei ta kanzuo zuihao de xiuxi shijian Dinner after BEI he regard best DE rest time ‘After dinner is regarded by him as the best rest time.’ Since the phrases in (19) containing yi—hou ‘after’ are noun phrases, McCawley concludes that yi-qian and yi-hou are nouns. 4.2.3 Problems with G & P’s and McCawley’s analyses G & P (1996) argues that yi-qian/yi-hou ‘before/after’ are postpositions, while McCawley (1992) argues that they are nouns. In this section, I will examine their analyses and I will support McCawley’s analysis as a better choice than G & P’s. G & P’s argument for the temporal clause not being a noun phrase is that there is no complex NP effect when topicalization occurs, as shown in (7), repeated here as (20). V (20) [[Zhei-zuo fangzi], [postpplni mait,] yi-qian]] yinggai qu zhao yi-ge this-CL house you buy before should go seek one-CL gongzhengren. notary public ‘This house, before you buy it, you should consult a notary public.’ (G & P 1996, 283) As we have discussed in Chapter 3, however, topicalization is not a good test to show whether or not a phrase is a complex NP. Furthermore (20) shows that a definite NP topicalization is a non-movement topicalization, which was shown in Chapter 3. 112 However, a bare noun phrase can not undergo topicalization out of an adjunct clause, as (2 1 ) shows. (21) a. Ni mai fangzi fl-gian yinggai qu zhao yi-ge gongzhengren. you buy house before should go seek one-CL notary public ‘Before you buy a house, you should consult a notary public.’ b. *Fanggi, ni mai t, fl-gian yinggai qu zhao yi-ge gongzhengren. house you buy before should go seek one-CL notary public ‘The house,, before you buy it,, you should consult a notary public.’ Now consider (22). (22) Speaker A: Wo xiang mai zhe-zuo fangzi. I want buy this-CL house ‘I want to buy this house.’ Speaker B: Ni yao mai e yi-qian, yinggai qu zhao yi-ge gongzhengren. You want buy before should go seek one-CL notary public ‘Before you buy the house, you should consult a notary public.’ (22) shows that (20) is a non-movement topicalization. In (22), after speaker A mentions zhe-zuofangzi ‘this house’, which is the topic in (20), speaker B can utter the sentence with a null object without an overt antecedent (the antecedent can be realized in the discourse), and the sentence has the same meaning as (20). (20) has zhei-zuofangzi ‘this house’ as the topic with a null object, and this shows that (20) could be the combination of a topic with a null object instead of an extraction of the topic out of the embedded clause. Furthermore, G & P, following Chomsky’s (1986a) ECP, claim that the Postpois a lexical head blocking movement of the subject as in (10), repeated here as (23). 113 (23) *Huiyi zai t, j ieshu yi-qian/yi-hou wo jin cheng meeting ZAI finish before/after I enter town ‘Before/after the meeting was over, I went downtown.’ (G & P 1996, 284) This argument does not seem very convincing. First, as a similar situation to the that- trace effect, that is not a lexical head, but a functional head. That is, assuming Chomsky (1986a), a functional head blocks government; whether or not a lexical head blocks government seems to need firrther examination. Second, what is considered to be ungrammatical due to government blocking as in (23) becomes grammatical if huiyi ‘meeting’ is the main subject as shown in (24). (24) Huiyi zai e kaishi bujiu yi-hou jiu liuhui le meeting ZAI start not-long after then miscarry LE ‘The meeting broke up soon after it started.’ The acceptability of (24) shows that the unacceptability of (23) may be due to subject competition since huiyi ‘meeting’ is right before zai, which is a typical subject position; however, the position after yi-hou has another subject, wo ‘I’, and the result is ungrammatical. Therefore, the unacceptability of (23) may not be due to any island effect, but to a problem of having two subjects at once7. 7 Assuming that Chinese word order is subject + adverb + verb phrase, and phrases containing zai are adverbials, then NPs preceding zai are usually subjects. Evidence from the following contrast shows the preference that NPs (including DP and bare NP) preceding zai are subjects. (The comma indicates that the NP preceding it is a topic.) (1) a. ?Nei-ben shu wo kanwan 1e yi-hou yanj ing hen bu-shufu that-CL book I read-finish LE after eye very uncomfortable ‘That book,, after I finished reading it,, my eyes becarrre very uncomfortable.’ b. *Nei-ben shu Q1 wo kanwan le yi-hou yanj ing hen bu-shufu that-CL book ZAI I read-finish LE after eye very uncomfortable Intended to mean: ‘That book,, after I finished reading it,, my eyes became very uncomfortable.’ 114 Now let us examine McCawley’s analysis. McCawley showed that yi-qian/yi-hou behave like localizers in that they act as adpositions as well as nouns, and the positions in which they can occur, after ba and the subject position of bei constructions, seem to show that they are nouns. However, the position after ba and the subject position of bei constructions are very similar to the positions held by after dinner as shown in the English gloss in (19), repeated here as (25). (25) a He regards after dinner as the best rest time. b. After dinner is regarded by him as the best rest time. It is assumed that the subject position in English is a position only for noun phrases; however this may not always be the case, as shown in (25). In (25a), after dinner occurs after the verb regard and is the subject of the small clause after dinner as the best rest time. In (25b) after dinner is in the subject position of the passive construction. However, after dinner in both sentences is a prepositional phrase, but not a noun phrase. Now let us look at (19), repeated here as (26). c. ??Nei-ben shu, g1 wo kanwan le yi-hou yanj ing hen bu-shufu that-CL book ZAI I read-finish LE after eye very uncomfortable Intended to mean: ‘That book,, after I finished reading it,, my eyes became very uncomfortable.’ (ii) Na-liang chezi, (zai) women chu le chehuo yi-hou jiu bei mai le. that-CL car ZAI we have LE car-accident after then BEI sell LE ‘The car,, after we had a car accident, it, was sold.’ In (ia), Nei-ben shu ‘that book’ is a topic, and the main subject is yanjing If an NP before zai is not forced to be interpreted as subject, (ib) should not be unacceptable since it is possible to have topic before zai as shown in (ii). Having a comma after nei-ben shu as in (ic), it makes the sentence better; however, it is still not very good. The contrast between (ic) and (ii) shows that when zai occurs, a topic is preferred to be a subject. See Chapter 5 for more discussion. 115 (26) a. Ta ba wanfan yi-hou kanzuo zuihao de xiuxi shijian. he BA dinner after regard best DE rest time ‘He regards after dinner as the best rest time.’ b Wanfan yi-hou bei ta kanzuo zuihao de xiuxi shijian. Dinner after BEI he regard best DE rest time ‘After dinner is regarded by him as the best rest time.’ According to Sybesma (1999), noun phrases after ba should be analyzed as subjects of small clauses. For example, in (26a), the small clause is wanfan yi-hou zuihao de xiuxi shijian ‘after dinner as the best rest time’. Since kanzuo ‘regard’ is an unaccusative verb and cannot assign case to wanfan yi-hou, wanfan yi—hou has to move to the position after ba to get cases. (26b) is a passive sentence which is like its English translation, and wanfan yi-hou occurs in the subject position. Wanfan yi-hou in both (26a) and (26b) is the subject, which behaves like its English counterpart shown in the translation. Ifonly NPs can occur after ba and be the subject of bei constructions, it seems reasonable to treat wanfan yi-hou as an NP. However, the similarity between the English and Chinese in (26) may reveal a possibility that wanfan yi-hou may not be a noun phrase. If it is true cross-linguistically that adpositional phrases can be subjects, it is not clear why the following sentences are ungrammatical. (27) a. *Ta ba zai wanfan yi-hou kanzuo zuihao de xiuxi shijian. he BA at dinner after regard best DE rest time ‘He regards after dinner as the best rest time.’ b. *Zai wanfan yi-houbg ta kanzuo zuihao de xiuxi shijian. ZAI dinner after BEI he regard best DE rest time ‘After dinner is regarded by him as the best rest time.’ 8 Kanzuo ‘regard’ is a combination of kan ‘look’ and zuo ‘as’, which is a resultative verb, i.e., an unaccusative verb according to Sybesma (1999). 116 (27) shows that the prepositional zai phrase is not able to occur in case position such as after ba (27a) or in the subject position of the passive sentence (27b). Li (1990) argues that Chinese prepositional phrases behave differently from English prepositional phrases since English, but not Chinese, allows prepositional phrases to occur in case positions. The following sentence further supports this fact. (28) a. (*Zai) jia haoxiang shi ta wancheng duoshu gongzuo de difang. ZAI home seem is he finish most work DE place ‘At home seems to be where he gets most of his work done.’ b. At home seems to be where he gets most of his work done. From (28), we can see that in a raising structure, English PPs can undergo raising to a subject position. Since Chinese PPs cannot, they do not behave like English PPs. Although yi-qian/yi-hou ‘before/after’ seem to behave like their English counterparts and are able to occur in the regard construction, we should also be aware of other syntactic differences between English and Chinese. Given this evidence, it seems more likely that yi-qian/yi-hou phrases are NPs. In summary, G & P’s argument only depends on yi-qian/yi-hou ’s clauses not showing the Complex NP effect, but topicalization in Chinese is not sufficient to prove that there is extraction going on, and therefore is not an argument to say that yi-qian/yi- hou clauses are not noun phrases. On the other hand, McCawley has shown that yi- qian/yi-hou phrases do behave like noun phrases, and I also have shown that Chinese PPs behave differently from their English counterparts. Thus, we cannot claim that because 117 before/after are adpositions, therefore yi-qian/yi—hou ‘before/after’ in Chinese are also adpositions. It seems more reasonable to treat yi-qian/yi-hou ‘before/afier’ as nouns. 4.3 Localizers Recall that Chinese before/after clauses can not only be expressed by yi-qian/yi- hou, they can also be expressed by zhi-qian/zhi—hou, and qian/hou ‘before/after’, and qian/hou are also localizers. This makes the behavior of yi-qian/yi-hou very much like localizers. In this section, I will discuss the syntactic category of localizers. I will first discuss the behavior of localizers, and then review previous studies regarding localizers and argue that localizers in fact are nouns that take the noun phrases that they attach to as complements. 4.3.1 Introduction Localizers are divided into short form localizers and long form localizers. Some short form localizers are given in (29). (29) a. shang ‘top’ b. 1i ‘inside’ 0. qian ‘before/front’ d. hou ‘after, back’ A long form localizer consists of a short form localizer and a noun which means side, as in (30). 118 (30) a. shang-mian ‘top side/surface’ b. li-mian ‘inside’ 0. qian-mian ‘front side’ (1. hou-mian ‘back side’. According to Li and Thompson (1981 ), a locative phrase has the following structure: (31) zai noun phrase- (localizer) (L & T 1981, 390) That is, a typical locative phrase consists of zai, a noun phrase, and a localizer, such as shang in (29a) and shang mian in (30a). The following are examples of locative phrases. (32) a. Shu z_ai zhuozi shang (mian). book ZAI table top-side ‘The book is on top of the table.’ b. Ta zai tushuguan li (mian) kanshu. he ZAI library in side study ‘He is studying in the library.’ c. Ta zai yuanzi gig (mian) wanr. he ZAI yard front side play ‘He is playing in front of the yard.’ Although the locative phrases consist of zai, a noun phrase, and a localizer, as shown in (32), locative phrases that have definite reference do not require localizers to occur (Liu 1998), as in (33). 119 (33) W0 zai jia/xuexiao/tushugpan (li) kan shu. (li is the localizer) I ZAI home/school/library in see book ‘I read at home/at school/ in the library.’ However when a locative phrase is in subject position, the localizer must occur, as shown in (34). (34) a. Jia 1i you liang ge ren. home inside has two CL people ‘There are two people at home.’ b. *Zai jia you liang ge ren. ZAI home has two CL people c. ?Zai jia li you liang ge ren’. ZAI home inside has two CL people d.*Jia you liang ge ren. home has two CL people As to the constituency of the zai + NP + localizer, Liu (1998) shows that zai and the NP do not form a constituent, but the NP and the localizer form a constituent, as shown in (35). (35) a. Wo zai zhuozi shang he y_iz shang dou fang le yi-ben shu. I ZAI table top and chair top all put LE one-CL book ‘I put a book both on the table and on the chair.’ 9 Although (34c) is not good with zai in the subject position, the following sentence is fine: (i) Zai jia li you hang ge ren, zai tushuggn 1i you 51 go ren ZAI home in has two CL people at library in has four CL people a. ‘At home, there are two people, at the library, there are four people.’ b. ‘There are two people at home and there are four people in the library.’ (1) can be translated as (ia) and (ib), but (ia) seems to be a better translation since (i) is contrasting the number of people in two different places; therefore the two places are emphasized and are in topic positions, but not subject positions. 120 b. Wo zai zhuozi he u'zi shang dou fang-1e yi-ben shu. I ZAI table and chair top all put-LE one-CL book ‘I put a book both on the table and on the chair.’ c. *Wo gai zhuozi he zgi vizi shang dou fang 1e yi-ben shu. I ZAI table and ZAI chair top all put LE one-CL book ‘I put a book both on the table and on the chair.’ (35a) shows that the two NP + localizer combinations can be coordinated. (35b) shows that two NPs can also be coordinated, while (35c) shows that two zai + NP combinations cannot be coordinated. Therefore, (35) shows that in a locative construction of zai + NP + localizer, zai + NP is not a constituent, but NP + localizer is a constituent. The following data also support this result. (36) a. Ta zai nar wanr? He ZAI where play ‘Where is he playing?’ b. *Ta nar qian-(mian) wanr? He where front-(side) play ‘Where is he playing?’ 0. Ta zai yuanzi qian (mian) wanr. He ZAI yard front (side) play ‘He is playing in front of the yard.’ (36a) and (36b) both are the possible questions for the answer in (360). Nar in (36a), which means where, corresponds to yuanzi qian (mian) ‘front yard’ in (36c), but the nar ‘where’ in (36b) does not corresponds to zai yuanzi ‘in the yard’ in (36c). That is, in questions, involving the combination of zai + NP + localizer, only NP + localizer is a 121 constituent and is a noun phrase that can be substituted for by a question, but zai + NP is not. Turning to the NP + localizer combination, as we see in (32), nouns can take either long form or short form localizers. However, Li and Thompson (1981) observed that the long forms often go with disyllabie nouns, and the short forms often go with monosyllabic nouns, as in (3 7). (37) a. Ta zai men gig wanr. he ZAI door front play ‘He is playing in front of the door.’ b. *Ta zai men qian-mian wanr. he ZAI door front-side play ‘He is playing in front of the door.’ c. Ta zai fangzi pang-bian wanr. he ZAI house beside-side play ‘He is playing beside the house.’ (1. *Ta zai fangzi pang wanr. he ZAI house beside play ‘He is playing beside the house.’ In addition to this difference, the short form and long form localizers behave differently syntactically. The long forms can follow de, but the short form can only follow zhi 10. This is shown in (38). (38) a. zhuozi shang table top ‘on the table’ ‘° Zhi is considered to be de in Classical Chinese (Gao 1957; Xu 1997; Simpson , 1999, 2000). Although zhi is not commonly used in modern spoken Chinese, NP+ zhi + short form is a legal combination in classical form. 122 b. zhuozi z_hi shang table ZHI top ‘on the table’/ ‘above the table’ c. zhuozi shang-mian table top-side ‘on the table’ (1. Zhuozi Q shang-mian table DE top-side ‘on the table’ e. *zhuozi de shang table DE surface ‘on the table’ f. *zhuozi zhi shang-main table ZHI top-surface ‘on the table’ g *zhuozi de zhi shang-(mian) table DE ZHI top side ‘on the table’ (3 8) shows that when insertion takes place in between the NP + localizer combination, de can only precede the long form and zhi can only precede the short form. (3 8g) shows that de and zhi can not co-occur. F urtherrnore, yi as in yi-qian/yi-hou ‘before/after’, although not a classical form of de, can also occur between the NP and localizer combination and have a different meaning from phrases with zhi and de. In (39), for example, the location that the localizer refers to is above the table, but not on the table. (39) a. zhuozi yi shang table YI top ‘above the table’ 123 b. *zhuozi yi zhi/zhi yi shang table YIZHI/ZHI YI top Intended to mean: ‘on/above the table’ c. *zhuozi yi de /de yi shang table YI DE/DE YI top Intended to mean: ‘on/above table’ Another difference between the short form and long form is that the long form is a full-fledged noun in modern spoken Chinese, but the short form is not, because it cannot occur in all argument positions, as shown in (40). (40) a. Zhuozi shangZShang mian you beizi. table top /top side has cup ‘There are cups on the table.’ b. *Shang you beizi. top has cup c. Shang-mian you beizi. top side has cup ‘There are cups on the top.’ Because of these differences, in the literature, the long forms are categorized as nouns. Because the short forms do not behave like full-fledged nouns, their category is controversial. The syntactic category of the short forms has been argued to be as a postposition (Ernst 1988), a noun (Li 1990), and a clitic (Liu 1998)1 ‘. " There is also a P’ analysis (Troike & Pan 1994) which takes zai and the noun as a constituent and localizers as adpositions. As shown in (27), zai and the noun cannot form a constituent; I will not further review this approach. 124 Ernst (1988) argues that short form localizers should be postpositions because they cannot stand alone in an argument position, and they cannot follow de, which usually is used to test noun-hood. Li (1990) argues that short form localizers are nouns because the NP + localizer combination is able to occur in argument positions. Liu (1998) argues that both the postposition and the noun analyses have their drawbacks; therefore, she proposes that short form localizers are clitics that do not take syntactic positions. In the following section, I will review these three analyses. 4.3.2 Localizers as postpositions Ernst (1988) argues that although both the postposition analysis and the noun analysis have their drawbacks, the postposition analysis is a better choice. That is, if we choose the noun analysis, according to him, we have to adopt the idea that an isolating language like Chinese has obligatorily bound nouns that cannot refer to actual places. They have to depend on other objects to determine places for them to refer to. However, if we choose the postposition analysis, we will have to say that Chinese has head-final adpositions although its verbs and prepositions precede their objects. Under this analysis we must also accept the unusual subcategorization that a preposition zai has to subcategorize a postposition phrase as in fli zhuozi Mg ‘on the table top’. Nevertheless, Ernst argues that the postposition analysis is a better choice. The following is a summary of his arguments. Ernst says that we cannot categorize the short forms as nouns just because the long forms are nouns. He emphasizes that the short forms do not behave like nouns as 125 their long form counterparts do. In fact, there are a number of ways in which the short forms do not behave like nouns. First, they can not follow de: (41) a. chezi de shang-mian car DE top-side b. *chezi de shang car DE top ‘on top of the car’ De is an element in Chinese that can be followed by nominal elements only. (41) shows that the short form cannot follow de. Second, short forms cannot stand alone in argument positions by themselves, as shown in (42b). (42) a. xia-mian you henduo shu under side has many book ‘There are many books underneath.’ b. *_xi_a_ you henduo shu under has many book ‘There are many books underneath. ’ Third, the element preceding a short form must be a noun phrase, and not an adj ectival verb phrase, such as zangzangde ‘dirty’, as in (43). (43) a. Wo pengdao 1e chezi de zhangzangde shang-mian. I touch-to LE car DE dirty top side ‘ Itouched the dirty top of the car.’ 126 b. *Wo pengdao 1e chezi dc zhangzangde shang. I touch-to LE car DE dirty top Finally, Ernst argues that NP+ short form is a relational phrase and it occurs in subject or object positions only when a relational reading is required. According to him, the relational phrases refer to ‘a place defined by position relative to some other object’ (1988, 229). That is, relational phrases refer to places according to the context, but not a constant place. This is shown as in (44). (44) a. *Chezi shang shi mutou zuode. Car top is wood make ‘The top of the car is made of wood.’ b. Chezi de shang-mian shi mutou zuode. Car DE top-side is wood make ‘The t0p of the car is made of wood.’ c. Ta tai ai, kan-bu-dao zhuozi shang. He too short see-not-to table top ‘He is too short; he cannot see on top of the table.’ (1. Under the bed is a lousy place to sleep.’ The subject in (443) requires a reading that refers to an actual place of the car, but not a relational place of car. Since the NP+ short form is relational, it cannot be in a subject position which does not allow a relational reading. (44b) shows that the long form, however, can refer to a part of the car that is made of wood because the long form has the function to refer to places that do not rely on other objects. In (44c) the NP + short form can be in an object position because this sentence is not about the actual place of the table, but the person’s height. That is, object in (44c) requires a relational reading, therefore 127 (44c) is grammatically correct with a NP + short form in the object position. Ernst also draws attention to a parallel phenomenon between English and Chinese relational readings of adpositions. He shows that the gloss in (44c) and the sentence in (44d) both contain a relational locative phrase, like the Chinese NP+ short form, which occurs in object and subject position respectively. This relational function of short forms leads Ernst to argue against the noun analysis because, according to him, it is not acceptable for a noun to have a relational function. In summary, for Ernst, the noun analysis and the postposition analysis both have problems. In the noun analysis, it would be more difficult to explain why short form localizers fail most of the tests for being a noun and why these nouns are bound forms and are relational, but not referential. He claims that the postposition analysis is better because adpositions cannot stand alone and cannot follow de, which are also properties of short form localizers. Ernst concludes that although it is marked to allow a postpositional phrase to be the object of a preposition and allow a head-initial language to have postpositions, these phenomena are also found in other languages and therefore, their markedness is tolerable”. 4.3.3 The noun analysis Li (1990) argues that Chinese is an SVO language and is prepositional in nature. Short form localizers are therefore nominal expressions, and not postpositions. She shows that NP + localizer has the same distribution as other noun phrases in that they '2 Ernst suggests that English also allows P+PP, for exanrple, from out of the darkness (J ackendoff 1973). Also, English notwithstanding or aside might be analyzed as postpositions, for example: (i) His parents’ disapproval notwithstanding, he married the girl. (ii) Table aside, there is a vase. 128 occur in subject positions, object positions, and pre-de positions, i.e., positions that PPs cannot occur in. The following sentences are evidence for this claim. (45) a. Yizi xia hen ganjing. subject chair under very clean ‘Under the chair is clean.’ b. Ni xian jiancha u’zi xia. object you first examine chair under ‘You examine the area under the chair.’ c. fl'zi xia de mao de N chair under DE cat ‘the cat under the chair’ (45a) and (45b) show that NP + short form can occur in both subject and object positions. She further emphasizes that (45c) is very important in that the pre-de position allows only noun phrases, adjectives and clauses, as shown in (46a, b, c), but not prepositional phrases, as shown in (46d). (46d) indicates that the preposition cannot be in pre-de position unless a verb is added to form a relative clause, as shown in (46c)l3 . '3 Ernst (1988) points out that some PPs can occur in pre-de position, as shown in (i) and (ii): (i) dui guojia de re-ai towards country DE love ‘Love of (one’s) country’ (p. 239) (ii) guanyu zheijian shi de wenti about this-CL matter DE problem ‘the problem with this matter’ (p. 240) The difference between (i) and (ii) on one side and (46e) on the other is that the pre-de elements in (i) and (ii) are complements of the head nouns, while (46c) is a relative clause. It seems that hua ‘words’ cannot take a PP complement, while re-ai ‘love’ and wenti ‘problem’ are able to take a complement; however examining the lexical properties of hua ‘word’ is beyond the scope of this dissertation. Moreover, PPs that can be predicates like zai phrases can form a relative clause. Therefore, it may be more precise to say that 129 (46) a. pengyou de fangzi (noun phrase) friend DE house ‘fiiend’s house’ b. haokan de fangzi (adjective) good-looking DE house ‘good-looking house’ c. ta zhu de fangzi (clause) he live DE house ‘the house where he lives’ (1. *dui ta de hua (prepositional phrase) to him DE word ‘words to him’ e. dui ta shuo de hua (clause) to him say DE word ‘words said to him’ Li also shows that NP + localizer does not occur in positions where prepositional phrases occur and prepositional phrases do not occur in places where noun phrases occur as shown in (47). (47) a. *Ta bu neng jia li gongzuo. he not can home in work ‘He cannot work at home.’ b. Ta bu neng zai jia li gongzuo. he not can at home in work ‘He cannot work at home.’ (48) a. (*Zai)xing kong xia shi shuijiao de hao difang. at star sky under is sleep DE good place ‘Under the stars is a good place to sleep.’ non-predicate PPs cannot occur in pre-de phrases alone to form relative clauses, instead of saying that PPs cannot occur in pre-de positions. 130 b. Ta cong (:21) men de hou big lai. he from at door DE back side come ‘He came from behind the door.’ (47) shows that NP + localizer cannot occur in adverbial positions where prepositional phrases can occur. (48) shows that zai cannot occur in argument positions where NP + localizer can occur. In short, Li argues that NP + localizers are in complementary distribution with PPs and have the same distribution as noun phrases. Chinese is an SVO language and has only prepositions, but not postpositions. Therefore, she supports the noun analysis for short form localizers. 4.3.4 Analysis of the short form localizers as clitics By comparing the arguments of the noun analysis and the postposition analysis for short form localizers, Liu (1998) concludes that they both have disadvantages and need other solutions. In order to capture short form localizers’ properties and the way that they interact with other syntactic categories, she proposes that the short forms do not belong to any category; they are simply clitics that have no syntactic positions. She agrees with Ernst that short forms cannot be categorized nouns because they cannot stand alone and they cannot follow de. Moreover, some of the NP + localizers cannot be in argument positions. However, she also claims that the localizers cannot be postpositions because, as Ernst points out, the postposition analysis does not follow the SVO parameter, and since Chinese is basically an SVO language, it should have prepositions instead of postpositions. Furthermore, the postposition analysis fails the coordination test, as shown in (49). 131 (49) a. Xiaoming zai 'ia li he xuexiao dou bu tinghua. ZAI home inside and school all not behave ‘Xiaoming does not behave either at home or at school.’ b. *Houflan he zai gianman dou you yi-ke juzi shu. back yard and at front yard all exist one-CL orange tree ‘There is an orange tree both in the back yard and in the front yard.’ The postpositional analysis takes zai to have either an NP or a postpositional phrase as its complement. If zai+ NP and NP + postposition (localizers) are both adpositional phrases, then they should behave the same syntactically. (49a) shows that NP+ localizer can coordinate with the noun phrase xuexiao ‘school’, but (49b) shows that the prepositional phrase zai qianyuan ‘in the front yard’ cannot coordinate with the noun phrase houyuan ‘baek yard’. This test shows that the prepositional phrase and the postpositional phrase seem to behave differently. Due to the drawbacks of both analyses (noun and adposition analyses), Liu proposes that short form localizers are clitics and do not have a syntactic category. Semantically, they carry a locative feature. According to her, with the locative feature that the short form localizers carry, the NP + short form only occurs in places where a locative phrase is required, such as an adverbial position, an argument position, and a topic position, shown as in (50), (51), and (52). (50) a. Ta fang le yi-ben shu zai nei-zhang zhuozi shang. (Adverbial position) he put LE one-CL book at that-CL table top ‘He put a book on that table.’ b. Wo zai jie shang yujian 1e yi-ge pengyou. I at street top meet LE one-CL friend ‘I met a friend on the street.’ 132 (51) Wu —li you yi-zhi mao. (Argument position) house inside exist one-CL cat ‘There is a cat in the house.’ (52) Laowang nar wo bu qu 1e. (Topic position) there I not go LE ‘As for Laowang’s place, I will not go anymore.’ Since these short form localizers do not have a syntactic category, they do not interfere with the syntactic category of the NP that they attach to; therefore, the NP + short form remains a noun phrase. Liu claims that this analysis covers all the characteristics of localizers. Since they are clitics, they do not stand alone and do not follow de. Since they are clitics and do not have a syntactic position, their combination with NP is able to conjoin with nouns as shown in (49a), and they can occur in argument positions where a locative phrase is required. Unlike the postposition analysis, there is no typological problem about their differences from verbs and other prepositions. There is also no problem with their failing the noun tests. 4.3.5 Problems of Ernst’s analysis As Ernst (1988) and Liu (1998) observe, both the postposition analysis and the noun analysis have their disadvantages. Emst’s observation is true in that short forms do in fact fail most of the tests for noun-hood. Although he agrees that the postposition analysis does not follow Chinese word order typology, the postposition analysis seems to him to be a better choice. Li (1990) also makes important points in her noun analysis in that NP + short form is able to occur in subject, object, and pre-de positions, as shown in 133 (45), and in places where prepositional phrases cannot occur, as shown in (46) and (47). However, Ernst argues against Li by claiming that a full-fledged noun should not be relational. Those NP + localizer combinations that can occur in subject or object positions are places that require a relational locative phrase. Also, it is not sufficient to claim that NP + short form is a noun just because it does not occur in some adverbial positions. Emst’s NP tests provide a strong argument against the noun analysis for short forms. However, if we further examine his analysis, we will see that his account, categorizing short forms as adpositions, is also not a good analysis since they also fail adpositional tests. Ernst claims that relational locative phrases can occur in argument position only when the position requires a location reading. He gives evidence from English, as shown in (44d), to show that it is adpositional phrases that can have this function and concludes that since Chinese NP + short form can occur in subject position, it must behave like its English counterpart and therefore also like adpositional phrases. However, Ernst does not notice that English relational locative phrases can occur in subject positions as well as in normal PP positions, but Chinese NP+localizer combinations do not occur in adpositional positions, as shown in (53) and (54). (53) a. (*Zai) Chuang xia bei renwei shi cang shen de hao difang. ZAI bed under BEI consider is hide body DE good place ‘Under the bed is considered to be a good place to hide.’ b. Wo xihuan cang *(zai) chuang xia. I like hide ZAI bed under ‘I like to hide under the bed.’ 134 (54) 3. W0 kan-bu-dao (*zai) zhuozi shang. I see-not-to ZAI table top ‘I cannot see on top of the table.’ b. *(Zai) zhuozi shang, wo kan-bu-dao”. ZAI table top I see-not-to Intended to mean: “On top of the table, I cannot see (the thing).’ The English glosses in (53a) and (54a) show that both under the bed and on top of the table, which Ernst calls relational phrases, can be in the subject position and the object position. In addition, the glosses in (53b) and (54b) also show that under the bed and on top of the table can occur in regular PP positions. However, (53a) and (54a) show that the NP+ localizer in Chinese, although it is able to occur in the subject and object position, cannot occur in a regular PP position without zai, the preposition. What (53) and (54) show is that NP + localizer is different from Emst’s relational phrase examples in English. The NP + localizers in (53) and (54) do not behave the same both in argument position and in regular PP position. However, the glosses in English show that the PPS under the bed/on top of the table can occur in argument position as in (533) and (54a), and also in the regular PP position as in (53b) and (54b). The ungrammatical Chinese PP with zai in (53a) and (54a) show that PPs in subject and object position are not allowed, while (53b) and (54b) show that NP + localizer cannot be in regular PP positions. The complementary distribution between PP and NP + localizer shows that NP + localizers are consistently behaving like noun '4 The null object in (54b) is understood in discourse. Although (54a) is acceptable without zai, it has a different meaning and structure from (54a) in that it means “1 cannot see the place which is on top of table”, as in (i). That is zhuozi shang ‘on the table’ is the object of kan-bu-dao ‘cannot see’ (i) Zhuozi shang wo kanbudao table top I see-not-to ‘I cannot see on top of the table.’ 135 phrases. Li (1990) also points out that NP + localizers do not occur in PP position, and PPs do not occur in positions where NP + localizers occur, as shown in (47) and (48), repeated here as (55) and (56) for convenience. (55) a. *Ta bu neng jia 1i gongzuo. he not can home inside work ‘He cannot work at home.’ b. Ta bu neng zai jia 1i gongzuo. he not can at home inside work ‘He cannot work at home.’ (56) a. (*Zai)xing kong xia shi shuijiao de hao difang. star sky under is sleep DE good place ‘Under the stars is a good place to sleep.’ b. Ta cong (1gp) men do hog biag lai. he from at door DE back side come ‘He came from behind the door.’ Ernst draws attention to cases like (55), and argues that although some post-subject positions only allow prepositional phrases, it is not sufficient to claim that NP + localizer cannot occur in post-subject position and therefore is a noun. He gives the following sentences to illustrate this point: (57) a. (Zai) qiuchang shang you ren ti qiu. At ball-field on there-be people kick ball ‘On the field some people were playing soccer.’ b. *Tamen qian-main fang le ji ben shu”. they front side put LE a-few CL book ‘They put a few books in front.’ '5 If tamen ‘they’ is used as a modifier of qian-mian ‘front side’ the sentence is fine. But in (57b) tamen is not a modifier of qian- mian, and the sentence is not good without zai preceding qian-mian. 136 c. *Tamen fang le yi da dui dongxi qian-mian. they put LE a big pile thing front side ‘They put a big pile of things in front.’ (Ernst 1998, 227) (57a) shows that both PP and NP + localizer can occur in sentence initial position. But (57b & c) show that, without zai preceding the locative qian-mian ‘front side’, the sentences are ungrammatical, and this might imply that the positions (after subjects and verbs) only allows adpositional phrases, but not noun phrases. Ernst is against this conclusion because there are cases where noun phrases, such as mingtian ‘tomorrow’, can occur in adverbial positions, i.e., in the post-subject position. Therefore, it is not sufficient to claim that all localizers are nouns just because NP + short form does not occur in the adverbial position. I agree with Emst’s idea that the fact that NP + localizers are not able to occur in adverbial position does not mean that they are nouns. However, it is well known that both nouns and adpositions can occur in post-subj ect positions in Chinese. Following this prerrrise, the implication is that whatever cannot occur in post-subject position is neither an adposition nor a noun. However, to claim that the short forms are neither category does not seem correct. Now, let us examine Emst’s examples in (5 7) again. The post subject position in (5 7b) is the position where Ernst argues that both an adposition and a noun are able to occur. With this in mind, the ungrammatical (57 b) is not a good example to argue that NP + short form is a noun, since both adpositions and nouns can occur here. However, in (57c), the predicate fang ‘put’ requires an adposition complement to follow the noun yi da dui dongxi ‘a big pile of things’, but not a noun complement. Therefore, the position following yi da dui dongxi ‘a big pile of things’ is a 137 good place to test for adposition. Since qian-mian ‘front side’ is along form localizer, it is a noun and therefore cannot occur in (57c). If NP + short form is an adpositional phrase, we would predict that an NP + short form is able to occur after yi da dui dongxi ‘a big pile of things’ in (57c). However, (58) shows that without zai, the NP + short form, men qian ‘door front’ cannot occur in the adposition position. (58) * T amen fang le yi da dui dongxi *(zai) men gian. they put LE a big pile thing at door front ‘They put a big pile of things in front of the door.’ Furthermore, (57a) may be acceptable with zai at the sentence initial position, but (53a) and (56a) show that zai is not allowed in subject position, and actually in some situations, the presence or absence of zai affects interpretation". Therefore, we must consider whether zai is really optional in cases where it seems to be optional”. As discussed above, preposition + NP + short form is consistently unable to occur in subject position. Additionally, (53), (54), (56a), and (5 8) show that in an argument position, NP + localizer and Preposition + NP + short form are in complementary distribution. This is more important than the data from post subject adverbial positions, since these positions allow both adpositional phrases and noun phrases as in (56b) and (57b). '6 The following sentences illustrate this fact. (i) Qiang shang wa-le yi-ge dong. wall top dig-LE one-CL hole ‘There is a hole dug in the wall.’ (ii) Zai giang shang wa-le yi-ge dong. ZAI wall top dig-Le one-CL hole ‘Sorrreone dug a hole in the wall.’ (Wen 1957, 25 ) '7 We will have further discussion of the optionality of zai in Chapter 5. 1 3 8 Emst’s arguments that categorize short forms as postpositions are not sufficient. Ernst knows that both the noun analysis and the adposition analysis have their problems, but since the short forms fail most of the noun tests and they cannot stand alone and have to follow NPs, he concludes that they are adpositions. However, as shown above, short forms also fail most of the adpositional tests, and therefore they clearly are not adpositions. A final problem with Emst’s analysis comes from his argument that a noun cannot be relational. (59) shows that this is not true. (59) a. The top of the table is made of wood. b. On t_op of the table is a vase. It is well known that English does not allow singular bare nouns. On top of the table is a relational phrase in (59b). Since singular bare nouns are not allowed, if we follow Emst’s analysis for the short form localizers, we would have categorized top as an adposition. Finally, let us reexamine Emst’s example of the relational sentence in (45c), repeated here as (60). (60) Ta tai ai, kanbudao zhuozi shang. He too short see-not-to table top ‘He is too short; he cannot see on top of the table.’ (60) is equivalent to the following sentence with the long form following zhuozi ‘table’. 139 (61) Ta tai ai, kanbudao zhuozi shang-mian. He too short see-not-to table top-side ‘He is too short; he cannot see on top of the table.’ If the NP + short form in (60) is relational and (61) has the same meaning as (60), does this mean that the NP + long form is also relational? In (60) and (61), both zhuozi shang and zhuozi shang-mian mean ‘on top of the table’ and the positions they occur in are relational in that it is the top of the table that the short person cannot see. The sentence is about the height that the person cannot see and not about the actual top of the table. If Ernst is correct that a noun can not be relational, then the function of long forms as able to refer to relational places and also as real nouns would be difficult to account for. In summary, if NP + localizers are adpositional phrases, they should behave like normal adpositional phrases. However, as shown in (53) and (54), the NP + localizer does not behave like its English counterpart PP. Furthermore, although the inability of the NP + short form to occur in adverbial position does not mean that the NP + short form is a noun, the complementary distribution between the prepositional phrase and the NP + localizer should not be a coincidence. Moreover, whether or not a noun is relational should not be taken as a test for noun-hood. The English top is relational; however, it is a noun, and the same fact holds for long form localizers in Chinese: they are nouns and have a relational function. Being relational is not a sufficient reason to exclude short forms from the noun category. Finally, it is not valid to consider only the adposition and noun analyses and then categorize short forms as adpositions by showing them not to be nouns without testing whether they are adpositions. In the next section, I 140 will show that the short forms in fact have many noun properties, and are therefore not adpositions. 4.4 The analysis Having discussed the problems of Emst’s analysis, we can see that the adposition analysis actually does not present a convincing argument supporting short forms as postpositions. However, the argument that short form localizers are nouns is still controversial. Li’s evidence that short forms are nouns is that they combine with NPs and show up in argument positions, and most of the time, NP + short form is in complementary distribution with PPs. However, whether or not the short forms are nouns is still an issue. It needs to be further explained why the short forms carry noun properties, but they fail most noun tests. In this section, I further investigate the behavior of the short form localizers, and argue that short forms are nominal clitics. The reason to pursue the idea of short forms as nominal clitics is that the grammar should allow words that are developing fi'om their classic forms to modern forms to have a transition stage. There are many monosyllabic nouns in Chinese which are in this stage in that they cannot stand alone and have to attach to other elements to exist in noun phrase positions. Li and Thompson (1981) have shown that many disyllabie words have in fact developed from monosyllabic forms”. The co-occurrence of short form and long form localizers could signify a transition stage. Assuming that short forms are in this transition stage, short forms then are losing their noun properties. Claiming that short forms do not have any '8 Words like this are like mutou ‘wood’ and xiezi ‘shoe’ in modern form; they both has their classical form mu ‘wood/tree’ and xie ‘shoe’ respectively. X ie may still be used in spoken language, but mu is only used in classical Chinese. 141 syntactic category seems to ignore the noun properties they still have. In order to characterize short forms, it is necessary to accept their being clitics and still possessing noun properties. In the next section, I explain why it is preferable to analyze short forms as nominal clitics. I first show the difference between the short forms and the long forms, and look for the short form’s noun properties independent from their attachment to NPs. Counter to Liu (1998), I argue that it is not the case that they are clitics that do not have syntactic positions. 4.4.] Short form localizers are nouns that carry clitic properties In this section, I examine the properties of short form localizers and argue that short form localizers are not adpositions, but nouns that carry clitic properties. First, let us review Chao’s (1968) description of short form localizers. According to Chao, short form localizers are shang ‘up’, xia ‘down’, qian ‘front’, hou ‘back’, nei ‘inside’, Ii ‘inside’, wai ‘outside’, zuo ‘left’, you ‘right’, dong ‘east’, nan ‘south’, xi ‘west’, and bei ‘north’. These localizers, except Ii, occur mostly in certain phrases based on classical Chinese and are free words in this situation. Consider the examples in (62) (62) a. Shang you tiantang, xia you Su-Hang. Top has Heaven below has Su-Hang ‘Above, there is heaven, below there are Su-zhou and Hang-zhou.’ b. Zuo ye bu shi, you ye bu shi. left also not is, right also not is Literally: ‘Neither the left nor the right is right.’ ‘There is no way to please a person.’ 142 Chao further claims that these localizers, when preceding bian ‘side’, mian ‘side’, or tou ‘side’, become long form localizers, which are the modern spoken style. Also, short forms can not follow or precede de, while long form localizers can. Moreover, yi (as in yi-qian ‘before’) and zhi (as in zhi-qian ‘before’) combine easily with short forms, since yi and zhi come from the classical Chinese. Next we should reexamine the properties of localizers, and we will see that Ernst and Liu’s argument against the noun analysis is actually just a matter of different lexical selection. Consider (38) and (39), repeated here together as (63). (63) a. zhuozi shang table top ‘on the table’ c. zhuozi shang-mian table top surface ‘on the table’ e. *zhuozi de shang table DE top ‘on the table’ g. *zhuozi de zhi shang(mian) table DE ZHI top side ‘on the table’ i. *zhuozi yi zhi/zhi yi shang table YI ZHI/ZHI YI top ‘on/above the table’ b. zhuozi z_hi shang table ZHI top ‘on/above the table’ (1. zhuozi d_e_ shang-mian table DE top-surface ‘on/above the table’ f. *zhuozi zhi shang-main table ZHI top-surface ‘on/above the table’ h. zhuozi fl shang table YI top ‘above the table’ j. *zhuozi yi de/de yi shang table YI DE/DE YI top ‘on/above the table’ (63) shows that long forms and short forms behave very similarly in that they both are relational and that they can refer to an area depending on the object to which they relate. However, why are long forms full-fledged nouns in modern spoken Chinese, and short forms not? Long forms are the combination of short form localizers and the following words: bian ‘side’, mian ‘side/surface’, and tau ‘side/end’. Without short forms, when bian, mian, and tau attach to noun phrases, they only refer to part of the object. With the combination of short forms and bian, mian, and tau, long form localizers have both a relational firnction and a referential function. That is, the function that long forms have is a result of short forms plus bian, mian, and tau. Long forms are nouns, can follow de and can occur in argument positions because of bian, mian, and tau, so they are full-fledged nouns in modern spoken Chinese. Another difference between short forms and long forms is that short forms and long forms do not allow the same elements to be inserted inside the NP+ localizer combination. Long forms allow dc to be inserted between NP and localizer, as shown in (63d), and short forms allow only zhi and yi, as shown in (63b& h). This is as Chao (1968) described. Since zhi and yi are from classical Chinese, they go well with short forms. As mentioned before, zhi was the classical form of de, and can only be followed by a noun. De is a good element to test for noun-hood, but not being able to follow de is not sufficient to claim that an item is not a noun. For example, zhi means the body parts that are at the end of our hands and feet, namely fingers and toes respectively. It is definitely a noun because it can be modified by numbers and occur in argument positions, as shown in (64) 19. '9 The same characteristic also applies to nian ‘year’. Nian is a noun; it has to be modified by numbers without classifiers, and it follows de only when a demonstrative is preceding it, as shown below: (i) *wo qu ni jia de nian. (ii) wo qu ni jia de nei nian. I go you home DE year I go you home DE that year ‘the year I went to your home’ ‘that year I went to your home’ 144 (64) Yi-zhi shou you wu zhl’. one-CL hand has five fingers ‘A hand has five frngers.’ Zhi ‘finger/toes’ is a noun. It cannot stand alone, but must attach to some element to be in an argument position or follow de. For instance, it has to attach to shou ‘hand’ to mean fingers or jiao ‘feet’ to mean toes and be followed by tou to be able to follow de or appear in an argument position, as shown in (65)”. (65) a. *Zhe shi wo de zhi. This is I DE fingers ‘This is my finger.’ b. Zhe shi wo de shou-zhi. this is 1 DE hand-finger ‘This is my finger.’ c. Zhe shi wo de zhi-touz'. this is I DE finger-TOU ‘This is my finger.’ d * Zhl shi ren shenti de yi bufen. finger/toe is human being body DE one part ‘Fingers/toes are part of our body.’ e. Shou-th shi ren shenti de yi bufen. fingers is human-being body DE one-part ‘Fingers/toes are part of our body.’ Zhi’ ‘fingers/toes’ is just one of the examples of bound nouns that cannot follow de. Nouns that have developed to be bound forms are common in Chinese”. 2° According to Li & Thonrpson, tou, although it does not have any meaning, like zi as in zhuo-2i ‘table’, is a suffix that is used to attach to a group of bound nouns. More examples are: man-tau ‘Chinese steamed bun’, and mu-tou ‘wood’ 2' Although zhl can mean both frngers and toes, (650) has the core meaning of referring to fingers. 145 The behavior of zhi ‘fingers/toes’ and the short forms is very similar in that they cannot follow de, cannot stand alone and have to attach to some element to exist in modem spoken Chinese”. It is necessary for words to be nouns in order to occur after de or in argument positions, but a word’s inability to occur after de is not a sufficient reason to claim that the word is not a noun as the sentences in (65) show. That is, de can be used to confirm whether or not an element is a noun, but using the test to show that an element is not a noun would incorrectly rule out words like zhi ‘fingers/toes’. I argue that short forms are like zhi ‘fingers/toes’ in that they are nouns that are becoming clitics and still carry their noun property from their classical forms. Short forms cannot follow de, but they can follow zhi, which is the classical form of de as shown in (63), and words after zhi have to be nouns. The following sentence further confirms that the short forms can be used in an argument position in a coordinate structure. (66) Zhe-dong wuzi, shang you diao-long, xia you shi-shi, zhen this-CL house top has carved-dragon below has stoned-lions really qipai. magnificent ‘On top of the house, there are carved-dragons and under the house, there are stone lions. It is really magnificent.’ 2‘ More examples are like er ‘ear’ and chi ‘teeth’: they cannot stand alone and cannot follow de. (i) ??Zhe shi wo de er. (ii) Zhe shi wo de er-duo. (iii) *Zhe shi wo de chi. (iv) Zhe shi wo de ya-chi. Thisisl DEear thisis I DEear ThisisIDE tooth thisis I DE tooth ‘This is my ear.’ ‘This is my ear.’ ‘This is my tooth.’ ‘This is my tooth’ 23 The difference between the short forms and zhi ‘fingers/toes’ is that short forms cannot be modified by numbers. This is not surprising, for relational words are not physical; therefore they cannot be counted. Further zhi is not pronounced as a neutral tone, but the short forms are in most dialects. However whether 01' not the short forms are pronounced as neutral tone varies from dialect to dialect. 146 According to Chao, when the short forms are used as free forms, they mostly occur in certain fixed phrases that are based on classical Chinese. (66) shows that the structure of these certain phrases is somewhat fixed but the structure can be used with different short forms within different contexts", as shown in (66); this kind of structure is a coordination. Actually, those phrases that Chao describes that allow short forms to occur in subject positions are all coordinate structures. (67) further shows that the short forms do still carry their noun properties and that their coordination can occur after de, but coordination of two prepositions cannot occur after de, as shown in (68). (67) Zhe zhang zhuozi de shang(he1xia dou shi mutou zuo de. this CL table DE top and under all is wood make DE ‘This table’s top and bottom are made of wood.’ (68) a. Ta wang xuexiao zou qu. he toward school walk go ‘He is going to school.’ b. Ta cong xuexiao zou lai. he from school walk come ‘He is coming from school.’ c. *Ta de wang he cong xuexiao zou lai zou qu rang ren .....tou hun. he DE toward and from school walk come walk to let people head faint ‘His walking to and from school makes me feel like fainting.’ 2‘ Liu (1998) considers (i) an idiom, and therefore, it does not show that the short forms are able to stand alone, however, this kind of construction with other localizers is also possible and quite productive in formal or classical Chinese. If I change (66) with other localizers as shown in (ii), the sentence is still fine. (i) Shang you tiantang, Li; you Su-Hang. Top has heaven beneath has Su-Hang ‘Above there is Heaven, below there are Suzhou and Hangzhou.’ (ii) Zhe-dong wuzi, shang you diao-long, zuo you shi shi, zhen qipai. this-CL house top has carved-dragon left has stone lions really magnificent ‘On top of the house, there are carved-dragons and to the left of the house, there are stone lions. It is really magnificent.’ 147 From (66) and (67) we can see that, when short forms occur in subject position and follow de, the environment is of a coordinate structure. This implies that short forms have noun properties; however, they are losing these properties. This limits the environments they can occur in as an argument. Liu (1998) argues that short forms are clitics that do not occupy any syntactic position in the phrase structure. However, the fact that short forms can be conjoined and follow de is not compatible with a pure clitic analysis. As (68) shows, the coordination of two prepositions cannot follow de. It must be that short forms still possess their noun properties and that their coordination is able to follow de. They do not have any noun properties, do not have syntactic positions, and are pure clitics. It is not possible for two pure clitics to conjoin and become a noun. As we have seen in section 4.3, NP + localizer combinations do not behave like regular adpositions. They are not able to occur in regular adpositional phrase positions, but only in post subject positions where both a noun phrase and an adpositional phrase can occur. Based on what we have observed about short form localizers, in order to capture the properties of short forms, I would like to propose that short forms are nouns that have clitic properties. The reason to pursue this analysis is that short forms do in fact possess noun properties as we have discussed above; however, their transition to clitics makes their noun properties limited to a certain structure, occurring only in coordination structures. Pure clitics do not coordinate, so short forms are unlikely to be pure clitics. Instead, they are becoming clitics; therefore, they cannot stand alone, fail most of the noun tests in modern spoken Chinese, and have to attach to some element to be in an NP position. Nevertheless, they are nouns from classical Chinese, so they can follow zhi, as in (63b), and their coordination can follow de, as in (67). Short forms occur when zhi and 148 yi show up, and long forms occur when de shows up. In order to characterize these properties, acknowledging their nominal position in the syntactic structure of locative phrases is preferable. 4.4.2 The structure of NP + localizer We have come up with the analysis that short forms are nouns that carry clitic properties. Now we should consider what the structure of a locative phrase is. Before investigating the structure of NP + localizer in general, we should examine the status of NP + short fornr and NP + long form to see if they are compounds or phrases. Furthermore, assuming that Chinese is head-initial, I will investigate the structure of locative phrases and base this analysis on the structure of de phrases provided by Simpson (1999, 2000). I will argue that both long forms and short forms take the noun phrases that they attach to as complements and that the noun phrases move to a higher position to precede the localizers. Now we should examine if NP + localizers and long forms are compounds. To test if an element is a compound, Duanmu (1997) suggests one of the tests suggested in Huang (1984b), Conjunction Reduction. Although single words can be coordinated, parts of compound words cannot be coordinated, as shown in (69)-(71)25 . (69) a. jiu de shu gen xin de shu old DE book and new DE book ‘old books and new books’ b. jiu de gen xin de shu old DE and new de book ‘old and new books’ ’5 (69) is from Duanmu (1997, 137). (70) is from Huang (1984 b, 61) 149 (70) a. huo-che gen gi-che fire-car and gas-car ‘train and automobile’ b. *huo gen gi che fire and gas car (71) a. blackbirds and bluebirds b. *black and blue birds Now consider (72) which contains a NP + localizer combination. (72) a. Zhuozi shang he yjzi shang dou ge fang-zhe yi ben shu. table on and chair top all each put-ZHE one CL book ‘There is a book on the table and the chair.’ b. _Zflrozi he flzi shang dou ge fang-zhe yi ben shu. table and chair top all each put ZHE one CL book ‘There is a book on the table and on the chair.’ (72b) shows that separating NP, zhuozi ‘table’, and the short form, shang ‘top/on’, does not alter the meaning in (72a). This means that NP + localizer is not a compound. Now consider (73), which focuses on testing the long forms. (73) a. zhuozi shang mian gen xia mian table top side and down side ‘on the table and under the table’ b. *zhuozi shang gen xia mian table top and down side ‘on the table and under the table’ 150 (7 3) shows that long forms are words, and separating the short forms from mian in (73b) creates an ungrammatical sentence. Therefore, the NP + localizer combination is a phrase, and long forms are compounds. In the following, I will therefore treat short forms and long forms the same, as the head of NP + localizer. As discussed above, the NP + localizer combination can contain zhi/yi and de (short form with yi/zhi, long form with de). Although yi is a bit different, the distinction between the presence or absence of zhi/dc inserted between the NP + localizer combination is subtle, as shown in (63). According to Chao (1968), the presence or absence of de in a locative phrase will result in a different meaning, perhaps implying that the locative phrase without de is a compound”. However, as shown in (72), NP + localizer is not a compound, and Duanmu (1997) has argued that de insertion is not a reliable test for compound words; therefore, I will treat locative phrases with and without yi/zhi/de both as phrases. Next, I will follow Simpson (1999, 2000) and take yi, zhi and de to be the same functional item and assume that the structure of a locative phrase is similar to the de structure, as shown below. (74) a. b. DP DP /\ /\ D’ D’ /\ A /\ 41 DO NP ['10 NP YI/ZHI N0 DP DB N0 DP A /\ A Shang zhuozi N0 N0 zhuozi ‘on’ ‘tiIble’ Shang mian ‘table’ ‘on’ ‘side’ I \ 26 For exarrrple, inserting de in qiang shang ‘on the wall’ changes the meaning to ‘above the wall’. 151 Since localizers are relational, they require something for them to relate to. That is, localizers are head nouns that select a DP”, such as zhuozi ‘table’ in (74), as their complement. The NP + localizer phrase is a DP whose head takes the NP (projected by the localizers) as a complement. The D0 in the higher DP requires that its specifier must be filled, following Simpson (1999, 2000); as a result, the complement of the localizer moves to fill the Spec of DP, in process similar to that undergone by complement clauses, as discussed in Chapter 2. When yi/zhi occurs in Do, the short form occurs. However, when de occurs in Do, the long form shows up. Furthermore, if D0 is not filled with yi/zhi/de, the Do position is empty and, therefore, there is no difference between NP + short form and NP + long form. The occurrence of these functional items (yi/zhi/de) will interact with the occurrence of localizers and determine the meaning of the whole locative phrase. This structure is consistent with other Chinese noun phrases in that the head noun goes to the right of its complement”. Now we should examine how yi can be equated with zhi and de and occur in D0, especially since yi and zhi/de do not have a close historical relation and yi does not have the same meaning as zhi and de. However, in the following, I will show that yi and zhi are in fact similar in many ways. 27 For convenience, and also following L &T, I have been using NP + localizer to refer to locative phrases; however, this NP can be as big as a DP, for example, (i) nei-go zhuozi shang that-CL table on ‘on that table’ 28 An audience member at NACCL 14 pointed out that the structures in (74) have the DP recursive problem in that (74) cannot prevent a localizer head noun from selecting a DP that contains another localizer as complement, which is ungrammatical, e.g. ’zhuozi shang zuo “table top lefi’. Actually, this can be avoided naturally for semantic reasons. That is, a relational head does not select a relational phrase as a corrrplernent, since it needs a referential element for it to relate to a position. Since both the localizer and the DP containing a localizer are relational, the DP recursive problem is avoided. 152 Simpson (1999, 2000) argues that de is an enclitic definite determiner, and that de developed from the demonstrative zhi. Actually, zhi does possess the property of being a determiner since it is not only a demonstrative, but also a pronoun. Nevertheless, there is no evidence in the literature to show that zhi is a demonstrative but not a definite determiner. The following sentences list common usages of zhi in classical Chinese. (75) a. Yi-ri pu zhi, (from Mengzi, Gaozi Shang) one day put-it-under-the sun ZHI, shi-ri han zhi ten-day make-cold ZHI ‘Expose it (a plant) to the sun for one day, and let it be in cold weather for ten days.’ b. Du zhi yu you yan. (from Hanshu, Caoshen chuan) Figure ZHI want has words ‘The emperor, figures that he, would like to say something.’ c. Zhi er chong you he ru (from Zhuangzi, Xiaoyaoyou) ZHI two animals what know ‘These (The) two small animals know what?’ (What do the two small animals know?) d. xiao da zhi yu small big ZHI law-case ‘large and small law cases’ e. Zhi ma zhi mu da er bu wei zhi ma da (from Mozi Xiaoqit) ZHI horse ZHI eye big then not say ZHI horse big ‘If the horse’s eyes are big, one does not say that the horse is big.’ (75a) and (75b) show that zhi can be used as a pronoun. (750) shows that zhi is used as what is usually called ‘demonstrative’. (75d) shows that zhi rs used like de, and (75e) shows that what is called the demonstrative zhi (the zhi preceding ma ‘horse’) and the zhi (the zhi preceding mu ‘eye’), which is used like de, can co-occur in a phrase. Apparently, 153 zhi is a determiner since it is used as a pronoun. (75c) even shows that zhi is more like a definite determiner, than a demonstrative. First, demonstratives often come in sets. That is, there is one for referring to near distance and another for a long distance, for example, zhe ‘this’ and no ‘that ', respectively, both in Modern spoken Chinese and English. However, there are no words that pair with zhi to show distance. Second, (75c) is a general statement that states a generic truth, that one does not call a horse that has big eyes a big horse. A demonstrative would not be appropriate to function as a generic indicator, and therefore a definite determiner seems more likely to be used in this context”. Now let us turn to the relationship between zhi and yi. As I have shown, zhi and yi, when combined with short form localizers, have very similar functions; for example, Jiang zhi dong ‘Yangtze river zhi east’ and Jiang yi dong ‘Yangtze river yi east’ both refer to an area in relation to the east part of the Yangtze River. The difference is that the phrase with zhi contains the meaning of the small area next to the river, while the phrase with yi denotes that the place starts from the river and the whole area which is east of it. Zhi and yi have similar functions when used with short form localizers. In addition, historically, yi behaves similarly to zhi syntactically. In classical Chinese, according to ’9 Zhi also has the following functions. (i) a. Huang-fir zhi er zi si yan. Huang-fir ZHI two son die YAN(exclamation) ‘Huang-fu and his two sons died!’ b. Shang-jun yu zhi ta-guo. Shang-jun want ZHI other—country ‘Shang-jun wants to go to other countries.’ (ia) shows that zhi can be used as a conjunction connecting Huang-fit and er zi ‘two sons’, and (ib) shows that zhi can be used as a verb. 154 He, et. al. (1985), yi was used as a verb, a preposition, a conjunction and a question pronoun, as shown in (76). (76) a. HuanGong jiu he zhuhou, bu yi_ bing one”. HuanGong nine unite dukes not use soldiers battle-vehicle ‘Huangong united the Dukes nine times without using force.’ (yi is used as a verb) b. J un-zi bu fl yan ju ren, bu yi ren fei educated-man not with-words recommend person not with-person disregard yan31. words ‘An educated man should not recommend a person because of what the person said, and should not disregard a person’s words because of his (low) social status.’ (vi is used as a preposition.) c. Zhu yu zuo wen fl ji zhi”. Bid me make essay YI record it ‘Ask me to write an essay to record it.’ (yi is used as a conjunction.) (1. Yu yi cheng zhi? At YI contain ZHI(it) ‘Where is it contained?’ (yi is used as a question pronoun) (76a) shows that yi can be a verb, since it can be preceded by bu ‘not’ and is followed by the objects bing che ‘soldiers and battle-vehicle’. (76b) shows that yi can be a preposition, since it is followed by a noun such as you ‘words’/ren ‘person’ and occurs in the adverbial position that is after the subject and followed by a verb. (76c) shows that yi can abe a conjunction that connects two verb phrases, zuo wen ‘write essays’ and ji zhi ‘record it’. (76d) shows that yi can be a question pronoun, and it not only can mean what, 3° From “Lun Yu, Xianwen” (Lun Yu is Confucius’ words that were collected by his students.) 3 ' From “Lun Yu, Weilinggong” 155 it can also mean where and how. He et. al. (1985) take yi in yi-qian/yi—hou as a conjunction. They claim: ‘yi precedes lai ‘come’, wang ‘toward’, shang ‘up’, xia ‘down’, dong ‘east’, xi ‘west’, nan ‘south’, bei ‘north’, etc. to give a further explanation in time, place and range.’ (p. 693). The following are examples of this claim. (77) a. Cong ci fl wang yong-feng huanhao. from now YI toward forever love-each other ‘From now on, love each other forever.’ b. You shan yi shang wu liu 1i you xue yaoran fi‘om mountain YI up five six miles has cave deeply ‘Five or six miles up the mountain, there is a deep cave.’ c. Yuji fl nan shu yu Yue, Qiantang yi bei shu yu Wu guo Yuji YI south belongs to Yue, Qiantang YI north belongs to Wu Country ‘The area from Yuji’s southern border toward its south belongs to Yue, the area from Qiantang’s northern border toward its north belongs to Wu Country.’ From (76) and (77), it is obvious that yi is not a verb when it is used with localizers, since NP + yi + localizer does not form a sentence. So yi + localizer is very likely a preposition, conjunction or pronoun; however, yi-qian/yi—hou can occur after a preposition, as shown in (78). (78) W0 zai fl-gian jiu shuo-guo 1e, zhe-jian shi shi xing bu tong de. I at before then say-GUO LE this-CL matter is work not through de ‘1 have said before that this is not going to work.’ Therefore, He et. al. (1985) could be right that yi in yi + localizer is a conjunction. However, we should examine whether yi in yi + localizer is really a conjunction, since, as ’2 From “Gu Wen Guan Zhi, Yu Yang Lou J i” 1 5 6 I will show in (81), yi+ qian/hou can occur at the beginning of the sentence and at the end of a clause to form a temporal phrase. In addition to its role in temporal phrases, after a speech, a speaker can utter (79a), or to state one’s opinion, a speaker can utter (79b), seems to show that yi is not a conjunction. (79) a. Yi-shang shi wo geren de yi-xie yijian. YI-top is I each-person De one-CL opinion ‘The preceding is my own opinion.’ b. Yi-xia shi wo geren de yi-xie yijian. YI-below is I each-person De one-CL opinion ‘The following is my own opinion.’ In (79), yi-shang refers to what the speaker just said and yi-xia refers to what the speaker is going to say. (79) shows that yi-shang/xia are in subject position, which is an argument position that requires a DP. The following yi + localizer also has this property. (80) A: Changjiang fl-bei you shenme difang? Yangtze-river YI-north has what place ‘What places are located to the north of the Yangtze river?’ B: Yi-bei you Xiang, Chuan deng sheng. YI-North has Xiang, Chuan etc. province. ‘To the north is Xiang, Chuan province etc.’ In (80), after A asks the question, B can answer without Yangtze-river. If yi is a conjunction connecting Yangtze-river and north, it would be surprising that yi + bei ‘north’ can substitute for the whole phrase Changjiang yi—bei and occur in the subject 157 position. Therefore, the data suggest that yi cannot be a conjunction or preposition when followed by localizers. Instead, it is very likely to be a pronoun. Now coming back to zhi and yi, zhi and yi do not behave similarly in many aspects in classical Chinese; however, syntactically, they both could be verbs, conjunctions, and pronouns. Also, when zhi/yi occur with other localizers they also behave quite similarly to each other. Furthermore, when zhi/yi occur with qian/hou, they have the same meaning and the same filnction, as discussed before and as shown in (81). (81) a. Wo yi-qian/zhi-qian bu zhidao ta shi Zhongguo-ren. I before/ before not know he is Chinese-person ‘Before, I did not know that he is Chinese.’ a’. Yi-qian/Zhi-qian wo bu zhidao ta shi Zhongguo-ren. before/before I not know he is Chinese-person ‘Before, I did not know that he is Chinese.’ b. Wo dao jia yi-qian/zhi—qian ta zai chi fan ne. I anive home before/before he ZAI eat meal NE ‘Before I came home, he was eating.’ Therefore, when combined with short form localizers, zhi and yi are very likely to have the same syntactic category. De is an enclitic determiner, and de’s historical counterpart zhi behaves like a definite determiner. NP + de + long form is interchangeable with NP + zhi + short form. Now let us examine the following sentences. (82) a. Dajai chifgn de gLan-top, ni bie chi dianxin. everyone eat-meal DE front-side you don’t eat desert ‘Don’t snack before we all have dinner.’ (Chao 1968, 119) 158 b. Dajia chifgn (2110M ni bie chi dianxin. everyone eat-meal ZHI before you don’t eat desert ‘Don’t you eat any snack before we all have dinner.’ c. _Dgrji_a chifan (vil-aia_n, ni bie chi dianxin. everyone eat-meal YI before you don’t eat desert ‘Don’t you eat any snack before we all have dinner.’ The three sentences in (82) have the same meaning. The long form localizer qian-tau corresponds to the short form qian. Since de corresponds to zhi, yi must also correspond to de and zhi; otherwise the parallel meaning and structure of the three sentences will be lost. Also, it would not make sense to analyze yi in (82c) differently from zhi and de in (82a) and (82b). Additionally, if yi were a definite determiner, it would be easier to explain why yi + short form can occur in subject position since Deterrniner + Noun is a DP. In summary, long forms are compounds and their function is like short forms in that they are heads that subcategorize a DP complement to form an NP. Also, the locative phrase is headed by a D0 and takes the NP formed by a localizer and its complement as a complement to form a higher DP. Since the Spec of the higher DP requires its Spec to be filled by an XP, the complement of the localizer has to move to Spec of the higher DP. Furthermore, de, zhi, and yi should be analyzed in the same fashion since they firnction the same in temporal phrases, as shown in (82). 