A Survey of Ellipsis in Chinese

Home , Answer ellipsis, Catena (linguistics), Ellipsis (linguistics), Gapping, Pseudogapping

Timothy Osborne Junying Liang Zhejiang University Zhejiang University Hangzhou Hangzhou China China [email protected] [email protected]

see which ellipsis mechanisms are and are not Abstract present. The analysis of ellipsis pursued below is de- Much work on ellipsis has been conducted pendency-based, and the catena unit plays a cen- using data from English, and many widely tral role in the account. A catena is a word or acknowledged types of ellipsis exist in Eng- combination of words that are linked together by lish. The extent to which the named ellipsis dependencies (Osborne et al. 2012). Ellipsis me- mechanisms exist in other languages is, chanisms in English have been shown to elide though, often not clear. This manuscript surveys ellipsis in Mandarin Chinese using a catenae. The survey seeks to determine the extent dependency-based approach to syntax. It to which the catena is also the central unit for a probes to see which ellipsis mechanisms exist theory of ellipsis in Mandarin. in Mandarin. The survey demonstrates that This contribution thus pursues three goals: 1) gapping, stripping, pseudogapping, sluicing, provide an initial exploration of ellipsis in Man- and comparative deletion do not exist in darin, 2) determine the extent to which the catena Mandarin (or are highly restricted) and that unit can serve as the basis for a theory of Manda- VP-ellipsis, answer ellipsis, and N-ellipsis are rin ellipsis, and 3) consider what can be learned all arguably present. Furthermore, zero ana- about ellipsis in general from a comparison of phora is frequent in Mandarin, whereas it is ellipsis mechanisms across English and Manda- absent from English (or highly restricted). The catena unit is pillar of the account, since rin. the elided material of ellipsis is a catena. A word of caution is appropriate concerning the dependency hierarchies assumed for Manda- 1 An inventory of ellipsis mechanisms rin below. To our knowledge, many basic aspects of Mandarin sentence structure have not yet been The study of ellipsis recognizes numerous dis- worked out in theoretical detail from a DG pers- tinct types. The following mechanisms are pective. Basic questions about the dependency among the most commonly acknowledged: status of sentence-final particles, coverbs, de-constructions, classifiers, etc. have not been 1. Gapping debated from a DG perspective. Thus the validity 2. Stripping of many of the structures posited below is taken 3. Pseudogapping for granted. Future explorations into the depen- 4. Sluicing dency structures of Mandarin may motivate cor- 5. Comparative deletion rections to the dependency hierarchies for Man- darin posited below. 6. VP-ellipsis 7. Answer ellipsis 2 Gapping, stripping, pseudogapping 8. N-ellipsis Gapping, stripping, and pseudogapping have 9. Null complement anaphora been thoroughly explored (e.g. Jackedoff 1971, 10. Zero anaphora Kuno 1976, Stump 1977, Levin 1986, McCawley Excepting zero anaphora, these mechanisms oc- 1998). The following three sentences illustrate cur in English, and most of them are present in gapping, stripping, and pseudogapping in Eng- languages related to English. The extent to which lish: they exist in languages more distant from English is often not clear, however. This contribution surveys ellipsis in Mandarin Chinese, probing to

