The interpretation of this type of NO is through antecedent, which is exactly the binder of the variable.

Chapter Four

Concluding Remarks and Theoretical Implications

In this thesis, I have examined the phenomenon of null objects in Chinese. Other than variables, I have shown the empirical justification for identifying two more types of NOs in Chinese: pros and null cognate objects. Suppose that NO is formally licensed in Chinese by V. If the fundamental requirement for empty categories is that their content be recoverable, the three NOs must be recovered through different types of ways.

Three modules in grammar are suggested here to contribute to the identification requirement for Chinese NOs. This preliminary account is intended to sketch out the semantic and syntactic characteristics of the three types of NOs, and the role pragmatic principles play in their recovery. Let us consider different contributions of the three modules. The relationship between a topic-bound variable and the topic is regulated by syntax through topicalization operation; therefore, the variable observes Binding Principle C and island conditions. That is to say, once a topic (either overt or covert) is established, any other interpretation than topic will turn out to be bizarre.

The interpretation of this type of NO is through antecedent, which is exactly the binder of the variable.

For the second type, null pronominal object, while syntax only requires pro

observe Binding Principle B, a pro object can escape island conditions and it may

allow various interpretations. How to determine the most relevant interpretations is

linked to discourse/pragmatics. Therefore, the interpretation of this type of NO is

mainly through pragmatics which can account for cases where the reference of a null

object is not determined by syntactic principles but is rather inferred from context.

Turn to the third type of NOs: the null cognate object. When the hearer cannot find any specific or identified referent in the context, the lexical semantics of a given verb, whether transitive or the so-called intransitive, allows basic cognate interpretation of the complement. We also argue for the syntactic presence and status of null cognate objects. Pragmatic principles also allow recipients to identify this type of null objects more precisely than pure cognate interpretations.

Our proposal for three different types of NOs and their different ways of recovery of interpretation are expected with different layers of a clausal structure, as shown below:

(1)

CP

C IP Pragmatics

I VP

V OBJ (NP/N)

1 2 3

This system deals with the results and various components of the structure

contributing to the interpretation of the object. After we decide the discourse topic

within pragmatics, the discourse topic is located in [Spec, CP] (if we assume the least

complex structure of CP). Line one represents the topic-bound variable NOs. Now let us look at line two, if the element in [Spec, IP] is the subject, we may think that subject contributes interpretation to objects in an opposite way. As Binding Principle B regulates, the object cannot coindex with its subject. Since syntax does not further decide the reference of pros, we need to resort to another level, pragmatics, to determine the most likely interpretation. Line three represents null cognate objects. It can be seen that the lexical verb contributes the cognate interpretation to the objects.

Furthermore, pragmatics also helps in the process of interpretation.

We would like to conclude by considering the implication derived from this thesis on null object constructions in Chinese. We have argued for the existence of different types of NOs, including variables, pros, and NCOs. However, we have never discussed the syntactic category of NCOs, since they obviously do not belong to any of the four ECs Chomsky identifies. Rizzi calls these NOs generic “pros”, after ruling out the other three empty categories; however, he does not provide evidence supporting the syntactic status of “pros”. That is to say, although Rizzi uses the label, generic pros, he does not discuss if there is any other possible syntactic category NCOs may belong to. In fact, Cummins & Roberge (2005) suggest NCOs are bare nouns, but they do not provide evidence for their assumption, either.

Nevertheless, we do observe many differences between NCOs and referential pros.

The first difference between pros and NCOs is their binding properties. Recall that a referential pro is subject to Binding Principle B. However, the NCO is undefined with respect to binding theory; therefore, one cannot discuss the binding condition of a NCO. Rather, its properties follow from its inherent semantics.

The second difference is specification of number. It is expected that NCOs are

unspecified for number. Therefore, singular or plural interpretation of the NCO is not available. The only entailment sentence (2) yields is that Zhangsan was engaged in land exploring event today. No information is available about the number of lands that Zhangsan explored.

(2) Zhangsan zheng tian kancha ___

Zhangsan whole day explore

‘Zhangsan explore (lands) whole day long.’

However, singular/plural interpretation is directly available when a pro occupies the same position, through coindexation with the antecedent.

The third difference, suggested by Roeper and de Villiers (1995), is that nonspecific nominals (NCOs) allow extraction (3a), while specific nominals (pros) don’t (3b):

(3) a. Q: How

does John like [

advice from home t

]?

A: With kindness.

b. Q: How

does John like t

[

the advice from home]?

A: Very much. (Roeper 2000)

Similar effects are observed in Chinese cognate objects and pronouns.

(4) a. Zhangsan xihuan he [sheme jiou]?

Zhangsan like drink what liquor

‘Zhangsan likes to drink what liquor?’

b. *xioutou yiwei mei ren kandao [chuan sheme yifu de ta]?

thief think no man see wear what clothes DE he

‘?The thief thought no one see him wearing what clothes?’

It is a widely accepted assumption, following Huang (1982), that wh-phrases undergo QR at LF in Chinese. The wh-phrase sheme in (4a, b) has to be extracted out of the nominal phrase in both cases in order to be interpreted. Such extraction is possible out of a cognate object in (4a), but is ungrammatical out of a pro in (4b).

From the above discussion, it seems plausible to argue that NCOs are different from pros. That is to say, both referential NOs and nonreferential NCOs are true syntactic arguments but occupy different object positions in syntax in terms of syntactic categories, contra Rizzi. However, we will leave the issue of specific syntactic category of pros and NCOs to further research. But based on the Projection Principle, the different syntactic status of NOs must result in different projections.

