an object clitic projection in mandarin chinese

41
BONNIE CHIU AN OBJECT CLITIC PROJECTION IN MANDARIN CHINESE* This paper argues that the clausal structure of Mandarin Chinese contains an object clitic projection (SuoP). The discussion is based on the syntactic behavior of the particle suo, which occurs when syntactic movement of an accusative NP has occurred. Suo is triggered when the movement has proceeded through SpecSuoP. Drawing the evidence from suo, I argue that Chinese inflectional structure contains Case projec- tions: NomP (nominative Case projection) and SuoP (accusative Case projection). The presence of these projections has far-reaching consequences for the analyses of various syntactic constructions, such as relative clauses, the bei-construction, and topicalization structures. An unified analysis is given to these constructions. 1. INTRODUCTION In the recent development of syntactic theory, various elements, such as Agreement and clitics, have been argued to be syntactically active heads and are treated as X~ elements, projecting XP categories (Pollock (1989), Chomsky (1989), Sportiche (1992)). The question arises as to what happens to these projections in languages, such as Chinese, that do not show overt realizations of these categories in general. Chinese lacks agreement marking, as well as clitics found in Romance languages, and the absence of such mor- phology makes it difficult a priori to argue for these categories in Chinese. In this paper, I argue that in spite of its impoverished morphological marking, the inflectional structure of Chinese contains projections which exhibit properties of clitics observed in Romance languages. Discussion will be based on the behavior of the particle suo, which occurs in relative clauses: (1) [xiaotou suo meiyou touzou de] thief SUO not-have steal DE neixie shoushi zai zuozi-shang those jewelry at table-top 'The jewelry that the thief didn't steal is on the table.' suo is a remnant from Classical Chinese. It is sometimes analyzed as a 'relative pronoun' by scholars of Classical Chinese (e.g., Wang (1962), Liu (1990)). Its present day usage often gives a 'literary flavor' to the sentences it appears in. 1 I will present specific properties associated with suo (section 2). suo will be shown to occur in a position between the subject Journal of East Asian Linguistics 4, 77-117, 1995. 1995 Kluwer Academic Publishers. Printed in the Netherlands.

Upload: bonnie-chiu

Post on 10-Jul-2016

221 views

Category:

Documents


1 download

TRANSCRIPT

B O N N I E CHIU

AN OBJECT CLITIC PROJECTION IN

M A N D A R I N CHINESE*

This paper argues that the clausal structure of Mandarin Chinese contains an object clitic projection (SuoP). The discussion is based on the syntactic behavior of the particle suo, which occurs when syntactic movement of an accusative NP has occurred. Suo is triggered when the movement has proceeded through SpecSuoP. Drawing the evidence from suo, I argue that Chinese inflectional structure contains Case projec- tions: NomP (nominative Case projection) and SuoP (accusative Case projection). The presence of these projections has far-reaching consequences for the analyses of various syntactic constructions, such as relative clauses, the bei-construction, and topicalization structures. An unified analysis is given to these constructions.

1. INTRODUCTION

In the recent development of syntactic theory, various elements, such as Agreement and clitics, have been argued to be syntactically active heads and are treated as X ~ elements, projecting XP categories (Pollock (1989), Chomsky (1989), Sportiche (1992)). The question arises as to what happens to these projections in languages, such as Chinese, that do not show overt realizations of these categories in general. Chinese lacks agreement marking, as well as clitics found in Romance languages, and the absence of such mor- phology makes it difficult a priori to argue for these categories in Chinese.

In this paper, I argue that in spite of its impoverished morphological marking, the inflectional structure of Chinese contains projections which exhibit properties of clitics observed in Romance languages. Discussion will be based on the behavior of the particle s u o , which occurs in relative clauses:

(1) [xiaotou s u o meiyou t o u z o u de] thief SUO not-have steal DE

neixie shoushi zai zuozi-shang those jewelry at table-top

'The jewelry that the thief didn't steal is on the table.'

s u o is a remnant from Classical Chinese. It is sometimes analyzed as a 'relative pronoun' by scholars of Classical Chinese (e.g., Wang (1962), Liu (1990)). Its present day usage often gives a 'literary flavor' to the sentences it appears in. 1 I will present specific properties associated with s u o (section 2). s u o will be shown to occur in a position between the subject

Journal o f East Asian Linguistics 4, 77-117, 1995. �9 1995 Kluwer Academic Publishers. Printed in the Netherlands.

78 BONNIE CHIU

and negation, which I will analyze to be triggered by the syntactic movement of an accusative NP proceeding through SpecSuoP. I will present evidence for the postulation of a nominative Case projection, NomP, following Sportiche (1992), and argue that the subject is in the Spec of this position (section 3). The analysis of s u o gives insight into Chinese clausal struc- ture and has consequences for the analyses of various syntactic constructions in the language. I will discuss these constructions, particularly relative clauses, b e i - c o n s t r u c t i o n s , and topicalization structures. Relative clauses are proposed to be formed either by movement or base-generation of a null operator. The same mechanism will be shown to be also at work in be i -

constructions. Topicalization is argued not to involve movement but rather involves base-generation of a null operator, similar to the case of a left- dislocated structure. Finally, s u o provides further evidence pertaining to certain controversial issues in Chinese syntax, such as whether raising verbs exist in the language and, if so, which verbs are raising verbs, and whether there is restructuring, aad, if so, which verbs trigger restructuring.

2. s v o

2.1 . D i s t r i b u t i o n o f suo

s u o occurs in clauses containing de . 2 There are two types of de-clauses: relative clauses and complement clauses of a noun, shown by (2) and (3), respectively. Tang (1976) observes that s u o occurs in relative clauses, but

a. [Lisi mai __ de] neixie shu Lisi buy DE those book

'the books that Lisi bought'

b. [Lisi s u o mai __ de] neixie shu Lisi SUO buy DE those book

(3) a. [women choucuo jinfei de] wenti (Tang (1976)) we collect funds DE question

'(lit.) the question (problem) that we collect funds'

b.*[women s u o choucuo jinfei de] wenti we SUO collect funds DE question

s u o can therefore only appear in relative clauses. However, not all types of relative clauses can contain s u o , as I will show below. Two factors

not in complement clauses of a noun.

(2)

AN O B J E C T C L I T I C P R O J E C T I O N IN M A N D A R I N C H I N E S E 79

determine suo's occurrence in relative clauses. First, it occurs only where

the relativization site contains an empty category; thus a resumptive pronoun cannot be present. Second, it is restricted to appearing in clauses whose relativized elements correspond to direct objects.

In relative clauses, a resumptive pronoun obligatorily occurs when an indirect object, the NP in a ba-phrase, the complement NP of a PP, or a

[Lisi song *(ta) yiben shu de] Lisi send him one book DE

neige ren (indirect object) that person

'the person that L sent a book'

(5) [Lisi ba *(ta) da-le de] neige ren (ba-phrase) Lisi BA him hit-LE DE that person

'the person that Lisi hit'

(6) [Lisi gen *(ta) zhu-guo] de neige ren (PP) Lisi with him live-GUO DE that person

'the person that Lisi lived with'

(7) [Lisi renshi *(tade) baba] de neige ren (genitive NP) Lisi know his father DE that person

'the person that Lisi knows his father is very rich'

When relativization applies to a direct object, however, a resumptive pronoun can be optionally present (Li and Thompson (1981)). 3

(8) [Lisi renshi (ta)] de neige ren (direct object) Lisi know him DE that person

'the person that Lisi knows'

suo may not occur in relatives which contain resumptive pronouns, as the following examples show:

(9) *[Lisi suo song ta yiben shu] L SUO send him one book

de neige ren (indirect object) DE that person

'the person that L sent a book'

genitive NP is relativized:

(4)

80 BONNIE CHIU

(lO) * [Lisi ba ta suo da-le] de neige ren Lisi BA him SUO hit-LE De that person

'the person that Lisi hit'

(ba-phrase)

(11) *[Lisi gen ta suo zhu-guo] de neige ren (PP) Lisi with him SUO live-GUO DE that person

'the person that Lisi lived with'

(12) *[Lisi suo renshi tade baba hen you qian] Lisi SUO know his father very have money

de neige ren (genitive NP) DE that person

'the person that Lisi knows his father is very rich'

(13) * [Lisi suo renshi ta] de neige ren (direct object) Lisi SUO know him DE that person

'the person that Lisi knows'

When a direct object is relativized, s u o only co-occurs with a gap, not with a resumptive pronoun (cf. (13) and (14)):

(14) [Lisi suo r e n s h i ] de neige ren Lisi SUO know DE that person

'the person that Lisi knows'

Let us next address the question whether having a gap constitutes a suffi- cient condition for s u o ' s occurrence. When a subject or an adjunct is relativized, a gap is present, and no resumptive strategy is available. Despite the occurrence of a gap in these cases, suo cannot appear.

(15) [ _ mai-le (*suo) neixie shu] buy-LE SUO those book

de neige ren DE that person

'the person who bought those books'

(subject)

(16) [Lisi (* suo) da-le Zhangsan __ ] Lisi SUO hit-LE Zhangsan

'the reason that Lisi hit Zhangsan'

de yuanyin DE reason

(why)

AN OBJECT CLITIC PROJECTION IN MANDARIN CHINESE 81

(17) [Lisi (*suo)

Lisi SUO

'the place that

(18) [Lisi (*suo)

Lisi SUO

'the time that

(19) [Lisi (*suo)

Lisi SUO

kandao Zhangsan __ ] de difang

see Zhangsan DE place

Lisi saw Zhangsan'

kandao Zhangsan see Zhang

Lisi saw Zhangsan'

da Zhangsan __ ] hit Zhangsan

(where)

__ ] de shihou (when) DE time

de fangfa (how) DE method

'the way that Lisi hit Zhangsan'

Thus, the data examined so far show that suo can appear only when a direct object has been relativized. 4

2.2. suo's Occurrence: Case- or Theta-Related?