271 Proceedings of the Third International Conference on Dependency Linguistics (Depling 2015), pages 271–280, Uppsala, Sweden, August 24–26 2015. (1) Should or should (6) líkāi-le líkāi-le I call you call you me Jō Lǐ yě Should I call you, or should you call me? *Jō líkāi-le, Lǐ yě líkāi-le. Jo leave-le Li also leave-le. (2) Intended: ‘Jo left, and Li also.’ Should or should (7) *Jō bìxū gōngzuò, Lǐ ye bìxū gōnzuò. I call you call Jo has.to work Li also has.to work. Jo her Intended: ‘Jo has to work, and Li too.’ Should I call Jo, or should you call her? Noteworthy about these failed attempts at gapping and stripping is the fact that Mandarin lacks (3) She should call me more than a direct equivalent to and for coordinating claus- should es. Perhaps the absence of such an element is a she call factor limiting the distribution of gapping and stripping, since these mechanisms are widely you acknowledged as occurring only in the she should call you. non-initial conjuncts of coordinated clauses. The following attempt at pseudogappng in Example (1) illustrates gapping, example (2) Mandarin also fails: stripping, and example (3) pseudogapping. Gap- ping and stripping occur in coordinate structures. (8) Nǐ yīngāi xué fǎyǔ, Pseudogapping can appear in subordinate clauses you should study French in the absence of coordination, but the pseudogap yīnggāi must find an antecedent – it cannot take a post- cedent. nǐ yě xué The elided material should…call in (1) and (2) déyǔ is a catena, and the word call in (3) is also a ca- *nǐ yě yīnggāi xué déyǔ. tena, a one-word catena. The fact that should you also should study German immediately dominates call is what makes the combination should…call a catena. The exam- Intended: ‘You should study French, and ples therefore deliver a sense of the importance you should study German, too.’ of the catena unit for the theory of ellipsis. There The data just produced demonstrate that gap- are, however, many details of the dependency ping, stripping, and pseudogapping are types of hierarchies shown in (1–3) that can be over- ellipsis that are either absent from Mandarin, or looked here, since they are not important for are much more restricted than in English. The surveying ellipsis in Mandarin. fact that examples involving both gapping and Turning to Mandarin, we see that these ellip- stripping are bad is not surprising since the two sis mechanisms are generally not possible. The are widely viewed as involving the same one el- following attempts at gapping fail: lipsis mechanism. Concerning the absence of pseudogapping (4) diǎn-le diǎn-le from Mandarin, however, the fact that it is not possible is more revealing. Pseudogapping be- Tā kāfēi tā chá haves like VP-ellipsis in certain ways, and like *Tā diǎn-le kāfēi, tā diǎn-le chá. gapping in other ways. It behaves like s/he ordered coffee s/he ordered tea VP-ellipsis mainly insofar as it is licensed by an Intended: 'He ordered coffee, and she tea.’ auxiliary verb just like VP-ellipsis, and it is like gapping insofar it involves a true “gap” with a (5) *Jō xǐhuān dàngāo, Lǐ xǐhuān qiǎokèlì. remnant, whereby the remnant must stand in Jo likes cake. Li likes chocolate contrast to the parallel constituent in the antece- Intended: ‘Jo likes cake, and Li chocolate.’ dent clause. Thus the absence of pseudogapping The following attempts at stripping in Mandarin verifies to an extent the insight that pseudogap- also fail: ping is at least somewhat related to gapping,

