Properties of Mandarin Reflexives

1.1 Animate NP antecedents for ziji and taziji

In Tang’s study of Chinese reflexives of 1989 (Tang 1989), it is claimed that Chinese reflexives can have only animate NPs as antecedents. While for Pan’s study of 1995 (Pan 1995), inanimate NPs can also be the antecedents of the Chinese reflexives. The examples are listed below:

1. Dan jian [NPyi dao jingqiao de baishi gongqiao_i] zai jinzhi de shuimianshang

   but see one CL beautiful DE white-stone arch-bridge at still DE water-surface

   touxia ziji_i de daoying.

   throw-down self DE reverse-shape.

   ‘Suddenly I saw a beautiful white stone arch-bridge throwing its own mirror image on the water.’

2. [NP Mei yige gongyuan_i], dou you ziji_i de dongtian.

   every one park all have self DE winter

   ‘Every park has its own winter.’

As for Chinese compound reflexives, like taziji, Pan also thinks that they can have inanimate NPs as antecedents. Take 3 for example:

3. Keyi renwei, [NPZhonggongi], yao baochi Xiang Gang de xianxing zhidu bubian,

   may think Chinese-party want maintain Hong KongDE current system not change,

   bingbu jinjin shi xinxuelaichao  de zhengzhi konghua,

   not only is prompted-by-sudden impulse DE political empty word

   er shi you [ ta zijii ] de liyi yaoqiu de.

   but be have he-self DE interest demand DE

   ‘One may think that it is not just an empty political promise prompted by a sudden impulse that the Chinese party wants to maintain the current system of Hong Kong, but that the party has its own interest and demands to do so.’

Now, the question naturally occurs whether Chinese reflexives really can have inanimate NPs as their antecedents. In Chinese, as in English, reflexives can be used under three circumstances. The first of these is to indicate by person or automatically. For by person, only the animate NPs can be used, for example, in Chinese we have

4. [NPTa_i] Ziji_i Lai le.

   He himself come PEF.

   ‘He himself has come’.

As can be seen from the English version of the Chinese sentence, himself also indicates by person. For the meaning in the automatic sense, however, inanimate NPs are normally used, for example in Chinese, we have sentence like

5. [NPMen_i] Ziji_i Kai le.

   The door itself open PEF.

   ‘The door opened automatically’.

And in English, the sentence The door opened itself also means ‘The door opened automatically.’

In this category, both in Chinese and in English, reflexives can refer back to inanimate NPs. Syntactically analyzed, reflexives in all these examples function as the adverbial modifiers of the main verbs of the sentences. According to Kroeger (2005), these kinds of reflexives can also be defined as emphatic pronouns. And strictly speaking, the reflexives in all these examples and their antecedents have not established any kind of thematic relations.

The second circumstance is where there exists a thematic relation between the reflexives and the antecedents. Both in Chinese and in English, we have sentences like

6. [NPTa_i] bu Xiangxin Ziji_i.

   He neg believe self.

   ‘He did not believe himself’.

7. He did not believe himself.

In both of these examples, the thematic relation of agent and patient between reflexives and antecedents has been established. The present study deems that due to this thematic relation, in this category of reflexive structures, the antecedents must be animate NPs both in Chinese and in English. For when the reflexive is used, it implies that the action indicated by the verb is performed with clear orientation. And this clear orientation is closely related to the intention of the doers. Only animate NPs may have intention. As for this category of thematic relation reflexives, Chinese does have some sentences like

8. [NPYueliang_i] na wuyun lai zhegai ziji_i.

   Moon take cloud come cover self’.

   ‘The moon covered herself with dark clouds.

9. [NP Lazhu_i] ranshao le ziji_i, zhaoliang le bieren.

   candle burn PEF self, light PEF others.

   ‘The candle burned itself and illuminated others’.

In both of these cases, it seems as if in Chinese sentences, thematic relation reflexives do have inanimate antecedents. But in fact, this is the metaphysical use. In both of the two cases, the agents Yueliang and Lazhu have been personified and thus can be considered animate. This is true for English as well. The English version of Chinese Example 8 is a particularly good one to illustrate this. In this sentence, the use of the preposition with indicates that dark clouds functions as instrument while usually only animate NPs can collocate with the instrumental theta roles.

The third category concerns the Chinese possessive structure formed by reflexives. The examples of 1 and 2 both belong to this category by means of the Chinese character DE indicating the possessive relation. For this category, since a thematic relation between the antecedents and the reflexives has not been established, these reflexives can have inanimate NPs as their antecedents. For example, in sentence 1, it is impossible to say *Baishi gongqiao touxia ziji, since white-stone arch-bridge is an inanimate NP, it naturally can’t intentionally throw itself down. But Baishi gongqiao touxia zijide daoying is completely acceptable since what is thrown down is not itself, but its own reverse shape.