A question immediately arises: what counts as a direct object? Is it an accusative Case-marked element or a thematic direct object that triggers the occurrence of suo? Let us turn to unaccusative constructions, which provide us with an answer.

If the relevant factor conditioning suo's occurrence is theta-related, suo

should be able to appear in unaccusatives. If accusative Case is the relevant factor, suo should not be able to show up in unaccusatives, given the standard assumption that unaccusative verbs do not assign accusative Case. The following examples show that suo cannot appear in unaccusative structures.

(20) [neixie ren lai-guo ] those people come-GUO

'Those people have come.'

(21) [ _ _ lai-guo ] de neixie come-GUO DE those

'those people who came'

(22) * [suo lai-guo ] de SUO come-GUO DE

'Those people who came'

ren people

neixie ren those people

This strongly suggests that licensing is not directly linked to theta roles but to accusative Case. 5

82 BONNIE CHIU

An apparent problem for the preceding generalization (accusative-Case being the deciding factor for suo's occurrence) comes from bei-construc-

tions, the so-called passive construction in Chinese, where suo can appear when the subject is relativized.

(23) Lisi bei Zhangsan ma-guo Lisi BEI Zhangsan scold-GUO

'Lisi has been scolded by Zhangsan.'

Relativization of the surface subject NP, Lisi, which is the thematic object, allows suo to appear: 6

(24) [__ bei Zhangsan suo ma-guo de] BEI Zhangsan SUO scold-GUO DE

neige ren that person

'the person who has been scolded by Zhangsan'

The data from bei-constructions appears to suggest suo's occurrence is theta- related. However, independent evidence suggests that in bei-construcfions

the verb retains the ability of assigning accusative Case, in contrast to verbs in passive constructions of other languages, such as English (Li (1990), Goodall (1990), Chiu (1993)). In view of this fact and other properties of bei-constructions, some linguists have proposed that there is no accusative Case absorption of verbs in bei-constructions (Feng (1991), Chiu (1993), Li (1994)). If this is correct, then, suo's occurrence in bei-

constructions is Case-related as well. suo's occurrence in the bei-con-

struction will be discussed further in section 3.4. To summarize, we have established that suo only occurs in relative

clauses containing a gap and that suo may not appear in relative clauses containing a resumptive pronoun. Furthermore, suo is restricted to rela- tivization of a direct object. I suggest, drawing on data from unaccusatives, that suo's occurrence is Case-related, not Theta-related. Bei-constructions

in which suo can occur when a subject NP is relativized are suggested to be accusative-Case-related as well. I therefore conclude that accusative Case is the relevant factor conditioning suo's occurrence.

2.3. Position o f suo in the Clause

I will assume the following inflectional structure for Mandarin Chinese, adopting the proposal in Chiu (1993): 7

AN OBJECT CLITIC P R O J E C T I O N IN M A N D A R I N CHINESE 83

(25) TP

VP

In this section, I consider suo ' s occurrence with respect to the structure in (25). To begin, suo must precede negation. There are two types of negation, bu and m e # y o u (lit. 'not-have'), in Mandarin Chinese. In general, the former is used in negating a non-past or stative event whereas the latter is employed when a past event is negated, suo must precede both types of negation.

(26) a. [Lisi suo bu c h i de ] neixie dongxi Lisi SUO not eat DE those thing

'the things that Lisi doesn't eat'

b.*[Lisi bu suo c h i de ] neixie dongxi Lisi not SUO eat DE those thing

(27) a. [Lisi suo meiyou ch i_ de ] neixie dongxi Lisi SUO not-have eat DE those thing

'the things that Lisi didn't eat'

b.*[Lisi meiyou suo ch i_ de ] neixie dongxi Lisi not-have suo eat De those thing

Secondly, suo occurs lower than both S-level adverbs and the subject. There are two positions in which S-level adverbs can appear: preceding or following the subject NP. In either case, suo must follow both the subject and the S-level adverb.

(28) a. [Lisi dagai suo renshi_ de ] neixie ren Lisi probably SUO know De those people

'the people that Lisi probably knows'

b. [Lisi (*suo) dagai r e n s h i de ] neixie ren Lisi SUO probably know DE those people

(29) a. [dagai Lisi suo renshi _ de ] neixie ren probably Lisi SUO know DE those people

b. [dagai (*suo) Lisi renshi_ de ] neixie ren probably SUO Lisi know DE those people

84 BONNIE CHIU

When an S-level adverb, such as dagai 'probably', follows the subject NP, suo must occur to its right, as shown in (28). If the S-level adverb precedes the subject NP, suo must follow both the subject NP and the S- level adverb as well, as shown in (29).

Since suo is higher than negation and lower than S-level adverbs, it should appear between S-level adverbs and negation, as is indeed the case:

(30) a. [Lisi dagai suo bu r e n s h i de ] neixie ren Lisi probably SUO not know DE those people

'the people that Lisi probably doesn't know'

b. [Lisi dagai suo meiyou mai__ de ] neixie shu Lisi probably SUO not-have buy DE those book

'the books that Lisi probably didn't buy'

Thirdly, suo appears higher than manner adverbs. Since negation precedes manner adverbs and suo precedes negation, it follows that suo should precede both negation and manner adverbs. This is indeed the case.

(31) a. [Lisi suo meiyou henhende s h a s i de ] neixie ren Lisi SUO not-have cruelly kill DE those people

'the people that Lisi didn't cruelly killed'

b. [Lisi meiyou (*suo) henhende s h a s i de] Lisi not-have SUO cruelly kill DE

neige ren that person

c. [Lisi meiyou henhende (*suo) s h a s i de ] Lisi not-have cruelly SUO kill DE

neige ren that person

The data presented above is summarized in (32):

(32) NP-subject S-level-adv S U O Neg manner-adv Verb NP-object

suo occupies a sentence-internal position which is higher than negation but lower than S-level adverbs.

A few words are in order regarding the position of suo with respect to bei and the NP which follows it in bei-construct ions (bei NP). suo must appear after bei NP in bei-constructions.

AN OBJECT CLITIC PROJECTION IN MANDARIN CHINESE 85

(33) *Akiu suo bei Lisi ma-guo Akiu SUO BEI Lisi scold-GUO

'Akiu has been scolded by Lisi.'

(34) Akiu bei Lisi suo ma-guo Akiu BEI Lisi SUO scold-GUO

'Akiu has been scolded by Lisi.'

Since bei NP occupies a position lower than negation and modals, it follows that suo in bei-constructions does not occur in the same position:

(35) Lisi (*suo) meiyou bei Akiu ma-guo Lisi SUO not-have BEI Akiu scold-GUO

'Lisi has not been scolded by Akiu.'

(36) Lisi meiyou bei Akiu (suo) ma-guo Lisi not-have BEI Akiu SUO scold-GUO

That suo seems to occupy different positions in bei-constructions will be discussed in section 3.4.

Various questions arise with respect to suo: in what position does it occur? What kind of element is suo? Is it an XP, like a relative pronoun, or some kind of X ~ head? If it is a head, what kind of head is it? Before addressing these questions, I first examine other distributional properties of suo.

2.4. Locali ty: Subjacency

Consider the distribution of suo with respect to the head noun of the relative clause, suo can appear if an object of an embedded clause is relativized (I limit myself here to tensed embedded clauses; infinitives will be examined in section 4):

(37) wo renshi [Lisi yiwei [Ak~u suo xihuan] de ] I know Lisi think Akiu SUO like DE

neige nuhaizi that girl

'I know the girl who Lisi thinks that Akiu likes.'

Interestingly, suo mus t remain the clause containing the gap and cannot occur in any of the higher clauses:

86 BONNIE CHIU

(38) *wo renshi [Lisi s u o yiwei [Akiu xihuan] de ] I know Lisi SUO think Akiu like DE

neige nuhaizi that girl

(39) *wo s u o renshi [Lisi yiwei [Akiu xihuan] de ] I SUO know Lisi think Akiu like DE

neige nuhaizi that girl

In this respect, s u o differs from relative pronouns, which must move to the highest CP in English sentences parallel to (37). For example, in (37), the relative pronoun, w h o , in the gloss must move to the highest CP.

There is evidence that clauses in which s u o occurs involve syntactic wh-movement. The evidence comes from the observation that sentences containing s u o show Island effects. Huang (1990) discusses relative clauses and observes that Island Constraints can be violated when the island from which extraction takes place is in the subject position. 8 The Island Cosntraints considered by Huang are the Sentential Subject Constraint (SC (40)), the Complex NP Constraint (CNPC (41)), the Left Branch Constraint (LBC (42))andtheAdjunctConstraint (AC (43)).

(40)

(41)

(42)

(43)

[[Lisi kan __ ] zui heshi de ] shu (SC) Lisi read most appropriate DE book

'the book that it is most appropriate for L to read'

[[__i zu __j de ] ren i hen duo de ] rent DE people very many DE

neidong fangzij (CNPC) that house

'the house that the people who rented it are many'

[___ baba hen you qian de ] neige ren (LBC) father very have money DE that person

'the person whose father is very rich'

neige ren, [yinwei ni piping-le __ ] person because you criticize-LE

wo hen bugaoxing (AC) 9

I very unhappy

'That person, because you criticized (him), I am very unhappy.'

