Chapter 3 Properties and Previous Analyses of the Chinese RNR
3.2 Two previous works on Chinese RNR
3.2.1 Cheng (2012)
Cheng argues for the MD approach based on ACD, licensing of reciprocals bici ‗each other‘ and plural marker -men. He first claims that ACD24 in Chinese RNR poses a problem
24 According to May (1985), VP-deletion is licensed by an identity relation that holds at LF and only QR generates a well-formed LF representation that licenses ACD. For instance, the quantificational DP every book
80
for the ATB-movement approach. Consider (46), where the RNR target is Wangwu song de
dongxi ‗the thing that Wangwu sent‘. Only the strict reading (i.e. the empty pronoun pro
which co-indexes with its antecedent), not the sloppy reading, is allowed, indicating that the overt movement does not occur. That is, the movement approach would wrongly predict the availability of the sloppy reading in (46).(46) Zhangsan yuanyi song ziji-de xiaohai, danshi Lisi bu yuanyi Zhangsan willing.to send self-GEN child but Lisi not willing.to song ziji-de xiaohai [NP Wangwu song e de dongxi]
send self-GEN child Wangwu send DE thing
‗lit. Zhangsan is willing to send self‘s child, but Lisi is not willing to send self‘s child the thing that Wangwu sent.‘ (strict *sloppy) (Cheng 2012)
= ‗Zhangsan is willing to send Zhangsan‘s child the thing that Wangwu sent Zhangsan’s
child, but Lisi is not willing to send Lisi‘s child the thing Wangwu sent Lisi’s child.‘
≠‗Zhangsan‘ is willing to send Zhangsan‘s child the thing that Wangwu sent Wangwu‘s
that Bill did in (i) has been raised to an adjoined position which dominates VP. Thus, the deleted VP2 is not contained within its antecedent VP1 as in (ii) and the problem of an infinite and uninterpretable VP is solved.
The well-formed LF representation is shown in (iii): John bought every book that Bill bought.
(i) John [VP1 bought every book that Bill did [VP2 e]]
(ii) [IP [DP every book that Bill did [VP2 e]]i [IP John [VP1 bought ei]]]
(iii) [IP [DP every book that Bill did [VP2 bought ei]]i [IP John [VP1 bought ei]]]
Just as the above English ACD sentences, Cheng (2012) observes that Mandarin Chinese counterpart in (iv) allows only a strict reading, too. Since VP2 is contained in VP1, direct copying of VP1 to VP2 is unacceptable, blocking the sloppy reading of the e. Yet, we can insert an empty pronoun pro to the position of e which is co-indexed with ziji-de xiaohai ‗self‘s child‘, deriving the strict reading. Note that if the DP is preposed to the sentence initial position, both strict and sloppy readings are available, as in (v). Also note that English allows covert movement in syntax while Chinese needs overt movement in syntax to solve the problem of regression in ACD.
(iv) Zhangsan [vP1song [VP1ziji-de xiaohai Lisi [vP2 song [VP2 e de dongxi]
Zhangsan send self-GEN child Lisi send DE thing
‗Zhangsan sent his child the thing that Lisi sent Zhangsan‘s child.‘ (strict *sloppy) (Cheng 2012) (v) Lisi song e de dongxi, Zhangsan (ye) song ziji-de xiaohai
Lisi send DE thing Zhangsan also sent self-GEN child
‗lit. The thing that Lisi sent, Zhangsan also sent self‘s child‘ (strict sloppy) (Cheng2012)
81
child, but Lisi is not willing to send Lisi‘s child the thing Wangwu sent Wangwu‘s child.
However, the ellipsis in (46) involves a DP ziji-de xiaohai ‗self‘s child‘, different from VPs or quantifiers usually found in ACD (see Soh 2003). In other words, (46) is not a legitimate instance of ACD and thus cannot be used to argue against the ATB-movement approach.
Cheng‘s second and third claims are against the PF-deletion approach. Just as Otaki‘s (2008) observation in Japanese RNR,25 bici ‗each other‘ in Chinese RNR cannot be licensed in each conjunct as the contrast between (47a) and (47b).
(47) a. Zhangsan yong shou er Lisi yong qiubang da-le bici
Zhangsan with hand while Lisi with bat hit-Asp each.other
‗Zhangsan hit Lisi with hand, while Lisi hit Zhangsan with a bat.‘ (Cheng 2012) b.* Zhangsan yong shou da-le
bici
er Lisi yong qiubang25 Otaki (2008, 2011) claims that there is an underlying plurality in the RNR target: the elements α and β that occupy the same syntactic position in different RNR conjuncts are interpreted as if they form a conjunction „α and β‟ underlyingly (c.f. Moltmann 1992). For example, the reciprocal otagai is licensed in (i) without a plural antecedent in each conjunct, showing that Masa and Ken form a conjunction ‗Masa and Ken‘ at certain level of representation. The reciprocal cannot be licensed when it is pronounced in each conjunct as in (ii), against the PF-deletion approach.