So, in a word, as for whether the Chinese reflexives can have inanimate NPs as their antecedents, it depends on different syntactic situations. For both the adverbial or emphatic use and possessive use of reflexives, reflexives can have inanimate NPs as their antecedents, while for the thematic relation use of reflexives they cannot, except for metaphorical use. And the corpora study of Chinese reflexives proves this to be true. Consulting the modern Chinese corpus of Beijing University, 239 samples are to be found where reflexives have inanimate NPs as their antecedents. Among these 239 samples, 126 are metaphysical use of thematic relation reflexives, occupying 52% of the total. 67 are adverbial or emphatic use of reflexives, occupying 28% of the total. 46 are possessive use of the reflexives, occupying 20%. And this also shows the frequency of these different uses. No exception has been found to these uses in this corpus.

For compound reflexives in Chinese, since they cannot be used as adverbial or emphatic pronouns, only the remaining two situations are to be briefly discussed here. For the thematic relation compound reflexives (and in this case, compound reflexives ought to strictly follow the BCA, as will be discussed in detail later), only the animate NPs can be used except for metaphorical use, the same case as with simple reflexives. This is shown in Example 3, where Zhonggong (Chinese communist party) is personified as an animate NP, and this point is made much clearer in this sentence by the use of the compound reflexive ta ziji (he-self - spelled in Chinese as “他/她自己”) instead of ta ziji (itself – spelled as “它自己”). And for the possessive use of compound reflexives, since they establish no thematic relation with antecedents, the antecedents can also be inanimate NPs.

1.2 Long-distance antecedents for reflexives

It is well known that ziji can be bound across the local subject, violating BCA (Chomsky 1981), as exemplified by 10:

10. [NP John_i] zhidao [NPBill_j] xihuan ziji_i/_j.

    John know Bill like self

    ‘John knows that Bill likes him /himself.’

So, in this sentence, reflexive ziji can be either locally bound by Bill or long-distance bound by John. The long-distance binding feature of Chinese reflexives is a popular topic for syntacticians. The present study will later ponder over this in more details when the syntactic properties of simplex reflexives and compound reflexives are to be involved. 1.2 will cover only an interesting aspect which was debated in former studies, that is, whether the compound reflexives in Chinese can be bound across local subjects. Xu (1993) claimed that the compound form cannot be bound across the local subject, as shown in 11 below:

11. [NPJohn_i] zhidao [NPBill_j] xihuan taziji_j.

    John know Bill like he-self

    ‘John knows that Bill likes himself.’

In sentence 11, which is different from 10, the compound form can only be locally bound by Bill. In contrast, Hu (2002) has shown that the compound form taziji can also be bound across the local subject:

12. [NPJohn_i] shuo [NPnaben shu_j] hai le taziji_i.

    John say that book hurt PERF he-self

    ‘John said that that book hurt him .’

In this sentence, John is the antecedent of taziji, and in this sense, Hu deems that taziji can be long-distance bound. Hu’s example is interesting as well as crucial to the understanding of the binding concept in Chinese. For English, John said that that book hurt himself is absolutely unacceptable, since the use of himself violates BCA. English syntax predetermines that the locality in BCA must be understood from the perspective of linear distance. Thus, as a NP closest to himself linearly, the book cannot properly bind the reflexive. This is not the case for Chinese syntax. Compared with linear distance, semantic content plays a vital role in defining the concept of locality. As has been discussed, the compound reflexive taziji in sentence 12 is in the framework of thematic relation, it must have an animate NP as its antecedent, and in this sentence the closest animate NP is John. What the present study aims to reveal through this example is that, different from English BCA, local binding for Chinese takes not only the closest linear distance of c-commanding, but also the semantic content of the reflexive and antecedent into consideration. Chinese local binding implies that a reflexive should be bound by an animate NP within the closest linear distance of c-commanding.

1.3 Subject orientation

In the literature it is claimed that the simplex form ziji can only choose subjects as its antecedents, as exemplified by the following sentence (from Battistella and Y. Xu, 1990):

13. [NPJohn_i] songge_i [NPBill_j] ziji_i de shu.

    John give Bill self DE book.

    ‘John gave Bill a book about him.’