4.5 Temporal clauses and their structure Recall that the rationale of discussing localizers is the fact that qian/hou ‘before/after’ are localizers. They are not just localizers that pick out relational spaces 159 and mean in front/back of some objects or places, but also pick out relational time and mean before/after a certain time. Since qian/hou have this property of being both a spatial and a temporal indicator, in this section I will apply the structure of localizers as the structure of the Chinese before/after clauses. I will further show that Chinese before/after clauses are noun phrases that fit the structure proposed. As we discussed before, in a temporal clause, qian, hou, yi-qian/yi-hou, and zhi- qian/zhi-hou all occupy the same position, i.e., the position after the temporal clause and before the main clause. All the forms have the same meaning when they occur with temporal clauses. Qian/hou are also localizers, and we have discussed the category of localizers and the structure of NP + localizer. Based on the usage of qian/hou both for space and time, it should be that both the spatial and temporal usages of qian/hou have the same structure. In the following, I will give further evidence to support this claim. First, they both have the structure of zai + XP + qian/hou, as shown in (83). (83) a. zai zhuozi qian ZAI table front ‘in front of the table’ b. ?zai mingtian qian ZAI tomorrow before ‘before tomorrow’ c. ?zai ni lai qian ZAI you come before ‘before you come’ Second, the XP in zai+XP+ qian/hou can be as big as a DP that contains a clause. 160 (84) a. Wo zai Lisi chang xuexi de tushugm qian da lanqiu. I often study DE library front play basketball ‘I play basketball in front of the library that Lisi usually studies in.’ b. Wo zai wo he Lisi vuehui de na-duan shijian qian da lanqiu. I I and date DE that-CL time before play basketball ‘I play basketball before the time when Lisi and I date.’ Third, zai + XP is not a constituent, but XP + localizer is a constituent, as shown in (85). (85) a Zhangsan zai zhuozi gian, zai yizi qian dou ge fang-1e yi-ben shu. ZAI table front ZAI chair front both each put-LE one-CL book ‘Zhangsan put a book in front of the table and a book in front of the chair.’ b. *Zhangsan zai zhuozi zai yizi qian dou ge fang-1e yi-ben shu. ZAI table ZAI chair front both each put-LE one-CL book ‘Zhangsan put a book in front of the table and a book in front of the chair.’ 0. Zai wo zuowan gongke hou, ni zuowg fan ian ZAI I do-fmish homework after, you do-finish meal before zheli shi yi-tuan zao. here is a-CL mess ‘After I finished doing homework and before you finished making dinner, it was already a mess.’ d.*Zai wo zuowan gongke, Zai ni zuowan fan qian ZAI I do-finish homework, ZAI you do-finish meal before zheli shi yi-tuan-zao. here is a-mess ‘Before I finished doing homework and you finished making dinner, it was already a mess.’ Finally, in light of Chao (1968), the following two sentences are parallel. (86) a. Dajai chifan degq_i_an-tou. ni bei chi dianxin. everyone eat-meal DE before you don’t e at desert ‘Don’t snack before we all have dinner.’ 161 chifan (zhi) qia_n, ni bei chi dianxin everyone eat-meal ZHI before you don’t eat desert ‘Don’t snack before we all have dinner.’ That is, it is not only yi-qian/zhi-qian/qian that can form the temporal phrase; de + long form can do so as well. If the noun analysis for localizers is correct, then the structure of the Chinese before/after for wo huijia yi-hou/zhi-hou ‘after I went home’ is likely to be as shown in DP /\ D, /\ DO NP | /\ zhi/yi NO CP ‘ A hou wo huijia ‘after’ ‘I went home’ That is, hou in (87) takes the clause ‘wo huijia ‘I went home’ as a complement and the clause moves to Spec of DP to fulfill the Spec head requirement. As for a phrase like (88) the structure is as shown in (88b). a. shangge xingqi-er yi-qian Tuesday before ‘before last Tuesday’ 162 D° NP l /\ zhi/yi NO DP A qian shangge xingqi-er ‘ before’ ‘last Tuesday’ J That is, the temporal localizers qian/hou take either a CP or a DP as their complement. If (87) and (88) are correct, they support McCawely’s conclusion that yi-qian/yi-hou phrases are noun phrases and they can occur after ba and be in the argument position of bei constructions. Although McCawely shows this only for NP + yi-qian/yi-hou, the following sentences show that the same facts hold for clause + yi-qian/yi—hou. This supports the idea that they are noun phrases. (89) a. Ta ba taiyang xia-shan yi-hog kanzuo shi zuihao de xiuxi shijian”. he BA sun descent-mountain after regard be best DE rest time ‘He regards after the sun sets as his best resting time.’ b. Taiyang xia-shan yi-hou bei kanzuo shi zuihao de xiuxi shijian. the sun descent-mountain after BEI regard be best DE rest time ‘After the sun sets is regarded as the best resting time.’ c. Taiyang xia-shan vi-hou shi zuihao de xiuxi shijian. the sun descent-mountain after is best DE rest time ‘After the sun sets is the best resting time.’ ’3 Native speakers’ judgment for this sentence is varied, and I found that not all before/after clauses are able to occur after ba. I have no explanation for this situation. However, it is certain that before/after clauses 163 It has been argued that arguments need to be DPs (Longobardi 1994). Chinese ba requires a nominal argument after it. (89a) shows that the yi-hou clause can occur after ba. (89b& c) show that the yi-hou clause can be in an argument position, further confirming that yi-qian/yi-hou clauses are noun phrases In summary, qian/hou function as localizers as well as time-denoting elements. Since they function the same in both spatial and temporal phrases, their structures should be the same. NP +yi-qian/yi-hou are noun phrases, and so is clause + yi—qian/yi-hou as shown in this section. The structure of Chinese yi-qian/yi-hou clauses is similar to the structure of NP+ localizer. The difference is that the former take either a clause or a noun phrase as their complements, while the latter take a noun phrase as their complement. Since yi/zhi are determiners, yi/zhi + qian/hou ‘before/after’ can be used as noun phrases and occur in subject position. 4.6 Comparing yi-qian/yi-hau with before/after In Chapter 3, I showed how Geis (1970) argued that when-clauses are relative clauses. Similar to what I was discussed in Chapter 3, in what follows, I will present Geis’s argument that the clauses that follow before/after are in fact relative clauses. Furthermore, I will examine whether yi-qian/yi-hou clauses behave the same as before/after. Finally, I will compare the differences between before/after and yi-qian/yi- hou clauses, and I conclude that yi—qian/yi-hou and before/after are different syntactically, although yi-qian/yi-hou behave similar to before/after in light of Geis (1970). are definitely able to occur in subject position when shi ‘be’ is the verb. Please note that shi does not behave like English be in that it does not co-occur with adjectives except when an emphasized tone occurs. 164 First of all, according to Geis, one piece of evidence to support the fact that clauses that follow before/after behave like relative clauses is that extraction out of these clauses observes the Complex NP constraint, as (90) and (91) show. (90) a. John departed before Mary kissed the other boy. a’. *Which other boy, did John depart before Mary kissed t, ? b. John dashed for the closet after his father screamed at his brother. b’ *Whose brother did John dash for the closet after his father screamed at t, ? (91) a. John departed before the moment at which Mary kissed the other boy. a’. *Which other boy, did John depart before the moment at which Mary kissed t,? b. John dashed for the closet after the time at which his father screamed at his brother. b’. *Whose brother, did John dash for the closet after the time at which his father screamed t,? (90a & b) and (91a & b) do not involve extraction, while (90a’ & b’) and (91 a’ & b’) involve wh-movement out of the clauses after before/after and violate an island constraint. Since (90) and (91) have the same meaning and relative clauses in (91) observe the Complex NP constraint, Geis believes that before/after clauses in (90) must be relative clauses that also observe the Complex NP constraint”. Second, before/after clauses can be pronominalized by then/that, as (92) shows. 3’ It should be noted that this argument is not very strong, since movement out of adjuncts would be blocked by the CED (Condition on Extraction Domain) (Huang 1982). 165 (92) a. John anived before Bill was fired, and George anived before then,, too. b. John departed after Bill left, and George departed after M, too. (93) a. John arrived before the moment at which Bill was fired, and George arrived before then,, too. b. John departed after the moment at which Bill left, and George departed after M, too. (92) shows that both Bill was fired and Bill left can be substituted by then and that respectively. (93) contains relative clause after before/after, has the same meaning as (92), and shows the same effect as (92). Therefore, Bill was fired and Bill left in (92) are likely to be noun phrases. Third, according to Geis, assuming that these clauses are adjuncts of a deleted noun at some abstract level consistent with the idea that before/after are prepositions that are subcategorized only for noun phrases. Furthermore, as with when-clauses, Geis shows that there is time adverbial movement from before/after clauses that is like adverbial movement in the relative clause counterparts, as shown in (94) and (95). (94) a. Joan left before Harry told her to (leave). b. Joan left after Harry requested her to (leave). (95) a. Joan left before the moment at which Harry told her to (leave). b. Joan left after the moment at which Harry requested her to (leave). 166 According to Geis, (94a) can mean that Joan left before Harry told Joan the time she should leave, and it can mean that Joan left before the time that she was told to leave. The situation also applies to (94b): it may be the case that Joan left after Harry made the request, or it may be the case that Joan left after the time that she was told to leave. (94)’s relative clause counterparts in (95) show the same readings as in (94). Thus, Geis suggests that the sentences in (94) must have adverbial movement that is like the adverbial movement underlying at which in (95), and this adverbial moves from either the first embedded clause or the most embedded clause. Furthermore, this adverbial movement can be blocked by the Complex NP constraint. (96) a. Joan left before Harry told her of his desire for her to leave. b. Joan left after Harry made his request for her to leave. (p. 129) The sentences in (96) are not ambiguous like (94) and (95) in that there is only one reading for each sentence in (96). (96a) means that Joan left when Harry told her about his desire, while (96b) means that Joan left after Harry made his request. The sentences in (96) do not show two readings because the time adverbials in the most embedded clauses are blocked by the two Complex NPs, i.e., of his desire for her to leave and of his request for her to leave. This suggests that there must be an adverbial movement from the most embedded clause; thus, the complex NPs in the two sentences block the reading fiom the most embedded clauses. 167 Now we should turn to yi-qian/yi-hou clauses to examine whether or not yi- qian/yi-hou clauses behave the same as Geis’s analysis of their English counterparts. First, let us examine if yi-qian/yi-hou clauses have relative clause counterparts. (97) a. Lisi zai Zhangsan chifan yi-qian/yi-hou kanshu. ZAI eat-meal before/after reading ‘Lisi read before/after Zhangsan eats/ate.’ b. Lisi zai Zhangsan chifan de shijian yi-qian/yi-hou kanshu. ZAI eat-meal DE time before/after reading ‘Lisi read before/after the time at which Zhangsan eats/ate.’ (97) shows that the yi-qian/yi—hou clause in (97a), without de shijian ‘DE time’ as its head noun, has the same meaning as the yi-qian/yi-hou clause in (97b), which has de- shijian as its head noun. If we follow Geis, (97a) is likely to have a deleted head noun like de-shijian in (97b) at some abstract level”. Now let us examine if clauses before yi-qian/yi-hou can be pronominalized. (98) a. Zhangsan zai Lisi likai, yi-qian/yi-hou likai, ZAI leave before/after leave Wangwu ye zai na-ge shihou, yi-qian/yi-hou likai. also ZAI that-CL time before/after leave ‘Zhangsan left before/after Lisi left, and Wangwu left before/after that time too.’ b. Zhangsan zai Lisi likai, de shike yi-qian/yi-hou likai, ZAI leave DE moment before/after leave Wangwu ye zai na-ge shihou, yi-qian/yi-hou likai. also ZAI that-CL time before/after leave ‘Zhangsan left before/after the moment at which Lisi left, and Wangwu left before/after that time too.’ 168 As I mentioned in Chapter 3, Chinese does not seem to have a pronominal form corresponding to then; however, we can still test the substitutability of temporal phrases by using the DP na-ge shihou ‘that time’. In (98), the yi-qian/yi-hou clauses behave similar to their English counterparts in that the clauses before yi—qian/yi—hou can be substituted for by the DP na-ge shi-hou ‘that time’. Also, (983) has the same meaning as (98b), which has an overt head noun de-shijian. In light of Geis, this seems to suggest that the clause before yi-qian/yi-hou in (98a) is a relative clause that has a deleted head noun. Now let us consider if yi-qian/yi-hou clauses also have two readings that involve time adverbial movement, as discussed above with before/after clauses. (99) a. Zhangsan zai [ta shuo [ta yao likai]] yi—qian likai“. ZAI he say he want leave before leave ‘Zhangsan left before he said he would leave.’ b. Zhangsan zai [ta shuo [ta yao daoda]] yi-hou daoda. ZAI he say he want arrive after arrive ‘Zhangsan arrived after he said he would arrive.’ Both (99a) and (99b) have two readings. (99a) can mean Zhangsan left before he made the claim that he would leave (the first embedded clause reading) and it also can mean he left before the time that he planned to leave (the most embedded clause reading). (99b) 3’ Chinese wh-elements do not perform overt movement; thus I cannot present the movement effect like (90) and (91) in English. Also, movement out of adjuncts is always illegitimate; therefore, the movement effect in (90) and (91) cannot show that before/after clauses violate the Complex NP constraint. 3" I have asked many native speakers regarding sentences that have the two-reading effect involving yi- qian/yi-hou like (99). Sonre speakers agree that there are two readings, but some do not and think there is no most embedded clause’s reading. However, when I translate Geis’s sentences into (99), there are two readings. Zhangsan’s leaving before the time he planned is more reasonable (the most embedded clause reading) than the time before he made the claim, which is the first embedded clause reading. Pragmatics may influence speaker’s judgement. If this is the reason for the two readings, the Complex NP constraint should not prevent the most embedded clause’s reading. Evidence in (101) shows that the most embedded reading is subject to the Complex NP constraint. 169 can mean he arrived after he made the claim that he would arrive. It can also mean he anived after the time he planned to arrive. Now let us insert de shihou before yi-qian/yi- hou in (99) and get (100), which now contains relative clauses and has the same meaning as (99). (100) a. Zhangsan zai [[ta shuo [ta yao likai]] de shihou] yi-qian likai. ZAI he say he want leave DE moment before leave ‘Zhangsan left before the moment at which he said he would leave.’ b. Zhangsan zai [[ta shuo [ta yao daoda]] de shihou] yi-hou daoda. ZAI he say he want arrive DE moment after anive ‘Zhangsan anived after the moment at which he said he would arrive.’ (100a) has the same meaning as (99a); it can mean Zhangsan left before he made the claim about his leaving or it can mean that he left before the time he planned to leave. (100b) means the same as (99b); Zhangsan arrived after he made the claim that he would arrive or Zhangsan arrived after the time he planned to arrive. Now I examine if the most embedded clause’s reading disappears if a complex NP is between the first embedded clause and yi-qian/yi-hou. If so, this phenomenon further implies that there is time adverbial movement going on in yi-qian/yi-hou clauses, and pragrnatics is not the reason for the long distance readings. (101) a. Zhangsan zai [ta tichu [[ta yao likai do] qingqiu]] yi-qian likai. ZAI he raise he want leave DE request before leave ‘Zhangsan left before he made the request that he would leave.’ b. Zhangsan zai [ta tichu [[ta yao likai de] qingqiu]] yi-hou likai. ZAI he raise he want leave DE request after leave ‘Zhangsan left after he made the request that he would leave.’ 170 Please note that the structure of (101) is different from that of (100), although the de + NP in both clauses occurs before yi-qian. The main verb, likai ‘leave’ in (101a) and (101b), can only associate with the verb in the first embedded clause and not with the verb in the most embedded clauses. Thus, syntax does play a main role in getting the two readings in that there is time adverbial movement involved. This also shows that yi—qian/yi-hou clauses behave the same as before/after in the aspect of time adverbial movement. From what we have discussed, it seems that yi-qian/yi-hou clauses behave the same as their English counterparts with respect to Geis’s analysis. This seems to imply that qian/hou take DPs, but not CPs. However, this cannot be valid, although yi-qian/yi- hou and before/after seem to behave similarly in light of Geis. Geis’s analysis of before/after clauses seems to make sense, but the construction of before/after clauses with and without an overt head noun may not have the same structure. That is, (90a) and (91a), repeated here as (102), although they mean the same, do not have the same structure. (102) a. John departed before Mary kissed the other boy. b. John departed before the moment at which Mary kissed the other boy. If all of the clausal phrases that before/after take are considered relative clauses simply just because they have the same meaning as their relative clause counterparts (e. g. (102a) has the same meaning as (102b)), then it is difficult to explain why some prepositions do not take clausal phrases, such as during, as shown in (103). 171 (103) a. *John left during Bill was eating. b. John left during the time that Bill was eating. In (103a), during takes a clause. In (103b) during takes a relative clause (a DP). Both (103a) and (103b) mean the same thing; however, (103a) is ungrammatical. This indicates that the clause after during in (1033) cannot be a relative clause, and this is simply evidence that during cannot take a clause. Along the same lines, before/after can take either a clause or a relative clause, and it should not be the case that clauses that before/after take are relative clauses because they have the same meaning as relative clauses that before/after take. Therefore, the analysis for qian/hou seems correct in assuming that they take either a clause or a noun phrase. Nevertheless, yi—qian/yi-hou do not behave exactly the same as before/after. Now consider the following sentences: (104) a. John left before Bill. b. John left before last Monday. (105) a. Zhangsan zai Lisi yi-qian likai. ZAI before leave ‘Zhangsan left before Lisi.’ b. Zhangsan zai shang-go xingqi-yi yi-qian likai. ZAI last-CL Monday before leave ‘John lefi before last Monday.’ (106) a. John left before anyone. (Geis 1970, 140) b. *John left after anyone. 0. John left before/after everyone. 172 (107) a. *Zhangsan zai renheren yi-qian likai. ZAI anyone before leave ‘Zhangsan left before anyone.’ b. *Zhangsan zai renheren yi-hou likai. ZAI anyone after leave ‘*Zhangsan left after anyone.’ 0. *Zhangsan zai meigeren yi-qian/yi-hou likai. ZAI everyone before/after leave ‘Zhangsan left before/after everyone.’ In (104) and (105), yi-qian/yi-hou seem to behave similarly to before/after in that both before/after and yi-qian/yi-hou can take general noun phases. However, the contrast in (106) and (107) shows that yi-qian/yi—hou are different from before/after in that yi- qian/yi-hou cannot take some quantifiers that before/after can take. Now compare (108a) and (108b). (108) a. John died before/ilLfter 3 o’clock. b. Zhangsan si *(zai) san-dian n’gian/fl-hou. die ZAI three-o’clock before/after ‘Zhangsan died before/after three o’clock.’ (108b) is a sentence with a verb that requires a complement to indicate the result of the event, i.e., a resultative complement, the time that John died. A resultative complement can be a verb, adjective or adposition, but cannot be a noun phrase. If yi-qian/yi-hou are adpositions like before/after, it is not clear why zai is required in (108b). The data from (106) to (108) show that yi-qian/yi-hou and before/after are not exactly the same. (108b) further shows that yi-qian-yi-hou phrases are not PPs like their English counterparts, and thus cannot follow the verb as resultative complements. 173 To conclude this section, yi-qian/yi-hou clauses and before/after clauses both behave similarly in that the clauses they take can be pronominalized and they have the same meaning counterparts in relative clause forms that yi-qian/yi-hou and before/after take. Furthermore, they both have two readings to associate with the main verbs when the clauses include an embedded clause. Also, when the most embedded clause is embedded in a noun phrase, the two readings disappear. The variable occurrence and absence of the two readings suggests that there is a time adverbial movement involved, following Geis. However, although yi-qian/yi-hou behave similarly to before/after clauses along the line of Geis, they do not behave exactly the same as before/after. 4.7 Revising the structure of yi-qian/yi-hau clauses Having discussed the differences between yi-qian/yi-hou and before/after, I have concluded that yi-qian/yi-hou clauses also involve adverbial movement in the sense of Geis (1970). In this section, I will review Larson (1990) and Munn (1991) to show how this time adverbial movement works in more recent literature. Then I will revise my structure of yi-qian/yi-hou clauses and show how an operator movement works in the structure I provided in section 4.5. Since both de shihou clauses and yi-qian/yi-hou clauses involve operator movements, I explain the difference between a de shihou clause structure and a yi—qian/yi-hou clause structure regarding movement. I further give evidence to show that de shihou clauses are relative clauses, while yi-qian/yi-hou clauses are noun complement clauses, although both involve movement. Larson (1990), following Geis’ time adverbial movement analysis, proposes that the two-reading effect occurs because before/after are prepositions that assign case to the 174 time adverbial from the first and most embedded clause. (109) contains the data that Larson is concerned with. (109) a. I saw Mary in New York before [she claimed that [she would arrive] ]. b. I encountered Alice after she swore that she had left. (p. 170) The main verb in (109a) can be associated with the first embedded clause and mean I saw Mary in New York before Mary made the claim of her arrival; also the main verb can be associated with the most embedded clause and mean I saw Mary before the time that she specified she would arrive. (10%) also has two readings: i.e., I encountered Alice after she made the statement and I encountered Alice after the time she specified she had left. Larson analyzes (109) as (110). (110) a. I saw Mary in New York [PP before [cPlO, she claimed [CP2 that she would arrive ]t,]]. a’. I saw Mary in New York [pp before [Cp,O, she claimed [on t, that she would arrive t, 1]]. b. I encountered Alice [pp after [cm 0, she swore [sz that she had left] t,]]. b’. I encountered Alice [PP after [cpl 0, she swore [CP2 t, that she had left t,]]]. According to Larson, a trace must have case in order to be interpreted as a variable bound by an operator; otherwise the null operator, 0, will bind nothing and cause vacuous quantification (Chomsky 1982, Koopman and Sportiche 1982). Since before/after can assign case to the null operator, the traces will also get case through the chain between 175 the operator and trace. The operator that is associated with the first embedded clause, as in (110a &b), and the operator that is associated with the lower embedded clause, as in (110 a’ & b’), are casemarked by before/after, and two readings are available. Evidence for this, according to Larson, comes from the fact that before/after can take NP complements, but while cannot take NPs, and therefore there is no two-reading effect, as shown in (111) and (112). (1 1 1) a. John left after the party. b. *John left while the pm. (112) a. I saw Mary in New York before she claimed that she would arrive. b. I didn’t see Mary in New York while she said she was there. (p. 174) According to Larson, (112a) shows two readings, as I have discussed above, while (112b) only shows one reading: I didn 't see Mary during the time when she was uttering something about her presence in New York. This is the case because, according to Larson, while does not casemark NPs, as (l 1 1b) shows, and thus it cannot casemark null operators as before/after do; therefore (112b) cannot show the reading in which the main verb is associated with the verb phrase in the most embedded clause. Munn (1991) argues that the two-reading effect cannot be derived from casemarking. Munn points out that, according to Larson (1987), in before/after constructions, Antecedent Contained Deletion (ACD) is possible as (114a) shows because before/after assign case to the subject left after the deletion. However, Munn (1991) questions why until/since can take NPs, can license the two-reading effect as 176 shown in (113), and yet cannot casemark ACD in (114b) and (114c). Therefore, ACD is not a good argument. (113) a. I couldn’t leave until John said I could leave. b. I haven’t been there since I told you I was there. (114) a. John left before/after [lp 3111 [p e]]. b. *John sang a song until Bill. b’. John sang a song until Bill did. c. *John has been singing since Bill. c’. John has been singing since Bill has been. Munn suggests that the difference between before/after and since/until is a matter of a semantic difference in the temporal operator, which is similar to the difference between when and while. The difference between when and while, according to Munn, is that when picks out points in time, and while picks out durational events, and this is the reason why when clauses have two readings. Thus, he proposes the following structure for before/after clauses. (115) [ppOi[p'bCf0rC/afier[[p ...t,... D] That is, for Munn the operator moves to Spec of PP and before/after take [P as their complement, instead of CP as Larson claims. In summary, Larson (1990) argues that before/after take CP as a complement and the temporal operator moves to Spec of CP to get case from before/after. On the other 177 hand, Munn thinks that the temporal operator in before/after clauses does not need case and he proposes that the operator moves to Spec of PP and before/after take only an IP. Nevertheless, the purpose of this section is to show how time adverbial movement is analyzed in more recent literature; it is not this thesis’s intention to argue support for either analysis. Now I shall turn to yi-qian/yi-hou clauses. Recall that a structure of yi-qian/yi-hou clauses is provided as follows: (116) DP /\ D’ /\ D° NP | /\ zhi/yi NO CP | A hou wo huijia ‘after’ ‘I went home’ Since a null operator is a wh-movement, the only place for the operator in (116) to move to is the Spec of CP as shown in (117). 