Syntax and Semantics, Volume 33 279 Copyright 2000 by Academic Press Long distance Reflexives All rights of reproduction in any form reserved.

0092-4563 /00 $30.00

WHY THE BLOCKING EFFECT? HAIHUA PAN Department of Chinese, Translation, and Linguistics City University of Hong Kong

1. INTRODUCTION

This chapter discusses the blocking effect observed in long-distance (LD) bound bare reflexive ziji in Mandarin Chinese. Unlike the symmetrical unlike-person blocking claimed in the literature (Huang & Tang 1991, Xue, Pollard, and Sag 1994), this chapter argues that (a) the blocking effect of ziji is not sym-metrical: first and second person pronouns can block third person noun phrases (NP) from long distance binding ziji , though third person NPs do not necessarily block first or second person pronouns from long distance binding ziji ; and (b) other grammatical functions fill ed by first and second person pronouns, not just subjects or NPs contained in the subject, can induce the blocking effect. The chapter claims that long distance bound ziji points to the carrier of belief, and reconstructs this notion as self-ascription. The blocking effect is explained by appealing to the fact that only first and second person pronouns are obligatory self-ascribers, and thus can block long distance binding of ziji by third person NPs if they intervene between the potential third person NP and the reflexive ziji , while third person NPs do not necessarily block ziji from being long distance bound by first/second person pronouns. The rest of this chapter is organized as follows: after describing the basic facts concerning the blocking effect in Section 2, I argue in Section 3 that the blocking effect is not symmetrical, and the crucial factor is not the local subject or the unlike person feature conflict, as suggested in the literature. Rather it is

Haihua Pan

280

the asymmetry between first/second and third person noun phrases that plays a crucial role in the blocking effect, and syntactic functions other than subject can also induce the blocking effect. In Section 4, I briefly discuss limitations of previous analyses employing feature agreement checking, closeness and poten-tial binders, logophoricity, and perspectivity. In Section 5, I f irst argue that long distance bound ziji is a de se anaphor, using Lewis’ terminology, and is thus constrained by self-ascription. Then I propose a condition to account for the properties of long distance bound ziji , and show that my account can better ex-plain the blocking effect than can previous analyses. Before concluding the chapter in Section 7, I discuss some relevant issues in Section 6, especially the differences between self-ascription and logophoricity. 2. THE BLOCKING EFFECT

Chinese bare reflexive ziji ‘ self’ exhibits the so-called blocking effect: long-distance binding of ziji is possible only if all i ntervening subjects agree in person features; otherwise, the long distance binding is blocked, as exempli fied in (1) (Y. H. Huang 1984, Tang 1985, 1989).1 (1) a. Woi juede ni j dui ziji * i/j mei xinxin.

I think you to self not confidence ‘I think you have no confidence in yourself/*me.’

b. Ni i juede woj dui ziji * i/j mei xinxin ma? you think I to self not confidence Q ‘Do you think I have no confidence in myself/*you?’

c. Zhangsani juede wo/ni j dui ziji * i/j mei xinxin. Zhangsan think I/you to self not confidence ‘Zhangsan thinks I/you have no confidence in myself/yourself/*him.’

d. Zhangsani zhidao Lisi j dui ziji i/j mei xinxin. Zhangsan know Lisi to self not confidence ‘Zhangsan knows that Lisi has no confidence in him/himself.’

Since they do not agree in person feature with the local subjects, the matrix sub-jects in (1a), (1b), and (1c) cannot be the antecedents of ziji . By contrast, the matrix subject in (1d) can be the antecedent of ziji , since there is no person fea-ture conflict between the local subject and the matrix subject. The blocking effect is the motivation for requiring successive cyclic movement of ziji at LF in the movement analysis of ziji (Battistella 1989, Cole, Hermon, and Sung 1990, Huang & Tang 1991, Li 1993, Cole & Sung 1994, Cole & Wang 1996, etc.).

Why the Blocking Effect?

281

However, the blocking fact is more complicated than presented in Tang (1989). Huang and Tang (1991) point out that an intervening sub-commanding NP with different person features can also induce the blocking effect, as shown by the following contrast: first/second person NPs but not third person NPs con-tained in a subject block third person matrix subjects from binding ziji .

(2) a. Zhangsani renwei Lisi j de jiao’ao hai-le ziji i/j.

Zhangsan think Lisi DE arrogance harm-Perf self ‘Zhangsani felt that Lisi’sj arrogance harmed himi/j.’

b. Zhangsani renwei woj de jiao’ao hai-le ziji * i/j. Zhangsan think I DE arrogance harm-Perf self ‘Zhangsani felt that my arrogance harmed him* i/me.’