272 enough so that if a language disallows gapping speakers in a main clause, the acceptability and stripping, then it will also disallow pseudo- judgments are less robust: gapping. (11) A: Jō xǐhuān mǒu gè rén. Jo likes certain CL person 2 Sluicing ‘Jo likes a certain person.’ Sluicing (Ross 1969, Merchant 2001) typically xǐhuān elides everything from a clause except an interrogative expression (wh-element), e.g. Jō Shuí ? (9) They are hiding something, but a. B: – Jō `xǐhuān Shuí? ? won’t b. B: – Jō Xǐhuān shuí? Jo likes who they say what (12) A: Lǐ zhèng cáng zài mǒu gè dìfang. Li now hide in certain CL place are ‘Li is now hiding in a certain place.’ they hiding g cáng 1 they won’t say what they are hiding. Lǐ zhèng zài The clause introduced by what is sluiced, that is, Nǎr the string they are hiding is elided. Sluicing is a ? frequently occurring type of ellipsis mechanism, a. B: – Lǐ zhèng cáng zài Nǎr? and it exists in most if not all Indo-European Li now hide in where languages. b. B: – ?Lǐ zhèng Cáng zài nǎr? Checking to see if sluicing exists in Mandarin, Li now hide in where the data are not entirely clear. Consider the fol- While there is a preference for the b-questions, in lowing examples: which the verb is repeated, the a-questions are (10) Tā xǐhuān mǒu gè rén, dàn not clearly bad. This situation clouds the picture, s/he likes certain CL person, but since the marginal a-questions look like the ‘S/he likes a certain person, but’ sluicing in direct questions that is frequent in zhīdào those languages that have sluicing. One might, however, assume that what has actually been wǒmen bù shì elided from the a-questions is the auxiliary shì shuí ‘be’. On such an account, such examples would, a. wǒmen bù zhīdào shì shuí. strictly speaking, not count as instances of sluic- we not know be who ing as it is commonly understood. ‘we don’t know who it is.’ Further data speak more clearly against the presence of sluicing in Mandarin. Cases of b. *wǒmen bù zhīdào shì shuí. so-called multiple sluicing are bad in Mandarin. we not know be who Multiple sluicing occurs when the sluiced clause Example (10a), in which the verb shì ‘be’ ap- contains two or more wh-remnants. The follow- pears, cannot, strictly speaking, be interpreted as ing example illustrates multiple sluicing in Eng- sluicing because sluicing typically elides the do- lish: minate verb in a clause. When the dominant verb (13) A: Somebody has a crush on somebody? is indeed elided (here shì), the result is bad, as illustrated with example (10b). This fact suggests has that sluicing is not present in Mandarin. Who crush Example (10b) is an attempt at sluicing in a a on subordinate clause. When sluicing occurs across whom

1 The hierarchical status of what as the root of the B: Who has a crush on whom? object clause, the dashed dependency edge, and the g The sluiced clause contains the two wh-remnants, subscript follow the approach to discontinuities pre- sented by Osborne (2014). The particularities of this who and on whom, identifying it as an instance of analysis are not relevant to the account of ellipsis. multiple sluicing.

273 Multiple sluicing is impossible in Mandarin: Checking to see whether comparative deletion is present in Mandarin is difficult to do. The (14) A: Yǒu rén xǐhuān mǒu gè rén. construction used to express comparison in exist person likes certain ge person Mandarin is of a much different nature than in ‘Somebody likes somebody.’ English. The elements being compared in Man- xǐhuān darin must be subjects, and the dimension along Shuí shuí which they are compared must appear as the main predicate, e.g. B: *Shuí xǐhuān shuí? ‘Who likes whom?’ (17) Diǎn-le píjiǔ de rén bǐ order-le beer de people than This attempt at multiple sluicing is quite bad. The example cannot be rendered in terms of the diǎn-le pútáojiǔ de (rén) gèng duō. verb shì, unlike examples (11a) and (12a). This order-le wine de people more many confirms that sluicing as it is commonly unders- 'More people ordered beer than ordered tood in English and related languages does not wine.’ exist in Mandarin. A number of accounts of sluicing-like data in The English translation employs a type of ad- Mandarin have acknowledged that what at times junct clause (than ordered wine) to express the looks like sluicing is in fact a different mechan- comparison, whereas its Mandarin counterpart ism, this mechanism being called pseudosluicing needs relative clauses (diǎn-le píjiǔ de ‘who or- (see for instance Wei 2004, and Adams and Ta- dered beer’ and diǎn-le pútáojiǔ de ‘who ordered mioka 2014). Pseudosluicing involves the aux- wine’) to express the comparison. iliary shì – but at times shì can be omitted. The Due to the quite different syntactic means for analysis of pseudosluicing put forth in the litera- expressing comparative meaning across the lan- ture (Adams and Tamioka 2014) is that it in- guages, it is difficult to acknowledge the pres- volves zero anaphora; a subject pronoun has been ence of comparative deletion in Mandarin. Given dropped, e.g. …wǒmen bù zhīdào (tā) shì shuí, lit. the lack of solid evidence in favor of the exis- ‘we not know it be who’ – more about zero ana- tence of comparative deletion, we conclude here phora below in Section 8. that it does not exist in Mandarin. The absence of sluicing in Mandarin is con- sistent with the absence of sluicing in wh-in-situ 4 VP-ellipsis languages in general (Merchant 2001: 84f.). VP ellipsis (Johnson 2001) occurs frequently in English. A non-finite verb phrase is elided, its 3 Comparative deletion content being retrieved from context, e.g. Comparative deletion (Bresnan 1975) elides a (18) have string of words that corresponds to focused ma- We visited terial in an antecedent clause, e.g. city (15) More men ordered beer than every have a. men ordered wine. b. *men ordered wine. they visited (16) We drank more beer than We have visited every city they have visited. a. they drank beer. Non-finite verb phrases consist of a non-finite b. *they drank beer. verb and all of its dependents. In this case here, These examples illustrate the manner in which just the nonfinite verb visited alone is elided be- men and beer must be elided. They must be cause it has no dependents. elided each time because their counterparts are VP-ellipsis occurs frequently in Mandarin as focused by the comparative element more in the well. As in English, it is typically introduced by a preceding clause. Thus comparative deletion oc- (modal) auxiliary verb. Li and Thompson curs obligatorily; it is unlike most other ellipsis (1981:182f.) classify the following verbs as aux- mechanisms in this regard, which occur option- iliaries: yīngaī ‘should’, yīngdāng ‘should’, gāi ally. ‘should’, néng ‘be able to’, nénggòu ‘be able to’, huì ‘be able to’, kěyǐ ‘be able to’, néng ‘be al- lowed to’, gǎn ‘dare’, kěn ‘be willing to’, děi

