Chapter Two
Literature Review
Mandarin Chinese has been an important language in the study of empty
categories (EC). Huang (1982) firstly identified three different types of ECs in Mandarin, namely, NP-traces, variables and Pros.1 Unfortunately, despite great contributions and insights from a vast literature, there are yet satisfactory analyses to answer the issues, such as what occupies the null object positions and how they are interpreted, and to answer them in a coherent manner.
Before presenting the proposal, we will first survey several important approaches to the null object construction (NOC): variable analysis (Huang 1982, 1984, 1987, Ai 2006), null epithet analysis (Huang 1991), Free Empty Category analysis (Xu 1986), VP/vP-Ellipsis analysis (Huang 1989, 1990, Hsu 2002), small pro analysis (Ai 2006), and True Empty Position analysis (Li 2007a, b). We will show that these analyses are still inadequate in some respects to accommodate the whole range of facts in terms of referential NOs. We have mentioned in Chapter One that none of the previous works discuss nonreferential NOs; therefore, this literature review involves only referential NOs in Chinese. In the following Chapter Three, we propose a modular account on NOC in Chinese to include both referential and nonreferential NOs.
2.1 Variable Approach
It is well known that Chinese being a topic-prominent language allows the object
1 Huang uses Pro to include pro and PRO. Because the distinction between pro and PRO is related to the issue of Case, which is not distinctive in Chinese, Huang disregards the distinction between the two and uses the term Pro to refer to both. However, the focus of this thesis is the EC in an object position, which is a cased position. Therefore, we will use pro instead of Pro throughout this thesis.
to be empty. Huang (1982) firstly states that when a topic NP appears in addition to a matrix subject, the NO is most naturally interpreted as bound by that topic as example (1) and (2) show.
(1) Neige reni, Zhangsan kanjian ei le.
that man Zhangsan see LE ‘That man, Zhangsan has seen.’
(2) Neige reni, Zhangsan shuo Lisi kanjian ei le.
that man Zhangsan say Lisi see LE ‘That man, Zhangsan said Lisi has seen.’
These NOs are variables which are derived by topicalization and are sensitive to island conditions. Huang provides evidence for variable analysis which can be derived from Subjacency and binding reconstruction effects. Subjacency predicts that if there is a relative island, topicalized NP will observe island effects. This is indeed borne out.
The unacceptability of (3) indicates that relative island effects are in fact observed in Chinese.
(3) *Li xiaojie, wo xiang renshi [[yuanyi qu e ] de nanren]
Li Miss I want know willing marry DE man ‘Miss Li, I want to know the man who is willing to marry (her).’
The second piece of evidence for a variable NO provided by Huang (1993) is that fronted predicates will obligatorily reconstruct. A reflexive contained in a fronted argument, as in (4), has to reconstruct to its original position since the reflexive has to be interpreted with the embedded subject to satisfy Binding Principle A. Sentences (4), in which the compound reflexive taziji is interpreted with the embedded subject,
illustrate this point.
(4) [ ta ziji ?*i/j/*k de wenzhang]g, Zhangsani renwei Lisij dou piping-guo tg. he self DE article Zhangsan think Lisi all criticize –Asp ‘His*i/j own article, Zhangsani thinks Lisij has made criticism.’
(Shyu 1995, (12))
Further embedding the referential expression in topic still observes reconstruction effects, as shown in (5):
(5) *Zhangsani gege de zhaopian, tai renwei ni dou hui xihuan (Shyu 1995, (19)) Zhangsan brother DE picture he think you all will like
‘Pictures of Zhangsan’s brother, he thinks you also will like.’
If we reconstruct the topic to its original position, it is clear why (5) is ungrammatical, due to the violation of Binding Principle C.
On the basis of topic construction, Huang further argues for the existence of null topics. According to Tsao (1977), the null topic is deleted under identity with a topic in a preceding sentence across discourse. The result of the deletion process is a “topic chain”. The e in (6) refers to someone whose reference is understood outside of the sentence in a discourse situation.
(6) Zhangsani shuo Lisij bu renshi e*i/*j. (Huang 1987, (19)) Zhangsan say Lisi not know
‘Zhangsan said that Lisi does not know (him).’
Since e is a variable, it is subject to Principle C which requires e not to be coindexed with the matrix subject.
Ai (2006) also recognizes and fortifies the existence of variable objects as one type of NOs which Ai identifies. According to Ai, variable NOs occur in the weak pragmatic control, as defined in (7) and exemplified in (8):
(7) Weak control: under which there is a salient linguistic topic, which is introduced by background utterances.
(8) [Zhangsan drives home in his new BMW]. (Ai 2006, (43)) Lisi [to his wife]: Qiao! Xinche-ai! Luse-er de!
look new-car green-color DE ‘Look! A new car! (It is) green!’
Wo yi-dian-er dou bu xihuan [NP Ø]
I one bit all not like ‘I do not like (it) at all.’
(it = Zhangsan’s new BMW.)
