Academia.eduAcademia.edu

Correlative Constructions in Chinese

1992

An analysis of correlative constructions in Chinese that: (1) gives a principled account of the distribution of correlative markers; and (2) offers an explanation for some puzzling facts about distribution of anaphoric pronouns is presented. It is suggested that previous research has misidentified instances of verb phrase coordination as zero-anaphora in such constructions, which raises a more general question about zero anaphora. It is suggested that better criteria for identification of zero-anaphora are required in languages like Chinese, which lacks corroborative evidence such as agreement morphology. It is concluded, then, that for at least a subset of Chinese, a better account is available without zero-anaphors. (MSE)

DOCUMENT RESUME ED 350 838 AUTHOR TITLE PUB DATE NOTE PUB TYPE EDRS PRICE DESCRIPTORS IDENTIFIERS FL 020 621 Harlow, Steve; Cullen, Connie Correlative Constructions in Chinese. Jun 92 19p.; In: Harlow, S. J. and Warner, A. R., Eds. York Papers in Linguistics, 16; see FL 020 620. An earlier version of this paper was presented at the autumn meeting of the Linguistics Association of Great Britain (1986). Reports Evaluative/Feasibility (142) MFO1 /PCO1 Plus Postage. *Chinese; Foreign Countries; *Grammar; *Language Patterns; Language Research; Linguistic Theory; *Pronouns; *Sentence Structure *Anaphora ABSTRACT An analysis of correlative constructions in Chinese that: (1) gives a principled account of the distribution of correlative markers; and (2) offers an explanation for some puzzling facts about distribution of anaphoric pronouns is presented. It is suggested that previous research has misidentified instances of verb phrase coordination as zero-anaphora in such constructions, which raises a more general question about zero anaphora. It is suggested that better criteria for identification of zero-anaphora are required in languages like Chinese, which lacks corroborative evidence such as agreement morphology. It is concluded, then, that for at least a subset of Chinese, a better account is available without zero-anaphors. (MSE) *****************************************************************1 %: *** Reproductions supplied by EDRS are the best that can be made from the original document. *********************************************************************** Correlative Constructions in Chinese* Steve Harlow and Connie Cullen University of York U S. DEPARTMENT OF EDUCATION Othce of Educational Research and improvement E UCATIONAL RESOURCES INFORMATION THIS TO REPRODUCE GRANTED BY "PERMISSION MATERIAL HAS BEEN CENTER (ERICA 4.) NS document has been reproduced as eCeived from the person or otgantzahOn ongtnattng it C Mmor changes have been made to improve reproduction ouahty PointS of new o oolnons stated m this dOcu ment do not necessanty represent olfictat OERI posthon or policy RESOURCES TO THE EDUCATIONAL CENTER (ERIC) INFORMATION BEST COPY AVAILABLE 2 CORRELATIVE CONSTRUCTIONS IN CHINESE* Steve Harlow and Connie Cullen University of York Chinese possesses a construction-type known as 'correlative constructions'. Such constructions are characterised by the presence of a pairs of morphemes marking the constituent halves of the construction. Examples (taken from Liu 1981) are as follows in which the markers of the correlative construction are italicised: ( I) Zhuxi yi chulai, qunzhong jiu gu-zhang chairman as-soon-as come-out, masses then applaud As soon as the chairman appeared the masses applauded (2) Suiran (3) An Xian nianji xiao, although An xian age small, to qie shemme dcm hui he however anything all can Although An Xian is young, he can do everything Yinwei Lta Niu sheng bingle, because big ox grow sick, suoyi to mei-lai Shang ke therefore he did-not-corue attend class Because Da Niu is ill he didn't come to class * An earlier version of this paper was read to the Autumn meeting of the Linguistics Association of Great Britain in 1986. The authors would like to thank participants at that meeting for their comments, in particular Elisabet Engdahl and Dick Hudson, neither of whom are responsible for our failure to take their advice. I Or