AN O B J E C T C L I T I C P R O J E C T I O N IN M A N D A R I N C H I N E S E 87

Interestingly, relative clauses with s u o behave differently, as the following sentences indicate.

(44) * [[Lisi s u o kan __] zui heshi de ] shu (SC) Lisi SUO read most appropriate DE book

'the book that it is most appropriate for L to read'

(45)* [[.--i s u o zu__j de] reni hen duo de] SUO rent DE people very may DE

neidong fangzij (CNPC) that house

'the house that the people who rented it are many'

When s u o occurs, SC and CNPC effects are observed, and these sentences become ungrammatical. This, together with the fact that s u o can appear only when the extraction site contains an empty category, presents strong evidence that relative clauses containing s u o are derived by syntactic w h -

movement. (For our purposes here, LBC is not relevant since the relevant cases involve object extraction. Relativization of an accusative NP out of an adjunct clause requires the presence of a resumptive pronoun, excluding SUO.)

The wh-Island Constraint (WHIC) is particularly interesting. Languages vary with respect to whether syntactic wh-movement can violate the w h -

island Constraint or not (cf. Italian vs. English). Thus, the question arises as to how Chinese acts in this respect. It is well known that in some cases, relativization can violate wh-islands in Chinese (Huang (1982)): 1~

(46) a. [Lisi xiangzhidao [Akiu weisheme mai-le __] de ] Lisi wonder Akiu why buy-LE DE

neiben shu that book

'the book that Lisi wonders why Akiu bought'

b. [Lisi xiangzhidao Lisi wonder

neige ren that person

'the person who Lisi wonder why _ bought the book.'

As (47)-(50) show, when s u o occurs, the WHIC is obeyed. H

[__ weisheme mai-le neiben shu] de] why buy-LE that book DE

88 BONNIE CHIU

(47) * [Lisi xiangzhidao [Akiu weisheme s u o mai-le __] de ] Lisi wonder Akiu why SUO buy-LE DE

neiben shu that book

'the book that Lisi wonders why Akiu bought'

(48) * [Lisi xiangzhidao [Akiu zenme s u o diu __] de ] Lisi wonder Akiu how SUO lose DE

neiben shu that book

'the book that Lisi wonders how Akiu lost'

(49) * [Lisi xiangzhidao [Akiu shenme-shihou s u o mai-le __] Lisi wonder Akiu what-time SUO buy-LE

de ] neiben shu DE that book

'the book that Lisi wonders when Akiu bought'

(50) * [Lisi xiangzhidao [Akiu zai nali s u o mai-le __] de ] Lisi wonder Akiu at where SUO buy-LE DE

neiben shu that book

'the book that Lisi wonders where Akiu bought'

In short, when s u o appears, Island effects are observed. This is strong evidence that s u o occurs only in relatives involving syntactic wh-movement. As far as I know, this type of relatives is the only clear case of a construction derived by syntactic wh-movement in Chinese. In other words, these rela- tives provide evidence that Chinese has syntactic wh-movement.

In this section, I have discussed the locality constraints which s u o must obey. First of all, the occurrence of s u o is clause-bound. That is, s u o must stay within the clause containing the gap of the relativization. Secondly, the s u o construction manifests Island effects.

3. ANALYSIS

We have established the following properties of s u o :

(51) a. s u o occurs in relative clauses derived by syntactic wh-movement. b. suo occurs in the clause containing the extraction site. c. Only relativization of an accusative NP licenses s u o .

AN OBJECT CLITIC PROJECTION IN MANDARIN CHINESE 89

d . s u o occupies the following position sentence-internally:

NP-subject S-level-Adv S U O Neg VP-level-Adv Verb

How should these properties be accounted for? Let us begin with the syn- tactic status of s u o . Is it an XP, like a relative pronoun, as some traditional

grammarians of Classical Chinese have assumed? The answer must be "no" because, in contrast to relative pronouns in English, which must move to the highest SpecCP, s u o must remain within the clause containing

the extraction site:

(52) a. the man who you know Mary saw

b.*the man you know who Mary saw

If s u o is not an XP, is it a head? The answer appears to be "yes". s u o

shares many properties with Romance clitics, such as those pointed out in Kayne (1975) for French clitics. Like clitics, s u o is not a complex expres-

sion, and it cannot be modified or stressed, even in contrastive situations. These facts suggest that s u o , like clitics, is an X ~ element.

Suppose that s u o occurs as the head of some projection, which I will call SuoP. Where is SuoP located? If s u o is realized as the head of SuoP, the position of s u o then reflects the position of SuoP. Recall from the discussion in section 2.3 that s u o occurs between S-level adverbs and negation, suggesting that SuoP must be higher than negation. In fact, evidence from the interaction between Tense-lowering and s u o shows that SuoP is higher than TP in the structure assumed for Chinese in (25).

In Chiu (1993), it is proposed that Tense is lowered to Asp, which occupies a position lower than negation in the clausal structure (cf. (25)). If SuoP is lower than TP and s u o resides in its head position, Tense lowering should be blocked by the presence of s u o : 12

(53) TP

l e SuoP

Suo'

s u o NegP

Neg AspP

Asp . . .

* t

90 BONNIE CHIU

However, s u o does not block Tense lowering, as the example below shows:

(54) [Lisi s u o song-le Akiu] de neiben shu Lisi SUO send-LE Akiu DE that book

'the book that Lisi sent Akiu'

This suggests that SuoP occupies a position higher than TP. We thus arrive at the following partial structure: 13

(55) SuoP

TP

AspP

VP

The structure in (55) has implications for the position of the subject. Because s u o must follow the subject, whether it is an NP ((56a)), or a pronoun ((56b)), the subject NP can no longer be regarded as being in SpecTP. The subject NP must therefore be structurally higher than SuoP.

(56) a. [Akiu s u o du-guo__] de neixie shu Akiu SUO read-GUO DE those book

'the books that Akiu read.'

b. [ta s u o du-guo__] de neixie shu he SUO read-GUO DE those book

Further evidence that the subject is higher than this projection comes from the interaction between s u o and d o u . d o u 'all' is often called a scope marker by Chinese linguists. It takes scope over an element, generally an NP, occur- ring to its left. L4

(57) zhexie xuesheng d o u du-guo neiben shu these students all read-GUO that book

'All of these students read that book.'

When d o u and s u o co-occur, s u o must precede dou: 15

(58) a. [tamen s u o d o u - d u - g u o __ ] de neiben shu tamen SUO DOU-read-GUO DE that book

'the book that they all read'

AN OBJECT CLITIC PROJECTION IN MANDARIN CHINESE 91

b.*[tamen dou suo d u - g u o ] de neiben shu tamen DOU SUO read-GUO DE that book

In (58), dou is related to the subject, tamen 'they', to mean 'all of them'. It must follow suo. According to the analysis proposed in Chiu (1990, 1993), dou must be incorporated into some verbal or inflectional projection. If dou could be incorporated into suo, which is in the head position of the projection, left-adjunction of dou to suo would yield the sequence dou

suo, which is impossible, as shown in (58b). It thus appears that the highest inflectional head that dou can be incorporated into is T, which is structurally lower than SuoP. Therefore, suo must precede dou.

Since overt subjects precede both suo and dou, subjects must be higher than TP and SuoP. Given these facts, I conclude that there is a projection higher than SuoP and that overt subjects occur in the Spec of this projection. What is this projection? I propose that this is the nominative Case projection, which I will call NomP, following Sportiche (1992). j6 Nominative Case is assigned by the head of NomP, Nom, to its Spec position via Spec-Head agreement. This amounts to saying that the subject, assuming that it is base-generated VP-internally, undergoes NP-movement to SpecNomP at S-structure, yielding the following structure:

(59) NomP

NP~ Nom'

Nom SuoP

TP

T NegP

Neg ~

Asp VP

t i V'

Now, what is SuoP? How is the appearance of suo in the head of SuoP triggered? Let us suppose that SuoP is the accusative Case projection; that is, the head of SuoP assigns accusative Case to its Spec position. Further- more, suppose that syntactic movement must move through SpecSuoP. suo

92 B O N N I E CHIU

is triggered in the accusative Case head (head of SuoP) when some element moves through SpecSuoP. Because the head of SuoP assigns accusative Case to its Spec, it forces SpecSuoP to be accessible only to elements receiving accusative Case, thus explaining the fact that suo 's occurrence is restricted to relativization of an accusative-Case-marked element. ~7

The fact that suo occurs in relative clauses and that a Subjacency effect is observed in these relatives indicates that syntactic wh-movement has occurred. Since no overt wh-element appears, and suo cannot be a wh-

element (i.e., a relative pronoun), I assume that a null operator, base- generated in the complement of the verb, undergoes syntactic wh-movement to SpecCP. The movement proceeds through SpecSuoP and triggers the appearance of suo on the head of SuoP.

We now have the structure in (59), in which nominative Case and accusative Case are assigned to positions higher than TP, namely SpecNomP and SpecSuoP, respectively. However, a problem arises with objects. At S-structure, the object NP must follow the verb. It is impossible for the object NP to move to SpecSuoP, giving rise to SOV order:

(60) *Lisi neixie shu du-le 1 8 Lisi those book read-LE

'Lisi read those books.'