However, as noted in Chou (1992) and L. Xu (1994), arguments other than subjects can also be the antecedents of ziji:

14. [NPHao dongxi_i] dou bei [NPta_j] na dao ziji_j wuli.

    good thing all by him take to self room-in

    ‘All the good things were taken to his own room by him .’

15. [NPTa_i] yizhi bei [NPwomen_j] dangzuo ziji_j de bangyang.

    He always by we take-as self DE example

    ‘He has always been taken as our example by us.’

16. Dui [NPLugang_i] laishuo, [PRO_j] rang bu rang ziji_i qu dou wusuowei.

    To Lugang come-say, let not let self go all not-matter

    ‘As for Lugang, it does not matter whether he can go or not.’

If all these examples are carefully examined, it is not easy to draw the conclusion that, for Chinese reflexives, there actually is a tendency for some NPs to enjoy the privilege of being the antecedents of reflexives over other NPs. However, in a simple sentence structure^[1], it is not the subject that prevails but the active agent. Suppose there is an active agent in a Chinese simple sentence, that agent will be the antecedent of the reflexive, and if there is none, the linearly closest NP with the semantic role of active agent will be the antecedent.

Take sentence 13 for example, John is the subject of the sentence and more importantly John is also the active agent of the event. And in contrast, Bill is syntactically the object of the sentence and linearly is closer to the reflexive, yet Bill functions as the passive goal of the event and consequently cannot be the antecedent of the reflexive. For example 14, ta is the active agent of the activity na and is closest to the reflexive, so ta is the antecedent. As has been analyzed, the use of ziji indicates the clear intention of the agent. But here hao dongxi (good thing), as the passive patient of the activity na (take), is inanimate and can not have this clear intention. Thus hao dongxi cannot be the antecedent of ziji. This is the same case for 15, only women can be the active agent of the activity dangzuo and thus is the antecedent of the reflexive. Example 16 is different, since an active agent cannot be found for the activity rang, though in formal analysis there is a PRO here, so the remaining possibility can only be that the closest NP is the antecedent of the reflexive.

In conclusion here, instead of claiming that the simplex reflexive ziji has the so-called subject orientation, the closest NP with the semantic role of active agent can be the antecedent of ziji.

1.4 The blocking effect

Tang (1985, 1989), quoting Y.-H. Huang (1984), claims that long-distance binding of ziji is possible only if all the subjects of the clauses intervening between the potential antecedent and ziji agree in person features; otherwise, long-distance binding is blocked. This is the so-called blocking effect.

17. [NPJohn_i] juede [NPBill_j] dui ziji_i/_j mei xinxin.

    John think Bill to self not confidence.

    ‘John thinks Bill has no confidence in him/himself.’

18. [NPWo_i] juede [NPni_j] dui ziji_j mei xinxin.

    I  think you to self not confidence

    ‘I think you have no confidence in yourself/*me.’

19. [NPNi_i] juede [NPwo_j] dui ziji_j mei xinxin ma?

    you think I to self not confidence Q

    ‘Do you think I have no confidence in myself/*you?’

20. [NPJohn_i] juede [NPwo_j/ni_j] dui ziji_j mei xinxin.

    John think I/you to self not confidence

    ‘John thinks I/you have no confidence in myself/yourself/*him.’

From 17–20, it can be clearly seen that in 17, ziji/John/Bill agree in third person singular feature and thus there is no blocking effect, while in 18–20, all reflexives and possible antecedents share no common person feature and, therefore, blocking effect occurs. However, the blocking facts are more complicated than presented in Tang (1989). Huang and Tang (1991) and Xue et al. (1994) show that possessive NPs and direct and oblique object NPs can induce blocking effect just as subject NPs can, like

21. [NPJohn_i] renwei [NPwo_j] de jiaoao hai-le ziji_i*/_j.

    John think I DE arrogance harm-Perf self

    ‘John felt that my arrogance harmed him*/me.’

22. [NPJohn_i] gaosu [NPwo_j] [NPBill_k] hen ziji_i*/_j*/k.

    John tell me Bill hate self

    ‘John told me that Bill hated him/*me/himself.’

23. [NPJohn_i] cong [NPni_j] nar tingshuo [NPBill_k] chang piping ziji_i*/_j*/k.

    John from you-there hear-say Bill often criticize self

    ‘John heard from you that Bill often criticized him/*you/himself.’

The present study examines all these examples and concludes that the so-defined blocking effect can be simplified in this way: In Mandarin complex and compound sentences where two or more than two events are involved, all NPs can be possible antecedents of reflexives if they share same person and number feature; if they fail to share these features, the blocking effect occurs.