in some cases an n-tuple. We restrict our attention in this paper to pairs, although the analysis that we propose should extend unproblematically to n-tuples. York Papers in Linguistics 16 (1992) 0 Steve Harlow and Connie Cullen 1-17 YORK PAPERS IN LINGUISTICS 16 In this paper we propose that the properties of these constructions are most satisfactorily accounted for if they are treated as being multi- headed; that is, both halves of the construction arc heads. The theoretical framework that we assume for concreteness is that of Generalised Phrase Structure Grammar (GPSG) as presented in Gazdar et. al. (1985), although we believe that the essentials of our analysis will transfer to any framework which exploits a feature decomposition of categories and a schema for constituent coordination. In effect, we arc proposing a theoretical reconstruction and justification of Chao's (1968) contention that these are coordinate constructions.2 A fruitful starting point for discussion of the properties of these constructions is a paper by Liu Feng-hsi (Liu 1981). Liu presents a more extensive range of examples than those provided so far. More specifically, the examples in (1-3) only represent a subset of the possible patterns involving the correlative markers yi.. jiu (as soon as...then), suiran...que (although...yet), and yinwei...suoyi (because...therefore). (4)-(7) show that correlative markers fall into four groups with respect to their permitted linear order relative to the subject: (4) yi qunzhong jiu gu-zhang chulai, Zhuxi chairman as-soon-as come-out, masses then applaud As soon as the chairman appeared the masses applauded b. *Yi zhuxi chulai, flu qunzhong gu-zhang a. 2 Chao (1968:793) lists the following set of correlative markers (transcription altered from GR to Pinyin [CJC/SAID: yue...yue !the more...the more), ye...yelyou...you (both...and), bu...bu (not...unless), yibian (X...while Y), huoshi...huoshi /haishi...haishi (either...or). suiran...dansiti (although...yet), yinwei...suoyi (because...therefore), jiran...jiu (inasmuch as...then), yaoshilyaoljiaru... jiu (if...then), chufei /chule... bu (unless...not). chufeilchule...cal (only if...then). bushi...jiushi (if not...then). zhiyao...jiu (so long as...then), shangqie...hequang (not even...how much more), budan...bingqie / erqie (not only...but also), ninke...yebu (would rather...than), yuqllyoude...ningkelburulhaishi (rather than...had better), jinguan...haishi (not matter if...still), zaildso...ye (no matter how...still), jiushiljishi...ye (even if...then), gang(cai)...jiu (just as...then), yi.. jiu (as soon asthen). Many of these pairs will be discussed below. 42 CORRELATIVE CONSTRUCTIONS IN CHINESE c. d. (5) a. *Yi zhuxi chulai, qunzhong jiu gu-zhang *zhuxi yi chulai, jiu qunzhong gu-zhang Suiran An Xian nianji xiao, although An Xian age small, Ia cite shemme dou hui he however anything all can b. c. d. (6) a. b. c. d. (7) Although An Xian is young, he can do everything An Xian suiran nianji xiao, to que shenme dou hui *Suiran An Xian nianji xiao, to que shenme dou hui *An Xian suiran nianji xiao, to que shenme dou hui Yinwei Da Niu sheng bingle, suoyi mama because big ox grow sick, therefore mother hen danxin very worried Because Da Niu is ill, therefore (his) mother is very worried Da Niu yinwei sheng bingle, suoyi mama hen danxin) *Yinwei Da Niu sheng bingle, mama suoyi hen danxin *Da Niu yinwei sheng bingle, mama suoyi hen danxin Bushi ni lai, jiushi wo w not-is you come then-is I go Either you come, or I go b. *Ni bushi lai, jiushi wo qu c. Ni bushi lai., wo jiushi qu d. *Ni bushi lai, wo jiushi qu a. The four classes of correlative markers have the following properties: A. in both halves, the subject must precede the correlative marker (4): yi...jiu. B. the correlative marker can either precede (5a) or follow (5b) the subject. In the secoid half of the construction the correlative marker always follows the subject: suiran...que. C. this pattern is like (B) in allowing both order of subject and correlative marker in the first half of the construction, but differs in 3 5 i 3 YORK PAPERS IN UNGUISTICS 16 the second half where the correlative marker must precede its subject (6): yinwei ...suoyi. D. the last class allows two possible orders in both halves, subject to a parallelism restriction: the order of correlative marker in the second half of the construction must be the same as that in the first half (7): bushi ...jiushi. The table in (8) summarises these ordering properties. (8) CM TYPE A SUBJ = (4a) Zhuxi CM yi SUBJ qunzhong jiu TYPE B i. = (5a) SUBJ An Xian CM Suiran CM suiran SUBJ An Xian SUBJ CM to clue SUBJ CM to quo CM Yinwei SUBJ Da Niu SUBJ Da Niu CM suoyi CM suoyi SUBJ mama SUBJ mama CM Bushi SUBJ Ni SUBJ ni CM bushi CM SUBJ wo SUBJ ii. = (5b) TYPE C i. = (6a) ii. = (6b) TYPED i. :: (7a) ii. = (7c) CM yinwei jir.:hi CM wo jiushi Ordering of subject and correlative markers is one property of correlative constructions for which an account is required. A second property, which is the major fa u; of Liu's paper, is the occurrence in these constructions of phon, togically empty pronouns.As is well-know, anaphoric pronouns ..an be either phonologically realised or phonologically empty in Chinese with a very considerable degree of freedom (cf. Chen 1984,Huang 1984, Li and Thompson 1979 and Xu 1986). The passage given as (9) below exemplifies som° of the 4 ar CORRELATIVE CONSTRUCTIONS IN CHINESE possibilities. It is taken from Chen (1984) and zeros are used to mark positions that have been analysed as being filled by empty NPs. (9) ge piqi Lao Qian you zhemme Lao Qian had this-kind-of a temperament O wen 0 pengyou yao shenme dongxi want what thing (he) ask (his) friend 0 jiu del gei 0 O like (he) immediately then must give (him) (it) O bu gei 0 0 (he) not give (him) (it) O jiu juede 0 shi qiaobuqi to (he) then felt O jitian (he) is despise him bu gaoxing (he) several-days not happy 1, 01d Qian had this kind of personality: if he asked his friend for something, the friend must give him it immediately; if he didn't, then Old Qian would feel that the friend must despise him and would be displeased for several days. Liu gives the following examples to illustrate the distribution of pronouns in correlative constructions.3 (10) is an example of Type A, where the subjects must precede the correlative markers. Here a zero pronoun is obligatory. (10) jiandao wo, (*la) jiu xiang wo Xiao Ming yi me me, she then to Xiao Ming as-soon-as see zhao shou wave hand As soon as Xiao Ming saw me, she waved at me 3 As usual, (X) denotes that X is optional; (*X) that X is not possible and *(X) that X is obligatory. The judgements given here are Liu's. Not all Chinese speakers share them. We will not discuss alternative judgements here, but will restrict our attention in this paper to the pattern of judgements given in the text. 5 YORK PAPERS IN LINGUISTICS 16 (11) is an example of Type B, in which alternative orders are possible in the first half of the construction, and the subject precedes the correlative marker in the second half. Here zero pronominals are optional when the subject of the first clause precedes its correlative marker, but obligatory otherwise. (11) An Xian nianji xiao, although An xian age small, shemme dou hui (ta) que he however anything all can Although An Xian is young, he can do everything nianji xiao, b. An Xian suiran An xian although age small, shemme dou hui ( *ta) que he however anything all can a. Suiran (12) is an example of Type C, in which again both orders arc possible in the first half, but in the second half, the subject follows the correlative marker. Here we find that possible zero pronominal occurrences are similar to those in (11), except that, when the first half is subject-initial, a zero pronoun is disprefcrred rather than impossible. (12) a. Yinwei Da Niu sheng bingle, because big ox grow sick, b. shang ke (ta) mei-lai suoyi therefore he did-not-come attend class Because Da Niu was ill, he didn't come to class Da Niu yinwei sheng bingle, suoyi ?(ta) mei-lai big ox because grow sick, therefore he did-not-come shang ke attend class Finally, in (13) we have examples of Type D, where there must be parallel orders in each half. Here an overt pronoun is preferred when it follows the correlative marker, but is impossible when it precedes. 