Since verbs do not raise in Chinese (Cheng (1989), Chiu (1993)), it must be that the object NP remains in its theta position. If this is correct, how does the object NP receive accusative Case assigned by Suo to SpecSuoP? I will assume that there is an expletive pro (e-pro) in SpecSuoP which is assigned accusative Case by Suo.

(61) NomP

Nom SuoP

e-proi Suo'

Suo . . . VP

V'

V N P i

Expletive pro forms a Case chain with the postverbal object NP, giving rise to a structure analogous to there-exp le t ive constructions in English: ~9

(62) There are three men in the room.

AN OBJECT CLITIC P R O J E C T I O N IN M A N D A R I N CH IN E SE 93

Like t h e r e - e x p l e t i v e constructions, the object NP raises to SpecSuoP at LF, due to the principle of Full Interpretation (Chomsky (1989)). If this is correct, an obvious question raised by this proposal is why object NP cannot move to its Case position, SpecSuoP, at S-structure, as subject NP does. One possible answer, suggested to me by Hilda Koopman, is that there is a (possibly universal) requirement that object NP be within the governing domain of the verb at S-structure. Since the verb in Chinese does not raise, and the object NP must remain within its governing domain, the object NP cannot move to SpecSuoP at S-structure. We thus expect s u o not to occur in simple sentences since s u o is triggered at S-structure by movement of some element through SpecSuoP. This is indeed the case:

(63) * Lisi s u o du-le neixie shu Lisi SUO read-LE those book

'Lisi read those books.'

The same account also explains the fact that s u o cannot occur when a resumptive pronoun is present. Because the resumptive pronoun must remain in the post-verbal position at S-structure, nothing is moved through SpecSuoP, and s u o therefore is not triggered. We thus derive the comple- mentary distribution between s u o and an overt resumptive pronoun.

But, why can s u o not be triggered by the expletive pro in SpecSuoP at S-structure in simple sentences or relatives with an object resumptive pronoun? Suppose that we interpret the Spec-Head relation as a sharing of some lexical index between the head and the element in the Spec position (cf. Chomsky (1986) on 'feature sharing' between head and Spec) and that s u o is realized on the head of SuoP when it shares the lexical index of the element in SpecSuoP. Expletives, by definition, lack lexical content and thus do not have a lexical index to be shared by the Spec and the head of SuoP. Consequently, s u o cannot appear in simple sentences because lexical object NPs and overt pronouns cannot reach SpecSuoP at S-struc- ture.

The situation is different when syntactic movement has occurred, as in relative clauses. Instead of an expletive pro, a wh-trace with a lexical index is in SpecSuoP. The head of SuoP shares the lexical index of the wh-trace, and s u o ' s occurrence is thus triggered. I will return to the analysis of relative clauses which do not contain s u o below.

Before proceeding to relative clauses, a few words are in order with respect to how NomP and SuoP are to be interpreted in a structure where Case is assigned by an Agreement projection (AgrS-P and AgrO-P), as proposed in Chomsky (1989) and Sportiche (1992). Nom and suo display 'split' properties: on the one hand, they resemble Agreement in that they

94 BONNIE CHIU

are Case assigners, while on the other hand, they occur in positions within a clausal structure where clitics in Romance languages may appear. Sportiche (1992) proposes a structure containing clitic projections, which he calls Accusative voice (for direct object clitic), Nominative voice (for subject clitic), etc. These positions are higher than all other inflectional projections:

(64) [clitic projections . . . [ggrS [inflectional p ro jec t ions . . .

[Ag~O [VP . . . ] ] ] ] ]

Under Sportiche's theory, clitics must move to the Spec position of their respective projections at LF to be licensed. These positions are associated with the interpretation of 'specificity' and thus derive the 'specific' nature of clitics.

s u o shares many properties with clitics, as mentioned earlier, and the position of SuoP is equivalent to that of the Accusative voice projection in Sportiche's structure. In addition, the interpretation of relative clauses with s u o also tends to be 'specific'. 2~ It thus appears that SuoP is the Accusative voice projection. By analogy, NomP, which occurs in a position higher than SuoP, is the Nominative voice projection. One piece of evidence supporting this analysis is the restriction that the subject must be specific in Chinese. According to Sportiche, 'specificity' is associated with the voice projections. Because the subject in Chinese occupies the Spec position of the Nominative voice projection at S-structure, only elements with a specific interpretation can occur as the subject, deriving the restriction on subjects being specific in Chinese.

I will thus continue to assume the structure proposed in (59), repeated here as ( 6 5 ) . 25

(65) NomP

Nom SuoP

Suo TP

T NegP

Neg ~ .

Asp VP . . .

According to the present proposal for Chinese clausal structure, the subject and object NP in a simple sentence, such as (66), occupy posi- tions in (67) at S-structure:

AN OBJECT CLITIC PROJECTION IN MANDARIN CHINESE 95

(66) Akiu du-le neixie shu Akiu read-LE those book

'Akiu read those books.'

(67) NomP

NP Nom'

Akiui Nom SuoP

e-proj Suo'

Suo TP

T NegP

Neg / ~

Asp VP

t i V'

V VP

V'

V NP I

neixie shuj

The subject NP, Akiu, having moved from its Theta position, is now in SpecNomP at S-structure. The object NP, which remains in its Theta position, is co-indexed with an expletive pro in SpecSuoP (e-pro) to which accusative Case is assigned. SpecNomP and SpecSuoP thus differ in terms of the elements that they can host at S-structure. This is illustrated in the chart in (68).

(68) SpecNomP SpecSuoP

lexical NP yes * expletive pro * yes overt pronoun yes *

At S-structure, SpecNomP is accessible to lexical NPs and overt pronouns but excludes expletive pro. In contrast, SpecSuoP is restricted to exple- tive pro and cannot host lexical NPs, overt pronouns.

96 BONNIE CHIU

To summarize, I have proposed that s u o is an X ~ element, realized in the head of the projection SuoP whose head assigns accusative Case to its Spec position. SuoP is located higher in the clause than TP. The occur- rence of s u o is triggered by movement through SpecSuoP at S-structure. Furthermore, a nominative projection (NomP) is postulated whose head assigns nominative Case to its Spec.

3.1 . R e l a t i v e C l a u s e s

According to the analysis proposed so far, the appearance of s u o in relative clauses indicates that syntactic wh-movement has occurred. Does the absence of s u o imply the lack of syntactic wh-movement? How should relative clauses be analyzed, either with or without s u o ? I turn to these questions next.

Recall that relativization of an object NP exhibits three options: with a gap and s u o (69), with a gap but without s u o (70), and finally with a (resumptive) pronoun but without s u o (71):

(69) [Akiu s u o renshi __ de ] neige ren Akiu SUO know DE that person

'the person that Akiu knows'

(70) [Akiu r e n s h i de ] neige ren Akiu know DE that person

'the person that Akiu knows'

(71) [Akiu renshi ta de ] neige ren Akiu know him DE that person

'the person that Akiu knows'

I will begin with the relatives containing s u o and a gap. We have seen earlier that relatives containing s u o obey Subjacency and that movement is there- fore involved in these cases. Since no over wh-element appears, I assume that the derivation involves movement of a null operation. The null operator, base-generated in the complement of the verb, undergoes syntactic w h -

movement to SpecCP. The null operator moves through SpecSuoP in the process, triggering the occurrence of s u o . The sentence in (72), thus, has the structure in (73).

(72) [Akiu s u o du-guo de ] shu Akiu SUO read-GUO DE book

'The books that Akiu has read'

AN OBJECT CLITIC PROJECTION IN MANDARIN CHINESE 97

(73) NP

CP NP

OPj C' shuj

NomP C

Akiu i Nom' de

Nom SuoP

tj Suo'

Suo TP [

S b /O j �9 ~ ~

VP

W l

V VP

V'

v tj 1

It should be noted that the movement of the null operator in the direct object position to SpecCP must proceed through SpecSuoP to avoid improper movement. SpecSuoP is the Case position of the direct object. The movement of the direct object to SpecSuoP is from a Caseless to a Case position, an instance of NP-movement. The subsequent movement, from SpecSuoP to SpecCP, is a case of wh-movement (from a Case position to a Caseless position), giving rise to a well-formed structure. However, if the movement moves the null operator directly to SpecCP, skipping SpecSuoP, a chain of two members in which neither one has Case would be formed (the direct object is Caseless, and so is SpecCP), thus resulting in illicit structure (the Caseless object trace is not A-bound, violating Principle A of the Binding Theory). Consequently, the movement of the null operator in (73) is forced to proceed through SpecSuoP.

Let us next turn to those relatives that do not contain suo. Recall that these relatives do not obey Subjacency, indicating that movement does

98 B O N N I E CHIU

not take place. Relatives without suo, thus, do not involve movement. There are two types of relative clauses to consider: one containing a gap and the other containing a (resumptive) pronoun. The sentence in (74) is an example of the former.

(74) [Akiu du-guo __ de] shu Akiu read-GUO DE book

'the books that Akiu has read'

Under our analysis, (74) cannot be derived by syntactic movement of the null operator, due to the lack of Subjacency effects. Instead, I propose that in (74), the null operator is base-generated in SpecCP position. Furthermore, there is a pro base-generated in the object position which is co-indexed with the null operator, yielding a left-dislocated construction:

(75) NP

CP NP

OPj C' shuj

NomP C

Akiu i Nom' de

Nom SuoP

e-proj Suo'

Suo TP

T NegP

Neg . ~

Asp VP

t i V '

J V VP

V'

V proj

AN OBJECT CLITIC P R O J E C T I O N IN M A N D A R I N CH IN E SE 99

At LF, the object pro is adjoined to the expletive pro in SuoP due to the principle of Full Interpretation (Chomsky (1988)), giving rise to the fol- lowing indexes among the relevant elements in (76). 22

(76) OPj [Akiui [proj-e-proj ti tj] (LF) I I

If this analysis is correct, syntactic movement is not involved in the for- mation of relatives in this case. It thus follows that there is no element passing through SpecSuoP in the syntax. Consequently, s u o cannot appear. In short, relatives containing a gap without s u o have a left-dislocation structure with a base-generated pro in the object position.