The most intriguing thing about the blocking effect is the so-called asymmetry of the blocking effect put forward by Bošković (2005) after his examination of the difference of the following two sentences:

24. [NPJohn_i] renwei [NPBill_j] de jiao’ao hai-leziji_i/_j.

    John think Bill DE arrogance harm-Perf self

    ‘John felt that Bill’s arrogance harmed him.’

25. [NPJohn_i] de xin biaoming [NPBill_j] haile ziji_j*/_j.

    John DE letter indicate Bill harm -Perf self

    ‘John’s letter indicates that Bill harmed himself.’

He concludes that the asymmetry of the blocking effect reveals itself in the differences between the antecedents of the two sentences. In sentence 24, the NP in possessive case is in the embedded clause and in this case, Bill forms no blocking effect and ziji in the embedded clause can have either John or Bill as its antecedent. In sentence 25, the NP in possessive case is in the main clause, Bill forms the blocking effect, and thus ziji in the embedded clause cannot have John as its antecedent.

Is this really the case? In order to verify the asymmetry of the blocking effect, a language survey was conducted. A questionnaire was delivered to 50 students, all of whom major in Chinese linguistics in one of the key universities in China and thus have an acute language sense for Chinese. 48 of all the questionnaires retrieved proved to be effective. In the questionnaire, sentence 25 is given and then three choices are listed, of which they were required to choose only one. The three choices were: A. ziji can refer back to Bill; B. ziji can refer back to John; and C. ziji can refer back to both John and Bill. Among the 48 effective results, 18 students chose A, 14 students chose B, and 16 students chose C, occupying 37.5%, 29.2% and 33.3% respectively. The chi-square test result is X2=0.50, p=.779>.05. This further indicates that there is no evident difference in statistical sense. The result of this language survey shows that in sentence 25, ziji can have either Bill or John as its antecedent. This experimental study indicates that there is no asymmetry for the blocking effect of Mandarin simplex reflexives.

1.5 The maximal clause effect

It is claimed in the literature that only local and matrix subjects are possible antecedents of ziji in the case where all intervening noun phrases have the same person feature. This effect is shown in 26 below (Tang 1989).

26. [NPJohn_i] zhidao [NPBill_j] juede [NPMark_k] dui ziji_i*/_j*/k mei xinxin.

    John know Bill think Mark to self not confidence

    ‘John knows that Bill thinks that Mark has no confidence in himself.’

The claimed priority of the third person matrix subject as an antecedent (in preference to intermediate subjects), when all the intervening NPs are third person NPs, is sometimes referred to as the maximal clause effect (Wang and Stillings 1984; Y.-H. Huang 1984; Battistella and Y. Xu 1990). But some researchers (Cole et al. 1990; Huang and Tang 1991; Pan 1995) believe that the intermediate subjects can be the antecedent of ziji.

So in a certain sense, the researchers did not agree upon whether the maximal clause effect existed. And the maximal clause effect is also contradictory to the conclusion drawn by the present study that all NPs no matter in what case can be possible antecedents of reflexives if they share the same person and number features. A language survey thus was also conducted as part of the present study to verify the so-called maximal clause effect. For the same 50 students, a questionnaire containing sentence 26 with two alternative choices was delivered to all. The choices were: A. ziji can refer back to all the men in the sentence and B. ziji can only refer back to the Mark. 47 students chose A, while only 3 students chose B. That is 94% versus 6%. The result of the language survey supports no maximal clause effect for reflexives and further confirms that the blocking effect conclusion made in the present study is correct.

1.6 Summary

This section serves as a brief review of the most important properties of Mandarin reflexives claimed by former studies by using theoretical analysis, linguistic corpus study and language survey. The conclusions are as follows:

1. Both simplex reflexives and compound reflexives in Mandarin, are required by the semantic principle to have animate NP antecedents.

2. Simplex reflexives in Mandarin exhibit the so-called phenomenon of long-distance binding while compound reflexives do not.

3. There is no subject orientation for Mandarin simplex reflexives. The closest NP with the semantic role of active agent can be the antecedent of ziji.

4. In Chinese complex and compound sentences where two or more than two events are involved, all NPs no matter in what case can be possible antecedents of reflexives if they share same person and number feature; if they fail to share this feature, the blocking effect occurs.

5. There is no maximal clause effect for Mandarin reflexives.

2.1 Introduction – from a pragmatic to a syntactic view of reflexives

Before analyzing the properties of Mandarin reflexives, it is intriguing to notice the reference differences between third person singular pronoun, simplex reflexive, and the compound reflexive by observing the following sentences.