178 (117) DP /\ A D, /\ D0 NP I /\ zhi/yi N0 CP | /\ hou O, C’ ‘after’ /\ C0 lP A Wot, huijia ‘I return-home’ l As (117) shows, the whole CP still moves to the Spec of DP; however, inside the CP, an operator movement occurs. This is different from the movement in de shihou clauses, for de shihou clauses involve head raising, while the movement in yi-qian/yi-hou is a null operator movement. The de shihou structure that I proposed in Chapter 3 is repeated here as (118) for convenience. (118) DP /l 1P D’ A l\ [ta t, huijia]m D0 CP ‘he went home’ | /\ A de NP C’ ‘DE’ A l\ shihou, c° 1P ‘time, A 179 As (118) shows, shihou undergoes direct movement out of the IP, and then the rest of the IP moves to Spec of DP leaving the wh-operator behind. The de shihou clause is a relative clause involving direct wh-movement, and yi-qian/yi-hou clauses are complement clauses that also have a null operator movement involved. In addition, a difference between de shihou clauses and yi-qian/yi-hou clauses can be illustrated as follows: (119) a. Ta lai de shihou wo bu zai jia. he come DE time I not ZAI home ‘I was not home when he came.’ b. Ta lai de nei-ge shihou wo bu zai jia. he come DE that-CL time I not ZAI home ‘I was not home at the time when he came.’ (120) a. Ta lai yi-qian wo bu zai jia. he come before I not ZAI home ‘I was not home before he carne.’ b. *Talai yi nei-ge qian wo bu zai jia he come YI before I not ZAI home 0. *Nei-ge ta lai yi-qian wo bu zai jia. that-CL he come before I not ZAI home (121) a. Ta lai yi-hou wo cai chu-men. he come after I then out-door ‘I left after he came.’ b. *Ta lai yi nei-ge hou wo cai chu-men. he come YI that-CL after I then out-door c. *Nei-ge ta lai yi-hou wo cai chu-men. that-CL he come after I then out-door (122) a. Dajia chifan de qian-tou ni bie chi dianxin. Everybody eat-meal DE before you don’t eat snack ‘Don’t snack before we all have dinner.’ (Chao 1968, 119) b. *Dajia chifan de nei-ge qian-touni bie chi dianxin. Everyone eat-meal DE that-CL before you don’t eat snack 180 c. *Nei-ge dajia chifan de qian-ton ni bie chi dianxin. That-CL everyone eat-meal DE before you don’t eat snack In Chapter 3, we have discussed that a relative clause allows pre-relative clause demonstrative and/or post-relative clause (post-RC) demonstratives. As shown in (119), de shihou clauses allow at least post-RC demonstratives, while yi-qian/yi-hou clauses in (120) and (121) do not allow any demonstratives to occur within them. Even (122), which contains a long form localizer that means before, also shows that a demonstrative is not allowed. This further confirms that the operator movement does not make yi- qian/yi-hou clauses into relative clauses as Geis (1970) suggests. 4.8 Conclusion In this Chapter, I have analyzed the structure of yi-qian/yi-hou clauses. Since qian/hou as localizers function the same as yi-qian/yi-hou in temporal clauses, I also examined localizers. In the literature, there are three main analyses for short form localizers: the noun analysis, the adposition analysis, and the clitic analysis. Here, I have argued that the short forms are nouns that have become clitics. Therefore, they do not behave like standard nouns in Modern spoken Chinese, and their noun properties are limited to certain environments. Interestingly, comparing the positions that nouns and adpositions occur in, NP + localizer can occur mostly in noun position and not in adposition position. Although short forms fail most noun tests in modern Chinese, in classical Chinese, they are free words and behave like nouns instead of adpositions. Short forms do not behave like standard nouns in modern Chinese because they have become clitics. In order to characterize their properties as nouns and clitics, it is 181 preferable to analyze them as nominal clitics. Since qian/hou are both location and time indicators, I have further shown that XP + localizers in locative phrases and XP + qian/hou in temporal phrases have parallel structures. Therefore, I conclude that they have the same syntactic structure, that qian/hou take an XP as a complement to form an NP, and that Do selects this NP as complement and requires its specifier to be filled. Thus, the XP has to undergo movement to fill the Spec of DP position. This D0 position can be empty or have zhi/de/yi occur in the position, following Simpson (1999, 2000). Since yi and zhi/de are not related in classical and modern Chinese, I have further examined the classical usage of zhi and yi. I have shown that zhi is very likely a definite determiner. Also, zhi and yi both can be verbs, conjunctions, or pronouns. Therefore, they possibly have the same syntactic category when combining with short form localizers. The sentences in (82) further confirm this claim that de, zhi, and yi should be analyzed in the same fashion, allowing a general account for these three items. Finally, I have also given further evidence that clause + yi-qian/yi-hou is a DP and that it can occur after ba and be the subject of bei constructions. In addition, I compare the differences between yi-qian/yi-hou clauses and before/after clauses in view of Geis (1970), and they seem to behave similarly. This also provides evidence that yi-qian/yi-hou clauses also involve operator movement. Nevertheless, when yi-qian/yi-hou and before/after take phrases involving quantifiers, they do not behave the same. Also, as shown in (108b), when yi-qian/yi-hou phrases follow verbs, zai is required, and this situation seems to imply that yi—qian/yi-hou do not behave like adpositions such as before/after. 182 Finally, since both de shihou clauses and yi-qian/yi-hou clauses involve movement, I further compare their differences and show that de shihou clauses are relative clauses involving head raising, while yi-qian/yi-hou clauses are noun complement clauses involving null operator movement. 183 Chapter 5: The optionality of zai in temporal clauses As I have discussed in Chapter 1, Chinese temporal clauses often contain zai. Zai is optional most of the time, however sometimes it is required. This inconsistent behavior of zai is puzzling. In order to understand what role zai plays in temporal clauses, in this chapter, I examine the optionality of zai regarding its behavior in temporal clauses. 5.1 The problems In Chapter 4, I have argued that both locative phrases and temporal phrases behave similarly in that they both occur in zai+ XP + localizer phrases]. Accordingly, this predicts that zai should behave the same in both locative and yi-qian/yi—hou phrases; however, zai seems to be optional with temporal phrases, while obligatory with locative phrases, as shown in (1). (1) a. Ta *(zai) wu gian da lanqiu. he ZAI house front play basketball ‘He plays basketball in hunt of the house.’ b. Ta *(zai) shang diannao ke de na—ge tushugw gian da lanqiu. he ZAI have computer class DE that-CL library front play basketball ‘He plays basketball in fiont of the library that holds the computer class.’ ’ In this chapter, I mainly discuss yi—qian/yi-hou phrases, i.e., (ZAI)+ CP/NP+ yi—qian/yi-hou; therefore terms such as, ‘zai phrases’ and ‘terrrporal phrases’ refer to yi—qian/yi-hou phrases; however the optionality of zai in yi-qian/yi-hou clauses also applies to other temporal phrases as I will discuss in section 5.6. For convenience sake, I use ‘temporal clauses’ refer to CP + yi—qian/yi-hou as well as CP + de shihou. 184 c. Tatzai] san dign (vii-qia_n da lanqiu. he ZAI three o’clock before play basketball ‘He plays basketball before three o’clock.’ (1. Ta (zai) shang diannao ke de na-duan shijianmrgian dalanqiu. he ZAI have computer class DE that —CL time before play basketball ‘He plays basketball before the time that the computer class starts.’ (1a) and (lb) contain locative phrases and zai is obligatory. (1c) and (1d) contain temporal phrases, and zai is optional. However, the use of zai in the following sentences shows that it is not the case that zai is always optional in temporal clauses. (2) a. Zhangsan ??(zai1Lisi chusheng u’gian jiu chuguo qu le. ZAI was-bom before then go-abroad go LE Intended to mean: ‘Before Lisi was born, Zhangsan had gone abroad.’ b. (zai) Lisi chusheng vi-g'an, Zhangsan jiu chuguo qu le. was-bom before then go-abroad go LE ‘Before Lisi was born, Zhangsan had gone abroad.’ (3) a. Meiguihua *(z_ai) yeshog 81 1e vi-hor_1 jiu diaoxie 1e. rose ZAI beast die LE after then wither LE Intended to mean: ‘The rose withered after the beast died.’ b. (Qi) yeshou si le yi-hoL Meiguihua jiu diaoxie 1e. at beast die LE after rose then wither LE ‘The rose withered after the beast died.’ In (2) and (3), the meaning in the (a) sentences is the same as those in (b). In the (a) sentences, the temporal clause occurs after the main subject and zai is not optional, while in the (b) sentences, the temporal clause is preposed, and the absence of zai does not influence the interpretation of the sentences. However, the following sentences show that zai is not only optional in the beginning of the sentences, as in (4b) and (5b). Zai can also 185 be optional when the temporal clause occurs after the main subject, as shown in (4a) and (5a). (4) a. Wo (zai) wanfan zuohao 1e vi-hou jiu chumen 1e. I ZAI dinner make-ready LE after then out-door LE ‘1 went out after dinner was done.’ b. (Zai) wanfan zuoh_2_10 le yi-hoLwo jiu chumen 1e. ZAI dinner make-ready LE after I then out-door LE ‘1 went out after dinner was done.’ (5) a. Zhangsan ?(z_ai) tiafli re 1e vi-hog cai mai lengqiji. ZAI weather hot LE after then buy air-conditioner ‘Zhangsan bought an air-conditioner after the weather became hot.’ b. (Zai[tiangi re le yi-hou Zhangsan cai mai lengqiji. ZAI weather hot LE after then buy air-conditioner ‘Zhangsan bought an air-conditioner after the weather became hot.’ Given that locative phrases and temporal phrases behave the same syntactically, as l have discussed in Chapter 4, it seems puzzling that zai is obligatory in locative phrases, but optional in temporal phrases. Therefore, exploring the asymmetry between locative phrases and temporal phrases can help us to understand the asymmetry more specifically. Supposing that an important function of prepositions is to assign case, the asymmetry between locative phrases and temporal phrases seems to be that locative phrases need case, while temporal phrases do not. This suggestion follows, if we ignore post-subj ect temporal phrases, which inconsistently omit zai. That is, if the assumption of the case requirement between locative phrases and temporal phrases holds, the problem which needs to be addressed is why zai is not consistently optional when the temporal phrase follows the subject. This inconsistency of the presence of zai seems to imply that this 186 situation is not determined wholly by the syntax. That is, there is some other reason that causes this inconsistency to arise. The rest of this chapter is organized as follows: in 5.2 I review Larson (1985) and Stroik (1992) to seek an explanation for the different behavior of locative phrases and temporal phrases regarding the use of prepositions. In 5.3, I investigate the behavior of zai in locative phrases and that zai is required in locative phrases. In 5.4, I examine the optionality of zai in yi-qian/yi-hou phrases. The inconsistency of the optionality of zai among similar sentences suggests that a processing account would explain the data more properly. In 5.5, I will explain why the occurrence of zai is inconsistent in the post- subject position in sentences containing temporal phrases. Finally, in 5.6, I examine post-subj ect positions containing de shihou clauses and show that de shihou clauses also show the same patterns regarding the optionality of zai. 5.2 Larson and Stroik Larson (1985) examines English bare-NPs, and points out that there is a group of bare-NPs that function as adverbial modifiers without accompanying prepositions or any other means that indicate that they are adjuncts. These NPs are called bare-NP adverbs. Examples of bare-NPs behaving as adverbial modifiers are as follows: (6) a. John arrived the Tuesday that I saw Max. b. John anived yesterday. c. You have lived someplace warm and sunny. 187 According to the Case Filter (Chomsky 1981), an NP needs case to be visible and to be theta-marked. However, the sentences in (6) do not have any indicators to show that the temporal NPs bear case. Larson proposes adverbial theta role assignment that “assign[s] an adverbial 0-role to a, where a is any phrase” (1985, 606). Therefore, a temporal phrase can bear a 0mm, while a locative phrase can bear a 91.66. As to the case of these bare-NP adverbs, Larson proposes that these bare-NPs contain head nouns that have a [+F] feature which assigns Oblique case to the bare-NPs. Since these bare-NP adverbs contain inherent Oblique case, they can be visible in adjunct positions. Stroik (1992) examines Larson’s proposal for the behavior of bare-NP adverbs, and points out that Larson’s Oblique case approach does not explain the fact that the temporal phrases in fact behave differently from locative phrases as shown in (7) and (8). (7) a. Mary will see John [one day]. b. Mary will see John [someday]. c. Mary saw John [the day that Bill died]. (1. Mary saw John [the day of his birthday]. e. Mary will see John [Monday]. (Stroik 1992, 269) (8) a. *Mary will see John [one place]. b. *Mary will see John [someplace]2. 0. *Mary saw John [the place that Bill died]. (1. *Mary saw John [the place of his birthday]. e. *Mary will see John [Madison]. (Stroik 1992, 269) 188 Stroik argues that temporal NPs behave differently from locative NPs in that temporal NPs are secondary predicates that do not need case, while locative NPs are not predicates and need to be case marked. According to Stroik, following Eng: (1985), “an adequate theory of the temporal interpretation of verbs requires that verbs take empty temporal NP arguments as their most proximate arguments” (p. 271). Thus, temporal bare-NP adverbs can be bound by this empty temporal NP-argument, which makes temporal bare-NP adverbs into predicates following Safir’s (1987) predicate principle which Stroik has revised with a condition of binding put forth by Rizzi (1990). The revised predicate principle is as follows: (9) Predicate Principle: A potential referring expression must be either (i) predicate, bound within its m- command domain or (ii) free. (p. 270) That is, when a sentence contains a temporal NP, it is bound by an empty temporal NP argument that is theta-marked by the verb, and thus a secondary predicate, following Rothstein (1985)3. However, locative phrases do not have an empty locative bare-NP to bind them, and thus are not predicates. Whether temporal phrases contain a [+F] feature that assigns obligatory case or they are predicates is not directly relevant. However, Larson’s examination of bare-NP adverbs and Stroik’s distinguishing locative phrases from temporal phrases support the idea that there is a group of adverbs that do not need case. In the following, I will 2 Some of the native speakers consider (8b) acceptable. 3 According to Rothstein (1985), there are two types of predicates: primary and secondary predicates. A secondary predicate takes a subject that is theta-marked by another lexical head. In the case of bare NP 189 examine both locative phrases and temporal phrases in Chinese and will argue that only locative phrases need zai for syntactic case, while temporal phrases do not need zai. Since the syntactic constraint does not apply to temporal phrases, I will try to explain why zai shows inconsistency in the post-subject position in sentences containing temporal clauses. 5.3 The optionality of mi in locative phrases In this section, I discuss the function of zai in locative phrases, and examine whether or not zai can be optional. There are cases where zai seems to be optional; however, these cases may not be cases of true optionality since other syntactic constraints may be involved. As I have discussed in Chapter 4, a typical locative phrase, according to Li and Thompson (1981) (henceforth, L & T), is zai + NP + (localizer) as shown in (10) 4' 5. (10) zai jia (1i) ZAI home inside ‘at home’ A locative phrase can occur as a predicate as shown in (11). temporal phrases, the empty temporal NP argument is theta-marked by the verb and also binds the bare NP adverb, and thus the bare NP adverb is a secondary predicate. ’ Localizers are optional when the preceding NP is a specific place, as discussed in Chapter 4. 5 The locative phrase can also occur after a verb indicating a result of the action caused by the verb preceding it. Since this post-verbal position is an argument position, it is not considered in this thesis. Example of a locative phrase occurring after a verb is as in (i). (i) Ta ba shu fang zai shuzhuo shang. he BA book put ZAI desk top ‘He put the book on the desktop.’ 190 (11) Zhangsan zai jia 1i. ZAI home inside ‘Zhangsan is at home.’ A locative phrase can occur as an adverbial phrase that it shows up in the post-subject and pre-verbal position as in 02)". (12) Ta zai jia li chi guo fan 1e. he ZAI home inside eat GUO meal LE ‘He has eaten at home.’ Also zai in (11) and (12) can not be omitted, as (13) shows. (13) a. Zhangsan *(zai) jia li. ZA home inside b. Ta *(zai) jia li chi guo fan 1e. he ZAI home inside eat GUO meal LE 6 According to Chen (1978) and Yang (1995), when there are no other aspect markers occurring in the sentences containing zai+ localizer phrase, zai can mark progressive aspect, as (1) shows. (1) Ta zai jia li yundong. he ZAI home inside exercise ‘He is exercising at home.’ However, there is no evidence that zai plays the role of progressive aspect when it occurs with a localizer phrase, since (i) can also mean he exercises at home. (ii) Ta yundong. (iii) Ta zai yundong. He exercises. he ZAI exercise ‘He is exercising.’ Or ‘He exercises.’ ‘He is exercising.’ As we can see from (ii) and (iii), when the bare verb phrase yundong does not follow zai, (ii) has both a progressive reading and a habitual reading. Only when zai occurs does it have the progressive reading. The situation in (i) is like (ii) in that the occurrence of zai in (i) does not exclude the habitual reading. 191 L & T (1981) show that a sentence with and without zai has different readings. (14) shows that the presence or absence of zai in a locative phrase results in a different reading, although it conveys basically the same information. (14) a. Shuzhuo shang dui 1e yi-xie shu. desk top pile LE a-few book ‘A few books were piled on the desktop.’ b. Zai shuzhuo shang dui 1e yi-xie shu. ZAI desk top pile LE a-few book ‘Someone piled a few books on the desktop.’ According to L & T, (14a) indicates a stative sense and is a presentative sentence, which is like an existential sentence. However, (14b) shows an activity, and the agent of the activity is understood from the context. Now let us add a pronoun in the beginning of (14) and get (15). (15) a. Tamen shuzhuo shang dui 1e yi-xie shu. they desk top pile LE a-few book ‘A few books were piled on their desktop.’ b. Tamen zai shuzhuo shang dui le yi-xie shu. they ZAI desk top pile LE a-few book ‘They piled a few books on the desktop.’ (15) shows the same results as L & T predicted: tamen ‘they’ in (15a) is not the agent of the piling; however, it is part of the whole noun phrase tamen shuzhuo-shang ‘their desk top’ and (15a) expresses a stative existential reading or a passive reading that the table top has some books placed on it. In (15b), tamen ‘they’ is the subject as well as the agent 192 in that tamen ‘they’ refers to the people who perform the act of piling the books. However, (16a) and (17a), although structurally similar to (14a) and (14b), are ungrammatical. (16) a. *Shuzhuo shang xiu le wu-bu diannao. desk top fix LE five-CL computers Intended to mean: ‘Five computers were fixed on the desktop.’ b. Zai shuzhuo shang xiu 1e wu-bu diannao. ZAI desk top fix LE five-CL computers ‘Someone fixed five computers on the desktop.’ (17) a. *Tamen shuzhuo shang xiu 1e wu-bu diannao. They desk top fix LE five-CL computers Intended to mean: ‘Five computers were fixed on their desktop.’ b. Tamen zai shuzhuo shang xiu-1e wu-bu diannao. They ZAI desk top fix-LE five-CL computers ‘They fixed five computers on the desktop.’ (l6) and (17 ) are parallel to (14) and (15) respectively; however, (16a) and (17a) are not allowed, while (14a) and (15a) are fine. The difference is due to the different verb classes. The verb in (14) optionally takes an agent subject, while the verb in (16) required an agent subject. (14a) and (15a) have the meaning that the desktop was influenced by someone by putting some books on the table, while in (16a) and (17a), the desktop is not influenced. Since the verb xiu ‘to fix’ requires an agent in its external structure, the lack of agent in the sentence causes the ungrarnmaticality of (16a) and (17a). In order for (16a) and (17a) to be grammatical, zai has to be inserted to make the locative phrase an adverbial phrase, as shown in (16b) and (17b), instead of inserting a subject noun phrase as in (15a) and (15b). The existence of zai in (14b), (15b), (16b) and 193 (17b) seems to indicate that locative phrases with zai are not subjects but adverbial phrases. The recoverability of the subject in (15b) from (14b) and (17b) from (16b) shows that locative phrases with zai cannot be subjects. That is, zai is not optional in the post-subject position, i.e. the adverbial position. Wen (1957) also mentions a similar situation and holds that zai can be used to test if a locative phrase is a subject or an adverbial phrase, as shown in (18). (18) a. Qiang shang wa le yi-ge dong. wall top dig LE one-CL hole ‘There is a hole dug in the wall.’ b. Zai qiang shang wa 1e yi-ge dong. ZAI wall top dig LE one-CL hole ‘Someone dug a hole in the wall.’ According to Wen (1957), the difference between (a) and (b) in (18) is that qiang shang ‘wall top’ in (a) is a subject and zai qiang shang ‘on the wall (top)’ in (b) is an adverbial phrase. If L & T and Wen are correct in their observations, the existential constructions in (19) should have different structures depending on whether zai occurs in the initial position. (19) a. (Zai) shuzhuo shang you liang-ben shu. ZAI desk top has two-CL book ‘There are two books on the desktop.’ b. (Zai) lou xia lai 1e liang-ge ren. ZAI stairs down come LE two-CL person ‘Two people arrived downstairs.’ 