An overt pronoun can occur instead of the small pro in (75), yielding the so-called resumptive pronoun strategy:

(77) [Akiu renshi ta de ] neige ren Akiu know him DE that person

'the person that Akiu knows him'

Relatives with resumptive pronouns do not obey Subjacency, nor can s u o

occur in these relatives. The analysis for (75) applies to these relatives, except that an overt pronoun, instead of a small pro, is base-generated in the object position.

Notice that not both strategies, base-generation of a pro and an overt pronoun, are available in all relatives. It appears that both strategies are available only in relativization of a direct object. In section 2.1, we saw that the resumptive pronoun strategy cannot be used when a subject or an adjunct is relativized. With relativization of an indirect object, the NP in ba-phrase, the object of a PP, the NP following b e i , or the genitive NP, the resump- tive pronoun strategy is obligatory. I assume that this distribution follows different licensing conditions on pros and overt pronouns.

To recapitulate, there are two ways of forming relative clauses, one via syntactic movement of an null operator and the other by base-generation of a null operator and a small pro or an overt pronoun. A base-generated null operator in SpecCP does not co-occur with s u o because the object pro (or overt pronoun) remains too low in the structure. These relatives have the left-dislocated structure. Relatives containing s u o involve syntactic movement of a null operator in the complement of SuoP to SpecCP. This movement must proceed through SpecAgrO-P and thus triggers the occurrence of s u o .

100 BONNIE CHIU

3.2. Topicalization

Let us next turn to the implications of the fact that suo is excluded from topicalization of accusative NPs:

(78) neiben shu, Akiu (*suo) mai-le_ that book A SUO buy-LE

'that book, Akiu bought.'

If topicalization in Chinese involves syntactic wh-movement, then the movement of the accusative NP, neiben shu 'that book', to the topic position should be able to trigger the occurrence of suo, as is the case of syntactic wh-movement of a null operator from the direct object position in rela- tives. It is well known, however, that there is evidence that topicalization in Chinese does not involve wh-movement. For example, topicalization out of an island in subject position does not show Subjacency effects, just like relatives clauses which do not contain suo:

(79) zheben shu, [Lisi kan __] zui heshi this book Lisi read most appropriate

'This book, it is most appropriate for Lisi to read.'

(80) neidong fangzi, [zu-guo __] de ren that house rent-GUO DE person

(sc)

hen duo (CNPC) very many

'That house, people wo rented it were many.'

(81) neige ren, [__ baba ] hen you qian (LBC) that person father very have money

'that person, whose father is very rich.'

(82) neige ren, [yinwei ni piping-le _] that person because you criticize-LE

wo hen bugaoxing (AC) I very unhappy

'That person, because you criticized (him), I am very unhappy.'

I propose that topicalization should be analyzed in the same way as non- movement relatives, namely as a left-dislocated construction: there is a base-generated XP in the topic position which is co-indexed with a pro within the clause at the level of LE 231 assume that topic occupies a position

AN OBJECT CLITIC PROJECTION IN MANDARIN CHINESE 101

higher than the NomE Example (83), an object topicalization construc- tion, thus has the partial structure in (84) at S-structure: 24

(83) neiben shu, Akiu m a i - l e that book Akiu buy-LE

'That book, Akiu bought.'

(84) CP

neiben shu j CP . . ,

NomP

Akiu i Nom'

Nom SuoP

e-proj Suo'

Suo TP

T NegP

Neg ~

Asp VP

t i V' J

V VP

V'

V proj

The structure in (84) is analogous to (75), a structure assigned to relatives which contains neither suo nor resumptive pronouns: there is a base- generated object pro co-indexed with an expletive pro in SpecSuoP. At LE the object pro is raised to SpecSuoP and adjoined to the expletive pro, yielding (85):

(85) [neiben shu]j, [Akiui [proj-e-proj] t~ tj] (LF)

I J

102 BONNIE CHIU

Under this analysis, suo cannot occur in object topicalization because the object pro is too low in the structure to trigger suo ' s appearance. Just as in simple sentences, suo never occurs when the object is pro:

(86) * Lisi suo du-le Lisi SUO read-LE

'Lisi read (it).'

In brief, topicalization in Chinese is proposed to be a left-dislocated structure, thus reducing topicalization to the distribution of pro at S-struc- ture, as in Cheng (1989). Because object pro at S-structure occupies a position lower than SuoP, suo cannot occur in topicalization structure.

3.3. S u m m a r y

I have proposed an analysis which treats suo as an X ~ head, realized as the head of SuoP when syntactic wh-movement has proceeded through SpecSuoP. suo occurs only in relative clauses because they are the only construction with wh-movement in syntax. Only relativization of an accusative NP triggers s u o ' s occurrence because SpecSuoP is assigned accusative Case. Since SuoP occupies a position higher than TP, elements which do not reach SpecSuoP by S-structure cannot trigger s u o ' s occur- rence. They include small pro, overt pronouns, and lexical NPs. At S-structure, object pro, overt pronouns, and lexical NPs remain in their base- generated positions (complement of the verb). They occupy a position lower than SuoP and are thus incapable of triggering suo ' s appearance, suo there- fore does not occur in simple sentences, relative clauses not involving movement, nor topicalization structures.

3.4. B e i - c o n s t r u c t i o n s

In this section, I will outline briefly the proposal on be i - cons t ruc t ions in Chiu (1993). This seemingly problematic construction for the current proposal on suo can in fact be accommodated quite straightforwardly.

In Chiu (1993), bei is proposed to be an X ~ category heading its own projection, BeiP, and taking a NomP as its complement, as in (87). 25

(87) BeiP

Bei'

Bei NomP

AN OBJECT CLITIC PROJECTION IN MANDARIN CHINESE 103

The reason behind such a proposal is that bei has the interpretation of 'being affected', often in an adversary manner (thus 'bearing some kind of hardship'). This somewhat 'verb-like' interpretation of bei ('bearing hardship') is a remnant from classical Chinese in which bei is a verbal category with the meaning 'to bear/receive' (as in 'to bear/receive hardship'). Though bei no longer exhibits verbal properties, the obliga- tory presence of bei in bei-sentences suggest that it is an X ~ category. In view of its historical development, this is not an unlikely scenario. It is therefore proposed that bei is an X ~ category and projects a phrasal category, BeiP.

Furthermore, the surface subject is suggested to be base-generated in SpecBeiP, capturing the interpretation that the subject NP is affected (bei) by the following proposition (NomP). Moreover, the verb in a bei-sentence retains it Case-assigning ability. Consequently, the thematic subject and object of the verb are base-generated in the same positions as their non-bei-sentence counterpart, namely SpecVP and the complement of V, respectively. Like its non-bei-sentence counterpart, the thematic subject undergoes NP-movement to SpecNomP to receive nominative case.

(88) BeiP

Bei'

Bei NomP

thematic subject~ Nom'

Nom

VP

t~ V'

V thematic-object

Given that bei takes as NomP as its complement, we expect NomP to have a full-blown structure, including SuoP. Thus, (89) under the present proposal would have the structure in (90).

(89) Akiu bei Lisi da-le Akiu BEI Lisi hit-LE

'Akiu was hit by Lisi.'

104 BONNIE CHIU

(90) BeiP

Akiu Bei'

Bei NomP

Lisi ~ Nom'

Nom SuoP

Suo p

Suo TP

Asp VP

h V'

V NP

I I da ?

The next question is what occupies the complement position of da 'hit ' in (90)? From the proposal outlined, it is an argument position to which accusative Case is assigned by the verb. It cannot be an NP-trace because it is a Case position, nor can it be PRO because it is a governed position. There are two possibilities left: a variable or pro. Chiu (1993) argues that either option is possible in bei-sentences. In bei-sentences with suo it is a variable, while in bei-sentences without suo it is a pro. In particular, the same strategies used for relative clauses are also at work in forming bei-

sentences. Adopting Feng's (1991) idea that there is a null operator in bei-constructions, Chiu proposes that there are two ways in which a bei-

sentence can be formed, analogous to the formation of relative clauses: the first is by movement, and the second is by base-generation of the null operator. The null operator in both scenarios, being empty, is in turn co- indexed with the surface subject to provide it with a range of interpretation. In the first case (movement scenario), the null operator is base-generated as the internal argument of the verb. It undergoes syntactic movement to NomP-adjoined position as shown in (91).

(91) [Akiui . . . bei [NomP OPi [NomP. �9 �9 [V ti]]] ] i I

AN O B J E C T C L I T I C P R O J E C T I O N IN M A N D A R I N C H I N E S E 105

This syntactic movement must proceed through SpecSuoP in the comple- ment of bei in (90), triggering the occurrence of suo and giving rise to a bei-sentence containing suo, analogous to the situation in relative clauses containing suo (see section 3.1). Under this analysis, the occurrence of suo in bei-construct ions is triggered by the null-operator movement from the internal argument position of the verb to the NomP-adjoined position. Under this analysis, the null-operator movement from the internal argument position of the verb to NomP-adjoined position should be able to trigger suo 's occurrence in bei-sentences . In other words, suo is expected to occur in bei-construct ions without relativization. This prediction is borne Out: 26

(92) zhongguo bei riben suo qinlue China BEI Japan SUO invade

'China was invaded by Japan.