86 CORRELATIVE CONSTRUCTIONS IN CHINESE (13) a. b. Bushi Li Si jicuole, jiushi ?(ta) wangle not-is Li Si misremember, then-is l forget Either Li Si misremembered, or else he forgot *Li Si bushi jicuole, wangle Li Si not-is misremember, he then-is forget In (14) is a table which is taken from Liu (1981) which summarises this correlation between ordering properties and zero anaphora. (14). TYPE A SUBJ CM SUBJ CM = (10) Xiao Ming yi (*ta) jiu TYPE B i. = (11 a) SUBJ An Xian CM SUBJ (*ta) SUBJ CM (ta) que saran que ZEROANAPHORA Obligatory Obligatory CM ii. = (11b) Suiran SUBJ An Xian TYPE C i.= (I2a) CM Yinwei SUBJ ii. = (12b) Da Niu SUBJ Da Niu CM yinwci CM suoyi CM suoyi SUBJ (ta) Pronoun preferred SUBJ (ta) Optional TYPE D i. = (i3a) SUBJ CM jiushi SUBJ SUBJ (ta) Pronoun preferred CM Bushi SUBJ ii. = (13b) Li Si Li Si CM bushi (*ta) CM Optional CM jiushi Obligatory The first two columns give the relative order of the subject and the correlative marker (CM) for the first clause, and the second par give the relative order for the second clause. The comments in the final column under the heading 'zero anaphora' give the status of phonologically empty pronouns in the subject position of the second clause. Liu (1981:200) provides the following a descriptive generalisation about this patterning: 7 9 YORK PAPERS IN LINGUISTICS 16 Zero anaphora is obligatory in parallel structures if the subject precedes the [correlative] marker; otherwise it is (15) optional. (In parallel structures with the subject following the [correlative] marker the pronoun is preferred.) To render the data supporting this generalisation more transparent, we present in (16) a reorganisation of Liu's table (14) in which we make the primary sorting key the status of the pronoun, and the secondary key the order of correlative marker and subject. In Liu's table, the primary key is the individual correlative markers; a consequence of the reclassification in (16) is that some correlative markers appear under more than we heading. This will be significant. (16) a b.i SUBJ CM CM SUBJ yi jiu suiran bushi que SUBJ c CM yinwei bushi A B.i Obligatory zero D.ii SUBJ CM crx suiran b.ii SUBJ PRONOUN CM CM CM yinwei suoyi SUBJ CM B.ii Optional C.ii Optional C.i Overtpreferrecl SUBJ SUBJ suoyi jiushi D.i We now have a generalisation about the distribution, but we do not as yet have an explanation. It is to this that we now turn. 10 8 CORRELATIVE CONSTRUCTIONS IN CHINESE The essence of our proposal is the claim that correlative constructions are in fact coordinate constructions, as Chao (1968) suggested. To make this claim specific, we provide in (17) a rule schema to define for these constructions. (17) V2 H[CONJ ao], H[CONJ a114 where a E ( <yinwei, suoyi>, <bushi, jiushi>,<suiran, que>, jiu>, <ye, ye>1 (17) is Gaidar et. al's Binary Coordination Schema (1985: 171), with a category value assigned to the mother and the lexical values of the Chinese correlative markers given as value:: of the CONJ feature.5 The essence of the GPSG analysis of coordinate structures is that all conjuncts are heads. This ensures (via the Head Feature Convention) that conjuncts share relevant properties, such as categorial identity. Hence, in any instantiation of (17), the two daughters will both be v2.6 4 V2 subsumes both S and VP. 5 The rule says nothing about the relative order of the daughters. This will be determined by a Linear Precedence (LP) statement. See Gazdar et. al. (1985: 172) for details. 6 A possible objection to our proposal that correlative constructions are in fact coordinate constructions might be that they do not possess a uniform semantics. To this we respond thtt the crucial property defining these constructions is multi-headedness and we use the term 'coordinate structure' to refer to this. It does not necessarily follow that such constructions share a common pattern of semantic interpretation. We would assume, although we will not attempt to justify it here, that the semantics of these constructions is compositional and that the coordinating morphemes themselves make a compositional contribution to the semantics of the construction as a whole. A second objection, due to Dick Hudson (personal communication) is that at