(3) a. Zhangsani renwei Lisi j neiyang zuo dui ziji i/j buli . Zhangsan think Lisi that-way do to self not-beneficial ‘Zhangsani felt that Lisi’sj acting that way didn’ t do him any good.’

b. Zhangsani renwei ni j neiyang zuo dui ziji * i/j buli . Zhangsan think you that-way do to self not-beneficial ‘Zhangsani felt that your acting that way didn’ t do him* i/you any good.’

Furthermore, direct objects and obliques, though not potential binders which are defined as c-commanding animate subjects in Huang and Tang (1991), can induce the blocking just as subjects can (Xue, Pollard, and Sag 1994).

(4) a. Zhangsani zhidao Lisi j gaosu-guo nik youguan ziji * i/j/ *k de gongzuo.

Zhangsan know Lisi tell -Guo you about self De work ‘Zhangsan knew that Lisi told you about his/*your work.’

b. Zhangsani shuo Lisi j gen nik tan-guo ziji * i/j/ *k de shi. Zhangsan say Lisi with you talk-Guo self De business ‘Zhangsan said that Lisi talked about his/*your business with you.’

c. Zhangsani renwei Lisi j cong wok nar tingshuo-le ziji * i/j/ *k de fenshu. Zhangsan think Lisi from I there hear-say-Perf self De score ‘Zhangsan thinks Lisi heard from me his/*my score.’

d. Zhangsani zhidao Lisi j zai wo/nik jia xi ziji * i/j/ *k de zhaopian. Zhangsan know Lisi at I/you home develop self DE photo ‘Zhangsan knew that Lisi was developing his picture(s) at my/your home.’

Sentence (4a) indicates that a second person object can induce the blocking ef-fect on the third person matrix subject, Zhangsan, though it does not do so on its co-argument, the intermediate third person subject Lisi. Sentences like (4b) and (4c) show that first and second person pronouns in an adjunct (wo/ni ‘ I/you’ ), though not c-commanding or sub-commanding ziji , can also induce the blocking

Haihua Pan

282

effect on the third person matrix subject. These sentences, as pointed out in Xue, Pollard, and Sag (1994), pose a serious problem for analyses based on I-to-I relations and Spec-head (Infl) agreement (Battistella 1989, Huang & Tang 1991, Progovaç 1992, 1993, Cole & Sung 1994), which claim that only c-commanding and sub-commanding subjects can induce the blocking effect.

Following Huang and Tang (1991), Xue, Pollard, and Sag suggest that the blocking effect in sub-commanding cases can be influenced by discourse factors. They claim that, while the blocking effect in the latter is subject to unlike-person blocking, that in the former is constrained by animate blocking, as exempli fied by the contrast observed in (5a) and (5b). The inanimate subject NP naben shu ‘ that book’ in (5a) does not block long distance binding of ziji , but the animate subject NP Lisi in (5b) does.

(5) a. [Zhangsani de xin] j biaoming naben shuk hai-le ziji i/* j/k.

Zhangsan DE letter indicate that book harm-Perf self ‘Zhangsani letter indicates that that book harmed himi.’

b. [Zhangsani de xin] j biaoming Lisik hai-le ziji * i/* j/k. Zhangsan DE letter indicate Lisi harm-Perf self

‘Zhangsan’si letter indicates that Lisi harmed himi/himself.’ c. [Zhangsani de xin] j anshi Lisik hai-le ziji i/* j/k.

Zhangsan DE letter hint Lisi harm-Perf self ‘Zhangsani letter hinted that Lisi harmed himi/himself.’

However, contrary to their claim, the sub-commander Zhangsan in (5c) can bind ziji , though there is an intervening animate subject Lisi. The difference between (5b) and (5c) is the main verb. As for the verb anshi ‘hint’ , we have the follow-ing: if Zhangsan’ s letter hints at X, then Zhangsan hints at X, but this is not nec-essarily true for the verb biaoming ‘ indicate’ . So only in (5c) Zhangsan can bind ziji . This contrast suggests that different verbs influence the binding possibilit y of ziji . Hence sub-commanding and c-commanding cases behave similarly, and should not be treated separately. 3. THE BLOCKING EFFECT IS NOT SYMMETRICAL

The characterization of the blocking effect so far implies that the blocking ef-fect is symmetrical; actually this seems to be the general consensus (Tang 1989, Huang & Tang 1991, Xue, Pollard, and Sag 1994, etc.). However, closer ex-amination reveals that it is not all that symmetrical.2 Actually, all the data pre-sented so far, i.e. (1)-(5), show only that intervening first/second person pro-nouns induce a blocking effect on third person NPs, while intervening NPs with

Why the Blocking Effect?