(93) zhexie shiqing bu neng bei Akiu suo liaojie these thing not can BEI Akiu SUO understand

'These things cannot be understood by Akiu.'

In the second case (the base-generation scenario), a null operator is base- generated in NomP-adjoined position, and a small pro is generated in the object position of da 'hit' (35), yielding a left dislocation structure:

( 9 4 ) [Ak iu i . . . bei [NomP O P i [NomP �9 �9 �9 [SuoP e -p ro i . . . [V p r o i ] ] ] ] ]

The base-generated small pro is co-indexed with an expletive pro in SpecSuoP for Case reasons under the current analysis. The small pro, in turn, is co-indexed with the null operator. Since there is no syntactic wh-

movement through SpecSuoP, suo is not triggered, deriving a bei -sentence

without suo, such as (23). Recall the difficulties posed by bei -sentences with respect to suo, that

is, the fact that relativization of a surface subject triggers suo ' s occur- rence and that the relative position of suo with respect to negation differs from the one observed in relatives (suo precedes negation in relatives but follows negation in bei-construct ions) . If the analysis on bei-construcf ions

in Chiu (1993) is correct, these problems can be explained quite straight- forwardly. It is not the relativization of the surface subject but rather the movement of a null operator which triggers suo ' s appearance, suo must follow negation in bei -sentences because SuoP is lower than BeiP, which is lower than N e g P . 27

To summarize, adopting the analysis on bei-construct ions in Chiu (1993), I have shown that the difficulties posed by this construction on the current analysis of suo receive a natural explanation. According to Chiu (1993),

106 BONNIE CHIU

the same strategies used in relative formation are also at work in the formation of the bei-construction. A null operator is base-generated either in the internal argument position of the verb or in the NomP-adjoined position. In the former case, the null operator undergoes syntactic movement, triggering the occurrence of suo, whereas in the latter, the null operator binds a small pro in the internal argument position of the verb.

4. SOME IMPLICATIONS

The syntactic behavior of suo provides us with a tool for investigating various clausal structures in Mandarin Chinese. In particular, the position of suo suggests that certain constructions, such as subject control struc- tures and raising structures involve restructuring. This in turn lends support to the existence of control and raising structures in Mandarin Chinese. In section 4.1, I discuss the different syntactic behaviors of suo in subject and object control structures. The controversy over whether Chinese has raising structures or not will be discussed in section 4.2.

4.1. Control vs. Restructur ing

Suo must remain in the tensed clause which contains the gap of the relative (cf. (37)-(39)). Let me now address the question of where suo occurs in infinitival complements. An issue raised by this question is whether Chinese has finite or non-finite clauses, and, if there are non-finite complements, what their distribution is. Due to the impoverished verbal morphology, the formal difference between these two types of clauses in Chinese is minimal. Despite the lack of formal differences, various linguists have shown that in Chinese clauses there exist distinctions parallel to those observed between finite and non-finite clauses cross-linguistically and have concluded that Chinese has both finite and non-finite clauses (Huang (1982), Li (1985), 1990), Tang (1990)). 28 For our purposes here, I will assume that Chinese has finite, as well as non-finite, clauses. As is the case for many languages, non-finite clauses are found in complements of control predicates such as quan 'persuade' and daying 'promise'. Let us then turn to suo's behavior in these control complements.

suo resembles a clitic (see discussion above). We know that in many languages, clitics climb in infinitival clauses which trigger restructuring. Given the parallel between suo and clitics, we expect suo to exhibit similar behavior in non-finite clauses. This expectation is borne out. suo occupies different positions in object- and subject-control structures. In object-control

AN OBJECT CLITIC PROJECTION IN MANDARIN CHINESE 107

structures, s u o must remain in the complement clause when the object of the complement is relativized (94): 29

(94) a. [wo quan Akiui [PROi s u o mai__] de ] I persuade Akiu SUO buy DE

neiben shu that book

'the book that I persuaded A to buy'

b.*[wo s u o quan Akiui [PROi mai__] de ] I SUO persuade Akiu buy DE

neiben shu that book

In contrast, s u o in subject-control structures must precede the matrix verb when relativization occurs:

(95) a. [Lisi i s u o daying [PRO i jiao __] de ] neimen ke Lisi SUO promise teach DE that class

'the class that Lisi promised to teach'

b.*[Lisii daying [PROi s u o j iao__] de ] neimen ke Lisi promise SUO teach DE that class

(96) a. [Lisi~ s u o dasuan [PROi mai __] de ] neiben shu Lisi SUO plan buy DE that book

'the book that Lisi plans to buy'

b.*[Lisi~ dasuan [PROi s u o mai __] de ] neiben shu Lisi plan SUO buy DE that book

That s u o must precede matrix verb in subject-control structures indicates that restructuring has occurred. The behavior of s u o thus provides evidence that subject-control structures in Chinese are restructuring configuration, but object-control structures are not.

4.2. R a i s i n g vs . S - l e v e l A d v e r b s

The syntactic behavior of s u o provides us with additional evidence that raising predicates exist in Chinese. The existence of raising verbs has long been a controversial issue. Predicates that could be considered as raising adjectives/predicates, such as k e n e n g 'likely', are treated as such by various

108 BONNIE CHIU

linguists (Huang (1982), Li (1985, 1990), Lin and Tang (1991), among others). Others regard them as S-level adverbs (Lu (1981), Zhu (1982) among others).

(97) Lisi keneng du-le neiben shu Lisi likely read-LE that book

'Lisi is likely to have read the book.'

(98) Lisi dagai du-le neiben shu Lisi probably read-LE that book

'Lisi has probably read the book.'

(Raising)

(S-level adverb)

The evidence for treating elements like keneng as S-level adverbs comes from their non-verb-like properties and their distribution. First, unlike most of the verbs, they never take verbal aspectual markers. Since adverbs do not take aspectual markings either, raising predicates and S-level adverbs are in this respect formally indistinguishable:

(99) *Lisi keneng-le du neiben shu (Raising) Lisi likely-LE read that book

(100) *Lisi dagai-le du neiben shu (S-level adverb) Lisi probably-LE read that book

Secondly, potential raising predicates seem to occupy the same positions as S-level adverbs in a sentence. Both raising predicates and S-level adverbs can occur in a pre-subject or a post-subject (but pre-verbal) position:

(101) keneng Lisi du-le neiben shu (Raising) likely Lisi read-LE that book

'Lisi is likely to have read the book.'

(102) dagai Lisi du-le neiben shu (S-level adverb) probably Lisi read-LE that book

'Lisi has probably read the book.'

Because of these shared properties between raising verbs and S-level adverbs, it is sometimes assumed that postulating the existence of raising verbs is unnecessary. The so-called raising verbs could simply be treated as S-level adverbs. Those linguists who recognize the existence of raising predicates consider the raising process to be optional in Chinese given that the subject can either precede or follow the raising predicate, as shown in (97) and (101). The subject of a raising predicate in Chinese can remain

AN OBJECT CLITIC PROJECTION IN MANDARIN CHINESE 109

in its base position and need not be raised to be matrix subject position. By contrast, raising in English is obligatory. See Li (1990) and Lin and Tang (1991) for further discussion.

Elements like keneng 'likely', however, also exhibit verbal properties, suggesting that they are not S-level adverbs. When the subject is raised to the matrix clause, thus preceding the raising predicate, the raising pred- icate can undergo A-not-A question formation, be negated, and serve as a short answer to a question. Those are properties associated with verbs,

Lisi keneng-bu-keneng du-guo neiben shu (A-not-A) Lisi likely-not-likely read-GUO that book

'Lisi is likely or not likely to have read the book?'

b. Lisi bu keneng du-guo neiben shu (Negation) Lisi not likely read-GUO that book

'Lisi is not likely to have read the book.'

c. Lisi keneng du-guo neiben shu ma? Lisi likely read-GUO that book Q

not S-level adverbs:

(103)a.

keneng (as short answer) likely

'Is Lisi likely to have read that book?' 'Yes.'

Interestingly, when the subject follows the raising predicate, it cannot undergo A-not-A question formation, but it can still be negated and serve as a short answer to a question:

(104)a.*Keneng-bu-keneng Lisi du-guo neiben shu (A-not-A) likely-not-likely Lisi read-GUO that book

'Is it likely or is it not likely that Lisi has read the book?'

b. bu keneng Lisi du-guo neiben shu (Negation) not likely Lisi read-GUO that book

'It is not likely that Lisi has read the book.'

c. keneng Lisi du-guo neiben shu ma? likely Lisi read-GUO that book Q

keneng (as short answer) likely

'Is Lisi likely to have read that book?' 'Yes.'

110 BONNIE CHIU

S-level adverbs remain excluded from these possibilities regardless of whether the subject precedes or follow them:

(105)a.*Lisi dagai-bu-dagai du-guo neiben Lisi probably-not-probably read-GUO that

shu (A-not-A) book

b.*Lisi bu dagai du-guo neiben shu (Negation) Lisi not probably read-GUO that book

c. Lisi dagai du-guo neiben shu ma? Lisi probably read-GUO that book Q

*dagai (as short answer) probably

'Has Lisi probably read that book?' 'Yes.'