least son e of these correlative mutters are, in fact, adverbs and not conjunctions (in particular ye 'also). While it is true that the label 'adverb' is applied in this way, it is also a fact that adverbs as a class arc notoriously difficult to define and the these items do not share the distribution of Chinese manner advcrbials, nor the distribution of, say, temporal adverbs. Lest translation equivalents muddy the decision on this point, note that whereas English also does have a characteristics adverb distribution ('Also, 9 .1 BB 1 COPY AVAILABLE YORK PAPERS IN LINGUISTICS 16 Thus far,. modulo the different lexical items involved, we are claiming that coordination in Chinese and English is essentially the same. We propose, however, that Chinese differs from English in two crucial respects. The first is that in Chinese the feature CONJ is a member of the set of head features, as specified in (18). (18) CONJ E HEAD This means that it will be subject to the Head Feature Convention with the result that we will get the effect of percolation of that feature and its value down through heads from the position at which it is introduced by the rules of the grammar. Secondly, the Chinese rules which spell out the conjunction itself, (19) and (20), making it a sister of a category identical to the mother, are, in contrast to their English counterpart, category specific.7 (19) S[CONJ a] ([SUBCAT a] ) , H[CONJ NIL] where a E (yinwei, suoyi, bushi, jiushi, suiran) (20) VP[CONJ a] --> ([SUBCAT a]), H[CONJ NIL] where a E yinwei, bushi, jiushi, ye, suiran, que, yi, jiu) Crucially there are different rules in (19) and (20) for sentences and for VPs. They differ specifically in respect of the particular coordinating conjunction that can be instantiated as a sister of the head. So, these rules specify that suoyi can only appear as a sister of S, and that ye, que, yi and jiu can only appear as sisters of VP. The remainder (yinwei, bushi, jiushi, suiran) can occur as sister of either VP or S. These rules, therefore, will licence the following trees: sentential coordination in (21), and VP coordination in (22) he speaks French'; 'He also speaks French'; 'He speaks French also'), this is not true of ye which can only appear in the pre-VP position specified by (20). 7 In fact, these restrictions would be enforced on instantiated features by Feature Coocurrence Restrictions. For simplicity of presentation we have chosen fto by-pass these details here. i2 10 we CORRELATIVE CONSTRUCTIONS IF CHINESE .---'s-S (21) SICONJ suoyi S[CONJ yinwei] CONJ CONJ I 1 suoyi yinwei S (22) VP NP VPICONJ jiu ... VP1CONJ yi ...) CONJ VP CONJ I I yi jiu que jiushi suiran bushi of Note, however, that although these rules preclude the appearance the S or VP, they do not restrict certain conjunctions as sisters to instantiation of the CONJ feature on an S or VP node. Indeed, the Head Feature Convention will ensure that mother and head daughters must (21) and share permitted feature instantiations. Ther.lore, in addition to coordinated following trees, in which the (22), the rules admit the category is S, but one or more of the coordinating conjunctions appears as a sister of VP. 11 YORK PAPERS IN UNGUIST1CS 16 (23) S[CONJ quej WON; suiran] NP CONJ VP[CONJ que CONJ suiran VP que Here the CONJ feature in both conjuncts is introduced at the S level by rule (17). They differ, however, in the level at which the conjunct themselves can be realised Suiran (although) is realised at the S level, but que (yet) is only realised at the VP level. The CONJ feature 'percolates' down the tree as a consequence of the Head Feature Convention. (24) S[CONJ suoyi: S[CONJ yinwei] VP[CONJ yinwei] CONJ NP CONJ yinwei suoyi Yinwei (because), but not suoyi(therefore), can also be realised at the VP level. Here too, the CONJ feature percolates down the tree as a consequence of the Head Feature Convention. 