194 Note that sentences (14a), (18a) and (19) are all existential sentences in that they all start out including a location, a non-transitive verb, and an indefinite noun phrase. A definite noun phrase is not allowed to occur after the verb as shown in (20). (20) a. *Shuzhuo shang dui 1e na—xie shu. desk top pile LE those book ‘Those books were piled on the desktop.’ b. *Qiang shang wa 1e na-xie dong. wall top dig LE those hole ‘There are those holes dug on the wall.’ c. *Shuzhuo shang you na-xie shu. desk top have those book ‘On the desktop, there are those books.’ d. *Lou xia lai 1e na-xie ren. stairs down come LE those people ‘There arrives those people.’ However, the difference of (19) from (14a) and (18a) is that the verbs in (14a) and (18a) are derived from transitive verbs, while the verbs in (19) are unaccusative. Therefore, we cannot insert an agent noun to test whether there is an agent subject; nor can we find a different meaning with and without zai. However, the subject position in (21) indicates that a real (rather than existential) subject position does not allow zai to occur. (21) (*Zai) shuzhuo shang shi xie zi de hao difang. ZAI desk top is write words DE good place ‘On the desktop is a good place to write.’ 195 (21) is not an existential sentence; the presence of zai in the subject position causes the ungrarnmaticality. This is further evidence to support Wen’s (1957) argument that when zai occurs in initial position, the phrase it occurs in is not a subject, but when zai does not occur, the NP + localizer is the subject. The contrast between (19) and (21) shows that in existential sentences with zai, the locative phrase is not a subject, while without zai, the locative phrase is a subject. Now examine (22). (22) a. Ta zai shuzhuo shang xie le wu-ge zi. he ZAI desk top write LE five-CL words ‘He wrote five words on the desktop.’ b. *(Zai) shuzhuo shang, ta xie-1e wu-ge zi. ZAI desk top he write-LE five-CL words ‘On the desktop, he wrote five words.’ (22) shows that a NP+ localizer combination cannot occur in topic position. The NP + localizer combination seems to need zai to license it when it is not in a subject position as (21) and (22) show. That is, since NP + localizer is a DP it needs case; it cannot stand alone in non-argument positions such as adverbial positions and topic positions. This suggests that in locative phrases, zai cannot be optional even if it seems to look so. In summary, zai in locative phrases cannot be optional. There seem to be cases where optionality exists; however, it is necessary to distinguish cases in which locative phrases with zai are in topic positions fi'om cases in which locative phrases without zai are in subject positions. 196 5.4 Zai in yi-qian/yi-hou phrases Recall the discussion in Chapter 4 that localizers like qian/hou are also temporal localizers and the position of yi-qian/yi-hou and zhi-qian/zhi-hou can be filled by qian/hou. That is, the structures of the locative phrases and the temporal phrases with yi- qian/yi-hou are similar. Therefore, I assume that temporal phrases with yi-qian/yi-hou have similar constructions to locative phrases as in (23). (23) zai + NP/CP + (yi)-qian/hou Like a locative phrase, a yi-qian/yi-hou phrase can be a predicate, as shown in (24). (24) Yinyuehui zai zhonggiu-jie fl-gian/fl-hou. concert ZAI moon-festival before/after ‘The concert will be/is/was before/after the moon festival.’ Also, yi-qian/yi—hou phrases can occur after the subject and before the verb, i.e., they can be an adverbial phrase as in (25). (25) Yinyuehui z_ai zhonggiu-iie (yi)-gign/hor_1_juxing. Concert ZAI moon-festival before/after hold ‘The concert will be/is/was held before/after the moon festival.’ 197 Moreover, unlike in the locative phrases, zai in an adverbial phrase as in (25) can be optional, as shown in (26). The only exception to this is when zai is part of a predicate as in (24). (26) Yinyuehui (zai) zhonggi -iie (vii—gLan/hogjuxing. Concert ZAI moon-festival before/after hold ‘The concert will be/is/was held before/after the moon festival.’ Nevertheless, as I have discussed in the beginning of this chapter, it is not the case that zai is consistently optional in each yi-qian/yi-hou phrase. Now we should discuss the circumstances of the optionality of zai in yi-qian/yi-hou phrases. The rest of the chapter is organized as follows: in 5.4.1, I argue that zai is not syntactically required. Since the optionality of zai is not due to syntactic constraints, I suggest that a processing account is an appropriate way to explain the situation. That is, the omission of zai can lead to a temporary ambiguity before the temporal clause is realized. In 5.4.2, I will show that there are a number of possible constructions for sentences starting with two noun phrases, and these possibilities effectively force zai to be present between the two NPs to avoid them. 5.4.1 Zai is not syntactically required in temporal phrases Consider a bare nominal temporal phrase such as mingtian ‘tomorrow’ occupying both the post-subj ect and pre-verbal positions, as shown in (27). 198 (27) a. Ta mingtian qingzhu shengri. he tomorrow celebrate birthday ‘He celebrates his birthday tomorrow.’ b. Ta zai mingtian qingzhu shengri. he ZAI tomorrow celebrate birthday ‘Tomorrow is the day that he celebrates his birthday.’ c. Mingtian ta qingzhu shengri. tomorrow he celebrate birthday ‘Tomorrow he celebrates his birthday.’ (1. *Zai mingtian ta qingzhu shengri. ZAI tomorrow he celebrates birthday ‘On tomorrow, he celebrates his birthday.’ In (27a) and (27b), we can see that there is a subtle difference between the sentence without zai and the sentence with zai; (27a) indicates that there is an event that will happen tomorrow, while in (27b) zai picks out tomorrow and emphasizes tomorrow to be the date that a certain event will take place. (27c) shows the adverbial bare noun mingtian ‘tomorrow’ does not need zai, and with zai in (27d) the sentence is not acceptable. This pattern consistently differs from locative phrases, which require zai even in preposed positions7. Now consider (28). (28) a. Zhe-jian shi (zai) yi-qian chengjing fasheng guo. This-CL matter ZAI before have-ever happen GUO ‘This matter has happened before.’ b. Zhe-jian shi (?zai) yi-hou hai hui fasheng. This-CL matter ZAI afterwards still will happen ‘This matter will happen again.’ 7 Although (27d) is not good, in the context of contrastive focus as in (i) this structure is fine. (i) Zai mingtian ta qingzhu shengri, zai houtian ta qingzhu jiehunj inianri ZAI tomorrow he celebrates birthday, ZAI the-day-after he celebrates marriage-anniversary ‘Tomorrow he celebrates his birthday; the day after tomorrow he celebrates his anniversary.’ 199 c. (Zai) yi-qian zhe-jian shi chenging fasheng guo. ZAI before this-CL matter have-ever happen GUO ‘This matter has happened before.’ (1. (Zai) yi-hou zhe-jian shi hai hui fasheng. ZAI afterwards this-CL matter still will happen ‘This matter will happen again.’ From (28), we can see that when yi-qian/yi-hou are by themselves, they do not need zai, either in post-subject positions or pro-subject positions. The data in (28) seem to confirm that bare time adverbials do not need case as I discussed in 5.1, which is unlike locative phrases. Also, recall that in the beginning of this chapter, I have shown that yi-qian/yi- hou clauses consistently do not need zai when the clauses are in the preposed positions, as (2) and (3) show, repeated here as (29) and (30) with additional examples. (29) a. Zhangsan *fzai) L_isi chusheng fl-gian jiu chuguo qu 1e. ZAI was-born before then go-abroad go LE Intended to mean: ‘Before Lisi was born, Zhangsan had gone abroad.’ b. (Zgi) Lisi chusheng yi-g'an, Zhangsan jiu chuguo qu 1e. ZAI was-bom before then go-abroad go LE ‘Before Lisi was born, Zhangsan had gone abroad.’ (30) a. Meiguihua *(zai) veshogu si-le yi-hgu jiu diao-xie 1e. rose ZAI beast die-LE after then wither LE Intended to mean: ‘The rose withered after the beast died.’ b. (Zai) yeshou si-le vi-hoL meiguihua jiu diao-xie 1e. ZAI beast die-LE after rose then wither LE ‘The rose withered after the beast died.’ (31) a. Wo *(zai) Lisi kan-dao Wangm yiflian jiu jin 1e cheng. I ZAI saw before then enter LE town ‘I went downtown before Lisi saw Wangwu.’ 200 b. (Zai) Lisikan-dao Wangfl yi-gian wojiu jin 1e cheng. ZAI saw before I then enter LE town ‘I went downtown before Lisi saw Wangwu.’ (32) a. Xiaowei ??(zai) pengyou lai-1e vi-hog jiu zhi gu wan. Xiaowei ZAI friend come-LE after just only attend-to fun ‘After her, fiiends came, Xiaowei, only cared about having firn.’ b. (Zai) pengyou lai-1e yi-hou Xiaowei jiu zhi gu wan. ZAI friend come-LE after Xiaowei just only attend-to fun ‘After her, fiiends came, Xiaowei, only cared about having fun.’ (33) a. Zhangsanjzai) zuowan gongke yj-hou jiu huijia le. zai do-finish homework after then went-home LE ‘After finishing his, homework, Zhangsan, went home.’ b. (zai) zuowan gongke fl-hou Zhangsan jiu huijia 1e. ZAI do-finish homework after then went-home LE ‘After finishing his, homework, Zhangsan, went home.’ (34) a. Lisi zai on e zuowan '-hou jiu hui-jia qu le. ZAI homework do-finish after then return-home go LE ‘After finishing his, homework , Lisi, went home.’ b. (Zai) gongke zuowan y_i-hou Lisi jiu hui-jia qu le. ZAI homework do-finish after then return-home go LE ‘After finishing his, homework, Lisi, went home.’ The additional sentences in (31), (32), (33), and (34) also support the idea that the presence of zai is not crucial syntactically. It is only when the temporal phrases occur after the matrix subject that the optionality of zai becomes a problem. This problem is possibly due to the fact that when temporal phrases occur after the matrix subject, the matrix subject and the first noun phrase of the temporal phrase are in a sequence that, without the presence of zai between the two noun phrases, causes a processing ambiguity. This assumption can be supported by the following sentences: 201 (35) a. Lisi (zai) erzi si 1e yi-hou jiu yijuebuzhen. ZAI son die LE after then collapsed ‘Lisi, collapsed after his, son died.’ b. Lisi zai erzi si 1e yi-hou jiu cheng-ming le. ZAI son die LE after then become-famous LE ‘Lisi, became famous after his, son died.’ b’. ??Lisi erzi 81 1e yi-hou jiu cheng-ming 1e. son die LE after then become-famous LE ‘Lisi, became famous after his, son died.’ Or ‘Lisi’s son, became famous after he, died.’ c. Lisi (zai) yi-ge erzi si le yi-hou you si 1e yi-ge erzi. ZAI one-CL son die LE after again die LE one-CL son ‘After one son died, Lisi’s other son also died.’ The sentences in (3 5) are supposed to have the same structure, since the verbs in the temporal clauses are all transitive; however, some require zai, but some do not. Apparently, the requirement of zai in some sentences is not because of case or any other syntactic constraints, since zai is not consistently required. (35a), (3 5b), and (35b’) have the same verbs in the temporal clause; however, (35a) is good without zai, but (35b’) is not acceptable. The inconsistent behavior of zai in (35) cannot be attributed to syntactic constraints. Additionally, in (35a), since Lisi’s son died, the only person who can collapse is Lisi, and therefore there is no question of which entity performs the action of the main verb. Moreover the presence of zai forces the interpretation that it is Lisi who is involved with the event in the main clause. In (35b’), both Lisi and erzi ‘son’ can be interpreted to be involved with the main verb, and this causes ambiguity, making it difficult to process the sentence. However, the occurrence of zai clearly assigns Lisi as the subject that performs the action with the main verb and there is no ambiguity as in the 202 sentence without zai. (3 So), with and without zai, has the same meaning, and this provides evidence that the optionality of zai is not a syntactic constraint". Based on what I have observed above, apparently the optionality of zai in post- subject positions is not because of syntactic constraints, but possibly due to the comprehender’s assumptions about the structure of sentences starting with two noun phrases as conflicting or compatible with the structure of sentences containing temporal phrases. That is, the comprehender imposes structure in the string of words. When comprehenders encounter two noun phrases at the beginning of the sentence, they impose a structure that has two noun phrases in the beginning of the sentence and if what they predict is compatible with a temporal construction, the sentence is fine; otherwise a garden path occurs, and it ends up requiring reanalysis. One of the famous garden path sentences is following: (36) The horse *(that was) raced past the barn fell. (Bever 1970) (3 6) shows that omission of that was causes the sentence to be unacceptable, although it is a grammatical sentence without that was. The reason that (36) is unacceptable is because the comprehender initially interprets the sentence as follows: (3 7) The horse raced past the barn. 8 1f the zai phrases are preposed, the first NP in the non-preposed sentence is consistently interpreted as the subject of the main verb for all the sentences in (35). 203 When comprehenders read (36), they impose the structure in (37) onto it, and are garden pathed when they encounter fell. (36) can be rescued by inserting then before fell, although it is a different structure. It has a very similar meaning to (36), as shown in (38). (3 8) The horse raced past the barn then fell. Another way to improve (36) is to bias the main verb as (39) shows. (39) The horse ?(that was) raced past the barn eats a lot. Similarly, the omission of zai has the same effect as that was in (36). Consider (40). (40) Zhangsan *(zai) Lisi chu le che huo yi-hou jiu qu kan ta. ZAI have LE car accident after then go see him Intended to mean: ‘Zhangsan went to visit him, after Lisi, had a car accident.’ (40) is not good after omitting zai because the initial part of the sentence is interpreted as (41) and the comprehender processes it as (41). (41) is a sentence starting with two noun phrases and the two proper names are interpreted as a coordination. (41) Zhangsan Lisi chu le che-huo. have LE car-accident ‘Zhangsan and Lisi had a car accident.’ 204 Since comprehenders interpret the beginning of (40) as (41), the disambiguation thus occurs when the comprehenders encounter the main verb. When comprehenders read the initial part of (40), they interpret it as after Zhangsan and Lisi had a car accident, and when they encounter the verb phrase qu kan ta ‘go to visit him/her’, they cannot get who went to see whom. That is, comprehenders suppose that since Zhangsan and Lisi had a car accident, they are not supposed to be the people visiting, they should be visited by someone else, and thus the comprehender does not know what the main subject is, and what ta ‘him/her’ refers tog. Now let us rescue (40) and get (42a) by changing the verbs: (42) a. Zhangsan Lisi chg le che-hug yi-hou jiu shoushang 1e. have LE car-accident after then injure LE ‘After [Zhangsan and Lisi], had a car accident, they, both were injured.’ b. Zhangsan Lisi z_ai chl_l 1e che-hl_1_o_ yi-hou jiu shoushang le. ZAI have LE car-accident after then injure LE ‘After [Zhangsan and Lisi], had a car accident, they, both were injured.’ c. Zhangsan Lisi (zai) chuL le che-Inn; yi-hou jiu shoushang 1e. ZAI have LE car-accident after then injure LE ‘After [Zhangsan and Lisi], had a car accident, they, both were injured.’ (1. Zhangsan *(zai) Lisi chp 1e che-h1;o_ yi-hou jiu shoushang 1e. ZAI have LE car-accident after then injure LE ‘After Lisi had a car accident, Zhangsan was injured.’ After we bias the verb to one that fits Zhangsan and Lisi being the main subject, the sentence is fine. Now let us add zai after the coordination and get (42b). The structure of (42b) is a regular Chinese word order that contains an adverbial starting with zai after the subject. Combining (42a) and (42b), we get (42c). That is, (42c) contains a temporal 9 There may be an interpretation of after Zhangsan and Lisi had a car accident, they went to see him, in which they refers to Zhangsan and Lisi; however this reading is difficult to get without a context for it. 205 adverbial in which zai is optional. (420) is like sentences such as (27) and (28), in that zai is optional; however, the structure of Zhangsan and Lisi is different from the one in (40). (42c) is fine and still contains a temporal clause. However, if we intend to make it mean the same as (42d), the absence of zai will lead a comprehender to get the wrong interpretation. Following what I have discussed, the assumption is that zai is optional. However, when zai is omitted in the post-subj ect position, it leaves two NPs in a sequence, and this situation resembles other structures, for example, structures starting with a coordination or a topic-subject pair, which leads the comprehender to assume a structure of sentences that starts with a sequence of two noun phrases. Ifthis assumption fits the sentence, it is fine as (42a) to (42c) show. If not, the comprehender encounters disambiguating information after the main verb, as (40) shows. Therefore, when yi-qian/yi-hou clauses occur after the main subject, including zai is preferred since omitting zai would cause ambiguity as shown in (40). This situation is similar to the English example in (36), in which that was provides disambiguating information and makes the sentence acceptable"). If this analysis is on the right track, it is important to know what constructions starting with two noun phrases are allowed to occur before yi-qian/yi-hou, as these structures may be hypothesized by comprehenders. loOmission of that in (i) also shows a temporary garden path with the verb spilled being the main verb; however the garden path is quickly recovered from after the main verb is encountered. This seems to be parallel to (ii) in that the omission of zai, although causing a temporary garden path with the first NP being the topic of the embedded clause, is recovered from after the main verb is encountered. (See section 5.4.2.2 for the discussion of topic readings.) (i) The coffee (that) spilled on the rug caused the stains. (ii) Na liang chezi (zai) wo xi 1e yi-hou jiu huai le That CL car ZAI I wash LE after then out-of-order LE ‘The car become out of order after I washed it.’ 206 5.4.2. Constructions starting with two noun phrases The possible structures in which two NPs can start a sentence in Chinese are either in coordination or a topic-subj ect relationship’ I . Coordination usually needs a conjunction, and not all noun phrases can be in topic-subject relations. In 5.4.2.1, I will examine what conditions allow two NPs to be coordinated without a conjunction, and in 5.4.2.2, I will discuss what constrains topic-subject relations, i.e. what can be and what cannot be in topic-subject relations. 5.4.2.1 Coordination In this section, I will introduce Chinese coordination and show the conditions that make a two NP sequence have a coordination interpretation without a conjunction. Chinese coordination with noun phrases usually requires he ‘and’ between the two noun phrases and sometimes requires dou ‘all’ after the coordinated elements, as (43) shows12 . (43) a. Shugui he zhuozi (dou) bu-jian 1e. bookshelf and table all not-see LE ‘The bookshelf and the table are both gone.’ 1' Sentences starting with a subject and a temporal adverbial (such as dates, days of the week, and holidays) are easy to process since adverbials like these are easily defined as adverbs. Therefore they don’t cause processing problems and will not be considered issues to be discussed here as a type of sentence starting with two NPs. '2 Gen ‘and’ also can be used to coordinate two noun phrases; however, it also means with, and without a context, it is difficult to tell if it means and or with, as shown in (i). Therefore, I don’t discuss it here. (i) Zhangsan gen Lisi chuqu banshi and/with go-out do-business ‘Zhangsan and Lisi went out to do some business.’ Or ‘Zhangsan went out to do some business with Lisi.’ 207 b. Na-zhang zhuozi he naxie yizi (dou) bu-jian 1e. that-CL table and those chair all not-see LE Intended to mean: ‘The table and those chairs have disappeared.’ c. Zhangsan he yizi ?(dou) bu-jian 1e. and chair all not-see LE ‘Zhangsan and the chair are both gone.’ (1. Yi-zi he Zhangsan ?(dou) bu-jian 1e. Chair all not-see LE Intended to mean ‘Zhangsan and the chairs both have disappeared.’ When the two noun phrases are two names or two related objects that are frequently combined, this situation can override the requirement of he ‘and’ and dou ‘all; both’ as in (44)- (44) a. Zhangsan (he) Lisi (dou) bei male. Zhangsan and Lisi both BEI scold LE ‘Zhangsan and Lisi were scolded.’ b. Zhuozi (he) yizi (dou) bu—jian le. table and chair all not-see LE ‘The table(s) and the chair(s) have disappeared.’ (44a) shows that when two proper names occur in the beginning of the sentence with a predicate, the interpretation is that the two proper names are in a coordinated relationship and undergo the same event. In (44b), the two objects are interpreted as two coordinated objects that have disappeared. Table and chair are closely related and usually come as a set, and zhuozi yi-zi ‘table chair’ is a commonly combined usage; therefore, without he and’ and dou ‘all’, the sentence can still be processed readily. 208 However, if the NPs are not names and commonly combined related objects, sentences lacking he ‘and’ and dou ‘all’ are not good depending on several variables, as shown below. (45) a. ?Shugui zhuozi bu-jian 1e. bookshelf table not-see LE ‘The bookshelf and the table have disappeared.’ a’. Shugui zhuozi dou bu-jian 1e. bookshelf table all not-see LE ‘The bookshelf and the table have disappeared.’ b. *Na-zhang zhuozi naxie yizi bu-jian 1e that-CL table those chair not-see LE Intended to mean: ‘The table and those chairs have disappeared.’ b’. ?Na-zhang zhuozi naxie yizi dou bu-jian 1e that-CL table those chair all not-see LE Intended to mean: ‘The table and those chairs have disappeared.’ c. *Zhangsan yizi bu-jian 1e. chair not-see LE Intended to mean ‘Zhangsan and the chairs both have disappeared.’ c’. ??Zhangsan yizi dou bu-jian 1e. chair all not-see LE Intended to mean ‘Zhangsan and the chairs both have disappeared.’ (I. *Yi-zi Zhangsan bu-jian 1e. Chair not-see LE Intended to mean ‘Zhangsan and the chairs both have disappeared.’ d’. ??Yi-zi Zhangsan dou bu-jian 1e. Chair all not-see LE Intended to mean ‘Zhangsan and the chairs both have disappeared.’ In (45a), although the sentence is acceptable, especially with the addition of dou in (45 a’), the bookshelf and the chair do not have a close relationship, and thus (45a) is not 209 as good as (44b) without dou. (45b) also contains table and chair, but the two NPs containing demonstratives are not commonly used. They are therefore unacceptable without he ‘and’ and dou ‘all’, and are not very acceptable even with the help of dou ‘all’ as in (45b’). (45c) and (45d) both contain two unrelated objects. In addition, the Zhangsan and yizi ‘chair’ sequence is usually interpreted as Zhangsan and his chair instead of a list of possible coordinated elements if he ‘and’ and dou ‘all’ do not occur, and is still not very good with the help of dou ‘all’, as shown in (45c’) and (45d’). In addition, for a pair of noun phrases to have a coordination reading without a conjunction, one NP must not be the antecedent of an overt or covert pronoun in the verb phrase. This can be shown by (46). (46) Zhangsan Lisi da 1e ta yi-dun. beat LE him one-CL ‘Zhangsan,, Lisi has beaten him, up.’ Or ‘Zhangsan, and Lisi, have beaten him,< up.’ (46) contains a pronoun in the verb phrase which can refer back to Zhangsan, but not to Lisi. Although there are two interpretations, the topic-subject interpretation is easier to get than the coordination interpretation. Therefore, in order to have a coordination reading straightforwardly without a conjunction, there should not be a pronoun occurring in the verb phrase. Thus, in order to ensure a coordination reading the two noun phrases have to be either two names or two related objects that are commonly used, and there must be no pronoun co-referent to one of the noun phrases occurring in the verb phrase. 210 5.4.2.2 Topic-subject relationship The second way in which two NPs can be interpreted is when the first NP is interpreted as a topic. In a sentence containing a topic, there must be some sort of relationship between the topic and the rest of the sentence that contains the topic; otherwise, the topic appears very odd. Chinese is a topic prominent language (Li and Thompson 1981). According to L & T, a topic has the following characteristics: (47) A topic is typically a noun phrase (or a verb phrase) that names what the sentence is about, is definite or generic, occurs in sentence-initial position, and may be followed by a pause or a pause particle (p. 87). Those characteristics are exemplified in (48) and (49). (48) W0 xihuan chi pingguo. I like eat apple ‘1 like to eat apples.’ (L & T 1981, 88) (49) a. Q! fl kan guo le. dog I see GUO LE ‘Dogs,, 1 have seen them,.’ Or ‘The dog,, I have seen it,.’ b. Nei-zhong douzi yi-jin sanshi kuai qian. that-CL bean one-catty thirty dollar money ‘That kind of bean, one catty is thirty dollars.’ (L & T 1981, 96) In (48), according to L & T, wo ‘I’ is definite, the sentence is about wo, and it can be followed by a pause, and is therefore a topic. 211 In (49a), there should be a resumptive pronoun or trace after guo that is coindexed with gou ‘dog’. In (49b), the subject yi-jin ‘one catty’ is a subset of the topic nei-zhang douzi ‘that-CL bean’; i.e., the topic and the subject are in a part-whole relationship. In this thesis, what is relevant are sentences with a topic and a subject because a temporal phrase in its post-subject position results in a two-NP sequence without zai. Also before/after clauses usually indicate a sequence of two events. Therefore, I will focus on sentences with topics that describe events but not states. Thus sentences like (49b), which does not express any event, will not be considered. In a topic-subj ect sequence sentence, the topic and the subject can have a number of different relations. The most common relations are those that indicate the same reference, possessor-possessee relationship, and agent-theme/theme-agent relations. In the possessor-possessee relation, clauses after the topic describe events relating to the possessee, but not to the possessor. In the agent-theme relation, the topic is the door of clauses after the topic, while in the theme-agent relation, the topic undergoes some event in clauses following it. In what follows, I will provide examples of these relations. First, a topic may be a co-referent with the subject of a sentence as in (50). (50) Zhangsan, ta, lai le. he come LE ‘Zhangsan,, he, himself, came.’ (50) is a sentence containing a topic, Zhangsan. Ta ‘he’ refers to Zhangsan, but not to other people. However, the reverse order is impossible, as shown in (51). 212 (51) *Ta, Zhangsan, lai le. he come LE In (51), not only can the two NPs not be coindexed, the sentence does not make sense at all. A second kind of topic-subject relation arises when the topic-subject is in a possessor-possessee relation as (52) shows”. (52) a. Xiaowei chezi huai le. Xiaowei car break LE ‘Xiaowei,, her, car is broken.’ b. Zhangsan erzi si 1e. son die LE ‘Zhangsan,, his, 8011 died.’ '3 Chinese possessor and possessee relationship is usually marked by using de, and de sometimes can be omitted as wo de meimei can be represented as wo meimei ‘my sister’. Although (52) may be translated as in (i). (i) a. Xiaowei chezi huai le b. Zhangsan erzi 51 1e Xiaowei car break LE son die LE ‘Xiaowei’s car is broken.’ ‘Zhangsan’s son died.’ and the translation in (52) and (i) may refer to the same event, unlike wo meimei, Xiaowei chezi ‘Xiaowei car’ and Zhangsan erzi ‘Zhangsan son’ are not legitimate constituents by themselves. This can be illustrated by the following contrast. (ii) a. Wo xihuan wo meimei b. *Wo xihuan Zhangsan erzi I like I sister I like son ‘I like my sister’ Intended to mean ‘I like Zhangsan’s son. c. *Wo kan guo Xiaowei chezi I see GUO Xiaowei car Intended to mean ‘I have seen Xiaowei’s car.’ Since Zhangsan erzi in (iib) and X iaowei chezi in (iic) are not constituents, it follows when they are in topic-subject potions, they do not have the meaning in (i) and they do not have an optional de as in wo de meimei ‘my sister’. 213 Again, the reverse order does not allow the possessee-possessor relationship as in (53). (53) a. *Chezi Xiaowei huai 1e. car Xiaowei broken LE Intended to mean: ‘That car, Xiaowei’s, is broken.’ b. *Erzi Zhangsan si 1e. p son die LE Intended to mean: ‘The son, Zhangsan’s, is dead.’ However, this kind of combination depends on how fiequent the possessor-possessee E relationship is in our world. It is difficult for (54a) to have the meaning that the maple trees belong to Zhangsan, although it is easy to get the car belongs to Xiaowei in (52). The reverse order is even worse as (54b) shows. (54) a. ??Zhangsan fengshu zhang gao le. maple-tree grow tall LE Intended to mean: ‘Zhangsan, his maple trees have grown tall.’ b. *Fengshu Zhangsan zhang gao 1e. maple-tree grow tall LE Intended to mean: ‘The maple trees, Zhangsan’s, have grown tall.’ A final topic-subj ect relationship is interpreted as agent-theme/theme-agent as in (49a). (55) contains some similar examples. (55) a. Zhangsan shu kanwan 1e. book read-finish LE ‘Zhangsan has finished reading the book.’ 214 b. ?Shu Zhangsan kanwan 1e. book read-finish LE ‘The book was finished by Zhangsan.’ As I have mentioned before, and as L & T argue, the sentence after the topic has to be about the topic. One way to create this relationship is to have part-whole relations, including possessor—possessee relations; another way is that the topic can be either the agent of the sentence after it, as in (55a), or the topic can be the theme of the sentence after it, as in (55b). However, a definite NP seems to be more readily accepted as a topic than a generic NP, as (56) shows. (53), (54), and (55b) also show a similar situation. (56) a. Na-ben shu Zhangsan hen xihuan. that-CL book very like ‘That book,, Zhangsan likes it, very much.’ b. (N a-ben) shu renren dou hen xihuan. that-CL book everyone all very like ‘(That book) Books, everyone really likes.’ c. ??Shu Zhangsan hen xihuan. book very like ‘The books,, Zhangsan really likes them,’ (56c) has a very similar structure to (56a), (49a), and (55b). The difference is that the verbs in (49a) and (55b) contain a resultative complement, guo (a perfective marker) for kan ‘see’ in kanguo ‘have seen’ in (49a), and wan ‘finish’ for kan ‘read’ in kanwan ‘finished reading’ in (55b). Moreover, since xihuan ‘like’ is not a verb with a resultative complement and shu ‘book’ is a generic bare NP, changing the subject to a universal quantifier rescues (5 6b) without a demonstrative. It seems that definite NPs including 215 proper names can be topics without much limitation, as I have shown in this section, while bare NPs have to have a resultative verb phrase predicate or have a universal quantifier as their subject in order to be a topic. In summary, sentences starting with two NPs are possibly a coordination subject or a topic followed by a subject. In order to be coordinated without conjunctions, the two NPs have to be names and commonly combined NPs, and there can be no co-referent pronouns involved in the verb phrase. In a topic-subject relation, the topic and the subject have to co-refer, or there must be a part-whole relation between the topic and some NP in the rest of the sentence, or the topic-subject must be in a possessor- possessee relation. Otherwise, the topic has to be either a theme or an agent associated with the rest of the sentence. Finally, it seems that bare NPs have more limitations on being a topic than definite NPs. 5.5 Explaining the problems As we have discussed before, zai is not optional with locative phrases, but is optional with temporal phrases. However, when zai phrases occur after the main subject, problems may occur because the omission of zai results in two NPs in a sequence, and this situation makes comprehenders impose an incorrect structure. If the structure imposed by a comprehender fits the sentence, the sentence is fine; if not, it is unacceptable. In this section, I will discuss how sentences starting with two NPs fit or do not fit the temporal structure. I will start from sentences starting with two coordinated NPs, then discuss topic-subj ect relation sentences. 216 Recall that Chinese coordination can leave out the conjunction and dou ‘all’ when the two NPs are names or two commonly combined NPs without a pronoun as the object of the verb phrase, as shown in (44), repeated here as (57), leaving out he ‘and’, and dou ‘both, all’. (57) a. Zhangsan Lisi bei male. Zhangsan Lisi BEI scold LE ‘Zhangsan and Lisi were scolded.’ b. Zhuozi yizi bu-jian le. table chair not-see LE ‘The table(s) and the chair(s) have disappeared.’ Zhangsan Lisi in (57a) and zhuozi yizi (57b) are understood as in coordination, and when they occur in a temporal sentence with yi-qian/yi-hou, ‘Zhangsan and Lisi’ and ‘table and chair’ are the subjects of the sentences as shown in (58). The structure is exactly as if zai is omitted after the second NP. (58) a. Zhangsan Lisi (zai) bei ma le yi-hou jiu ku le. ZAI BEI scold LE after then cry LE ‘After [Zhangsan and Lisi], were scolded, they, cried.’ b. Zhuozi yizi (zai) bu—jian 1e yi-qian shi bai zai keting 1i. table chair ZAI not-see LE before is put ZAI living inside ‘Before they, had disappeared, [the table and the chair], were put in the living room.’ Now consider (59). 217 (59) Zhangsan *(zai) Lisi bei ma 1e yi-hou jiu ku 1e. ZAI BEI scold LE after then cry LE Intended to mean: ‘After Lisi was scolded, Zhangsan cried.’ In (59), zai occurs after Zhangsan, and Zhangsan is the main subject, while Lisi is the subject of the yi-hou clause. After omitting zai, the sentence only has the reading of (58a), but not the intended meaning from (59). That is, the omission of zai in (59) results in a different reading, which is still a temporal sentence; therefore, zai is required in (59), not for syntactic reasons, but to prevent a different interpretation. A similar situation applies to (60). (60) Zhuozi *(zai) yizi mai hui-lai le yi-hou jiu fang zai keting 1i table ZAI chair buy back LE after then put ZAI living inside Intended to mean: ‘After the chair was purchased, the table was put in the living room.’ In (60), the occurrence of zai makes zhuozi ‘table’ be perceived as the main subject, while the absence of zai makes (60) have the meaning as in (58b), and thus a garden path occurs. However, when a sentence begins with NPs that can not be perceived as in coordination without he ‘and’ and don ‘all’, the sentence is unacceptable even if the sentence starts a larger sentence containing a temporal construction, as (61) shows. (61) a. ??Na-zhang zhuozi na-ba yizi mai-huilai 1e yi-hou jiu fang zai keting 1i. that CL table that-CL chair buy-back LE after then put ZAI living inside ‘After [that table and that chair], were purchased , they, were put in the living 9 room. 218 b. Na-zhang zhuozi zai na-ba yizi mai hui lai yi-hou jiu fang zai keting 1i. that-CL table ZAI that-CL chair not-see LE before is put ZAI living inside ‘Before that chair was purchased, that table was put in the living room.’ As I have discussed in 5.4.2.1, uncommonly combined NPs are not generally taken as in coordination. (61a) contains two NPs that are not commonly combined. The two-NP combination is not common and it is unclear how the two NPs are analyzed; when the comprehender encounters yi-qian, it is still difficult to determine whether the two-NP sequence caused by leaving out a conjunction or having omitted zai. The illegitirnately coordinated NPs cannot help the reanalysis that (61a) has an optional zai after the first NP when the comprehender encounters yi-hou. Thus, (61b) shows that zai is required to avoid the unacceptable NP combination. Recall (2a) and (3a), repeated here as (62a) and (63a). (62) a. Zhangsan ??(zai) Lisi chusheng yi-qian jiu chuguo qu 1e. ZAI was-born before then go-abroad go LE Intended to mean: ‘Before Lisi was born, Zhangsan had gone abroad.’ b. ??Zhangsan Lisi chusheng yi-qian jiu chuguo qu 1e. was-born before then go-abroad go LE Intended to mean: ‘Before Lisi was born, Zhangsan had gone abroad.’ (63) Meiguihua *(z_ai) yeshou si le vi-hou jiu diao-xie le. rose ZAI beast die LE after then wither LE Intended to mean: ‘The rose withered after the beast died.’ In (62a) with zai, the first proper name is the main subject, while the second proper name is the subject of the temporal clause. Without zai, as shown in (62b), the two proper names are interpreted as in coordination, and the sentence means before [Lisi and 219 Zhangsan], were born, they, had gone abroad, which is semantically illegitimate. However, when a comprehender reaches the main verb phrase, disambiguation should occur, but it seems to be difficult to reanalyze the sentence to get the intended meaning. In (63), with zai, meiguihua ‘rose’ is the main subject, and yeshou ‘beast’ is the subject of the embedded clause. Without zai, as shown in (63), the two NPs do not have a close relationship like table and chair, (recall (44b)), and are not names. It is difficult to understand what the relation is between meiguihua and yeshou; therefore, it is unclear what the meaning is by saying meiguihua yeshou si la ‘roses beasts died’. The two NPs might be perceived as an NP that is about a kind of yeshou ‘beast’ which has the property of meiguihua ‘rose’; however, it is unlikely. Even in the case of a coordination interpretation, [the rose and the beast], withered after they, died in (63) is difficult to obtain since it is not common for the two NPs to share the predicates of both the embedded clause and the main clause. More importantly, the bias of the predicate, i.e. die for the beast and wither for the rose, is supposed to help in perceiving that the second NP is the subject of the embedded clause. However, it seems that the garden path of the unacceptable NP combination is too strong to be reanalyzed. Now let us add demonstratives to the two NPs in (63) and have (64). (64) Na-duo meiguihua *(zai) na-zhi yeshou si le yi-hou jiu diaoxie 1e. that—CL rose ZAI that-CL beast die LE after then wither LE Intended to mean: ‘That rose withered after that beast die.’ 220 Similar to (63), the two NPs are not in legitimate in coordination without he ‘and’ and don ‘all’ since they are not closely related or commonly combined. Starting with an ungrammatical NP combination, (64) cannot be reanalyzed as an acceptable sentence. In summary, when a temporal clause occurs after the main subject, the absence of zai may lead NPs that are names or two objects that are commonly combined to be interpreted as in coordination. The coordination can still be the subject of the temporal mar sentence, but it has to semantically fit the main verb phrase; otherwise, it is difficult to process and results in an unacceptable sentence as shown in (62b). Nevertheless, once a comprehender reaches the coordination interpretation, the interpretation with zai after the first NP cannot be obtained. Finally, NPs that are not names or common combinations cannot be coordinated without he ‘and’ and dou ‘all’. When zai is omitted in the post- subject position, leaving a two-NP sequence that reads as an illegitimate NP combination, as in (613), (62), (63), and (64), the relationship of the NP combination cannot be determined and thus the sentence is unacceptable. Now we should turn to topic-subject relationships. Before discussing the possible sentences with topic that can appear in yi-qian/yi-hou clauses, I shall discuss how topic sentences can be compatible with temporal sentences. Consider (48), repeated here as (65). (65) W0 xihuan chi pingguo. I like eat apple ‘1 like to eat apples.’ (L & T 1981, 88) 221 According to L & T, a subject can sometimes also be a topic. (65) has wo ‘I’ as the topic as well as the subject. Since subjects in Chinese sentences often have this property, main subjects in temporal sentences can be topics as well as subjects as shown in (66) (The commas are used to emphasize the existence of the topics; they are not used else where. Specifically, the coma in (66b) illustrates that the main subject is also a topic.) (66) a. Na-liang chezi zai women chu 1e chehuo yi-hou jiu bei mai 1e. that-CL car ZAI we have LE car-accident after then BEI sell LE ‘After we had a car accident, the car was sold.’ b. Na—liang chezi, (zai) women chu le chehuo yi-hou jiu bei mai le. that-CL car ZAI we have LE car-accident after then BEI sell LE ‘The car,, after we had a car accident, it, was sold.’ c. Na-liang chezi women (zai) chu le chehuo yi-hou jiu ba ta mai 1e. that-CL car we ZAI have LE car-accident after then BA it sell LE ‘The car,, after having a car accident, we sold it,.’ (66a) is a regular temporal sentence which is composed of a subject, an adverbial phrase, and a verb phrase. However, the main subject can be separated from the rest of the sentence to be a topic as shown in (66b). The sentence is still grammatical and has the same meaning as (66a), and zai can also be omitted. With this in mind, we should predict that a grammatical topic sentence that involves events can occur before yi-qian/yi-hou, and the topic can also be the subject of the main predicate. In addition, as long as the whole sentence is about the first NP, the subject after the topic can be the main subject; in this case zai should be after the subject, as shown in (66c), instead of before the subject as in (66b). Nevertheless, a topic can be the main subject, for example in (66b), na-liang 222 chezi ‘that car’ can be a topic as well as a subject; however, if zai occurs, the NP before zai is generally perceived as a subject. This can be shown from the following sentences. (67) a. ?Nei-ben shu wo kanwan 1e yi-hou yanjing hen bu-shufir that-CL book I read-finish LE after eye very uncomfortable ‘That book,, after I finished reading it,, my eyes became very uncomfortable.’ b. *Nei-ben shu zai wo kanwan le yi-hou yanjing hen bu-shufu that-CL book ZAI I read-finish LE after eye very uncomfortable Intended to mean: ‘That book,, after I finished reading it,, my eyes became very uncomfortable.’ In (67a), nei-ben shu ‘that book’ is the topic, but not the main subject; however, after adding zai in (67b), the sentence becomes unacceptable because nei-ben shu ‘that book’ is interpreted as the main subject, and it competes for the subject position with yanjing ‘eye’. This shows that after adding zai, the first NP is generally interpreted as a subject, although the first NP can also be a topic”. Now we shall consider the possible topic- subject sentences that I discussed in 5.4.2.2, which are allowed to occur before yi-qian/yi- hou in a sentence that contains yi-qian/yi-hou clauses. First, consider a topic-subj ect that has the same reference. (68) Zhangsan, ta, lai le. he come LE ‘Zhangsan,, he, himself, came.’ (69) a. Zhangsan zai ta lai le yi-hou jiu yizhi hen bu gaoxing. ZAI he come LE after then all—the-time very not happy ‘Zhangsan, was very unhappy, after hep/,- came.’ '4 It is unclear why NPs before zai phrases, such as in (67b), are perceived as subjects, but not as topics. 223 b. Zhangsan ta lai le yi-hou jiu yizhi hen bu gaoxing. he come LE after then all-the-time very not happy ‘Zhangsan, was very unhappy, after he, came.’ (68) is a possible topic-subj ect sentence, and the two NPs can only be interpreted as having the same reference. As shown in (69a), with zai, Zhangsan is analyzed as the main subject, and ta ‘he’ is the subject of the embedded clause. Although to can refer back to Zhangsan, this reading is not as easy to get as the disjoint reading. In (69b), without zai, there is only one reading; ta ‘he’ can only refer to Zhangsan. Therefore, the omission of zai makes Zhangsan and to interpreted as in topic-subj ect relation as in (68). When the clause ta lai le ‘he came’ encounters yi-hou, it is possible to reanalyze ta lai le yi-hou ‘after he came’ as an adverbial clause. When the main predicate is encountered, it is still a property of Zhangsan and it is a legitimate sentence. Therefore, in order to get a disjoint reading, zai is necessary in (69a). Now let us examine the reverse order of Zhangsan and ta. (70) a. Ta zai Zhangsan lai le yi-hou jiu yizhi hen bu gaoxing. he ZAI come LE after then all-the-time very not happy ‘After Zhangsan, came, he, was very unhappy.’ b. ??Ta Zhangsan lai 1e yi-hou jiu yizhi hen bu gaoxing. he come LE after then all-the-time very not happy Intended to mean: ‘After Zhangsan, came, he, was very unhappy.’ In (70a), with zai, ta is the main subject, and Zhangsan is the subject of the embedded clause. However, when zai is omitted, as shown in (70b), it is unclear how the sentence can be interpreted, since Ta Zhagnsan lai le ‘he, Zhangsan came’ is not a grammatical 224 sentence with a topic. Thus, when zai is omitted after the first NP, the ungrammatical NP sequence cannot have the co-referent reading, nor can it help in reanalyzing the second NP as the subject of the embedded clause. Now let us examine the possessor-possessee sequence. (71) a. Zhangsan (zai) chezi huai le yi-hou jiu diu 1e gongzuo. ZAI car break LE after then lose LE job ‘Zhangsan, lost his, job after his, car broke down.’ b. Zhangsan chezi (zai) huai le yi-hou jiu bu-neng kai le. car ZAI break LE after then cannot drive LE ‘Zhangsan,, his, car, cannot be driven, after it, broke down.’ c. Chezi *(zai) Zhangsan diu 1e gongzuo yi-hou jiu huai le. car ZAI lose LE job after then break LE Intended to mean ‘After Zhangsan, lost his job, his, car broke down.’ d. ?Zhangsan chezi huai 1e yi-hou jiu bei mai 1e. car break LE after then BEI sell LE ‘Zhangsan, his car, was sold, after it, broke down.’ ‘Zhangsan, was sold after his, car broke down.’ In (71a), the presence or absence of zai does not influence whether chezi ‘car’ is interpreted as Zhangsan’s car. When zai is present, Zhangsan is the subject of the main clause, while chezi ‘car’ is the subject of the embedded clause. When zai is absent, Zhangsan is the topic as well as the subject of the main clause. Therefore, the topic- subject reading is compatible with the structure at the point that zai occurs after Zhangsan. (7la) again supports the idea that zai is optional. In (71b), chezi ‘car’ is interpreted as the main subject. The optionality of zai is not a problem here since the main predicate semantically biases the choice of subject to chezi ‘car’. However, without zai, the reverse order of Zhangsan and chezi in (71c) is not an acceptable NP sequence 225 for a possessor-possessee pair, and the sentence is very difficult to process. Now examine (71d). Although in many cases the name-car pair seems to have no ambiguity, the main predicate in (71d) still shows an ambiguity. Many of my informants readily have the reading of the car being sold; they also think there is another reading that Zhangsan is sold”. A similar kind of ambiguity to (71d) is even more obvious when the possessor and possessee are two animate NPs, as we have seen in (35). The sentences in (72), which I repeat from (35) all have same NP pair. (72) a. Lisi (zai) erzi si 1e yi-hou yijuebuzhen. ZAI son die LE after collapse ‘Lisi, collapsed after his, son died.’ b. Lisi zai erzi si 1e yi-hou jiu cheng-ming 1e. ZAI son die LE after then become-famous LE ‘Lisi, became famous after his, son died.’ c. ??Lisi erzi si 1e yi-hou jiu cheng-ming 1e. son die LE after then become-famous LE ‘Lisi, became famous after his, son died.’ Or ‘Lisi’s son, became famous after he, died.’ In (72a), due to the death of Lisi’s 8011, only Lisi can be the subject of the main predicate, yijuebuzhen ‘collapse’. Therefore, the presence of zai does not make any difference (with zai, Lisi is the main subject; without zai, Lisi is both the topic and the main subject). In (72b), there is only one reading, that Lisi is the person who became famous. If zai does not occur as in (72b) and we get (7 2c), although the initial part of the sentence is the same as in (72a), ambiguity still occurs. The ambiguity occurs because Lisi is perceived '5 There is a third reading of after Zhangsan ’s car broke down, someone sold it; however this reading is not as easy to get as the other two readings. 226 as a topic, and it can be associate with the main predicate. Meanwhile, since Lisi is a topic and the predicate has no semantic bias for choosing the appropriate subject, Lisi erzi ‘Lisi, his 8011’ can be associated with the main predicate, too. Now consider a proper name and bare NP pair in (73). (7 3) a. ??Zhangsan fengshu zhang gao 1e yi-hou jiu ba fengye maple-tree grow tall LE after then BA maple-leave cai xia lai. pick down come Intended to mean: ‘Zhangsan, after his maple trees, had grown tall, he picked their, leaves off.’ b. Zhangsan zai fengshu zhang gao 1e yi-hou jiu ba fengye maple-tree grow tall LE after then BA maple-leave cai xia lai. pick down ‘Zhangsan picked those maple trees’, leaves off, after they, had grown tall.’ Recall that the proper name-inanimate bare NP pair can be interpreted as a possessor- possessee pair only when possessees are closely related to the possessor or are common things that the possessors can have. That is, the possessees have to be things such as cars, computers, etc. that are common in our lives. However the pair in (733), i.e., Zhangsan fengshu, is not a common possessor-possessee pair, and thus (73a) cannot be generally understood as an acceptable possessor-possessee relationship in which Zhangsan can own the trees and pick the leaves off“. However, when zai occurs after Zhangsan as shown in '6 Note that de is usually required to denote the possessor-possessee relationship unless the possessor and the possessee are closely related. However, when the possessee contains a demonstrative, it does not seem necessary for the possessee to be a common object in our lives as (1) shows. In (i), without zai, those maple trees are interpreted as Zhangsan’s trees. This situation cannot be due to syntactic constraints; however, a pragmatic reason of presupposition due to accommodation can explain why (i) is fine. That is, the demonstrative introduces new information that forces a hearer to accommodate it. Therefore, the following sentence is acceptable whether or not zai occurs. 227 (73b), it is not necessary to interpret Zhangsan as having any relationship with fengshu ‘maple trees’, it simply denotes one event happens to the other. The situation in (73) is similar to what I have discussed for the coordination interpretation and the co-referent interpretation. When zai is omitted after the first NP, it leaves a two-NP sequence, which makes a comprehender seek an interpretation for the two-NP sequence, both from their combination relationship and their relationship with the predicate after them, and when the sequence cannot be interpreted, the sentence cannot be reanalyzed after encountering a temporal element. Finally, let us examine topic-subj ect pairs that have agent-theme/theme-agent relationships. (74) a. Zhangsan (zai) shu kan-wan 1e yi-hou cai likai. ZAI book see-finish LE after then leave ‘Zhangsan left after he finished the book.’ b. Shu ?(zai) Zhangsan kan-wan 1e yi-hou jiu bujian le. book ZAI see-finish LE after then not-see LE ‘The book disappeared after Zhangsan finished reading it”.’ (i) Zhangsan (zai) na-xie fengshu zhang gao 1e yi-hou jiu ba fengye cai-xia lai ZAI those maple-tree grow tall LE after then BA maple-leaves pick-down come ‘Zhangsam, after his, maple trees,- had grown tall, he, picked their, leaves off.’ (without zai) ‘Zhansan picked those maple trees’, leaves off, after they, had grown tall.’ (with zai) '7 In Chapter 3, I have argued that bare NPs cannot undergo movement out of adjunct clauses. (74b) is fine without zai because the topic shu is also the subject of the main clause. If it is not the main subject, the sentence will not be good, as shown in (ia); however with a definite NP in (ib), the sentence is acceptable although the topic is not the main subject. (1) a. *Shu wo kanwan le yi-hou yanjing jiu hen bu-shufu bookI read-finish LE after eye then very uncomfortable Intended to mean: ‘The book,, after I finished reading it,, my eyes became very uncomfortable.’ b.? Neiben shu wo kanwan 1e yi-hou yanj ing jiu hen bu-shufu that-CL book I read-finish LE after eye then very uncomfortable ‘ That book,, after I finished reading it,, my eyes became very uncomfortable.’ c. ? (zai) Zhangsan kan-wan 1e yi-hou shu jiu bujian le. ZAI see-finish LE after book then not-see LE ‘The book, disappeared after Zhangsan finished reading it,. ’ 228 c. Shu ??(zai) Zhangsan dao 1e yi-hou jiu bu-jian 1e. book ZAI Zhangsan arrive LE after then not-see LE Intended to mean: ‘The book, after Zhangsan arrived, disappeared.’ (74a) is an agent-theme combination, while (74b) is a theme-agent relationship. Whether zai occurs or not, both sentences have the same meaning, since with or without zai the topic can be the main subject, and there is no semantic bias for the main verb to choose shu ‘book’ as its subject. However, although (740) has the same meaning when zai is present, it is not good since shu Zhangsan dao 1e ‘books, Zhangsan arrived’ is not a legitimate sentence that starts with two NPs. Since (74c) starts out with an illegitimate (' topic-subj ect sentence, a comprehender cannot recover after he/she encounters yi-hou. That is, the illegitimate topic-subj ect interpretation hinders the comprehender from reanalyzing the sentence into a temporal clause. Now consider the following sentences. (75) a. Zhangsan ?(zai) tianqi bian re 1e yi-hou cai mai le lengqiji. weather become hot LE after then buy LE air-conditioner ‘Zhangsan bought an air-conditioner afier the weather became hot.’ b. Tianqi *(zai) Zhangsan mai 1e lengqiji yi-hou cai bian re. weather ZAI buy LE air-conditioner after then become hot ‘The weather became hot after Zhangsan bought the air-conditioner.’ Assuming what I have discussed in Chapter 3, definite NP topics can be topicalized through being coindexed with a moved null operator or a resumptive pronoun, while bare NP topics cannot do so. However, the bare NP in (74b) without zai can be in subject position and therefore it is in an A(rgument)— position, and the gap after the verb should not be a variable, but a resumptive pronoun. However, if shu is a topic, it is coindexed with the subject which forms a chain with the object position of kanwan ‘finish reading’ parallel to a parasitic gap construction. Since bare NP topicalization involves movement, shu must move to the topic position from the subject position as in (ic), and the gap in adjunct position is parasitic to the gap in subject position; therefore, no movement occurs in the adjunct position, thus (74b) is grammatical without zai. 229 (76) a. Zhangsan ?(zai) feng-ye bian hong le yi-hou yao qu ouzhou wan. ZAI maple-leave become red LE after will go Europe play ‘Zhangsan will travel to Europe after the maple-leaves turned red.’ b. F eng-ye *(zai) Zhangsan qu ouzhou wan yi-hou jiu bian hong 1e. Maple-leave ZAI go Europe play after then become red LE Intended to mean: ‘The maple leaves became red after Zhangsan went to Europe.’ Recall that I have discussed that definite NPs are more common topics than bare NPs. (75a) and (76a) are acceptable without zai; however, the reverse order of the NP pair in (75b) and (76b) is not acceptable without zai. This can be explained by the fact that a proper name-bare NP order is a plausible sentence with a topic, but a bare NP-proper name order is not, if they are not in a theme-agent relation. In (75a), when a comprehender reads Zhangsan tianqi ‘weather’, he/she takes Zhangsan as a topic and waits for a logical interpretation; therefore, the sentence is only acceptable, but not especially good according to many of my informants. However, when a comprehender hears tianqi ‘weather’ Zhangsan, he/she cannot take tianqi as a topic and therefore, the sentence is unacceptable at the very beginning. 5.6 Zai in rte-shihou clauses This chapter starts with the problem about the asymmetry between locative phrases and yi-qian/yi-hou phrases. In fact, the issue of the absence of zai in the post- subject position also exists in de-shihou clauses”. In the following, I examine the ‘8 This post-subject two-NP sequence ambiguity also applies to other temporal phrases, as shown in (i). Since the situation of ambiguity is similar to yi-qian/yi-hou clauses, I will not discuss it further. (i) Lisi *(zai) Zhagnsan lai de na-yi-tian chuguo qu 1e. ZAI come DE that-one-day out-abroad go LE Intended to mean: ‘Lisi went abroad the day that Zhangsan came.’ 230 optionality of zai in ale-shihou clauses by looking at sentences whose initial part can be interpreted as a coordination or a topic-subject relationship when zai is omitted. First let us examine the initial phrases that can be interpreted as in coordination as shown in (77). (77) a. Zhangsan *(zai) Lisi lai de shihou cai kanshu. ZAI come DE time then study Intended to mean: ‘Zhangsan studies only when Lisi comes.’ b. Zhangsan Lisi lai de shihou xia le women yi-tiao. come DE time scare LE us study one-jump ‘When [Zhangsan and Lisi], came, they, scared us’ In (77a), with zai, Zhangsan is the main subject; however, when zai is omitted, Zhangsan and Lisi are readily interpreted as in coordination, and when comprehenders encounter the verb phrase, they find that the coordination interpretation cannot be right and have to reanalyze the sentence. In (77b), when we semantically bias the verb phrase, the sentence becomes fine. Now examine the following sentences in which the two NPs can have the same reference in topic-subj ect relationship before comprehenders encounter de-shihou. (78) a. Zhangsan zai ta lai de shihou ba women xia le yi-tiao. ZAI he/she come DE time BA us scare LE one-jump ‘Zhangsan, scared us when he, came.’ b. . Zhangsanta lai de shihou ba women xia le yi-tiao. he/she come DE time BA us scare LE one-jump ‘Zhangsan,, when he, himself, came, be.- scared us.’ 231 In (783), the presence of zai indicates that Zhangsan is the main subject, and Zhangsan and ta ‘he’ have a disjoint reading"). Without zai in (78b), Zhangsan is perceived as a topic and has the same reference with ta ‘he’. Now we shall examine the possessor- possessee relations. (79) a. Zhangsan (zai) erzi si de shihou ku de hen shangxin. ZAI son die DE time cry DE very hurt-heart ‘Zhangsan, cried very sadly when his, son died.’ b. Zhangsan zai erzi chu chehuo de shihou shoushang le. ZAI son have car-accident DE time get-injure LE ‘Zhangsan, was injured when his, son had a car accident.’ 0. ??Zhangsan erzi chu chehuo de shihou shoushang 1e. son have car-accident DE time get-injure LE ‘Zhangsan, was injured when his, son had a car accident.’ Or ‘Zhangsan,, when his, son, had a car accident, he, was injured.’ In (79a), without zai, Zhangsan is still interpreted as the main subject, since Zangsan erzi ‘Zhangsan, his son’ died, and the only person who cries is Zhangsan. In (79b), the occurrence of zai makes Zhangsan the main subject; however, without zai, in (79c), the sentence is ambiguous in that either Zhangsan or Zhangsan erzi can be the main subject and is difficult to process. Now let us examine (80) in which a proper name and a bare NP have an agent-theme relationship. (80) a. Zhangsan (zai) yifu xiwan de shihou jiu hundao 1e. ZAI clothes wash-finish DE time then faint LE ‘Zhangsan, fainted when he, finished washing clothes.’ '9 As I have discussed before, with zai, the co-referent reading between the two NPs is possible when the name precedes the pronoun, however it is not reached as readily as the disjoint reading. 232 b. Yifu ?(zai) Zhangsan xi de shihou jiu po le. Clothes ZAI wash DE time then worn-out LE ‘The piece of clothing was worn out when Zhangsan washed it.’ c. Yifu ??(zai) Zhangsan chuqu de shihou jiu po 1e. Clothes ZAI go-out DE time then worn-out LE Intended mean: ‘The piece of clothing was worn out when Zhangsan went out.’ (803) and (80b) start with agent-theme and theme-agent relation respectively, and the omission of zai does not influence the sentence meaning. In (80c), the beginning of the sentence is not a legitimate topic-subj ect relation, and the first NP does not contain a demonstrative; therefore, the sentence is difficult to process. 5.7 Conclusion In this chapter, I first examined the asymmetry between locative phrases and temporal phrases with respect to the presence of zai . I assume that locative phrases need case, while temporal phrases do not need case. Stroik’s (1992) analysis that English temporal phrases are predicates and do not need case further supports my assumption that Chinese temporal phrases do not need case either. In order to confirm that in Chinese, locative phrases need case, I examine cases where locative phrases seem to be allowed without case. As a result of my investigation, when locative phrases appear not to have case, the locative phrases are actually in case positions (e. g. subject position). After confirming that locative phrases do need case, I focused my investigation on temporal constructions, mainly on yi-qian/yi-hou clauses. I found that zai can be optional in most cases except for in post-subject position, in which the occurrence of zai is not consistent. Therefore, I propose that the optionality of zai in 233 post-subj ect position is independent of syntactic constraints. It seems to behave similarly to that was in the horse (that was) raced past the barn fell in which the omission of that was causes a garden path. The occurrence of that was helps the parser to construct the correct structure. That is, when zai is absent, the comprehender tries to structure the two sentence-initial NPs into the sentence. If the imposed structure is a legitimate coordination or a legitimate sentence with a topic, and the comprehender does not encounter disambiguating information such as a verb phrase (that is, a verb phrase that does not semantically suit the imposed structure), the sentence is fine. However, these grammatical sentences without zai may have different meanings from sentences with zai. This is the case because the NP that immediately precedes zai is generally perceived as the main subject. Without zai, the main subject can be the two NPs in a coordination, or the subject can be either the first or the second NP in a topic-subj ect sentence, depending on the main verb. Nevertheless, when zai is absent and the two NPs are not a legitimate combination, the sentence cannot be reanalyzed as an acceptable sentence even after a comprehender reaches the temporal element yi-qian/yi-hou. Therefore, in order to avoid an incorrect analysis, including zai is preferred in post-subject position. Finally, I examined the post-subj ect de shihou clauses, and found similar patterns as in yi-qian/yi- hou clauses in post-subject position. Therefore, these phenomena generally occur in temporal phrases after the main subject, and the optionality of zai should not be attributed to syntactic constraints. Instead, it can be attributed to processing cues which help the comprehender avoid an unrecoverable garden path sentence. 234 Chapter 6: Conclusion In this thesis, I have discussed the structure of de shihou clauses, yi-qian/yi-hou clauses, and the use of zai in Chinese temporal clauses. I will now answer the questions that I stated in Chapter 1. These questions are repeated here as follows: 1. Is de shihou a single element? That is, can it be separated into smaller units? What is the structure of de shihou clauses? 2. Are yi-qian/yi-hou each a single elements? That is, can they be separated into small units? What is the structure of yi-qian/yi-hou clauses? 3. What is the fimction of zai in Chinese temporal clauses? What are the constraints on its occurrence? For the first question, de shihou cannot be a single element. One major reason to treat de shihou as a single element is that shihou cannot be modified, and therefore, is not a noun phrase. However, the fact that shihou cannot be modified cannot be taken as evidence to argue against its being a noun phrase because there are many noun phrases which cannot be modified. Other evidence that shows that de shihou is not a single element is that de and shihou can be separated by a demonstrative, similarly to other adjunct relative clauses’ head nouns. Finally, based on the structure of relative clauses that I have argued for in Chapter 2, I have shown that de shihou clauses are adjunct relative clauses. 235 Yi-qian/yi-hou are also not single elements. Yi, zhi, and de are all determiners. Yi and zhi both are determiners that still keep their classical use, unlike de, which has become semantically bleached as Simpson argues,. Furthermore, qian/hou are nominal clitics that take a noun phrase or a clause as their complement, and qian/hou can be used as yi—qian/yi-hou with either an NP or a CP preceding it; thus yi-qian/yi—hou are DPs, but not compound words. As for the structure of yi-qian/yi-hou clauses, qian/hou take a CP as their complement instead of moving fiom within the CP, and this CP moves to Spec of DP that has yi or zhi as its head, and thus is like a noun complement clause. With regards to the last question, zai in Chinese temporal clauses is optional both when temporal clauses are in pre-subject positions or post-subject positions. Nevertheless, zai is sometimes required in post-subject position to avoid a temporary parsing ambiguity caused by two noun phrases in a sequence. 236 References: Bever, T. (1970) The Cognitive Basis for Linguistic Structure, Wiley, New York. Boarwning, M. (1991) Null Operator Constructions, Garland, New York. Chao, Y. (1968) Grammar of spoken Chinese, University of California Press, Berkeley and Los Angeles. Chen, C.-Y. (1978) “Aspectual Features of the Verb and the Relative Positions of the Locatives,” Journal of Chinese Linguistics 6, 76-103. Cheng, L. (1986) “De in Mandarin,” Canadian Journal of Linguistics 31, 4, 313-326. Chomsky, N. (1977) “On Wh-movement,” in P. Culicover, T. Wasow, and A. Akrnajian ed., Formal Syntax, Academic Press, New York. Chomsky, N. (1981) Lectures on Government and Binding, Foris, Dordrecht. Chomsky, N. (1982) Some Concepts and Consequences of the theory of government and Bindng, MIT Press, Cambridge, Massachusetts. Chomsky, N. (1986a) Barriers, MIT Press, Cambridge, Massachusetts. Chomsky, N. (1986b) Knowledge of Language: Its Nature, Origin, and Use, Praeger, New York. Chomsky, N. (1993) “A minimalist Program for Linguistic Theory,” in K. Hale and S. J. Keyser, ed., The view from Building 20: Essays in linguistics in honor of Sylvain Bromberger 1-52, MIT Press, Cambridge, Massachusetts. Chomsky, N. (1995) The Minimalist Program, MIT Press, Cambridge, Massachusetts. Duanmu, S. (1997) “Wordhood in Chinese,” in J. Packard ed., New Approaches to Chinese Word Formation: Morphology, Phonology and the Lexicon in Modern and Ancient Chinese, Mouton de Gruyter, Berlin, Germany. Eric, M. (1985) Temporal interpretation, ms., University of South California. Ernst, T. (1988) “Chinese Post-positions?-Again,” Journal of Chinese Linguistics 16, 2, 219-244. Ernst, T. (1994) “Chinese Adjuncts and phrase Structure Theory,” Journal of Chinese Linguistics, 22,1,41-70. Gao, M.-K. (1957) Han Yu Yufa Lun [On Chinese Grammar], Kexue Publisher, Beijing. 237 Gasde, H. & W. Paul (1996) “Functional Categories, Topic Prominence, and Complex Sentences in Mandarin Chinese,” Linguistics 34, 263-294. Gies, M. (1970) Adverbial Subordinate Clauses in English, Doctoral dissertation, MIT, Cambridge, Massachusetts. Giusti, G. (1997) “The Categorial Status of Determiners,” in L. Haegeman, ed., The new comparative syntax, Longrnan. Greenberg, J. (1978) “Generalizations about numeral systems,” in J. Greenberg ed., Universals of Human Language Vol 3: Word Structure, 249-297, Stanford University Press. Grosu, A. (1988) “On the distribution of genitive phrases in Romanian,” Linguictics 26, 931-49. Hashimoto, A. (1971) “Mandarin syntactic structures,” Unicorn 8. He, L. et al. (1985) Gu dai han yu xu ci tong shi [A handbook for Archaic Chinese particles], Beijing publisher, Beijing. Hou, J. and C. Kitagawa (1987) “Null Operators and the Status of Empty Categories in Chinese,” Linguistic Inquiry, 18, 3, 18-523. Huang, C.-T. (1982) Logical relation in Chinese and the theory of grammar, Doctoral Dissertation, MIT, Cambridge, Massachusetts. Huang, C.-T. (1983) “On the Representation of Scope in Chinese,” Journal of Chinese Linguistics 11,1, 37-91. Huang, C.-T. (1984a) “On the Distribution and reference of empty pronouns,” Linguistic Inquiry 15, 531-574. Huang, C .-T. (1984b) “Phrase structure, Lexical Integrity, and Chinese Compound,” Journal of the Chinese Language Teachers Association, 19,2, 53-78. Huang, C.-T. (1989) “ Pro-Drop in Chinese: A generalized control theory.” in J aeggli and K. Safir ed., The Null Subject Parameter O. Hwang, J.-L. (1998) “A Comparative Study on the Grammaticalization of Saying Verbs in Chinese,” paper presented at the 7‘h Annual Meeting of the International Association of Chinese Linguistics/the 10th North American Conference on Chinese Linguistics, Stanford University, June. J ackendoff, R. (1971) “Gapping and Related Rules,” Linguistic Inquiry 2, 21-35. 238 1r- J ackendoff, R. (1973) “The Base Rules for Prepositional Phrases,” in S. Anderson and P. Kiparsky ed., A F estschrzft for Morris Halle, Holt, Rinehart, and Winston, New York. Kayne, R. (1994) The Antisymmetry of Syntax, MIT Press, Cambridge, Massachusetts. Koopman, and Sportiche (1982) “Variables and the Bijection Principle,” Linguistic Review 2, 2, 139-160. Larson, R. (1985) “Bare-NP adverbs,” Linguistic Inquiry 16, 335-391. Larson, R. (1990) “Extraction and Multiple Selection in PP,” The Linguistic Review 7, 169-182. Larson, R. (1987) “Missing Prepositions and the Analysis of English Free Relative Clauses,” Linguistic Inquiry 18, 239-266. Li, Y.-H. A. (1990) Order and Constituency in Mandarin Chinese, Kluwer, Boston. Li, Y.—H. A. (1999) Word order, Structure and Relativization. ms., University of South California. Li, Y.-H. A. (2001) The ba Construction, ms., University of South California. Li, C. & S. Thompson. (1981) Mandarin Chinese: A Functional Reference Grammar, University of Califomia Press, Berkeley. Lin, J ., K. Murasugi, & M. Saito. (2001) “Modification and Specification: A study of NO and DE,” paper presented at Theoretical East Asian Linguistics Workshop, Irvine, June. Liu, F. (1998) “A clitic analysis of Locative particles,” Journal of Chinese Linguistics, 26, 1, 48-70. Longobardi, G. (1994) “Reference to proper names: a theory of N-movement in syntax and Logical Form,” Linguistic Inquiry 25, 609-665. McCawley, J. (1992) “Justifying Part-of-speech Assignments in Mandarin Chinese,” Journal of Chinese Linguistics 20, 2, 211-245. Muller, G. (1996) “A constraint on Remnant movement,” Natural Language and Lingyistic Theory 14, 355-407. Munn, A. (1991) “Clausal adjuncts and temporal ambiguity,” in G.Westphal, B. Ao, H.-R. Chae, ed., Proceedings of the eighth Eastern States Conference on Linguistics 265-276. 239 Murasugi, K & M. Saito (1990) “no ' deletion in Japanese: A Preliminary Study,” Hajime ed., In Japanese/Korean Linguistics 1, 285-301. Ning, C. (1993) The overt syntax of relativization and topicalization, Doctoral dissertation, University of California, Irvine. Ning, C. (1996) “DE as a functional head in Chinese,” in UCI working papers in Linguistics 1, 63-79. Nishiyama, K. (1997) “Free order in Buginese Noun Phrases,” poster presented at NELS 28, University of Toronto. Pulleyblank, E. (1995) Outline of Classical Chinese Grammar, UBC Press, Vancouver. Rizzi, L. (1990) Relativized Minimality, The MIT Press Cambridge, Massachusetts. Rothstein, S. (1985) The syntax forms of Predication, Indiana University Linguistics Club, Bloomington, Indiana. Ross, C. (1982) “On the functions of Mandarin DE,” Journal of Chinese Linguistics 11, 2, 215-246. Safir, K. (1986) “Relative clauses in a theory of binding and levels,” Linguistic Inquiry 17, 663-689. Safir, K. (1987) “What Explains the Definiteness Effect?,” in E. Reuland and A. ter Meulen eds., The representation of (in)definiteness, MIT Press, Cambridge, Massachusetts. Shi, D. (1997) “Issues on Chinese Passive,” Journal of Chinese Linguistics 25, 1, 41-70. Simpson, A. (1999) “On the status of ‘modifying’ DE and the structure of the Chinese DE,” CSLI, Stanford. Simpson, A (2000) “Definiteness agreement and the Chinese DE,” ms., University of Southern California. Stowell, T. (1981) Origins of Phrase Structure, Doctoral dissertation, MIT, Cambridge, Mass. Stroik, T. (1992) “On the Distribution of Temporal and Locative NP Adverbials,” The Linguistic Review 9, 3, 267-284. Sybesma, R. (1999) The Mandarin VP, Kluwer Academic Publishers, Boston. 240 Szabolcsi, A. (1994) “The Noun Phrase,” in F. Kiefer and K. E. Kiss ed., Syntax and Semantics, 27, The Syntactic Structure of Hungarian. Tang, T. (1976) Guo2 yu3 yu3fa3 yan2jiu lun4ji2 [A Study of Mandarin Chinese Grammar], Students Books, Taipei. Tang, C-C. J. (1990) Chinese Phrase Structure and the Extended X'-Theory. Doctoral dissertation, Cornell University, Ithaca. Troike, and Pan (1994) “ In Defense of P’: the Status of Chinese Locative Particles,” in the proceedings of NA CCL, 6, 59-263. Tsao, P. (1997) “Complement and adjunct distribution and the two-place nominal in Chinese,” paper presented at the 9th NACCL, Victoria. Vergrraud, J .-R. (1974) French relative clauses, Doctoral dissertation, MIT. Wang, L. (1962) Gu dai han yu [Archaic Chinese], Zhonghua book store, Beijing. Wen, L. (1957) Chusuo, Shijian, he Fangwei [Places, Times, and Directions], Xin Zhishi Chubanshe, Shanghai. Wu, X.-Z. (2000) Grammaticalization and the Development of Functional Categories in Chinese, Doctoral dissertation, University of Southern California. Xu, D. (1997) Functional Categories in Mandarin Chinese, Netherlands. Yang, S. (1995) The Aspectual System of Chinese. Doctoral Dissertation, University of Victoria. 241 1|NM][lllfllflsjlljfllflfil