On the syntax and semantics of sluicing in Mandarin Chinese

2.2.1 The PF deletion approach

The PF deletion approach (e.g., Kim 1997; Lasnik 2001; Merchant 2001, 2004, 2006, 2008, 2013, 2015, 2019; Ross 1969; Sag 1976) states that sluicing is derived from overt movement of the wh-phrase followed by deletion of an IP at PF. Concretely, the wh-phrase is base-generated within an IP, from which it moves overtly to [Spec, CP] before spell-out. At PF, the complement of CP, namely, IP, is deleted, thus stranding the wh-phrase.

Merchant (2001) argues that the external structure of sluicing is an interrogative CP rather than an isolated wh-phrase serving as a complement to an embedding predicate. Furthermore, he presented four pieces of evidence to support his assumption that what appears to be a simple wh-phrase in isolation is actually a CP. These four pieces of evidence are the selection requirement of predicates, the case-matching requirement, the number agreement between a predicate and its subject, and the syntactic distribution of an NP and a CP.

Moreover, in the internal structure of sluicing, a wh-phrase undergoes overt movement from its base position to [Spec, CP] in order to check the strong [+WH] feature on C, and then the E feature, which is located in C, licenses the deletion of the complement of C, namely, IP, at PF, thus stranding the sluiced wh-phrase and deriving a well-formed sluicing.

2.2.2 The pseudo-sluicing approach

Other linguists disagree with Merchant’s (2001) claim that sluicing is derived from a full interrogative CP (e.g., Barros 2014; Erteschik-Shir 1977; Rodrigues et al. 2009; van Craenenbroeck 2010, 2012; Vicente 2008). They argue instead that sluicing like (10a) is not derived from (10b) but from an underlying structure in (10c) with deletion of the subject it and the copular is.

These linguists further posit that there is no “true sluicing” in English and that what appears to be “true sluicing” is actually derived from a cleft. The pseudo-sluicing approach seems to be reasonable because it conforms to the spirit of the Economy Condition, which requires that derivations and representations have to be minimal with no superfluous steps in derivations and no superfluous symbols in representations (Chomsky 1989), and it can also avoid the island repair issues which have been encountered by the PF deletion approach.

However, given that it is not always possible for the subject and the copular in a cleft to be deleted, the claim that sluicing is derived from an underlying structure of a cleft with deletion of the subject it, the copular and the presupposition part does not seem to hold water after all.

2.2.3 The LF copying approach

The LF copying approach (e.g., Chao 1987; Chung et al. 1995; Lobeck 1993, 1995) assumes that the wh-phrase in sluicing does not involve any overt movement but is instead base-generated in [Spec, CP]. Furthermore, there is no underlying structure following the wh-phrase throughout the entire derivation of sluicing. In other words, the base-generated wh-phrase is followed by an empty IP. However, at LF, a series of mechanisms is involved, namely, IP recycling, sprouting, and merger, in order to provide an appropriate question interpretation for sluicing. Therefore, the task of interpreting sluicing is to provide a question type meaning for it. If a legitimate LF can be constructed for sluicing, sluicing is interpretable and has an appropriate interpretation which is the same in all relevant respects as the interpretation of the corresponding constituent question. If sluicing is uninterpretable, the only reason might be that no legitimate LF can be constructed for it.

According to the LF copying approach, no wh-movement is involved in sluicing, and the approach can avoid the problems that the PF deletion approach has encountered. However, it is still defective, to some extent, because it fails to explain why the connectivity effect and preposition-stranding are observed in sluicing. Moreover, the LF copying approach has severed the connection between PF and LF, which are two closely-connected components after spell-out. Such a discrepancy is not expected under the PF deletion approach, because what has been sent to the PF component and the LF component is the same material, which is a full interrogative CP.

2.3.1 Movement approaches

Although Chinese is a wh-in-situ language (Huang 1982), wh-phrases could be fronted to the left periphery of a sentence in order to check other features rather than the [+WH] feature.

Accordingly, there are works in the literature (e.g., Chiu 2007; Murphy 2014; Wang 2002; Wang and Wu 2006) which posit that wh-phrases in Chinese sluicing undergo overt focus movement. In other words, it is the focus feature rather than the wh-feature that triggers overt movement of the wh-phrase. Furthermore, the Focus Phrase (FocP) is generated between CP and IP, and the copular verb shi ‘be’ is located in the head of FocP and is regarded as a focus marker.