1.4 12 IN CHINESE CONSTRUCTIONS CORRELATIVE jiu) SjCONJ (25) WON., NP /"N yij VP[CONJ yi I ji Li] NP VP VP CONS CONI jiu level, although at the VP feature CONS realised the both Feature are Again, of the Head as) andjiu(thcn) coordination. soon Yi (as consequence a sentence-level as a precede in Chineseaccounts they expressdown the tree that conjunctsVP sisters percolates their requirement Convention, (19) preceded that the above. subjects rules like Note also discussed that facts and two separateof coordinating for theorderingthat having their heads out spellings objected to this objection be for straightforwardly might response Our Now, it necessary category-specific and ad hoc. is in any case which are specific, and (20) of thing are indeed category is unmotivated kind conjunctions,out that this conjunctions is to point coordinating because Chinese, illustrates. as (26) ni dou qu Wo genlhe you all go going I and are both ye qi I and you qu, ni also go Wo ye you I also go both going and you are ni qu *Ye ...vo ye also you go also I going) you are both (1 and yi 13 K.; YORK PAPERS IN LLNOU1sT1Cs 16 In Chinese we have a situation in which NPs, for instance, take a different set of coordinating conjunctions to the verbal projections which we have been discussing above. The coordination rule for NPs will look like (27), (27) NP -.> H(CONJ ao] H (CONJ al I where a e (<NIL gen> <NIL he>) which defines NP coordinating conjunctions to be the morphemes gen and he. We therefore need category-specific conjunctions for Chinese anyway. Our proposed analysis of correlative constructions merely extends this to S and VP. with some degree of overlap between the two.8 The analysis thus far has only been concerned with the distribution of correlative markers. We have not yet touched on the issue of zero anaphora. Recall that, according to Liu's generalisation, we get several instances of obligatory zero anaphora when in both conjuncts the coordinating conjunctions occur after the subject. According to our analysis, the correlative markers occur in this position because they have been realised at the VP level. They arc therefore by the rules that we have given for VP coordination. So the simple answer the reason why there are no overt pronouns in the second conjunct in these cases is because there is no noun phrase in the second conjunct at all. These constructions are simply instances of the structure in (22) coordinated VPs. So the reason also why we do get overt pronouns in the other Note also that English coordinating conjunctions exhibit some degree of category-specificity. Both...anct is decidedly awkward with bare S's i. ?Both John is rich and Mary is fie,. while being fine with ii. Max stated both that John is rich and that Mary is rich Either...or on the other hand is good with both iii. Either John is rich or Mary is rich iv. Max stated either that John is rich or that Mary is rich 8 6 14 CORRELATIVE CONSTRUCIIONS IN CHINESE kinds of structures is because they all involve cases of sentential coordination? There is some interesting support for this position from a rather different kind of evidence. Chen (1984), a study of the distribution of Chinese zero anaphors in discourse, provides the following statistics for the occurrence of zero anaphora In the corpus: (28) Relative occurrence of zero an phora (Chen 1984:15) Syntactic position Numlx,r Percentage topic/subject direct object indirect object 43 75.4 11 19.3 3 5.3 A glance at these results reveals a striking asymmetry between the number of occurrences of topic/subject zero anaphors and the number of occurrences of their direct object and indirect object counterparts, with the former being by far and away the most frequently occurring kinds of empty categories. Why should there be this asymmetry? Our analysis It is valid to ask how one can be sure that one is dealing with VP 9 coordination, rather than sentential coordination, where the subject of the second conjunct is a zero-pronoun. This is a difficult question to answer conclusively. What can be demonstrated is that there is indeed evidence that Chinese actually allows VP coordination. On the assumption that zero- pronouns in Chinese are simply the covert versions of their overt counterparts and, in particular, that they serve the same anaphoric functions, the fact that there are coordinate structures such as (i) below which do not permit an oven pronoun in the second conjunct, should serve to argue that we have here a situation in