283

third person features do not induce such an effect for third person NPs. What is not shown is whether intervening third person NPs can block first/second NPs from binding ziji . For the unlike-person blocking hypothesis to hold in its most general form, intervening third person NPs should also induce a blocking effect on first/second person pronouns. Unfortunately, this is not supported by Chinese data.

First, although intervening first and second person subjects block long dis-tance binding of ziji , third person subjects do not necessarily block long distance binding of ziji by first or second person NPs, as exemplified by the following sentences.

(6) a. Woi zhidao Lisi j bu xihuan ziji ?i/j.

I know Lisi not like self ‘I knew that Lisi did not like me/himself.’

b. Ni i xiang mei xiang guo Lisi j conglai jiu mei xihuan guo ziji ?i/j? you think not think Guo Lisi never conj not like Guo self ‘Have you ever thought about the idea that Lisi never liked you/himself?’

c. Woi yizhi yiwei Zhangsanj xihuan ziji i/j, keshi wo cuo le. I so-far think Zhangsan like self but I wrong Prt ‘I always thought that Zhangsan liked me, but I was wrong.’

(7) a. Woi bu xihuan Lisi j guan ziji i/j de shi. I not like Lisi interfere self De matter ‘I don’ t like Lisi interfering in my (own) business.’

b. Ni i xihuan Lisi j guan ziji i/j de shi ma? you like Lisi interfere self De matter Q ‘Do you like Lisi interfering in your (own) business?’

c. Lisi i bu xihuan wo/ni j guan ziji * i/j de shi. Lisi not like I/you interfere self DE matter ‘Lisi does not like me/you interfering in my/your (own) business.’

d. Lisi i xihuan Zhangsanj guan ziji i/j de shi. Lisi like Zhangsan interfere self DE matter ‘Does Lisi li ke Zhangsan interfering in his (own) business?’

(8) a. Dangshi woi pa Lisi j zai lai zhao ziji i/?j de mafan, jiu gei-le Then I afraid Lisi again come find self DE trouble Conj give-Perf ta 100 kuai qian. him CL money ‘At that time I was afraid that Lisi would come to cause trouble for me again, so I gave him 100 dollars.’

b. Dangshi ni i pa Lisi j zai lai zhao ziji i/?j de mafan, jiu gei-le Then you afraid Lisi again come find self DE trouble Conj give-Perf

Haihua Pan

284

ta 100 kuai qian, dui-bu-dui? him CL money right-not-right ‘At that time you were afraid that Lisi would come to cause trouble for me again, so you gave him 100 dollars. Is that right?’

c. Dangshi Lisi i pa woj zai qu zhao ziji * i/j de mafan, jiu gei-le Then Lisi afraid I again come find self DE trouble Conj give-Perf wo 100 kuai qian. I CL money ‘At that time Lisi was afraid that I would go to cause trouble for him again, so he gave me 100 dollars.’

Although the local third person subjects are the preferred antecedents of ziji , the first/second person matrix subjects in (6a) and (6b) are also possible antecedents. The first/second person matrix subjects in (6c), (7a), and (7b) are equally ac-ceptable antecedents. The contrast between (7a)/(7b) and (7c) is a strong piece of evidence arguing for the claim that the blocking effect is not symmetrical. The pragmatic factor involved is biased toward the matrix subject, so one would expect that the matrix subject could be the antecedent for all occurrences of ziji in (7). However, the matrix subject is not a possible antecedent only in (7c). If the symmetrical unlike-person account is correct, all the long distance binding should be blocked except for (7d), which is not consistent with the facts. The binding patterns in (8) further support our position. Thus, the binding patterns observed in (6), (7), and (8) argue strongly against the symmetrical view.

Furthermore, the asymmetry between third person NPs and first/second per-son pronouns also shows up in the sub-commanding cases. Although an inani-mate NP does not block long distance binding of ziji in (9), inanimate NPs with a possessive can block the long distance binding of ziji by third person NPs only if the possessive in question has the first or second person feature, as shown in (10) and (11).

(9) Zhangsani shuo naben shu hai-le ziji i.

Zhangsan say that book hurt-Perf self ‘Zhangsani said that that book hurt himi.’

(10) a. Zhangsani zhidao wo/ni j de baogao hai -le ziji * i/j. Zhangsan know I/you DE report hurt-Perf self

‘Zhangsan knew that my/your report hurt me/you.’ b. Woi zhidao Lisi j de baogao hai -le ziji i/j.

I know Lisi DE report hurt-Perf self ‘I knew that Lisi’s report hurt me/him.’

c. Zhangsani zhidao Lisi j de baogao hai -le ziji i/j. Zhangsan know Lisi DE report hurt-Perf self

Why the Blocking Effect?