Under Ai’s analysis, the NOs in the weak pragmatic control are variables because the salient linguistic topic makes it possible for the reference of the NO to be recovered. In other words, the linguistic topic serves as a kind of NP antecedent for the NO. Therefore, the reference of the NO can be recovered by being co-indexed with the null topic, so the NO under weak pragmatic control is actually a variable as Huang (1984) argues for. Ai’s proposal of weak pragmatic control and variable NOs within the weak control does not a bit differ from Huang’s proposal of variable NOs.
Since variable NOs in a topic construction is well studied and widely accepted, this thesis will also assume a type of NOs as variables, which are traces left after the A’-movement.
Notice that in Lasnik and Stowell's (1991) study on A’-movement, when the A'-binder is referential, they call the EC a null epithet, instead of a variable. That is to say, if we adopt Lasnik and Stowell’s system, we may say that the topic construction in Chinese involves null epithets instead of variables. Nevertheless, we will use the term variable to cover the cases where the A’-trace is bound by a referential A’-binder.
However, one might question the null epithet analysis to Chinese NOs, since NOs are not subject to Strong Crossover (SCO),2 unlike the null epithet in Lasnik and Stowell’s, which is subject to SCO but not to Weak Crossover (WCO).3 Example (9) demonstrates that English epithets are subject to SCO but not to WCO.
(9) SCO
a. *Johni, I think hei told Mary to visit ei. (Lasnik and Stowell, (67c))
WCO
b. This booki, I expect itsi author to buy ei. (Lasnik and Stowell, (20c))
2 Strong Crossover is treated as a violation of Condition C of the Binding theory. Chomsky assumes that variables are treated as R-expressions with respect to condition C. In a SCO configuration like (ii),
(i) * whoi did you say hei made you visit ti
Condition C is violated if the pronoun is coindexed with the wh-phrase and its trace, since the pronoun A-binds the variable trace. Hence, the pronoun must have a distinct index to satisfy Condition C, thus excluding the bound construal.
3 Weak crossover is named by Wasow (1972) and can be exemplified by the following example.
(ii) *whoi did hisi mother see ti
Who is moved across his in (i). The fact that who is moved across his is referred to as 'weak crossover', because the ungrammaticality is not due to a Condition C violation as in the case of strong crossover.
If we assume null epithets for Chinese NOs, we expect the same contrast in (9) to be observed in Chinese cases. However, Xu (1986) points out that Chinese NOs are not subject to SCO, as (10) shows, while we do observe that Chinese NOs are not subject to WCO as expected.
(10) SCO
a. Zhangsani, tai shuo ni bu ken bang ei
Zhangsan he say you not willing to help
‘Zhangsan, he said you are not willing to help (him).’
b. WCO
zhe ge reni, wo jide tadei pengyou yijing da dianhua yaoqing ei le this CL man I remember his friend already make call invite LE ‘This man, I remember his friends have already called to invite (him).’
We believe that (10a) cannot be a counterexample, and the NO is indeed a trace of the A’-movement; therefore, we expect (10a) to show SCO effects, which is contra to the fact. However, in Chapter Three, we will argue for another derivation, derived by a covert pro. It is this second derivation that rescues (10a) from the SCO effects.
To conclude, this thesis agrees with Lasnik and Stowell on their null epithet analysis when the A' binder in an A’-movement is referential. However, as mentioned above, we will use the term variable to cover the cases where the A’-trace is bound by a referential A’-binder.
2.2 Null Epithet Approach
Huang (1991) assumes Lasnik’ (1991) theory of English (anaphoric) epithets, and argues that Chinese NOs may be considered to be the null counterpart of an epithet.
This proposal eliminates the need for a variable to be A’-bound by a topic (the variable approach). To argue for a null epithet object in Chinese, Huang shows that the NO shares important properties with the anaphoric epithet: (i) it may not be A-bound, (ii) it may be A’-bound, (iii) it need not be A’-bound for it can be coindexed with an argument as long as the argument does not c-command them. (11) illustrates these properties, respectively.
(11) a. *Zhangsan i yiwei [zhege bendan i hen youqian] (Huang 1991, (19a)) Zhangsan think this idiot very rich
‘*Zhangsan i thought this idiot i was very rich.’
b. Zhangsan i, wo bu xihuan zhege wangbadani. (Huang 1991, (19b)) Zhangsan I not like this bastard
‘Zhangsan i, I don’t like this bastard i.’
c. ruguo ni xiang jian Zhangsani, wo jiu pai ren qu zhao zhe xiaozii. if you want see Zhangsan I then send man go look this sissy ‘If you want to see Zhangsan, I will send someone to look for the sissy.’
(Huang 1991, (23))
NOs in Chinese have all three properties of an epithet as (12) shows.
(12) a. Zhangsani shuo Lisi bu renshi e*i/j. Zhangsan say Lisi not know
‘Zhangsan said that Lisi does not know (him).’
b. neige reni, Zhangsan shuo Lisi bu renshi ei. (Huang 1991, (21)) that man Zhangsan say Lisi not know
‘That man, Zhangsan said Lisi does not know.’
c. Ruguo ni bu xihuan zheben shui, jiu qing bie mai ei. if you not like this book then please don’t buy ‘If you don’t like this book, then please don’t buy (it).’