274 ‘must’, bìxū ‘must’, bìyào ‘must’, huì ‘will, Note that the English translations are unaccepta- know how to’. The next examples illustrate ble (because intend and want do not license VP-ellipsis in Mandarin: VP-ellipsis in English). Therefore what examples (21-22) illustrate is (19) Wáng yīngāi fàngsōng, that the elision of verb phrases is much less re- Wang should relax, stricted in Mandarin than in English. Apparently, yīngāi most any verb in Mandarin that takes a VP com- Lǐ yě fàngsōng plement can license VP-ellipsis, not just auxiliary verbs. Observe also that the elided material indi- Lǐ yě yīngāi fàngsōng. cated in each of the examples is a catena. Li also should relax ‘Wang should relax, and 5 Answer ellipsis Li should relax, too.’ The ellipsis mechanism associated with answer (20) Zhāngsān néng dú hànyǔ, fragments has been studied and debated in detail John can read Chinese (e.g. Morgan 1973, Merchant 2004). Answer ellipsis exists in Mandarin just as it does in English, néng although the questions that elicit answer frag- Mǎlì yě dú ments vary significantly from the questions in hànyǔ English insofar as all interrogative elements re- main in situ, i.e. they do not appear in Mǎlì yě néng dú hànyǔ. clause-initial position. Mandarin is a wh-in-situ Mary also can read Mandarin language in this regard. Despite this significant ‘John can read Chinese, and difference across English and Mandarin, Manda- Mary can read Chinese, too.’ rin has answer fragments that are similar to their counterparts in English. As in English, the an- These instances of ellipsis are closely similar to swer fragments in Mandarin are constituents (i.e. their English counterparts, as indicated with the complete subtrees), which means that the elided translations. VP-ellipsis therefore appears to be material has the status of a catena. quite similar across the two languages. The following examples illustrate the extent But while English and Mandarin both have to which the elided words of answer ellipsis in VP-ellipsis, the two languages differ in the fre- English are catenae: quency of the mechanism. VP-ellipsis occurs frequently in English, but is licensed by a rela- (23) Who are you waiting for? tively limited set of verbs, i.e. by auxiliary verbs am and the particle to. In Mandarin in contrast, I waiting VP-ellipsis occurs with auxiliary verbs as well as for with (what are designated in English as) control verbs. Thus VP-ellipsis is more widely available Susan in Mandarin than in English, e.g. I am waiting for Susan. (21) Wǒ xiǎng hē jiǔ, (24) Who is waiting for whom? I intend drink wine, is tā yě xiǎng hē jiǔ. Bill waiting s/he also intend drink wine for 'I intend to drink some wine; Susan *s/he also intends to drink some wine.’ Bill is waiting for Susan. (22) Tā yào chī fàn, The elided material in each of these two cases s/he wants eat meal has catena status, i.e. I am waiting for is a catena wǒ yě yào chī fàn. in (23), and is waiting is a catena in (24). I also want eat meal Switching to Mandarin, question-answer pairs in Mandarin also easily submit to analyses in ‘S/he wants to eat a meal; terms of catenae: *I also want to eat a meal.’