(i) Masai-wa te-de, (sosite) Kenj-wa batto-de otagaii+j-o nagut-ta Masa-Top hand-with and Ken-Top bat-with each.other-Acc hit-past
‗lit. Masa by hand, and Ken hit each other with a bat.‘ (Otaki 2011)
82
Besides, similar to the dummy plural marker (DPM)-tul in Korea (Chung 2004),26 the plural marker -men in Chinese can be licensed in RNR. Tamen ‗they‘ can refer to Zhangsan and Lisi in the well-formed (48a) but it cannot do so in the ill-formed (48b). Such a contrast is not captured under the PF-deletion approach.
(48) a. Zhangsan renwei, danshi Lisi que bu renwei, tamen
hen congming
Zhangsan think but Lisi whereas not think they very smart‗lit. Zhangsan thinks, but Lisi does not think, that they (Zhangsan and Lisi) are very
smart.‘ (Cheng 2012)
b. *Zhangsani renwei
tamen
ihen congming, danshi Lisi que
bu Zhangsan think they very smart but Lisi whereas not renweitamen hen congming
27think they very smart
‗Zhangsani thinks that they(Zhangsan and Lisi) are very smart but Lisi doesn‘t think that they (Zhangsan and Lisi) are smart.‘ (Cheng 2012)
However, we do not think that (48b) is legitimate evidence against the PF-deletion approach.
This is because independently, tamen ‗they‘ can refer to the people and Zhangsan is one of them as in (49).
26 Chung‘s (2004) examples are given below.
(i) John-un nonmwun-ul yelsimhi(*-tul) ilk-ess-ta John-Top article-Acc hard-DPM read-Past-DE
‗John read articles hard‘
(ii) John-un nonmwun-ul kuliko Mart-nun chayk-ul yelsimhi(-tul) John-Top article-Acc and Mary-Top book-Acc hard-DPM Ikl-ess-ta
Read-Past-de
‗John read articles and Mary read books hard.‘
27 Cheng (2012) notes that when the pronoun tamen ‗they‘ does not refer to Zhangsan, this example is well-formed, meaning ‗Zhangsan thinks that they (some people salient in the discourse) are very smart‘.
83
(49) Zhangsan rewei tamen hen congming Zhangsan think they very smart.
‗Zhangsan think that they are very smart.‘
Seen in (49), that tamen ‗they‘ can be licensed is in fact irrelevant to the property of Chinese RNR, and therefore cannot be as an argument to abandon the PF-deletion approach.
Cheng‘s proposal, following Otaki‘s (2008, 2011) null hypothesis (cf. Grosz 2007),28 argues for an MD approach to the licensing of relational modifiers.
(50) a. Zhangsan yong shou er Lisi yong qiubang da-le bici
Zhangsan with hand while Lisi with bat hit-Asp each.other
‗Zhangsan hit Lisi with hand, while Lisi hit Zhangsan with a bat.‘ (Cheng 2012) b.
(Cheng 2012)
According to Cheng, in (50a), the two singular DPs Zhangsan and Lisi first form a
28 Otaki (2008, 2011) argues that in the derivation of (i), ‗Masa and Ken‘ form a plural DP under a null &, denoting plurality and binding the reciprocal otagai.
(i) Masai-wa te-de, (sosite) Kenj-wa batto-de otagaii+j-o nagut-ta Masa-Top hand-with and Ken-Top bat-with each.other-Acc hit-past
‗lit. Masa by hand, and Ken hit each other with a bat.‘ (Otaki 2011)
84
constituent under the null &, denoting plurality and binding the reciprocal bici ‗each other‘.
And then the conjoined DP undergoes movement to distinct Spec of TPs to meet the
ECP/Case requirement. Finally, two rooted TPs are combined by er ‗while‘.29 However, we doubt the existence of a null &, which contains a constituent formed by the two subjects DPs in each conjunct. Specifically, we find a similar pattern between the Chinese RNR and the English RNR with respect to Grosz‘s (2014) anticollectivity: predicates or reciprocals which require a plural antecedent (e.g., meet, outnumber, each other) are unacceptable in the configurations as shown in (51).
(51) a.*[Sue‘s proud that Billi __ ] and [Mary‘s glad that Johnj ___ ] have finally ti/j met.