Notice that Huang’s null epithet approach is different from Lasnik and Stowell’s (1991) proposal. Huang uses the parallelism between the overt epithets and NOs as arguments, instead of the contrast between SCO and WCO in Lasnik and Stowell’s.
However, Huang’s using overt epithets as arguments is proved to be incorrect. Li (2002) and Liu (2004) point out that there are sentences where overt epithets may occur but NOs may not.
Furthermore, to argue (12c), where the antecedents are in an adjunct clause, involves a null epithet might also be problematic. As we have reviewed in the previous section, Lasnik and Stowell (1991) adopt null epithets only when the A’-binder is referential, but not when the antecedents are in an adjunct clause. Our suspicion turns out to be correct when examining the contrast of SCO and WCO. If sentences like (12c) indeed involve a null epithet, it is expected that these sentences are subject to SCO but not to WCO, which is contra to the fact.
(13) SCO
a. ruguo ni bu xihuan zhe ge reni dehua, wo jide tai shuo-guo ni jiu buyao if you not like this CL man I remember he say-Asp you then not
da dianhua yaoqing ei le make call invite LE
‘If you don’t like this man, I remember that he said you then don’t call to invite (him).’
WCO
b. ruguo ni bu xihuan zhe ge reni dehua, tadei pengyou keyi daiti ni da if you not like this CL man his friend can replace you make
dianhua yaoqing ei. call invite
‘If you don’t like this man, his friend can replace you to call to invite (him).’
We can see that in instances where the antecedents are in an adjunct clause, the NO cannot be a null epithet, because these instances are not subject to SCO. And we have mentioned in the previous section that the NOs that are not subject to SCO are actually pros, and this pursuit is actually supported by Lasnik and Stowell’s discussion. As Lasnik and Stowell notice both pronouns and epithets may corefer with an NP that precedes it in the discourse, they retain their distinctive behavior with respect to binding conditions. When an epithet “corefers with an NP that precedes it in the discourse, that NP may not occur in an A-position that c-commands the epithet”
(Lasnik and Stowell: 708). A pronoun is not subject to this imitation, but it still can corefer with an NP that precedes it in the discourse.
(14) a. *Billi thinks Mary told me to visit the guyi.
b. Billi thinks Mary told me to visit himi.
In this respect, epithets differ from true pronouns. This follows that epithets, unlike pronouns, must obey Condition C.
To sum up, NOs in topic constructions can be called variables as in Huang’s (1982, 1984, 1987) or as null epithets as in Lasnik and Stowell’s (1991). Huang’s (1991) null epithet analysis is only partly correct. Although Huang’s using overt epithets as
argument for null epithets is problematic, it is correct to argue that NOs in a topic construction is a null epithet. However, Huang’s proposal is incorrect when he tries to extend the null epithet analysis to instances with antecedents in adjunct clauses, which does not exhibit SCO effects. We acknowledge that there are NOs which can be neither variables nor epithets. Chapter Three will argue that this type of NO in fact is a covert pronoun subject to Binding Principle B.
2.3 Free Empty Category Approach
Huang’s variable analysis has generated many responses and controversies because of the complexity of the data. A different approach is pursued by Xu (1986), who provides the following acceptable cases directly contradicting the example in (6).
(15) haizii yiwei mama yao zeguai ei le.
child think mother will reprimand LE
‘The child thinks (his) mother is going to reprimand (him).’
(16) xiaotoui yiwei mei ren kanjian ei. thief think no man see
‘The thief thought nobody saw (him).’
In (15) and (16), NOs can be A-bound by the matrix subject. In view of these conflicting examples, Xu claims that the NO in Chinese is not identifiable with any of the four types of empty categories defined by Chomsky. He suggests that, instead, all empty categories, including NOs, are of the same type-a Free Empty Category (FEC), which is not specified for any pronominal or anaphoric features. Therefore, an FEC is not subject to any principled interpretive rules and is interpreted according to discourse/pragmatics. An FEC adds a new type into the theory of EC.
Xu’s claim is based on the following arguments where he shows that NOs cannot be variables, NP-traces, PRO nor pro.
(17) (i) cannot be variables bound by TOPICs. Relation between the topic and the gap is not subject to the strong crossover condition nor island condition, which is a
standard diagnostic for variable binding as (a) and (b) show.
(a) Zhangsani tai shuo ni bu ken bang ei
Zhangsan he say you not willing to help
‘Zhangsan he said you are not willing to help (him).’
(b) Zhe ben shui [NP[S du guo ei] de ren] bu duo this CL book read Asp. De person not much
‘Not many people read this book.’
(ii) cannot be an NP-trace, because it doesn’t need to be bound by antecedent in the same sentence.
(iii) cannot be PRO. It is found in the position of a verbal object and is therefore governed and case-marked.
(iv) cannot be pro because sometimes it behaves like a null anaphor:
(a) shei piping guo ziji? Zhangsan piping guo e.
who criticize Asp. self Zhangsan criticize Asp ‘Who has criticized himself? Zhangsan has.’