27. John rang Bill piping ta.

    John let Bill criticize him.

    ‘John asked Bill to criticize him.’

28. John rang Bill piping ziji.

    John let Bill criticize self.

    ‘John asked Bill to criticize him/himself.’

29. John rang Bill piping taziji.

    John let Bill criticize him-self.

    ‘John asked Bill to criticize himself.’

According to BCB, a pronoun can have an antecedent as long as the antecedent is not local or does not c-command the pronoun, thus ta in sentence 27 cannot refer back to Bill but to John. And the pronoun can also be free, and in that sense, ta in 27 can refer to all the third person instances except Bill. In sentence 28, ziji as a simplex reflexive can refer back to either Bill or John according to the long-distance binding principle. In sentence 29, taziji can only refer back to Bill according to BCA.

The observations with regard to 27 and 29 easily reveal that in Mandarin, the third person pronoun is complementary to the compound reflexive as far as the reference is concerned. This is the same case for the English, where the third person pronoun and reflexive form the complementary relation for reference. This observation may offer some insights into the different properties of Mandarin simplex and compound reflexives.

Now return to simple English sentences like

30. John criticizes him.

31. John criticizes himself.

The question posed here is what attributes the complementarity between him and himself for their reference. This is a simple question but significant for the understanding of the properties of reflexives both in Chinese and English.

In the rise and potential fall of reflexive pronouns of Mira Ariel’s Pragmatics and Grammar, she anticipated the grammaticalization process of the reflexives from a pragmatic perspective. She pointed out that a typical event normally concerns two participants and the two participants are different from each other, as in John criticizes Tom. Both the speaker and the listener can make a clear pragmatic inference to differentiate the two participants. But in the case that the two participants are the same, so as to clearly indicate the two participants are the same, there must be a pragmatic marker. The reflexives are this kind of pragmatic marker.

This is also true for sentence 30 and 31. In 30, the pronoun him, as the third person, has the de re feature, which implies that it can refer to anyone except for speaker and listener. There is the possibility that him can also refer back to the subject John. For the purpose of avoiding ambiguity and clearly indicating the two participants are the same, the reflexive form appears and the reference complementarity can be well explained.

Ariel made efforts to explain the reflexive phenomenon from the pragmatic perspective instead of the syntactic perspective. She deems that the grammaticalization process of the reflexives as pragmatic markers is the result of the mechanism of the salient discourse profile, while the generative grammar analyses of reflexives as “co-argument,” which obeys the BCA principle, might conceal the pragmatic motivation of the phenomenon.

This is not true. As Chomsky (1975, 1981, 1993) emphasized on several occasions, the semantic and pragmatic motivation is there but what syntax aims to do is analyze how the syntactic forms for those motivations are generated. Previous syntactic studies concerning reflexives focused on the binding conditions but scarcely touched on the generative process of the reflexives, or simply how reflexives derive their forms during the derivation. This is a problem that cannot be avoided when the properties of both Chinese and English reflexives are further studied.

Among few former studies concerning how reflexives derive their forms, some (Reinhart 2006) thought that the traditionally defined VP had different layers which could assign different cases to pronouns or NPs. There might exist a special designated layer which is in charge of designating reflexive forms to pronouns. The findings of the present study suggest that this is not correct. Quite different from cases which mainly reveal the relation between predicates and arguments and in turn can be determined by the V, v, or T element, the reflexives represent the relation between the two arguments and exhibit the co-arguments relation between external and internal arguments. And consequently, the idea of verb designation must be abandoned.

Ariel’s study of reflexives, though a completely pragmatic one, proves to be inspiring. Take wh-move for example, no one can deny wh-move is for pragmatic motivation, that is, to construct an interrogative sentence for the communication purpose of putting forward a question. However, for generative grammar, the pragmatic motivation, as a known truth, can be ignored and the syntactic form can be analyzed in a purely formal way, that is, wh is moved due to the edge feature of complementizer. Can the special case of reflexives be analyzed in the same way or is internal merge operation involved here?