It is observed that the presence of the copular verb shi ‘be’ is obligatory in front of wh-arguments, such as shei ‘who’ and shenme ‘what,’ while it is optional before wh-adjuncts, such as shenme shihou ‘what time.’ Wang and Wu (2006) attributed this asymmetry between wh-arguments and wh-adjuncts with respect to the presence of shi ‘be’ to a case requirement in that the presence of shi ‘be’ serves as a case-assigner for wh-arguments, and the absence of it is explained by the fact that wh-adjuncts do not need to be case-marked.

In addition to the focus movement approach, Fu (2014) proposed the topic movement account under the Cartographic Approach (Cinque 1999; Rizzi 1997). Fu (2014) agrees with what has been advocated by the focus movement approach, namely, that the overt movement of wh-phrases takes place in Chinese sluicing. However, deviating from the focus movement approach, she argues that the overt fronting of wh-phrases is triggered not by the [+WH] feature but by the [+Contrast/Aboutness] feature of the head of TopP. Furthermore, she classifies constructions with simplex wh-phrases, such as shenme ‘what’ and shei ‘who,’ into pseudo-sluicing, which involves a simple structure that contains an empty pronominal subject pro, the obligatory presence of the copular verb shi ‘be,’ and the wh-phrase.

2.3.2 The cleft and pseudo-cleft approaches

Kawabara (1996) and Kizu (1998) argue that sluicing in Chinese is derived from a cleft in which other constituents excluding the copula verb shi ‘be’ and the wh-phrase are deleted at PF, thus generating a well-formed sluicing.

However, sluicing in Chinese is different from a cleft in many ways. First, in a cleft, the subject can be focalized, but the object cannot. However, this asymmetry is not observed in sluicing because the wh-phrase can appear in both the subject and the object positions in sluicing.

Second, when the wh-phrase is an adjunct, the situation becomes even worse. According to the cleft approach, (11a) is assumed to be derived from (11b) with the deletion of the material on both sides of the sequence shi shenme shihou ‘be what time.’

Nonetheless, it is agreed that only what is constituent can be deleted (Jayaseelan 1990; Lasnik 1999; Merchant 2004). What is deleted in (11b) does not form a continuous constituent, which, accordingly, makes the deletion operation groundless. Moreover, Ince (2012) has claimed that no theory of ellipsis can naturally explain how an intermediate position in a string can be pronounced while phrases before and after this position are elided.

Third, since the presence of shi ‘be’ is obligatory in a cleft, the cleft approach is not able to explain the discrepancy between the obligatory presence of shi ‘be’ in sluicing before simplex wh-arguments and the optional one before wh-adjuncts.

2.3.3 The pseudo-sluicing approach

The pseudo-sluicing approach (Adams 2004; Adams and Satoshi 2012; Li and Wei 2014; Wei 2004, 2017) states that there is no true sluicing in Chinese. Instead, what appears to be true sluicing is in fact a pseudo-sluicing, which can be represented as a simple clause with an empty pro, a copular verb shi ‘be,’ and a wh-phrase. In addition, the pseudo-sluicing approach assumes that there is a pro subject in the predicative part of the embedded clause of pseudo-sluicing in Chinese. In spite of its implicitness, the pro subject can serve as a connection between the antecedent and the wh-phrase. Furthermore, according to the pseudo-sluicing approach, syntactic island effects are not observed. Since no wh-movement is involved in the analysis of sluicing in Chinese, no syntactic island has been crossed and hence no violation has been induced even though there does exist a syntactic island in the antecedent clause.

2.3.4 The LF copying approach

Liu (2006) argues that sluicing in Chinese should be analyzed via the LF copying approach. Specifically, in sluicing the wh-phrase in overt syntax is base-generated in [Spec, CP] following the embedding predicate in the second clause and it is followed by an empty IP. At LF, the operations of IP recycling, sprouting, and merger are involved to reconstruct the empty IP and provide a legitimate interpretation for sluicing. Therefore, the task of interpreting sluicing in Chinese is to derive a question type meaning for the interrogative CP. Liu (2006) follows Berman (1991) in claiming that a full constituent question is a tripartite construction, which includes three essential elements: interrogative operator, nuclear scope and restrictive clause. She divides sluicing in Chinese into two kinds: one with an overt correlate and the other with a covert one. As for the former, the operations of IP recycling and merger are required. And as for the latter, the operation of sprouting is also needed in addition to IP recycling and merger.

3.2.1 The syntactic constraints

Merchant (2001) posits that only those predicates that semantically select interrogatives and categorically select CPs allow sluiced wh-phrases in English. On the basis of the properties of predicates in sluicing, he argues that the external structure of sluicing in English is a CP.