(1986:88)
We have discussed in section 3.1 that example (ib) can be rescued by the pro
derivation; therefore, it cannot be viewed as a counterexample to argue against a variable analysis. Furthermore, (ib) indeed shows that some NOs cannot be analyzed as variables in consideration of island conditions. For (ii) and (iii), we agree with Xu that NOs cannot be NP-traces nor PROs, but we do not agree on (iv)—Xu’s objection to pros. According to Xu, e in the object position behaves very much like a pro, but it cannot be simply identified as a pro because sometimes it behaves like a null anaphor.
Thus in the above example, e is “not only coindexed with but also referentially dependent upon the subject” (1986: 88). Nonetheless, Xu also noticed that this coreference between object and subject in the above example is restricted to contextual manipulations.
(18) John ye da le [e]. (1986, (54)) John also hit Le
‘*John also hit.’
For (18), if no context is provided, the utterance is less natural. But it becomes perfect when a pragmatic context is provided:
(19) [a movie scene in which everyone has to slap himself]
The narrator: John ye da le [e]. (1986, (55)) John also hit Le
‘*John also hit.’
That is to say, anaphoric meaning only appears when (18) is in a restricted context or when (18) is used as a reply to the question as the sentence in (iv) shows.
Other instances of NOs, according to Xu, show typical properties of overt pronouns. For example, (20a) shows that the antecedent does not have to c-command
the object gap and (20b) shows that the object gap can take split antecedents. (20c) indicates that the object gap does not have to obey locality principles:
(20) a. [NP tadei mama yao zeguai [ei] le]. (1986, (35)) his mother will blame Le
‘His mother is going to reprimand him.’
b. Johni han Maryj yiwei mama yao zeguai [ei+j] le (1986, (36)) John and Mary think mother will blame Le
‘John and Mary think that mother will reprimand them.’
c. Johni yiwei [S Maryj zhidao [S mama yijing piping guo [ei] le]] (1986, (37)) John think Mary know mother already criticize Asp. Le
‘John thinks Mary knows that mother has already criticized him.’
Based on the above facts, Xu concludes that since the NO in Chinese can be both a pro and a null anaphor, it is an instance of a FEC, whose reference can only be recovered from the context in which it is used (where the context can be either linguistic or pragmatic).
However, Huang (1987) provides many arguments against the existence of an empty anaphor and that of an FEC. For example, the sentence in (iv) has the reading indicated only when it is used as a reply; it does not have that reading in general.
Furthermore, it seems that all the unacceptable sentences should be acceptable under this account because it is always possible to find favorable contexts for the intended readings. For example, the case of (iv), when uttered out of context and in the absence of any person named other than the subject, should have the NO bound by its own subject as its primary interpretation, a clearly false prediction. In addition, it is also
questioned by Li (2007b) and many others that Chinese does not distinguish different types of ECs. If all null arguments are FEC, how do we account for the fact that some NOs, such as variables are subject to island conditions while some NOs, such as pros, are not? Therefore, we do not agree with Xu on the indiscrimination of all null arguments. However, this thesis is greatly inspired by Xu’s examples and notices the need to pursue another type of NOs other than variables. This type of NO can be A-bound and allows more possible interpretations, and to determine among these possible interpretations is based on discourse/pragmatics. Therefore, we agree with Xu’s identification of pronominal properties, but we do not agree on empty anaphor, since the anaphoric meaning is not the most natural interpretation.
2.4 VP/vP-Ellipsis Approach
Instead of treating NOs universally as variables, Huang (1989, 1990), followed by Li (2002), argues that some NOs are not genuine null objects, but null VPs in disguise. It is argued that some NOC should be treated on a par with VP-Ellipsis (VPE)4 in English via an operation of V-to-Infl raising (Huang) or via a movement to
4 A typical instance of English VPE is given in (1), with do-support:
(1) Because Xiaoming likes this story very much, Xiaohua does [VP _____], too.
The Chinese counterpart of (1) can be formed in three ways: shi ‘be’- support, aux-support, or with the support of the main verb (that is, a NOC). The former two are shown in (2) and (3), respectively.
(2) shi-support (Shi construction)
Yinwei Xiaoming hen xihuan zhege gushi, Xiaohua ye shi.
because Xiaoming very like this story Xiaohua also be
‘Because Xiaoming likes this story very much, Xiaohua also does/likes this story very much.’
(3) aux-support (Aux construction)
vP (Li). The repeated verb is moved out of the VP to a higher position, followed by
the deletion of the relevant VP, then we have an instance of VPE. The claim is based on two similar properties between Chinese NOC and English VPE. One is the strict/sloppy reading as (21) and (22) show.
(21) John saw his mother. Bill does e , too.
Bill saw John’s mother, too.
Bill saw Bill’s mother, too.
(22) John kanjian le tade mama, Mary ye kanjian le e . John saw Le his mother, Mary also saw le
‘John saw his mother. Mary also saw e .’
John saw his mother, Mary also saw (John’s mother) John saw his mother, Mary also saw (Mary’s mother)
The property of strict and sloppy readings found in (22) as indicated in English
Yinwei Xiaoming hui hen xihuan zhege gushi, Xiaohua ye hui.
because Xiaoming will very like this story Xiaohua also will
‘Because Xiaoming will like the story very much, Xiaohua also will (like the story very much.)’