2.2 Different syntactic properties of Mandarin simplex and compound reflexives

The present study hypothesizes that the syntactic properties of reflexives are special from the derivational perspective. For John criticizes him, the first step of the derivation is the external merge of V and pronoun, thus we have the pair [criticize him]. The accusative form of the pronoun, as an unvalued feature, is valued by the verb and the pair then projects a VP. In contrast, for John criticizes himself, the first step of the derivation is the external merge of NP John and pronoun him, thus we have the pair [John himself]. Notice here both the nominative case of John and the reflexive form of himself have not been valued yet. Inside the pair, John and himself semantically establish the co-argument relation, which differentiates itself from a simple case of internal and external arguments. And syntactically, the two establish a governing relation in which NP John values the reflexive feature of the pronoun, and as a result, the reflexive himself gets its form. The reason that NPs value the reflexive feature of pronouns but not the opposite is the sound logic that only when the second thing appears can we say the two things are the same and thus the syntactic feature to show the same should always be with the pronoun, which merges later. The pair [John himself] projects a NP. As Elly, referring to Chomsky, has mentioned, pronouns cannot be heads because they are weak (Elly 2009). The position John is occupying now is not one where the nominative case of John can be valued. Then John moves to the edge position of the projected NP for further derivation. This hypothesis is reasonable in two senses. First, semantically, that [John himself] forms the pair can best represent the co-argument relation of the two. Second, syntactically, it cannot be expected that V, v, T, or C can value the formal feature of the reflexives, for under the framework of traditional generative grammar, they have all been assigned some other tasks. Then the only remaining possibility is the subject NP values the reflexive feature of the object pronoun. The pair [John himself] also forms a governing relation and John as a head can value the feature of the complement.

From the case of reflexives, it is safe to say the pragmatic motivation of marking the two same participants can only be realized in the process of syntactic derivation. The unique syntactic properties of reflexives are during the derivation, subjects and reflexives form the pair first and subjects are in charge of valuing the reflexive features of the pronouns.

Till now too much has been talked about the syntactic features of reflexives taking English as an example and it is time to shift focus to Chinese reflexives. As has been mentioned, as for reference, the Chinese third person pronouns and compound reflexives are complementary to each other. And the present study deems that the Chinese compound reflexives possess the same syntactic feature as the English reflexives.

Take John xihuan taziji for example. During the derivation, the first step is the external merge of the pair [John taziji] and the pair projects an NP. The NP John is in charge of valuing the reflexive feature of taziji and then for the consideration of implicit case valuing, John moves to the edge position of the NP. This way of derivation makes sense except for the possible challenge which might occur, that is, in Chinese, taziji as a compound reflexive is the combination of two independent free morphemes ta and ziji. Then how can three free morphemes of John, ta, ziji form a pair? This is not the case, as the following analyses illuminates, for the Chinese compound reflexive ziji is possibly not an independent morpheme but a suffix.

In Chinese, there are two forms of pronoun plus ziji structures, as the following two examples show

32. [NPTa_i] ziji_i yao qu.

    He self want go.

    ‘He himself wants to go.’

33. [NPJohn_i] piping [taziji].

    John criticize him-self.

    ‘John criticized himself.’

In sentence 32, ta and ziji are two independent morphemes. The main arguments for this are, first between ta and ziji, a focus marker shi (is) can be inserted; second some adverbs like youyi (intentionally) can be inserted. The two possible insertions show us in case 32 that ziji is an independent morpheme which functions more like an adverb since it can have a focus marker or another adverb. But this is never the case for the compound reflexive taziji in 33, since no other syntactic element can be inserted between ta and ziji. And thus in this case, ziji, as an independent morpheme, functions more like a suffix. This suffix indicating reflexive can be valued by subject NP.

The above analysis can also be applied to Mandarin first and second person compound reflexives. In conclusion, Mandarin compound reflexives are quite similar to English reflexives and possess unique syntactic features from the perspective of derivation. The closest c-commanding subject NP forms a pair with the compound reflexive and values the reflexive feature. Ziji in these compound reflexives functions more like a suffix and shows itself as a result of feature valuing.

This also helps to explain why Mandarin compound reflexives, similar to English but different from Mandarin simplex reflexives, do not have the long-distance binding as conclusion 2 in 1.6 indicates. As has been discussed, at the very beginning of derivation, NPs and reflexives form a pair and NPs are in charge of valuing the reflexive feature of the compound reflexives, especially the syntactic legitimacy of suffix ziji. Then they project an NP. Since the position occupied by NP is not a position where the case of NP can be assigned, NP has to move successively to the SPEC of T to be assigned a case. Since the co-argument relation of the NP and the compound reflexive has been predetermined ever since the pair was formed, the compound reflexive can only take that NP as the antecedent which is now at the closest c-commanding position where it can be assigned a case. In this sense, the so-defined local binding is strictly required for both Mandarin compound reflexives and English reflexives.

The Mandarin simplex reflexives are different. Syntactically, they follow the same derivational pattern as a simple object NP or pronoun. That is, in the first stage of derivation, V and ziji form a pair. Semantically, they are anaphors plus implicit pronouns. They are anaphors so they must have antecedents in the sentence. They are implicit pronouns so that during the process of referring back, they have to identify their pronoun feature.