275 (25) Tā shénme shíhou lái? Interestingly, however, N-ellipsis is limited in s/he what time come English. It occurs mainly just with possessive ‘When is s/he coming?’ determiners/pronouns (mine, yours, his, hers, its, ours, theirs) and cardinal and ordinal numbers lái (one, two, three, first, second, third, etc.). It does Tā Míngtiān not occur with most adjectives, e.g. *his big cat and her small cat. – Tā Míngtiān lái. In many languages closely related to English, s/he Tomorrow come however, N-ellipsis is much more productive. ‘He is coming Tomorrow.’ For instance, most adjectives can introduce (26) Nǐ bàifǎng-le shuí? N-ellipsis in German: you visit-le who (29) ‘Who did you visit? a. seine große Katze und ihre kleine Katze bàifáng-le his big cat and her small cat Wǒ Zhāngsān b. billiges Bier und teures Bier Wǒ bàifǎng-le Zhāngsān. cheap beer and expensive beer I visit-le John ‘I visited John.’ c. alte Lieder und neue Lieder (27) Tā bǎ wǒ de hùzhào fàng zài nǎr? old songs and new songs s/he ba I de passport put in where 'Where did s/he put my passport?’ English has to reach to one in such cases. That is, when the adjective at hand cannot introduce le N-ellipsis in English, the pronominal count noun fàng one is employed instead to reduce redundancy, Tā zài lǐ e.g. old songs and new ones. Mandarin is more like those languages that kǒudài more freely employ N-ellipsis (such as German). -de Pre-modifiers of nouns are typically immediately Nǐ followed by the clitic de in Mandarin, this clitic serving as a marker of a pre-modifier: – Tā fàng zài Nǐ de kǒudài lǐ le. s/he put in you de pocket in le (30) Wǒ xǐhuān tā de gǒu, ‘S/he put it In your pocket.’ I like s/he de dog, Examples like these illustrate best the potential xǐhuān of the catena concept for serving as the basis for tā gǒu theories of ellipsis. In each of these Mandarin -de examples, the elided material is discontinuous in the linear dimension, yet despite this fact, it wǒ qualifies as a catena each time. When the frag- tā xǐhuān wǒ de gǒu. ment answer is a complete subtree, the elided s/he likes I de dog material is necessarily a catena. Despite the dras- tic differences in syntactic structures across the ‘I like her/his dog, s/he likes mine dog.’ English and Mandarin examples, the elided ma- (31) Tā hē guì de píjiǔ, dànshì terial is a catena in both languages. s/he drinks expensive de beer, but 6 N-ellipsis hē tā píjiǔ Noun ellipsis (N-ellipsis, also called NP-ellipsis -de or NPE) elides a noun and often additional material that is adjacent to the noun, e.g. piányi (28) tā hē piányi de píjiǔ. s/he drinks cheap de beer a. his old cat and hers old cat b. the first talk and the third talk ‘S/he drinks expensive beer, but s/he drinks c. their photos of me and ours photos of me cheap beer.’