The two constructions, for a long time, have been treated as a unified ellipsis construction involving VPE/vPE (Huang 1987, 1989, 1991; Li 2002; Xu 2003; Wei 2007; Li 2007b). According to them, Aux construction is the same as VPE in terms of several diagnostics. Shi construction behaves highly similar to VPE only with a slight difference and the difference is suggested to result from the emptiness of Shi. However, Ai (2006) argues that Shi construction is not an instance of VPE. Instead it is an instance of deep anaphora, in the sense of Hankamer and Sag (1976). Soh (2007) also argues that Shi occupies a position above negation and licenses the ellipsis of a ΣP, which is headed by a negative marker or a zero affirmative marker.
translations is similar to English VPE construction as (21) shows.
Another property of NOC Huang observes which is similar to English VPE construction concerns the “locality effects” on the sloppy reading, as illustrated in (23) and (24).
(23) John saw his mother, and Mary knew that Bill did e , too.
a. ‘John saw his own mother, and Mary knew that Bill saw his own mother.’
(Sloppy reading)
b. ‘John saw his own mother, and Mary knew that Bill saw John’s mother.’
(Strict reading)
c. ‘*John saw his own mother, and Mary knew that Bill saw her mother.’
(False sloppy reading) (Huang, 1998, (23))
(24) John kanjian le tade mama, Mary zhidao Bill ye kanjian le e . John saw Le his mother, Mary know Bill also saw le
‘John saw his mother, and Mary knew that Bill also saw e .’
a. John saw his own mother, and Mary knew that Bill saw his own mother.
(Sloppy reading)
b. John saw his own mother, and Mary knew that Bill saw John’s mother.
(Strict reading)
c. *John saw his own mother, and Mary knew that Bill saw her mother.’
(False sloppy reading)
Compared with (23), (24) also has two readings; a strict one and a sloppy one. The sloppy reading, furthermore, is restricted to the embedded clause. In other words, there is a clause boundary for the sloppy reading.
However, Hoji (1998) and Xu (2003) argue against the VPE/vPE analysis, summarized to three main arguments here. Firstly, Hoji argues that neither of the two empirical bases, i.e. locality constraint on sloppy identity and sloppy/strict identity, is valid. Hoji uses Japanese examples to show that the “locality effects” on the “sloppy identity reading” as illustrated in (23), are independent of the properties of VPE and can be eliminated under appropriate contexts. For instance, the null object sentence in (25) allows the nonlocal soppy reading.
(25) a. John-wa zibun-no gakusei-o suisensita.
‘John recommended self’s student.’
b. Mary-ga Bill-ga [e] suisensita to omotteita.
‘Mary thought that Bill also recommended e.’
‘Mary thought that Bill also recommended (John’s student).’
‘Mary thought that Bill also recommended (self’s student).’
‘Mary thought that Bill also recommended (her student).’
(Hoji, 1998, (30))
If the locality of the sloppy reading in (25) is due to the properties of VPE, then the
“nonlocal sloppy reading” should be excluded.5
Furthermore, Hoji also shows that what some authors (for examples, Huang (1989, 1990) and Li (2002) for Chinese; Otani and Whitman (1991) for Japanese) call the sloppy reading is not a genuine sloppy identity reading, but what Hoji terms a
5 The locality condition on sloppy identity in an embedded clause has been questioned by many others.
Fiengo and May (1994:106) cite examples showing that it is not observed in some English cases. After Hoji (1998:136–138) points out that the locality effects in Japanese arise from independent semantic functions rather than the syntactic properties of VPE, Kim (2000) provides Korean examples supporting Hoji’s position.
sloppy-like reading, and that its existence is independent of VPE. The alleged sloppy reading arises owing to the pragmatics, suggesting that the NOC should be analyzed as a verb followed by a null object, rather than as a VPE structure.
(26) a. Johni-wa zibuni-o nagusameta. ‘Johni consoled himselfi’ *Billj-mo [e]j nagusameta. ‘Bill consoled (e).’
b. Johni consoled himselfi.
Bill did, too. ‘Billj consoled himselfj.’
(Hoji, 1998, (12))
The sloppy identity reading is not always available for the NOC, unlike its VPE counterpart in English, as indicated by the contrast between (26a) and (25b). Due to the status of null object, it is expected that NOs allow readings aside from the strict one and the sloppy one. Hoji (1998) provides Japanese examples to show the third reading. In Chinese, a third reading is also possible. Xu (2003) provides the Chinese sentence in (27), translated from Kim’s (2003) Korean example, with some minor changes to make the Chinese sentence more natural. In a context in which people know that Mike hit his son first and then Jeanne hit Mike, one can say:
(27) Mike xian da-le tade erzi, Jeanne cai da de. (Xu 2003, (17)) Mike first hit his son Jeanne then hit PARTICLE
‘Mike hit his son first, then Jeanne hit (somebody).’
In fact, as Kim observes, the reference of the NO is not limited to antecedents syntactically present in the preceding sentence. Suppose that Mike’s son and Jeanne’s daughter misbehaved so Mike punished his son and Jeanne her daughter. In such a context one can still use (27) with a NO referring to Jeanne’s daughter.
Now consider (28), the English VPE example.