Given the universality of natural languages, there are certain principles that all languages share. If predicates in English sluicing have the properties as mentioned, then we might assume that predicates in Chinese sluicing also have the same properties as their English counterparts do. Let us first figure out some of the predicates that frequently occur in Chinese sluicing, such as zhidao ‘know’ and jide ‘remember.’ These predicates, generally speaking, usually select an interrogative sentence as their complement when they are negated, as shown in (17).

All the predicates in the above examples take a complement CP. However, a distinction has to be made between a complement CP and a complement NP. Let’s take the predicate zhidao ‘know’ as an example.

The two sentences indicate that the predicate zhidao ‘know’ can take either an NP or a CP as its complement. However, the choice of a complement results in different interpretations of the relevant sentence. Specifically, when the predicate selects a complement NP zhege nanren ‘this man,’ it semantically selects an entity such that the entity refers to a particular person. When the predicate selects a complement CP zhege nanren shi shei ‘who this man is,’ it semantically selects the identity of a particular entity. Suppose there is a context where Mary met a man in the school library but most of her classmates do not know who this man is except John, who is asked the following question.

Since the expression ni ne ‘how about you’ questions whether John knows who this man is or not, a natural answer to (19) will be (18b) rather than (18a). In addition, John can even reply with a simpler answer, as in (20).

(18b) and (20) are semantically identical in that both convey the meaning of a state of an entity instead of an entity that is isolated from a previous linguistic context.

On the basis of the above, it is appropriate to claim that only those predicates which categorically select a complement CP and semantically convey a state of an entity or an event may appear in sluicing in Chinese.

3.2.2 The syntactic functions

The syntactic functions of sluicing are to convey a state of an entity or to describe an event. Let’s take (21) as an example.

It is indicated that the wh-phrase in (21a) is not an isolated constituent asking about which person but is part of a CP which describes the identity of this person such that Zhangsan hates this person. Likewise, the wh-phrase in (21b) does not inquire about any random reason but about the one for the event that Lisi broke up with Xiaohong. Therefore, the wh-phrases in the above sentences should have the same interpretation as a full interrogative CP.

Further supporting evidence comes from some constraints on the well-formedness of sluicing. The first constraint states that sluicing is only possible provided that there is an antecedent clause available, as in (22).

(22a) is a well-formed sluicing since the wh-phrase shenme shu ‘what book’ can locate its correlate yiben shu ‘a book’ in the antecedent clause, which functions as a presupposition and provides a linguistic context for the second part. Accordingly, (22a) is grammatical and acceptable and asks about the identity of the book that Xiaohong bought. However, in (22b), sluicing is out of the question simply because the sentence comes out of the blue and no linguistic context is available and the sentence is weird and unacceptable.

The second constraint concerns the relevance between the first part and the second part of sluicing. Sluicing is possible only when the two parts are semantically relevant.

In (23a), the first half is about a film and the second half is supposed to ask about the identity of the film that Xiaoming likes in order for sluicing to be possible. However, it inquires about the identity of a person. Thus, sluicing is ruled out. Likewise, in (23b), the first half expresses an event that Xiaohong bought an apartment and the second half might ask about the time, place, or reason why such an event has taken place. Nonetheless, the second half asks about the identity of an entity. Therefore, sluicing is also unavailable.

The two constraints on the availability of sluicing indicate that the functions of sluicing are related to the state or the identity of an entity or the description of an event. Moreover, sluicing should be interpreted in the same way as a full interrogative CP is.

3.5.1 The analysis of pseudo-sluicing

In terms of pseudo-sluicing, I follow the analysis proposed by Wei (2004, 2009, 2011, who argues that what appears to be sluicing in Chinese is actually pseudo-sluicing, which is a simple structure that is composed of an empty pro, a copular verb shi ‘be,’ and a wh-phrase. However, I also argue that only the two simplex wh-arguments, such as shei ‘who’ and shenme ‘what,’ and simplex wh-adjuncts that are associated with one of the attributes of the correlate in the antecedent clause, such as duo gui ‘how expensive,’ appear in pseudo-sluicing. Furthermore, in front of the two simplex wh-arguments, the presence of shi ‘be’ is obligatorily required. Nonetheless, in front of the simplex wh-adjuncts, the transitive verb you ‘have’ is used and its presence is optional since the following wh-adjuncts are intrinsically predicative. When you ‘have’ is present, it will have the same function as the dummy auxiliary do in English.