The syntactic and semantic features of Mandarin simplex reflexives can help us have a better understanding of the long-distance binding of simplex reflexives and the blocking effect. Take 34 for example:

34. [NPJohn_i] renwei [NPTom_j] xihuan ziji_i/j.

    John think Tom like self

    ‘John thinks that Tom likes him/himself.’

Since at the beginning of the derivation, ziji did not form a pair with any NP, it has not been predetermined which specific NP is the antecedent of the anaphor-ziji. However, ziji as an anaphor has to refer back to some NP in the sentence. In LF, the first NP it encounters during the process of referring back is Tom. Thus, ziji as an anaphor identifies its pronoun feature, that is, the third person feature. As has been mentioned earlier, the third person possesses a de re feature, which means it can refer to anyone except speaker and listener. Thus, ziji may refer further back to John, which is allowed by the third person pronoun de re feature of ziji.

This is the same case for sentence 35 and 36:

35. Wo renwei wo xihuan ziji.

    I think I like self

    ‘I think that I like myself.’

36. Ni renwei ni xihuan ziji ma?

    You think you like self Q?

    ‘Do you think that you like yourself?’

In both of these, ziji first encounters I and you and identifies pronoun feature separately and then can refer further back to the subjects of the main clause since they are the same.

In conclusion, the argument of Tang (1985, 1989) and Y.-H. Huang (1984) that long distance binding of ziji is possible only if all the subjects of the clauses intervening between the potential antecedent and ziji agree in person features is pertinent here. But for the third person, that is because of the de re feature, and for first and second person, that is because the pronouns are in fact all the same.

But consider 37:

37. [NPTom_i] renwei [NPwo_j] xihuan ziji_i/_j.

    Tom think I like self

    ‘Tom thinks that I like myself.’

When ziji first encounters wo, the first person pronoun, its pronoun feature has been identified as first person and cannot refer back to Tom since the mismatch of pronoun features will lead to the LF crash.

For the long-distance binding and blocking effect of Mandarin simplex reflexives, there are more hard nuts to crack.

38. [NPZhangsan_i]shuo [NPLisi_j] chang piping ziji_i/_j.

    Zhangsan say Lisi always criticize self

    ‘Zhangsan_i says that Lisi always criticizes himself/him.’

39. [NPZhangsan_i]shuo[NP wo_j] chang piping ziji*_i/_j.

    Zhangsan say I always criticize self

    ‘Zhangsan says that I always criticize myself/*him.’

As has been mentioned above, the Mandarin simplex reflexive, during the process of referring back, has to identify its implicit pronoun feature. For the third person in 38, because of its de re feature, ziji can refer back to both Lisi and Zhangsan. Thus here we have the long-distance binding phenomenon. For the first person in 39, because of its de se feature, ziji can only refer back to wo. Thus we have the blocking effect phenomenon.

Sample 40 and 41 show more complicated situations:

40. [NPZhangsan_i] dui [NPLisi_j] shuo [NPWangwuk] buneng lao piping ziji_j/_k.

    Zhangsan to Lisi say Wangwu cannot often criticize self

    ‘Zhangsan_i says to Lisi_j that Wangwuk cannot criticize himself /him_i.’

41. [NPZhangsan_i] dui [NPLisi_j] shuo [NPtak] buneng lao piping ziji_i/_j/k.

    Zhangsan to Lisi say he cannot quite often criticize self

    ‘Zhangsan_i says to Lisi_j that hek cannot criticize himself/him_i/_j.’

In the case of 40, ziji can refer back to both Wangwu and Zhangsan but not Lisi. Since Lisi, though possessing the third person feature, is not the active agent of the act of saying. As a result, besides Wangwu, the simplex reflexive ziji in this sentence can refer back to another closest NP, Zhangsan, which is the active agent of the act of saying. The point has been made clear in 1.3. The case of 41 seems to be more complicated. As has been discussed in case 40, ziji can refer back to both ta and Zhangsan. But here there might exist a special case in which ta, as the subject of the embedded clause, because of its de re feature can refer back to Lisi. In this situation, the sample sentence 41 can mean ‘Zhangsan says to Lisi that Lisi cannot criticize himself quite often.’ And this special case of ziji referring back to Lisi, which plays the semantic role of patient, does not violate the principle mentioned in 1.3, because ziji here actually refers back to ta. And 41 can also mean ‘Zhangsan says to Lisi that Lisi cannot criticize Zhangsan quite often’ and for this meaning, ziji is long-distance bound by Zhangsan.