276 When the noun phrase contains a classifier, The analysis here positions the classifier as a de- the de clitic is usually not employed, but rather pendent of the noun. This analysis may be con- the classifier alone introduces N-ellipsis: troversial, since an alternative analysis might position the classifier as head over the noun. As (32) Tā zuò-le dì yī liàng huǒchē, stated in the introduction, many aspects of Man- s/he took -le -st one CL train darin sentence structure have not yet been de- zuò-le bated in DG circles, so the analysis assumed here tā huǒchē is tentative. There is, however, one consideration that liǎng supports this preliminary analysis (i.e. the clas- èr sifier as a dependent of the noun). This consider- dì ation is the fact that the de marker can co-occur with the classifier, e.g. tā zuò-le dì èr liǎng huǒchē. ? s/he took -le -nd two CL train (34) Tā zuò-le dì yī liàng de huǒchē. s/he took -le -st one CL de train ‘S/he took the first train, and s/he took ‘She took the first train.’ the second train.’ While the co-occurrence of liàng and de is mild- (33) Tā tīngdào-le dì yī cì tánhuà, ly marginal, it is nevertheless good enough to s/he heard -de -st one CL talk, support the analysis shown in (32) and (33). The tīngdào-le de is serving its normal role as marker of a pre- tā tánhuà modifier, i.e. it helps identify dì yī liàng as a predependent of huǒchē. If huǒchē were a post- cì dependent of liàng, we would expect (34) to be èr bad, because in such a case, de would not be dì marking a pre-modifier of the noun. Otherwise, the clitic de occurs frequently and tā tīngdào-le dì èr cì tánhuà. in numerous varied environments. At times it s/he heard-le -nd two CL talk. even serves to nominalize clauses. When it does ‘S/he listened to the first talk, and s/he so, the result can at times be rendered with free listened to the second talk.’ relative clauses in the English translation, e.g.

(35) hǎo

dōngxī bǐ gèng .-de dōngxī xǐhuān -de Tā xǐhuān tā Tā xǐhuān de dōngxī bǐ tā xǐhuān de dōngxī gèng hǎo. s/he likes de things than s/he likes de things more good ‘What s/he likes is better than what s/he likes.

The two clauses what s/he likes in the transla- lish. The ability of de to serve as a nominalizer tion are free relative clauses. The clitic de serves makes N-ellipsis widely available. as a nominalizer in the second case, rendering the preceding clause a nominal. The noun 7 Null complement anaphora dōngxī ‘things’ can be interpreted as having been elided, as indicated in the tree. Null complement anaphora (Hankamer and Sag Many aspects of N-ellipsis in Mandarin are 1976, Depiante 2000) is a mechanism that elides not clear. The examples just produced suggest, a complement clause, to-phrase, or prepositional however, that N-ellipsis is a frequent occurrence phrase, e.g. in Mandarin, much more frequent than in Eng-

277 (36) Jim promised he would help, and edge VP-ellipsis as discussed above, and when the elided complement can be interpreted as a promised definite or indefinite noun phrase, an analysis in Bill also would terms of zero anaphora is available (see the next he help section). Thus the extent to which null complement anaphora is present in Mandarin is unclear. Bill also promised he would help. (37) Sam refuses to help, and 8 Zero anaphora