In (24a), since the wh-phrase shei ‘who’ is non-predicative, the presence of shi ‘be’ is obligatorily required in order to help it realize predication. Moreover, the empty pro is controlled by the indefinite nominal yiwei nansheng ‘a boy’ according to the Generalized Control Rule (Huang 1984). Accordingly, the second half of (24a) is interpreted as wo bu zhidao nage nansheng shi shei ‘I don’t know who that boy is.’ In a parallel fashion, in (24b), the presence of shi ‘be’ is also obligatory because the following wh-phrase shenme ‘what’ is non-predicative. The empty pro refers to the indefinite nominal yige liwu ‘a gift’ in the antecedent clause and the second half of (24b) is interpreted as wo bu qingchu nage liwu shi shenme ‘I am not clear what that gift is’.

Nonetheless, in the case of pseudo-sluicing that involves the auxiliary you ‘have,’ since the wh-phrase following you ‘have’ is predicative, the presence of you ‘have’ is optional.

In the above examples, the presence of you ‘have’ is optional since all the wh-phrases following it, namely, duo gui ‘how expensive,’ duo gao ‘how tall’ and duo weixian ‘how dangerous,’ are predicative. However, when you ‘have’ is present, it has no effect on the semantic interpretation of the relevant sentences but only adds an emphatic meaning to the following wh-phrase. In addition, since the empty pro has to be controlled by a nominal indefinite in the antecedent clause, the empty pro refers to the housing prices in (25a), the height of the younger sister in (25b), and the degree of danger in mountain-climbing in (25c) respectively. Therefore, it connects the wh-phrases with the attributes of the empty pros.

3.5.2 The analysis of typical sluicing

Apart from pseudo-sluicing, typical sluicing also exists in Chinese. Typical sluicing can occur not only in matrix clauses but also in embedded clauses. When it occurs in an embedded clause, the negation of the clause does not affect the grammaticality of sluicing. I posit that typical sluicing is closely related to focus. Specifically, complex wh-arguments undergo focus movement to [Spec, FocP] and wh-adjuncts are base-generated in [Spec, FocP].

Let us first discuss the derivation of typical sluicing with embedded clauses.

According to our hypothesis, in (26), the focus marker shi ‘be’ is base-generated in Foc and attracts the complex wh-phrase nawei nühai ‘which girl’ to move from the complement position of the predicate xihuan ‘love’ within IP to [Spec, FocP] in order to check the strong [+Foc] feature in a spec-head configuration. Shi ‘be’ can either move from Foc to Force or remain in Foc. If overt movement of shi ‘be’ takes place, the complement IP will be deleted at PF. This is because the Focus Condition on IP-ellipsis is already satisfied in the sense that the Focus-closure of IPE, which is represented as “Focus-clo(IP_E) = Ǝx. Xiaoming xihuan x ‘Xiaoming loves x,’” entails the Focus-closure of IP_A, which is expressed as “Focus-clo(IP_A) = Ǝx. Xiaoming xihuan x ‘Xiaoming loves x,’” and vice versa, thus deriving (26a) where shi ‘be’ appears in front of the wh-phrase. Furthermore, if shi ‘be’ stays in-situ, it will be deleted together with the complement IP at PF due to the Sluicing-Foc Assumption and the Extra Deletion Assumption, thus generating (26b), where shi ‘be’ is not present.

(27) is an example of contrast sluicing. The derivation of (27) is the same as that in (26), with the wh-phrase jizhi gou ‘how many dogs’ undergoing focus movement to [Spec, FocP] and with shi ‘be’ moving from Foc to Force or remaining in Foc. However, the Focus Condition on IP-ellipsis is satisfied a little bit differently than the one in (26). Concretely, the nominal expression wuzhi mao ‘five cats’ is regarded as a focus in the first half and jizhi gou ‘how many dogs’ is regarded as focus in the second half. Therefore, the Focus-closure of IPE can be represented as “Focus-clo(IP_E) = Ǝx. Xiaoming yangle x ‘Xiaoming raises x,’” and the Focus-closure of IP_A can be expressed as “Focus-clo(IP_A) = Ǝx. Xiaoming yangle x ‘Xiaoming raises x.’” Since the Focus-closure of IP_A and the Focus-closure of IP_E mutually entail each other, IP_E is licensed to be deleted at PF. With respect to shi ‘be,’ if it moves overtly, it will appear in front of the wh-phrase, as in (27a), but if it stays in-situ, it will be deleted along with its complement IP, as in (27b).

(28) is a typical sluicing that occurs in an embedded clause that does not involve negation. The derivation of (28) is the same as that of (26) and (27), where the complex wh-argument nabu dianying ‘which film’ has undergone focus movement to [Spec, FocP] and shi ‘be’ has moved from Foc to Force in (28a) but stayed in-situ in (28b), thus accounting for its presence and absence when the deletion operation takes place at PF.