(28) Mike hit his son first and then Jeanne also did.
There is a sharp contrast in interpretation between (27) and (28). In the latter only the strict reading and the sloppy reading are available. In other words, (28) is a VPE and therefore excludes a third reading, so (27) and (28) are not exact counterparts.
The difference on the availability of a third reading has two implications. One is that in NOC, the missing part after the stranded verb should be an empty argument, while the missing part for VPE is VP/vP. The empty argument, carrying nominal features, is expected to have stronger referentiality ability than an elided VP/vP.
The second implication is that although the structure is open to more interpretations in principle, not every sentence of the coordinate construction has three or more interpretations, particularly when an isolated sentence is cited with no context provided. Whether the strict reading, the sloppy reading or another reading is actually available is lexically and pragmatically determined. Comparing the following sentences, it is more difficult to get a strict reading or a sloppy reading but easier to get another reading.
(29) Wang xiansheng ti-le tade toufa he huzi, Li xiansheng ye ti-le (Xu 1986, (24)) Wang mister shaved his hair and beard Li mister also shaved
‘Mr. Wang shaved his hair and beard and Mr. Li also shaved.’
(30) Wang xiansheng ti-le tade toufa he huzi, Li xiaojie ye ti-le (Xu 1986, (25)) Wang mister shaved his hair and beard Li miss also shaved
‘Mr. Wang shaved his hair and beard and Miss Li also shaved.’
If (29) is uttered in isolation, it is most likely to be interpreted as describing a situation in which Mr. Wang and Mr. Li each shaved their hair and beard. Sentence (30) is structurally identical to (29) but referentially different. If Mr. Wang decided to become a monk and Miss Li a nun, (30) could be semantically true and grammatically correct on the understanding that the latter shaved her hair but not her hair plus beard.
In contrast to (30), the following sentence is not acceptable because the pragmatically appropriate interpretation is syntactically excluded by constraints on the interpretation of the VPE:
(31) * Mr. Wang shaved his hear and beard. Miss Li also did.
The second difference we summarize from Hoji’s and Xu’s discussion is adjunct inclusion. Adjunct-inclusion is regarded as one typical property of the VPE that adverbials in the second conjunct are required to be deleted along with the verb if they are identical to those in the first conjunct. When hearers interpret a VPE, they must recover the deleted adverbials.
(32) John saw Mary in Paris, and Peter did, too.
(33) John carefully cleaned his teeth, and Peter did as well.
The sentence in (32) does not have the interpretation that Peter saw Mary in a city other than Paris. Likewise, (33) is not used to describe the fact that though Peter did clean his teeth, he cleaned them carelessly.
However, this requirement on adverbial inclusion does not hold for the cases of the NOC.
(34) John zixide shua-le ya, Peter ye shua-le. (Xu 2003, (10))
John carefully brushed teeth Peter also brushed
‘John carefully brushed his teeth. Peter also brushed his teeth.’
(35) John mei tian shua san bian ya, Peter ye shua. (Xu 2003, (11)) John every day brush three times teeth Peter also brush
‘John brushes his teeth three times a day. Peter also brushes his teeth.’
The sentence in (34) does not exclude the possibility that Peter brushed his teeth carelessly.6 The one in (35) means that John brushes his teeth three times a day and Peter also brushes his teeth (though the number of times he does so isn’t specified).
The contrast between (34) and (33) with respect to the interpretation of the VP becomes even clearer when we add one more sentence:
(36) keshi Peter dou suibian shua. (Xu 2003, (12)) but Peter only carelessly brush
‘But Peter always carelessly brush (them).’
The third difference between NOC and VPE is the sensitivity to subcategorization.
Xu argues that NOC is sensitive to subcategorization by use of fanrong ‘boom’ and wennuan ‘warm’, cited from Hu and Fan (1995), and expand their examples to coordinate structures as follows:
(37) Jiu zhengce fanrong-le jinrong shichang, *xin zhengce ye fanrong-le.
old policy boomed money market new policy also boomed
‘The old policy caused the money market to boom and the new policy also did.’
6 A few remarks about the structure of the Chinese VP are made here. Most adverbials occur preverbally, with manner adverbs such as zixide ‘carefully’ in (33) and (34) staying closer to the verb than others. In (35) san bian ‘three times’ appears between the verb and its object. They are VP internal adverbials and must be elided along with the verb in the VPE.
(38) Tade hua wennuan-le womende xin, *nide hua ye wennuan-le.
his words warmed our hearts your words also warmed
‘His words warmed our hearts and your words also did.’
These sentences are ungrammatical. If NO in these sentences result from verb raising as the English VPE does, the construction should be insensitive to subcategorization as their English translations show.
To summarize, Chinese NOC cannot be treated as involving a VPE construction but only a NO, as suggested by the third reading. Some construction in Chinese may involve VPE (see note four), but if we have understood the difference between NOC and VPE, it is necessary to approach the two constructions with different analyses. A sentence with a stranded V is a NOC, in which readings other than the strict and sloppy ones are in principle available, but not always pragmatically possible.