refuses Zero anaphora (Kroeger 2005: 79ff.) typically Sue also to involves a null definite or indefinite pronoun or help noun phrase. English and Mandarin vary significantly concerning zero anaphora; zero anahora Sue also refuses to help. occurs frequently in Mandarin, whereas its oc- The predicates that license null complement currence in English is, if it exists at all, highly anaphora in English (e.g. ask, know, promise, restricted. The difference across the two lan- refuse, try) are limited. Similar predicates that guages is illustrated well using the answer to a one might expect to also license null comple- yes-no question: both the subject and the object ment anaphora fail to do so (e.g. imagine, intend, can be absent from the Mandarin answer: pretend, say, think, etc.). (40) A: Nǐ yuèdú kèwén le ma? Examples from Mandarin similar to (36-37) you read text le ma also allow ellipsis: ‘Have you read the text?’ (38) A: Nǐ zhīdào fāshēng shénme le ma? le you know happened what le ma ‘Do you know what happened?’ Yuèdú Wǒ tā zhīdào Wǒ bù le B: Wǒ Yuèdú tā le. I read it le lit. ‘Have read.’ fāshēng shénme In contrast, the direct English translation of this example is quite bad: B: Wǒ bù zhīdào fāshēng shénme le. I not know happened what le (41) A: Have you read the text? ‘I don’t know what happened.’ Have (39) Tā jùjué líkāi, I read s/he refuses leave it jùjué B: *I Have read it. wǒmen yě líkái The acceptability contrast across the two lan- wǒmen yě jùjué líkái. guages is due to the unrestricted nature of zero we also refuse leave anaphora in Mandarin, whereas zero anaphora may not exist in English at all. ‘S/he refuses to leave, and we also Further examples suggesting that zero ana- refuse to leave.’ phora is highly restricted in, or absent from, These two examples suggest that the similar English are given next: predicates across the languages allow for the (42) a. *He saw me, and she saw me, too. ellipsis of a complement clause or phrase. b. He saw me, and she saw me, too. However, concluding that Mandarin has null complement anaphora in the same way that (43) a. *I study Mandarin, and she studies it, too. English does is difficult. The difficulty is due to b. I study Mandarin, and she studies it, too. the fact that Mandarin seems to freely allow the In contrast, the Mandarin equivalents of these ellipsis of most all complements that can be eas- a-sentences are fine: ily recovered from context. When the elided complement is a verb phrase, one can acknowl-