(106)a.*dagai-bu-dagai Lisi du-guo neiben probably-not-probably Lisi read-GUO that

shu (A-not-A) book

b.*bu dagai Lisi du-guo neiben shu (Negation) not probably Lisi read-GUO that book

c. dagai Lisi du-guo neiben shu ma? probably Lisi read-GUO that book Q

*dagai (as short answer) probably

'Has Lisi probably read that book?' 'Yes.'

The distribution of suo provides further confirmation of the difference between raising predicates and S-level adverbs, suo must precede the raising predicate when the subject has raised to the matrix clause:

(107)a. [Lisi~ suo keneng [ti du-guo __ ] de ] neiben shu Lisi SUO likely read-GUO DE that book

'the book that Lisi is likely to have read'

b.* [Lisii keneng [t~ suo du-guo_] de ] neiben shu Lisi likely SUO read-GUO DE that book

AN OBJECT CLITIC PROJECTION IN MANDARIN CHINESE 111

But s u o cannot precede the raising predicate when the subject is not raised to the matrix clause. It must remain within the raising complement:

( 1 0 8 ) a . * [ s u o keneng [Lisi du-guo __] de ] neiben shu SUO likely Lisi read-GUO DE that book

'the book that Lisi is likely to have read'

b. [keneng [Lisi s u o du-guo __] de ] neiben shu likely Lisi SUO read-GUO DE that book

As for S-level adverbs, s u o must follow the S-level adverb regardless of whether the subject precedes or follows the S-level adverb:

(109)a.*[Lisi s u o dagai d u - g u o de ] neiben shu Lisi SUO probably read-GUO DE that book

'the book that Lisi has probably read'

b. [Lisi dagai s u o d u - g u o de ] neiben shu Lisi probably SUO read-GUO DE that book

( l l O ) a . * [ s u o dagai Lisi d u - g u o de ] neiben shu SUO probably Lisi read-GUO DE that book

b. [dagai Lisi s u o d u - g u o de ] neiben shu probably Lisi SUO read-GUO DE that book

As these examples show, s u o exhibits a different distribution in raising pred- icates and S-level adverbs. When the subject precedes the raising predicate or S-level adverb, s u o must occur to the left of the raising predicate but to the right of the S-level adverb, as shown in the contrast between (107) and (109). In this way, the distribution of s u o provides additional evidence for the existence of raising predicates in Chinese.

When the subject follows the raising predicate, as in (108), s u o appears in the raising complement and cannot occur in the matrix clause. In contrast, s u o must occur in the matrix clause when the subject is raised to the matrix, as in (107). This suggests that raising predicates, like subject-control predicates, trigger obligatory restructuring when the subject has under- gone raising, s u o therefore must occur in the matrix clause in (107). If the subject remains in its base-generated position, as is the case in (108), restructuring is not triggered, and s u o therefore occurs in the embedded complement.

s u o thus provides an additional piece of evidence that Chinese has raising predicates. The position s u o occupies differs depending on whether raising

112 B O N N I E CHIU

of the subject NP has occurred or not. This suggests that raising predi- cates, like subject-control predicates, trigger restructuring.

5. CONCLUSION

In this paper, it is shown that Mandarin Chinese contains a nominative Case projection (NomP) and an accusative Case projection (SuoP). Evidence for their presence comes from sentences containing s u o . s u o is proposed to be an X ~ element realized as the head of SuoP when an accusative NP has undergone syntactic wh-movement through SpecSuoP. Small pro, overt pronouns, and lexical NPs cannot trigger s u o ' s occurrence, due to the fact that their S-structure positions are lower than SpecSuoP. Relative clauses are argued to be formed either by movement or by base-generation of a null operator. If the movement option is chosen, then s u o is triggered since there will be movement through SpecSuoP. On the other hand, if the option of base-generating a null operator is chosen, thus yielding a left-dislo- cated structure, then s u o does not occur because movement does take place. Furthermore, the same mechanism operating in relative clause formation is shown to be at work in the b e i - c o n s t r u c t i o n as well. The seemingly problematic data posed by this construction, in fact, follow straightforwardly from the analysis on suo . Topicalization, which does not permit s u o , is construed as a left-dislocated structure, parallel to the structure given for relatives not containing s u o . Furthermore, I have shown that the distribution of s u o can shed light on other aspects of Chinese syntax, providing evidence that raising predicates exists in Chinese and that subject control, as well as raising, triggers restructuring.

I have argued that Chinese clausal structure contains NomP and SuoP. If this proposal is correct, it provides evidence for clitic projections, as proposed by Sportiche (1992), in the clausal structure of Chinese, in spite of its impoverished morphological marking. My proposal therefore lends strong support to the universal nature of clitic projections.

NOTES

* I would like to thank James Huang, Hilda Koopman, Audrey Li, Dominique Sportiche, Tim Stowell, and two anonymous reviewers for their comments and suggestions on the analysis. I am also grateful to the audience at the UCLA Syntax and Semantics seminar, the Chinese Linguistic Workshop at Linguistic Institute, UC Santa Cruz, the East Asian Linguistics Workshop at UC Irvine, and the Third Leiden Conference for Junior Linguists at the University of Leiden where various versions of the analysis were presented. i Inspite of its 'literary flavor', naive consultants have solid judgments about sentences containing s u o . All of the judgments in this paper were confirmed with other native speakers.

AN OBJECT CLITIC P R O J E C T I O N IN M A N D A R I N CHINESE 113

2 SUO can also occur in b e i - c o n s t r u c t i o n s , the so-called passive construction, in Chinese.

i. neixie shi bu neng b e i Lisi s u o lijie these thing not can BEI Lisi SUO understand

'These things cannot be understood by Lisi.'

s u o occurring in b e i - c o n s t r u c t i o n s will be discussed in section 3.4. 3 For some speakers, the presence of a direct object resumptive pronoun is not quite accept- able. In addition, the overt pronouns, ta 'he' or t a m e n ' they', are restricted to animate objects:

i. *[Akiu mai-guo ta de ] neiben shu Akiu buy-GUO him DE that book

'the book that Akiu bought it'

ii. [Akiu jian-guo ta de ] neige ren Akiu see-GUO him DE that person

'the person who Akiu saw him'

4 A somewhat similar conclusion is reached in Lu (1981), who defines s u o as an auxiliary which must be used with transitive verbs. 5 Unfortunately, resultatives and other constructions (such as ECM constructions) cannot be used to test whether licensing is linked to Case or theta-roles since they all require a resump- rive pronoun, s u o is precluded for this reason in these constructions. 6 Relativization of the external argument, which is the NP preceded by b e i , requires the appearance of a resumptive pronoun, s u o is therefore precluded.

i. [Lisi bei *(ta) ma-guo ] de neige ren Lisi BEI him scold-GUO DE that person

'the person that Lisi was scolded by'

ii. *[Lisi bei ta s u o ma-guo ] de neige ren Lisi BEI him SUO scold-GUO DE that person

7 For discussion and justification for such a structure, see Chiu (1993). s Island Cosntraints are obeyed when the Islands are in the object position. See Huang (to appear) for an account which suggests that these cases involve control structures. 9 In Huang (1990), examples illustrating violations of Adjunct Condition involve topical- ization. However, in relative clauses, the Adjunct Condition is obeyed. It is not clear how Huang's Generalized Control account can work here:

i. *[[yinwei ni piping-le ], wo hen bugaoxing] de neige ren because you criticize-LE I very unhappy DE that person

'the person that I am very unhappy because you criticized (him)'

Jo The lack of wh-island effect in (46) is explained under Huang's (1982) proposal that wh-movement takes place at LF in Chinese: wh-islands are not formed at S-structure when relativization occurs, hence the violation of WHIC in (46). '~ Interestingly, an unexplained asymmetry between argument and adjunct is observed. Sentences (47)-(50) become grammatical when the wh-element (in situ) is an argument, as in (i), as opposed to an adjunct.

i. [Lisi xiangzhidao [shei s u o mai-le _ ] de ] neiben shu Lisi wonder who SUO buyoLE DE that book

'the book that Lisi wonders who bought'

This appears to contrast with English: relativization of an adjunct wh-island, as in (ii), seems to be better than relativization of an argument wh-island (iii).