(Grosz 2014) b. [Sue‘s proud that___ ] and [Mary‘s glad___ ] that Bill and John have finally met.
(Grosz 2014)
According to Grosz (2014), the plurality formed by two singular DPs is not in syntax as in (51a) or at LF; instead, it is on the verbal agreement as in (51b). The structure is illustrated in (52), where the two singular DPs in different conjuncts show the simultaneously cumulative agreement with the plural predicate.
29 Cheng summarizes the mechanism of the MD approach as below:
a. &P2: ―Zhangsan, &2, Lisi‖ external merge b. VP: ―hit, each other‖ external merge
In addition, Cheng notes two potential problems for the analysis. One, if the two singular DPs are separate conjuncts of the coordinate structures, the CSC is violated. He assumes that the two singular DPs undergo movement at the same time, just like ATB-movement. Or, he hypothesizes that the CSC is assumed to be a PF phenomenon. Since the conjunction is null at PF-level, the CSC violation is avoided. Another problem has to do with the nature of Mandarin Chinese—sensitivity to island effects.
85
(52) […Xα[SG] ____ ] and […Zβ[SG]… _____ ] F[PL] tα/β Q (Grosz 2014)
Agreement
Likewise, the contrast can be seen in the Chinese examples (53) and (54).
(53) *[Zhangsan zhidao Xiaoming] er [Mali ye zhidao Xiaomei] hehao-le Zhangsan know Xiaoming and Mary also know Xiaomei reconcile-Asp
‗Zhangsan knows that Xiaoming and Mary also knows that Xiaomei are reconciled.‘
(54) [Zhangsan zhidao] er [Mali ye zhidao] Xiaoming han Xiaomei hehao-le Zhangsan know and Mary also know Xiaoming and Xiaomei reconcile-Asp ‗Zhangsan knows and Mary also knows that Xiangmin and Xiaomei are reconciled.‘
In the ill-formed (53), Xiaoming and Xiaomei are the subjects of each conjunct respectively and cannot license hehao ‗reconcile‘. On the other hand, in the well-formed (54), Xiaoming and Xiaomei are the subject of the same clause as hehao ‗reconcile‘ and can license hehao
‗reconcile‘. The contrast between (53) and (54) indicates that reciprocals must be c-commanded by a plural subject in the local domain. Regarding these two examples, Cheng‘s (2012) null& analysis would wrongly predict the well-formedness of (53) since
Xiaoming and Xiaomei form a plural antecedent under the null &.
Moreover, unlike island insensitivity in English RNR, Cheng notes that Chinese RNR shows island effects. See the comparison in (55).
(55) a. Zhangsan xihuan danshi Lisi bu xihuan [zhe-ben shu]
Zhangsan like but Lisi not like this-Cl book
‗Zhangsan likes, but Lisi doesn‘t like this book.‘ (Cheng 2012)
86
b. *Zhangsan renshi [ yi-ge [ t1 mai-le t2] de nuhai1 ] er Lisi renshi Zhangsan know one-Cl buy-asp DE girl and Lisi know [san-ge [ t3 du-le t2 ] de nanhai3 [ zhe-ben shu]2
three-CL read-asp DE boy this-Cl book
‗Zhangsan knows a girl who bought, and Lisi knows three boys who read this book‘
(Cheng 2012)
Sentence (55a) is grammatical whereas (55b), in which the rightward extraction out of the coordinate structure, is ungrammatical, indicating the relative clause island effects. However, the ungrammaticality of (55b) is not due to island effects as Cheng claims; rather, it is due to the violation of RER. The RNR target zhe-ben shu ‗this book‘ is not in the rightmost position of each conjunct at first place. Relative clauses cannot be used to test the island effects in Chinese RNR since such clauses in Chinese are head-final. In other words, it is not possible to extract an element out of the conjunct without violating the RER.
Summed up in Table 10, Cheng‘s analysis explains the licensing of reciprocals (i.e. bici
‗each other‘). Nevertheless, he does not discuss how the MD approach accounts for other properties of Chinese RNR. If he assumes the MD approach as those in the literature, the following properties may possibly be accounted for: RER, non-constituency, CSC, and RRC violation. Still, it is unclear how other characteristics are captured under the MD approach:
asymmetric island sensitivity, asymmetric P-stranding, sloppy reading, and asymmetric NPI licensing.
87
Table 10 Properties explained by Cheng’s MD approach
1. Right Edge Restriction (RER) 2. Non-constituency
3. Coordinate Structure Constraint (CSC) 4. Asymmetric island sensitivity ?
5. Asymmetric P-stranding ?
6. Strict/sloppy readings ?
7. Asymmetric licensing of NPI ? 8. Licensing of relational modifiers