which a pronouns are not permitted. The only remaining option is VP coordination. i. Mcigc rca ye dou hui shuo Hanyu, every man both all can speak Chinese (ta) ye dou hui shuo Yingyu (he) and all can speak English Everyone can speak both Chinese and English 15 BEST COPY MAE YORK PAPERS IN LINGUISTICS 16 provides the bisis for an explanation. Since Chen does not consider the possibility of VP coordination, all such potential instances VP coordination are treated as examples a zero subject anaphora. A plausible account of the much greater frequency of occurrence of zero subject anaphors is that more than one construction is involved. A combination of VP coordination with genuine zero discourse anaphora provides a double source for the number of apparent empty topic/subject positions. For object positions, by contrast, there is no coordination possibility which has the appearance of a zero pronoun; in these cases we are only going to get missing objects where there are, indeed, missing objects. So we suggest that the statistics which Chen has produced are the result of a kind of conspiracy between real empty NPs and the facts of VP coordinatior.. What we have tried to do in this paper is to outline an analysis of Chinese correlative constructions which not only gives a principled account of the distribution of correlative markers, but which also offers the prospect of an explanation for some puzzling facts about the the distribution of anaphoric pronouns. Our suggestion is that Liu and Chen have misidentified instances of VP coordination as zero-anaphora in such constructions. This leads us to raise a more general question about zero-anaphora. It seem clear to us that better criteria are required for the identification of zero-anaphors in languages like Chinese, which lack corroborative evidence (such as agreement morphology). There is a danger that the implicit criteria are simply that if what could be an NP position is empty, it ..iust in fact be occupied by a zero-anaphor. Sometimes (but not with respect to the data we have discussed in this paper), it even seems that the occurrence of an overt pronoun in an English translation is sufficient evidence to warrant the postulation of a zero-anaphor in Chinese. We have shown in this paper that, for at least a subset of Chinese, a better account is available without zeroanaphors.1° 10 We should point out that our analysis is certainly not the last word on the subject. While it is correct to say that when VP coordination is the only possibility oven subject pronouns are excluded and where sentential coordination is possible then overt subject pronouns are possible, there is a third situation for which our analysis does not provide a complete account. (as soon ss...then) type, we propose For correlative markers of the 1S 16 CORRELATIVE CONSTRUCTIONS IN CHINESE REFERENCES Chao Yuen-ren (1968) Grammar of Spoken Chinese. Berkeley: University of California Press. Chen Ping (1984) A Discourse Analysis of Third Person Zero Anaphora in Chinese. Bloomington: Indiana University Linguistics Club. Gazdar, Gerald; Ewan Klein; Geoffrey K. Pullum and Ivan A. Sag (1985) Generalized Phrase Structure Grammar. Oxford: Blackwell and Cambridge, Mass: Harvard University Press. Huang James C-T. (1984) On the distribution and reference of empty pronouns. Linguistic Inquiry 15.4:531 573 Li Charles N and Sandra Thompson (1979) Third person pronouns and zero- anaphora in Chinese discourse. In Talmy Givon (ed). Syntax and Semantics. Volume 12. Academic Press Liu Fcng -hsi (1981) Zero-Anaphora in Mandarin Chinese. CLS 17:197-204 Xu Lie-jiong (1986) Free empty category. Linguistic Inquiry. 17.1:75-93 that there is sentential coordination, but that the correlative markers are spelled out at the VP level. This is because they occur in post-subject position. The other aspects of our analysis would lead us to expect that, because the second conjunct contains a subject position, an overt pronoun should be possible there. This is not in fact the case: S Ne S NP. VP CONJ /\\VP VP CON! *la VP U Clearly, additional factors are at work here which serve to limit the permissible structure to VP coordination. 17 BEST COPY AVAILABLE