278 (44) Tā kàndào-le wǒ, Concerning the material that is elided, ellip- s/he saw-le me sis in Mandarin is like ellipsis in English insofar as the elided material is a catena. This aspect of tā yě kàndiào-le wǒ. ellipsis is especially evident with answer ellipsis, s/he also saw-le me which often elides non-string catenae. (45) Wǒ xuéxí hànyǔ, tā yě xuéxí tā. Finally, a comment about a possible genera- I study Chinese, s/he also studies it lization is in order. Four of the five ellipsis me- Furthermore, Mandarin even allows the absence chanisms that are not present in Mandarin (or of an indefinite noun phrase, i.e. what would be are highly restricted) involve the ellipsis of the equivalent to one in English: matrix predicate (gapping, stripping, pseudogapping, and sluicing). Mandarin hence seems (46) Tā xiě-le yī gè gùshì, in general to be less willing than English to elide s/he wrote one CL story the matrix predicate. On the other hand, it is tā yě xiě -le yī gè gùshì. much more willing to omit the arguments of s/he also wrote one CL story predicates (in terms of VP-ellipsis or zero anaphora). The reasons why these general differ- ‘S/he wrote a story, and s/he also ences across the languages exist is unknown, wrote one.’ however. The availability of zero anaphora in Mandarin means that Mandarin can omit most any subject References or object pronoun, noun, or noun phrase. In fact Peter Wang Adams and Satoshi Tomioka. 2012. its existence clouds the picture concerning other Sluicing in Mandarin Chinese: An instance of ellipsis mechanism. It is, for instance, difficult pseudo-sluicing. In Jason Merchant and Andrew to acknowledge VP-ellipsis and/or null com- Simpson (eds.), Sluicing: Crosslinguistic Pers- plement anaphora in Mandarin because what pectives, 219-47. Oxford University Press, Ox- looks like such ellipsis mechanisms may in fact ford, UK. be zero anaphora instead. Finally, whether or Ellen Barton. 1990. Nonsentential Constituents. not zero anaphora is a form of ellipsis is debata- John Benjamins, Philadelphia. ble. It seems, rather, to be the unmarked form of Joan Bresnan. 1975. Comparative deletion and con- anaphora in Mandarin. When tā ‘he/she/it’is or straints on transformations. Linguistic Analysis 1. some other proform is overt, it is in fact an em- phatic pronoun that serves a special discourse Johnny Cheng. 2011. Argument ellipsis in Chinese. rd North American Confe- role, namely that of emphasis. Proceedings of the 23 rence on Chinese Linguistics (NACCL-23), 224–240. 9 Concluding remarks Marcela Depiante 2000. The Syntax of Deep and This manuscript has surveyed ellipsis in Manda- Surface Anaphora: A Study of Null Complement rin. Gapping, stripping, pseudogapping, sluicing, Anaphora and Stripping/Bare Argument Ellipsis. and comparative deletion are either absent from Ph.D. thesis, University of Connecticut. Mandarin, or highly restricted. VP-ellipsis, an- Jorge Hankamer and Ivan Sag 1976. Deep and sur- swer ellipsis, N-ellipsis, and zero anaphora are face anaphora. Linguistic Inquiry 7, 3, 391–428. present in Mandarin. Whether null complement Ray Jackendoff. 1971. Gapping and related rules. anaphora is also present in Mandarin is unclear Linguistic Inquiry 2, 21–35. due to the overlap of the data in the area with Kyle Johnson. 2001. What VP ellipsis can do, and the data of VP-ellipsis and zero anaphora. Per- what it can’t, but not why. In Mark Baltin, M. and haps the most noteworthy difference in ellipsis C. Collins, The Handbook of Contemporary Syn- across English and Mandarin concerns the abil- tactic Theory, ed. 439–479. Blackwell Publishers, ity of Mandarin to omit complements and sub- Oxford. jects at will, as long as they can be easily re- Paul Koreger. 2005. Analyzing Grammar: An Intro- trieved from context. In contrast, English does duction. Cambridge University Press, New York. not elide complements (and subjects) so freely, but rather in order to do so, the requirements of Susumo Kuno 1976. Gapping: A functional analysis. VP-ellipsis, null complement anaphora, or some Linguistic Inquiry 7, 300–18. other ellipsis mechanism must be met. Nancy Levin. 1986. Main-Verb Ellipsis in Spoken English. Garland, New York.

279 Charles Li and Sandra Thompson. 1981. Mandarin Chinese: A Functional Reference Grammar. University of California Press, Berkeley. James McCawley. 1998. The Syntactic Phenomena of English. The University of Chicaog Press, Chi- cago. Jason Merchant. 2001. The Syntax of Silence. Oxford University Press, Oxford, UK. Jason Merchant. 2004. Fragments and ellipsis. Lin- guistics and Philosophy 27, 661–38. Jerry Morgan. 1973. Sentence fragments and the no- tion ‘sentence’. In Braj Kachruetal. (eds.), Issues in Linguistics, 719-751. University of Illinois Press, Urbana. Timothy Osborne. 2014. Type 2 rising: A contribution to a DG account of discontinuities. In Kim Gerdes, Eva Hajicová, and Leo Wanner, Depen- dency Linguistics: Recent Advances in Linguistic Theory Using Dependency Structures, 273–98. John Benjamins, Amsterdam. Timothy Osborne, Michael Putnam, and Thomas Groß. 2012. Catenae: Introducing a novel unit of syntactic analysis. Syntax 15, 4, 354-396. John Ross. 1969. Guess who? In R. Binnick, A. Da- vison, G. Green, and J. Morgan (eds.), Papers from the 5th Regional Meeting of the Chicago Linguistics Society. Chicago Linguistic Society, 252–86. Gregory Stump. 1977. Pseudogapping. Ms., Ohio State University. Ting-Chi Wei. 2004. Predication and Sluicing in Mandarin Chinese. Ph.D. dissertation, National Kaohsiung Normal University.

280