In addition, Pollard and Xue (1998) have discussed the so-called blocking effects induced by non-subjects, revealed by sample sentence 42.

42. Zhangsani gaosu wo_j Lisik juedui bu hui piping ziji*_i/*_j/_k.

    ’Zhangsan told me Lisi never criticizes himself.’

According to their analysis, ziji in the embedded clause cannot refer back to Zhangsan because of the blocking effect induced by the non-subject wo in the main clause. In order to verify whether this blocking effect can stand, a linguistic survey was conducted. Among the 48 effective results, 43 think ziji in 42 can refer back to Zhangsan, occupying 89.58%. This result shows that for most of the native speakers, a so-called blocking effect induced by non subject does not exist.

Another point worthy of mention here concerns the difference in number features for reflexives, as has been discussed by Lidz (2001). Sample 43 illuminates the point.

43. 1. Zhangsan_i shuo tamenj zai piping ziji_j/_j.

       ’Zhangsan_i says that they are criticizing themselves/him_i.’

    2. Tameni shuo Zhangsan_j zai piping ziji*_i/_j.

       ‘They say that Zhangsan is criticizing himself/*themselves.’

For Lidz, ziji in 43a can refer back to both taman and Zhangsan while ziji in 43b can only refer back to Zhangsan. Thus there exists asymmetry concerning the reference of ziji in these two sentences.

As the studies by Liu (2003) and Huang and Liu (2001) correctly make clear, the Mandarin simplex reflexive has a strong tendency to refer to NP possessing singular person feature. 43 is the epitome of this tendency. In 43a, during the process of referring back, when ziji first encounters the NP which possesses plural person feature, it is acceptable for ziji to refer to this NP and because of the de re feature of third person, ziji can refer further back to Zhangsan. In contrast, in 43b, during the process of referring back, when ziji first encounters the NP which possesses singular person feature, since it has the strong tendency to refer to singular person, it will not refer further back to the subject of the main clause tamen, which possesses the plural person feature. Even for 43a, the linguistic survey in which the participants were required to make a choice between “A. ziji can refer to tamen” and “B. ziji can refer to Zhangsan” shows that, among the 46 effective results, 31 chose B and 15 chose A. And this result also indicates ziji has a tendency to refer back to NP possessing singular person feature.

So, the asymmetry of ziji’s reference induced by the third person feature disagreement can be seen as an exception to the long-distance binding of ziji if it has the implicit third person pronoun feature. Another exception has been discussed by Huang and Liu (2001), as illustrated by the sample sentence 44:

44. [NPZhangsan_i] shuo [NPxiaotou_j] tou le ziji_i’de qianbao.

    Zhangsan say thief steal-Perf self’s wallet

    ‘Zhangsan said the thief had stolen Zhangsan’s wallet.’

In 44, according to the properties of simplex reflexive mentioned in this section, ziji, during the process of referring back, can refer back to both xiaotou and Zhangsan, since they are both the active agents. However, since the usual semantic inference is ‘thief will steal others’ wallets,’ ziji here has a strong tendency to refer to Zhangsan not the thief.

2.3 Summary

From the above analysis, a conclusion can be drawn that the Mandarin simplex reflexive and compound reflexive possess different syntactic features. For the simplex reflexive, it follows the normal derivational pattern and possesses both anaphor and pronoun features. As an anaphor, it should be bound by some antecedent, and as a pronoun, during the process of referring back, it has to identify its pronoun feature. This helps to explain the so-defined long-distance binding effect and blocking effect found in Chinese simplex reflexives. During the process of referring back, if the simplex reflexive identifies its implicit pronoun feature as third person, its de se feature make ziji refer back to all the active agents with the exceptions of person number disagreement and the special semantic inference. Thus we have the long-distance binding of the simplex reflexive. However if the simplex reflexive identifies its implicit pronoun feature as first and second person, their de re feature make ziji only refer to that first and second person. Thus we have the blocking effect. For the compound reflexive, it functions more like the English reflexive and follows a unique derivational pattern, though pragmatically it can be analyzed as a pragmatic marker to indicate the two participants are the same. Syntactically, it is a pure anaphor and during the derivation, it forms a pair with its antecedent first and this antecedent helps to value the reflexive feature of the compound reflexive and gives the suffix ziji to pronouns. The antecedent cannot be properly assigned a case in that position and has to move to the closet T SPEC position to be assigned proper implicit case. This helps to explain why, for Mandarin compound reflexives, the BCA has to be strictly followed.