114 B O N N I E C H I U

ii. * the bookl which I wonder who bought t~

iii. ? the book~ which I wonder where John bought t~

~2 This is reminiscent o f the situation in which tense lowering is blocked by negation in Chinese. Because le, the tense marker, cannot be lowered due to the presence of the negation bu, sentences such as (i) are ungrammatical.

i. *Lisi t~ bu da-le~ Akiu Lisi not hit-LE Akiu

Instead, another construction, involving an aspectual head, you ' have ' , must be employed, deriving the following grammatical sentence:

ii. Lisi mei you da Akiu Lisi not have hit Akiu

For a detailed account, see Chiu (1993). ~3 This analysis differs from Chiu (1991), where I proposed that this projection, which I then called AgrO-P, intervenes between TP and NegP, as in (i):

i. [n, �9 �9 �9 [^g,o-e [AgrOTy01, [^~O SUOI [NesP �9 �9 �9 [AspP �9 �9 �9 [VP �9 �9 -]]]]]]

This structure cannot be maintained due to the fact that Tense lowering is not blocked in suo-sentences, suggesting that suo is not in a position lower than TP. ~4 For further discussion on dou, see Huang (1982), Chiu (1990, 1993), Cheng (1991), and Li (1991). t5 dou in (60) takes the subject NP, not the object NP, as its antecedent NP. It is impos- sible for dou to take scope over the object NP when suo is present:

i. *[ta suo dou-du-guo__] de neixie shu he SUO DOU-read-GUO DE those books

' those books that he all read'

This reflects the observation that, like clitics, suo cannot be modified. See Chiu (1993) for further discussion of dou. ~6 This account of NomP differs from Sportiche (1992) in that nominative case under her theory is assigned by AgrS, not by the head of NomP. The question of whether or not Chinese has AgrS will not be taken up here, but see Chiu (1993) for further discussion. ~7 I will return to how accusative Case is assigned in sentences without syntactic wh- movement below. ~s The sentence in (65) is grammatical with the interpretation 'As for Lisi, those books, he read' , where both subject and object NP are taken to be topics. ~9 However, unlike there-expletive constructions in English, there is no definiteness effect in simple Chinese sentences. 20 My intuition is that relative clauses with suo have 'specific ' interpretation though syn- tactic evidence of such an intuition is not apparent. Huang (1982) points out that a relative clause preceding the numeral-classifier sequence gives rise to a referential or specific reading of the NP, while one that follows does not. In both cases, suo can occur:

i. Akiu du-guo [Lisi xie (suo) de ] sanben shu Akiu read-GUO Lisi write SUO DE three book

'Akiu read the three books that Lisi wrote. '

ii. Akiu du-guo sanben [Lisi (suo) xie de ] shu Akiu read-GUO three Lisi SUO write DE book

"Akiu read three books that Lisi wrote. '

AN OBJECT CLITIC PROJECTION IN M A N D A R I N CHINESE 115

The distinction between (i) and (ii) may be one of definiteness, instead of specificity. In (i) where the relative clause precedes the numeral-counter sequence, it has the interpretation that Lisi only wrote three books, and Akiu read all three of them. It is a 'definite' and specific reading. In contrast, (ii), where the relative clause follows the numeral-counter sequence, has the reading that Lisi wrote more than three books, and Akiu read three of them, an 'indefinite' but specific reading. Since we understand very little about how the notions of 'specificity', 'definiteness', and 'referentiality' are realized in Chinese, it is not clear to me, except at an intuitive level, how to relate suo to a 'specific' interpretation of a relative clause. Further work is necessary in this area of Chinese syntax. 2~ If NomP and SuoP are to be regarded as the Nominative and Accusative voice projec- tions, are there also AgrS and AgrO, as in (60), which assign nominative Case and accusative Case to their respective Specs? The answer is not clear. Given that Chinese lacks any agreement markings, we have no strong empirical reason to postulate an AgrS-P or an AgrO-P in the clausal structure of Chinese. However, there is some evidence which suggests the presence of an AgrS-P, as well as an AgrO-P, in Chinese. It is based on the observa- tion that two positions, accessible only to subject and object empty pronominals (small pro) are present between SuoP and VP where we expect AgrS-P and AgrO-P to be. The evidence comes from the syntactic behavior of dou. Dou, when taking a subject or an object small pro as its complement, undergoes local movement to a position between SuoP and VP. In Chiu (1993), I suggest that this local movement is due to the fact that subject and object small pro are licensed in SpecAgrS-P or SpecAgrO-P positions respectively. If this is correct, then, we have evidence that some projection analogous to AgrO-P is present in Chinese clausal structure. (See also Cheng (1989), who observes that overt pronouns and small pro do not have the same distribution.) Since the presence of an AgrS-P and AgrO-P projections is not crucial to the discussion on suo, I will not include them in the clausal structure here. 22 The rule of Predication at LF will identify the index of the head noun, shu 'book', with that of the null operator. See Chomsky (1982: 13) for details. 23 Cheng (1989) also proposes a non-movement (left-dislocated structure) analysis for topicalization. Based on the observation that possible pro-drop positions are parallel to possible topicalized positions, she argued that topicatization yields a left-dislocated structure whereby the topic binds an empty pronominal pro in the base position. 24 I assume that the topic position is adjoined to CP. 25 See Feng (1991) and Li (1994) who also treat bei as a verbal head with its own projec- tion. 26 A reviewer points out that suo is more available in relatives than in bei-constructions, a fact that the current analysis, which attempts to make parallel between relatives and bei- constructions, does not reflect. I have no insightful explanation for this tendency at this point. 27 In bei-sentences, negation must precede bei and its following NP.

i. Akiu meiyou bei Lisi ma Akiu not BEI Lisi scold

'Akiu was not scolded by Lisi.'

ii. * Akiu bei Lisi meiyou ma Akiu BEI Lisi not scold

28 Not all Chinese linguists agree on such a distinction. See Lee (1986) for an opposing view. 29 There appears to be one exception; the object control predicate, yaoqiu 'to ask', seems to allow suo to precede the matrix predicate as well, as shown in (i) and (ii).

116

i.

ii.

BONNIE CHIU

[Lisi yaoqiu [Akiu~ [PRO~ suo mai] de ] neixie shu Lisi ask Akiu SUO buy DE those book

'those books that Lisi asked Akiu to buy'

[Lisi suo yaoqiu Akiu~ [PRO~ mail de ] neixie shu Lisi SUO ask Akiu buy DE those book

REFERENCES

Cheng, Lisa L.-S. (1989) "Aspectual Licensing of pro in Mandarin Chinese," ms., MIT. Cheng, Lisa L.-S. (1991) On the Typology of wh-questions, PhD dissertation, M1T. Chiu, Bonnie (1990) "A Case of Quantifier Floating in Mandarin Chinese," paper presented

at Northeast Conference on Chinese Linguistics, Cornell University, Ithaca. Chiu, Bonnie (1991 ) "'Suo: an Object Clitic in Mandarin Chinese," in proceedings of the Third

Leiden Conference for Junior Linguists, Leiden. Chiu, Bonnie (1993) The Inflectional Structure of Mandarin Chinese, PhD dissertation, UCLA. Chomsky, Noam (1982) Some Concepts and Consequences of the Theory of Government

and Biding, MIT Press, Cambridge, Massachusetts. Chomsky, Noam (1986) Barriers, MIT Press, Cambridge, Massachusetts. Chomsky, Noam (1989) "Some Notes on Economy of Derivation and Representation," in

L. Laka and A. Mahajan (eds.), Functional Heads and Clause Structures, MIT Working Papers in Linguistics, vol. 10, pp. 43-74.

Feng, S.-L. (1991) "The Passive Construction in Chinese," ms., University of Pennsylvania. Goodall, Grant (1990)"Mandarin Chinese Passives and the Nature of Case Absorption," paper

presented at the Third Northeast Conference on Chinese Linguistics, Cornell University, Ithaca.

Huang, C-T, James (1982) Logical Relations in Chinese and the Theory of Grammar, PhD dissertation, MIT.

Huang, C.-T. James (1990) Chinese Syntax and Logical Form, ms., University of California, Irvine.

Kayne, Richard S. (1975) French Syntax, MIT Press, Cambridge, Massachusetts. Lee, Thomas (1986) Studies on Quantification in Chinese, PhD dissertation, UCLA. Li, Charles and Sandra Thompson (1981) Mandarin Chinese Functional Reference Grammar,

University of California Press, Los Angeles. Li, H.-J. Grace (1994) "Bei: An Affected Head - The Bei-Constmction Revisited," unpub-

lished M.A. thesis, National Taiwan Normal University. Li, Y.-H. Audrey (1985) Abstract Case in Chinese, PhD dissertation, University of Southern

California. Li, Y.-H. Audrey (1990) Order and Constituency in Mandarin Chinese, Studies in Natural

Language and Linguistic Theory, Kluwer Academic Publishers, Dordrecht. Li, Y.-H. Audrey (1991) "Indefinite WH in Mandarin Chinese," manuscript, University of

Southern California. Lin, J.-W. and J. Tang (1991) "Modals in Chinese," paper presented at the Third Northeast

Conference on Chinese Linguistics, Cornell University, Ithaca. Liu, Y.-J. (1990) "Suo Ji Xin Lun [Suo revisited]," Zhonguo Yuwen, Vol. 2. Lu, S.-X. (1981) Xiandai Hanyu Babai Ci [Eight Hundred Phrases of Modern Chinese],

Shangwu Publisher, Beijing. Pollock, J.-Y. (1989) "Verb Movement, Universal Grammar, and the Structure of IP,"

Linguistic Inquiry 20, 365-424. Sportiche, Dominique (1992) "Clitic, Voice and Spec-Head Licensing," ms., UCLA. Tang, C.-C. Jane (1990) Chinese Phrase Structure and the Extended XTyOl h-theory, PhD

dissertation, Cornell University.

AN OBJECT CLITIC PROJECTION IN MANDARIN CHINESE 117

Tang, T.-C. (1976) "Focusing Constructions in Chinese: Cleft Sentences and Pseudo-cleft Sentences," in T.-C. Tang, R. L. Cheng, and Y.-C. Li (eds.), Studies in Chinese Syntax and Semantics, Universe and Scope: Presupposition and Quantification in Chinese, Student Book Co., Taiwan.

Wang, L. (1962) Gudai Hanyu [Classical Chinese], Zhonghua Shuju Publisher, Beijing. Zhu, Dexi (1982) Yufa Jianyi [A Brief Lecture on Syntax], Shanwu Yingshu Guan.

Received 22 August 1993 Revised 30 June 1994

Department of English National Taiwan Normal University 162 East Hoping Road Taipei Taiwan t22011 @ cc.ntnu .edu. tw