Therefore we would be led to expect two different positions of NOs: one for referential NOs and the other one for non-referential NOs.

Our observation conforms to Basilico’s (1998) and Panagiotidis’ (2002) discussions, both of whom specify two positions for objects. An object can be merged directly in the Transitive Phrase (a functional category) position or within the inner V.

Two object positions corresponding to two interpretations (thetic vs. categorical; see Kuroda 1972, Cardinaletti 2002). The thetic interpretation (inner V) gives emphasis to the event involving the object but does not single out the object. The categorial interpretation (outer position) ascribes a property to the object. This distinction conforms to our identification of referential and non-referential NOs in Chinese.

The second implication comes from the advantage of the modular account. It

provides a mechanism for revealing some important differences and similarities

Chapter Three, we have seen that Chinese has null cognate objects. English and French are also argued to have null cognate objects. In other studies, Chung (1984) and Cole (1987) report that Thai and Korean are like Chinese in allowing null-topic bound variable NOs. According to Cole (1987), Imbabura Quechua, like Chinese, allows pronominals, but Imbabura Quechua, does not allow null variable object, since the language fails to permit null topics.

In view of all the differences and similarities among Chinese and many other languages, we generalize a possible universal implication (Greenberg 1963, 1966) of NOs in the world languages:

(5) Universal implication of null objects

Null Cognate Object < Pronominal Object < Null topic-bound variable Object

(5) indicates that if a language allows variable objects bound by a null topic, this very language also allows pronominal objects and null cognate objects. If a language allows pronominal objects but not variable objects bound by a null topic, this very language also allows null cognate objects. If a language only allows null cognate objects, but not pronominal objects, this same language will also disallow null-topic-bound variable objects.

Chinese is the representative of the first type, which allows all variable, pronominal and null cognate objects. Since Cole (1987) also argues that Korean and Thai also allow null topic-bound variables and pronominal objects, we expect that the two languages also allow null cognate objects. This prediction is born out in several previous works (Huffman 1973, Baker and Stewart 1999 Zlatev and David 2003 for

1

This universal implication of null objects does not include variable objects bound by an overt topic,

since there is no enough data to make the judgment for the time being.

Thai; Martin 1992, Kang 1997 for Korean). Tuvaluan also confirms this generalization. Besnier (2000) reports that Tuvaluan allows three types of null objects:

(a) NCOs, (b) pro, and (c) topic-variable. Imbabura Quechua represents the second type, which allows pronominal pro but not null-topic-bound variable pro. Similarly, we expect to find null cognate object in Imbabura Quechua and the expectation is also born out (Cole 1982, Hornstein 2006). English represents the last type, which only allows null cognate object, but not pronominal object nor null-topic-bound variable object (Cole 1982, Cummins and Roberge 2005).

It should be noted, though, that the problem of object omission is indeed a very complex one, with ramifications and implications for different grammatical areas.

Unfortunately, we have not been able to tackle all of them at this stage. It seems necessary, therefore, to continue studying in greater detail these interactions in the hope of finding more systematic motivations for the role of object omission in language use.

Specific further work is needed on the similarities and differences between overt pronouns and covert pronouns. Previous studies on this issue focus the discussion on their phonological difference. However, there appear more than phonological differences in Chinese. Consider the following paradigm:

(6) a. Zhangsan shuo Lisi bu renshi ___.

Zhangsan say Lisi not know ‘Zhangsan says Lisi does not know e.’

b. Zhangsan shuo Lisi bu renshi ta.

Zhangsan say Lisi not know him

Although the NO in (6a) can have the interpretation Zhangsan, this interpretation seems to be disfavored. The favored interpretation in (6a) is the third person. By contrast, although the pronoun in (6b) can refer to the third person, the favored interpretation is actually Zhangsan. We suggest that the difference it is because (6a) can be derived by two derivations—the variable and the pro. However, more detailed discussions are required.

Recall that we have implied NCOs are different from referential NOs in terms of syntactic projections. It is suggested that NCOs is located within the V. The expected question is how it can be a controller or a binder if it occupies a very low position.

Related to the previous remaining issue is the issue of transitivity. Transitivity is suggested to involve a course of grammaticalization. Verbs that were once considered

“intransitive” seem to be allowed to appear with a direct object nowadays, regardless of the verb’s semantic content or the semantic relationship between verb and object.

(7) a. ta feichang-de reai yinyue he chuangzuo yinyue, zhe ci ta he very love music and compose music this time he

dai-zhe tade nu you laidao yige hubian yiqilai zhao yinyue de bring-Asp his girl friend to one lake together search music DE linggan, suiran zuihou zhi chenggong le yishou ge …

inspiration though finally only succeed LE one song

‘He loves music and composing very much. This time, he brings his girlfriend to a lake district in search of inspiration of composition, though at last he only

succeeds composing one song. ’

b. Ren keyi shibai yijian shi, yige xiangmu, yige touzi，dan bu keyi

people can fail one thing one item one investment but not can

zai zuo ren shang shibai at conduct people on fail

‘People can fail one thing, one item, one investment but cannot fail conducting themselves. ’

The two verbs, chenggong ‘to succeed’ and shibai ‘to fail’ have been classified into

“intransitive” verbs for a long time. However, we may observe more and more

instances of the two verbs appearing with direct objects, especially in the language of

the young. Can this be applied to our discussions on NCOs? That is to say, NCOs

were once the implicit arguments absorbed and the verbs were once ‘intransitives’. It

is because of the grammaticalization of transitivity that NCOs are now syntactically

active. This possibility suggests the issue that NOs are in the middle of a language

change and grammaticalization. The three interesting issues deserve further studies.