Nonetheless, in (29) and (30), the simplex wh-adjunct shenme shihou ‘what time’ and the complex wh-adjunct yong shenme fangfa ‘with what methods’ are assumed to be base-generated in [Spec, FocP] and are associated with their implicit correlates, namely, mouge shihou ‘some time’ and yong mouzhong fangfa ‘with a certain method,’ in the antecedent clauses respectively. Likewise, shi ‘be’ can remain in-situ or move from Foc to Force. Since the Focus Condition on IP-ellipsis is satisfied in (29) and (30) with IP_A and IP_E mutually entailing the Focus-closure of each other, IP_E is licensed to be deleted at PF. Accordingly, in (29a) and (30a), when shi ‘be’ has undergone head movement, it survives the deletion operation and appears in front of the wh-phrases, while in (29b) and (30b), where shi ‘be’ remains in-situ, it is deleted along with the complement IP.

So far, we have discussed that the two simplex wh-arguments, shei ‘who’ and shenme ‘what,’ and the simplex wh-adjuncts that are associated with one of the attributes of the correlate in the antecedent clause appear in pseudo-sluicing. Moreover, complex wh-arguments, complex wh-adjuncts, and simplex wh-adjuncts appear in typical sluicing. However, there is one exception. It is observed that the simplex wh-adjunct zenmeyang ‘how’ cannot occur in either typical sluicing or pseudo-sluicing.

It is indicated in (31a) and (31b) that the simplex wh-adjunct zenmeyang ‘how’ can appear neither in typical sluicing nor in pseudo-sluicing. In contrast, (32a) and (32b) have the same interpretation as that conveyed in (34) by attaching the wh-adjunct zenmeyang ‘how’ to the VP tongguo le mianshi ‘pass the interview’ and by converting the wh-adjunct into a PP yong shenme fangfa ‘with what methods’ respectively. The ungrammaticality of zenmeyang ‘how’ in typical sluicing can be explained by our analysis. According to our assumption that wh-phrases are regarded as focus and they either undergo focus movement to [Spec, FocP] or are base-generated in [Spec, FocP], it is impossible for the wh-adjunct zenmeyang ‘how’ to be focused in Chinese, as in (33).

It is shown in (33a) that zenmeyang ‘how’ can stay in-situ and be preceded by the focus marker shi ‘be.’ However, the ungrammaticality of (33b) indicates that zenemeyang ‘how’ cannot be preposed and occur in a focus position at the left periphery of an IP. Therefore, zenmeyang ‘how’ cannot occur in typical sluicing where a wh-phrase is assumed to undergo a focus movement and occur in a focus position. Furthermore, the ungrammaticality of zenmeyang ‘how’ in pseudo-sluicing, as in (33b), is due to the fact that the empty pro cannot locate a possible correlate in the antecedent clause and therefore no connection can be formed between the correlate, the empty pro, and the wh-adjunct. Therefore, the wh-adjunct zenmeyang ‘how’ can occur neither in typical sluicing nor in pseudo-sluicing.

With respect to typical sluicing with matrix clauses, Merchant (2004) argues that fragmentary utterances have fully sentential syntactic structures and are subject to ellipsis. Therefore, (34a) is supposed to be derived from (34b), with the overt movement of the NP and the deletion of the constituents following it.

In (34b), the NP Mary has moved from the complement of the predicate love to [Spec, FocP] with a subsequent deletion of IP at PF, thus generating the fragmentary sentence.

On the basis of the approach adopted by Merchant to deal with fragmentary utterances, I also assume that in typical sluicing that appears in matrix clauses, the wh-phrase does not stand alone but is derived from a full-fledged syntactic structure which involves movement and deletion.

In (35), the complex wh-argument nawei nühai ‘which girl’ seems to be an independent fragment, but, according to our assumption, it is part of a fully sentential structure out of which the wh-argument has moved with subsequent deletion of the rest of constituent at PF. Specifically, shi ‘be’ is base-generated in Foc and attracts the wh-argument to undergo focus movement to [Spec, FocP]. Besides, shi ‘be’ can either stay in Foc or move from Foc to Force. Therefore, shi ‘be’ will be deleted along with IP if it stays in-situ, as in (35a), while it survives deletion and appears in front of the wh-argument if it has undergone movement, as in (35b).

In a parallel fashion, in (36), the wh-adjunct shenme difang ‘what place’ is not an isolated element. We assume that it is base-generated in [Spec, FocP] followed by an IP. Shi ‘be’ is base-generated in Foc and can move further to Force or remain in-situ. At PF, the IP, together with whatever is in Foc, is deleted, thus producing the well-formed syntactic structure in (36).