2.5 Small Pro Approach
We have already seen Ai’s (2006) claim for topic-bound variable objects appearing in the weak pragmatic control in section 2.1. In contrast with weak pragmatic control, Ai identifies strong pragmatic control as a scenario in which the background utterances do not contain a salient linguistic topic. Ai further argues that in the strong pragmatic control context appears another type of NOs. According to Ai, this type of NO is sensitive to the [Animacy] feature as shown below:
(39) [Zhangsan drives home in his new BMW] (Ai 2006, (37)) Lisi [to his wife]: # Wo yi-dian-er dou bu xihuan [NP Ø]
I one bit all not like ‘I do not like (it) at all.’
(it = Zhangsan’s new BMW.)
(40) [Zhangsan walks home with his newly adopted husky] (Ai 2006, (38)) Lisi [to his wife]: Wo yi-dian-er dou bu xihuan [NP Ø]
I one-bit all not like ‘I do not like (it) at all.’
(it = Zhangsan’s newly adopted husky)
According to Ai, the [Animacy] feature of the object does play a role here. Under strong pragmatic control, the NO has to be recovered purely from the pragmatic context. [+Animate] NOs can recover their references fully from the pragmatic context, but [-Animate] NOs cannot. It is widely adopted that overt pronouns cannot usually represent inanimate entities, which also possess the [+Animacy] feature.
Therefore, Ai argues that NO in strong pragmatic control is an instance of pronoun, which is also sensitive to [Animacy] feature in the same way as overt pronouns are.7
Ai’s study of elliptical predicate construction in Chinese supports a mixed approach in analyzing NOC, which greatly inspires the current thesis. However, questions arise with the validity of Ai’s diagnostics. Ai proposes to differentiate the pro object from variable object based on the pragmatic context. Based on [+Animacy]
feature, Ai argues for the existence of pro objects, but few native speakers can sense the distinction between (39) and (40) by [+Animacy] feature. Imagine under the same context, if both Lisi and his wife look at Zhangsan’s new BMW or point at it, (39) is completely acceptable to many native speakers. Eye contact and gestures can not be
7 Ai does not discuss GCR at all in his argument for pro analysis. Object pros are expected to be ruled out by the conspiring requirement of GCR and Binding Principle B. We will take up this issue in Chapter Three.
counted as a linguistic topic to be a weak control situation;8 therefore, (40) remains to be strong pragmatic control. And Animacy feature seems to be the most important basis for Ai’s argument for pro object. The invalidity of this feature, therefore, weakens Ai’s proposal. The current thesis will also argue for a type of NOs that must be identified as pros. But identification of this type of NOs is not based on the sensitivity to Animacy feature in a certain pragmatic context.
2.6 True Empty Position Approach
Li (2007a, b) proposes that NO in Chinese is a True Empty Position (TEP), which is ‘an empty position devoid of any material, coming into existence only to fulfill subcategorization requirements’ (Li 2007b:37). In other words, a TEP exists because the empty position is not filled by any item from the Lexicon, including empty pronouns. This makes TEP a last resort strategy when the subcategorization requirement requires an object to be present but no element from the recognized inventory of empty categories can occupy the position. The missing part is truly empty—a TEP. The proposal of TEP is intended to provide a unified account to all NOC, Aux construction and Shi construction (see note four). For instance, in an Aux construction, if an auxiliary subcategorized a verb phrase as its complement, the elided complement is a VP; it is not a nominal category and is not a variable or a pro in the standard sense, which is a nominal category. On Li’s analysis, it is a TEP with VP form. We will soon examine whether or not TEP works for NOC.
How is such a TEP interpreted? Li suggests that the interpretative requirement is met by copying at LF the materials from a linguistic antecedent or the
8 Ai follows Xu and Langendeon (1985) to argue that a discourse topic is the topic which is linguistically mentioned or implied previously by a speaker.
discourse/pragmatic context, if a linguistic antecedent is not available. Just as Xu’s (1986) FEC, TEP resorts to pragmatics and therefore allows freer interpretations. For example, Li notes that although (41), as Huang argued, can most naturally be interpreted as coreferent with a discourse topic, (41) can allow the possibility of coreferential NO with subject. (42) contains such examples.
(41) Zhangsani shuo Lisij bu renshi e*i/*j. (Huang 1987, (19)) Zhangsan say Lisi not know
‘Zhangsan said that Lisi does not know (him).’
(42) a. haizii yiwei mama yao zeguai ei le. (= (17a)) child think mother will reprimand LE
‘The child thinks (his) mother is going to reprimand (him).’
b. xiaotoui yiwei mei ren kanjian ei. (= (17b)) thief think no man see
‘The thief thought nobody saw (him).’
c. xiaotoui yiwei jingcha bu renchi ei. thief think police not know
‘The thief thought that the police did not know (him).’
Generally speaking, the thief is afraid to be seen. The person who scolds child is always the mother. (42c) shows that even we use the same verb as (41), it is possible for the NO to corefer with the matrix subject. Li then concludes that these sentences are basically grammatical but pragmatics favors certain interpretations so strongly that they make certain other interpretations difficult to obtain.
In arguing for the emptiness of TEP, Li provides evidence which comes from the Aux construction. Specifically, the behavior of relativization from within the missing VP in the Aux construction shows that the missing phrase is indeed empty syntactically. Consider the following sentence: (Li 2007b:35)
(43) [[ta bu hui xihuan de] xuexiao] gen [[women bu hui *(xihuan) de] xuexiao]
he not will like De school as we not will like De school shi yiyang de
be the same De
‘The school he will like is the same as the school we will not like.’
(43) involves a relativization of an argument by directly raising the phrase from the original position inside the relative clause to the position occupied by the NP (the
‘Head’), modified by the relative clause. (44) shows this relativization.
(44) argument relativization with a gap (Li 2007b:36) [[CP [IP….[NP ti ]…][Head NP ] i]
…direct NP movement to Head
In (44), an argument from within the relative clause is raised to the Head position directly, leaving a trace in the original position. The relativization of an object in (43) must consider the object inside the VP, which makes a missing VP impossible, as shown by the unacceptability of (43). By use of the interaction of Aux construction and relativization, Li suggests the missing VP is indeed syntactically empty.
However, although Li’s argument shows that Aux construction may involve TEP, this does not show that NOC involves TEP. Specifically, we would like to examine if the NO can be relativized. If the NO can be relativized, it is not truly empty, but a null
argument that is relativized. Consider the following example and its syntactic structure in (45):
(45) a. [wo xihuan t] de e] shi Jenny.
I like De be Jenny ‘(The girl) who I like is Jenny.’
b. [[wo xihuan ___] de ___ ] shi Jenny.
Li (2007b: 43) argues that when [XP + de] contains a YP that is not overt, i.e., [XP + de + [YP___]], the YP is a TEP.9 However, when a relative clause has an empty head, there is evidence that movement is involved, and the YP/TEP is raised to the head position. Consider (46).
(46) *[[yinwei Zhangsan xie-le e] wo hen bu manyi] de (na feng xin) because Zhangsan write-Asp I very not satisfied DE that CL letter ‘Because Zhagnsan wrote (the letter) with which I am not satisfied.’
Suppose (46) has an empty head. If movement were not involved, (46) would contain two TEPs, and we would expect no island effects due to the lack of movement.
However, we observe island effects in (46) no matter the head na feng xin ‘that letter’
is overt or not. This shows that relativization in (46) involves movement. Therefore, (45) involves a covert head moved from inside the island. If a missing object is indeed empty as Li argues, no object could exist to be relativized. That is, object relativization is not supposed to be possible in a NOC. However, example (45) and (46) show that the NO can be relativized; this is counter to Li’s argument for the
9 In Li’s (2007b) study, de within nominal expressions occurs with a “modifying” expression, such as a relative clause, an adjective, and a possessor.
emptiness of the position after stranded V and also counter to the arguments for TEP analysis.
Since we have seen that NO is not as empty as Li argues, it is better to approach NOC with a different approach. Li’s discussion on the properties of NOS sheds some light on our search of another approach. Specifically, Li notices similar properties between TEP and pro. However, Li does not allow a pro in the NOC in order to maintain Huang’s proposal of Generalized Control Rule (GCR), which requires pro to be controlled by the nearest c-commanding NP. So if a NO in an embedded clause is a pro and can be coindexed with subjects in the matrix clause, GCR cannot be maintained. However, Chung (1984) has shown that in Chamorro, null objects pattern differently from variables and must be instances of small pro. Furthermore, GCR is severely questioned ever since it is proposed (Chapter Three will discuss its inadequacy in detail). Therefore, if we do not adopt the controversial GCR, the most possible candidate for this type of NOs we have been discussed in Chapter Two, which cannot be variables, must be pros. Hence, this thesis agrees with Li that some NOs after the stranded V involve a base-generated null argument. However, they are not TEPs; rather, we will argue that this type of referential NOs is a pro (if they are not variables). Furthermore, we also agree with Li that the interpretation of these NOs need to resort to pragmatics/discourse for freer interpretation possibilities.
2.7 Summary
Summing up, this chapter has concentrated on previous studies on referential NOs in Chinese. The review presents a very mixed picture of NOs. Some NOs are A’-bound while some NOs are A-bound. Some NOs have interpretations through the topic antecedents, while some NOs depend on pragmatics for interpretation. It has
been demonstrated that these referential NOs do not always correspond to a variable bound by a covert or overt A’-binder. In some cases, NOs within an island can refer to elements outside the island, and in some cases, NOs can be A-bound. We also present arguments against this type of NOs being analyzed as a VPE in disguise. Specifically, this type of NO portrays typical properties of pronouns, such as adjunct exclusion, sensitivity to subcategorization and referential possibility. Furthermore, this type of NOs is subject to discourse for its interpretation.
Other than referential NOs, we have seen in Chapter One that there is another type of NOs that are not referential shown by the lack of antecedents. As we can see in the literature review, this type of NOs has not been studied by any of the previous analyses in Chinese. Therefore, none of the previous analysis can accommodate the whole range of facts in NOC in Chinese. It is concluded that there is a necessity to make another proposal not only to incorporate their insights and to avoid their problems, but also to accommodate various different types of NOs in Chinese. In Chapter Three, we propose a modular account to approach this issue.