285

‘Zhangsan knew that Lisi’s report hurt himself/him.’ (11) a. [Zhangsani de hua] j anshi [Lisik de xin] l hai-le ziji i/* j/k/* l.

Zhangsan DE speech imply Lisi DE letter harm-Perf self ‘Zhangsan'si words imply that Lisi’s letter harmed himi/himself.’

b. [Zhangsani de hua] j anshi [nik de xin] l hai-le ziji * i/* j/k/* l. Zhangsan DE speech imply your DE letter harm-Perf self

‘Zhangsan’si letter implies that you harmed himi/himself.’ c. [Ni i de hua] j anshi [Lisik de xin] l hai-le ziji i/* j/k/* l.

you DE speech imply Lisi DE letter harm-Perf self ‘Youri letter implies that Lisi harmed himi/himself.’

Thus, the animate feature of the intervening subjects is not the correct factor to differentiate the long distance binding possibilit y of ziji in sub-commanding cases, as claimed in Xue, Pollard, and Sag (1994). Rather it is the person feature, especially the contrast between first/second person pronouns and third person NPs, just like the c-commanding cases. Hence, we think that there is no need to differentiate the so-called different types of blocking effect---unlike person blocking for c-commanding cases and animate blocking for sub-commanding cases---and they should be dealt with by the same principle in the semantic and discourse modules.

Sentences like (12) (from Li 1993) further support our non-symmetrical view of the blocking effect: although the local subject is animate, the first and second person pronoun possessors in (12a) block the long distance binding of ziji by the matrix subject, while the third person possessor in (12b) does not.

(12) a. Baoyüi yiwei wode/nidej xueshengk bu xihuan ziji * i/* j/k. Baoyü think my/your student not like self

‘Baoyü thinks that my/your student does not like himself.’ b. Baoyüi yiwei Lisi j de xueshengk bu xihuan ziji i/* j/k.

Baoyü think Lisi DE student not like self ‘Baoyü thinks that Lisi’s student does not like him/himself.’

Therefore, the correct generalization for the blocking effect is that only interven-ing first/second person pronouns induce the blocking effect, while third person NPs do not necessarily do so, and first/second person pronouns in other gram-matical functions, not just subjects or those contained in a subject, can block the long distance binding of ziji . Thus, the blocking effect is more sensitive to dif-ferent person NPs than to local subjects. This descriptive generalization indi-cates that any analysis crucially depending on the properties of local subjects, e.g. the LF head movement analysis of ziji , will fail to explain the blocking ef-fect of long distance bound ziji .

Haihua Pan

286

4. PREVIOUS ANALYSES AND THEIR LIMITATIONS

Most of the previous analyses of ziji employ feature compatibili ty checking to explain the blocking effect, while Huang and Tang (1991) use the concepts of closeness and potential binder with feature compatibil ity in their account of ziji . In the first two subsections following I will discuss the problems for these two approaches.3, 4 In Section 4.3 I will point out the problems for the accounts based on logophoricity and perspectivity.

4.1 Feature Compatibility Checking

In order to explain the blocking effect, most of the previous GB analyses em-ploy feature percolation and subject-head agreement (Battistella 1989, Cole, Hermon and Sung 1990, Progovaç 1992, 1993, Cole, Hermon and Sung 1993, Cole and Sung 1994, Cole and Wang 1996, etc.). In these accounts it is assumed that subjects play a crucial role. The blocking effect is explained as follows: once ziji gets its person feature from the local subject or the possessor contained in the subject if the subject is inanimate, the potential antecedent will be checked against this feature. If there is a feature conflict, then we have the blocking ef-fect. Otherwise, long distance binding is allowed.

Although feature agreement checking can explain some of the blocking ef-fects of ziji , the following problems remain. First, feature agreement checking is symmetrical, since neither will a third person local subject agree in person fea-ture with a first/second person pronoun, nor will a first/second person pronoun with a third person NP. Thus, this account predicts that the blocking effect will be observed for both first/second person pronouns and third person NPs if the intervening NP has a different person feature. Although this prediction is true for the potential candidate with a first person feature and the intervening NP with a second person feature, or the other way round, it is not true for first/second person pronouns when the intervening NP has a third person feature. In the latter case the blocking effect is not always observed, and long distance binding is possible, as shown in (6), (7a), (7b), (8a), and (8b).

Second, inanimate subjects without possessors could be problematic for this approach, since, unlike the claim made in Tang (1989), ziji can have inanimate antecedents (Pan 1995, 1997), as exempli fied in the following.5 (13) [Meige gongyuan] i dou you (ta)-ziji i de tedian.

every-CL park all have (it)-self DE special-feature ‘Every park has its (own) special feature.’

Why the Blocking Effect?

287

Since inanimate subjects can be the antecedents of ziji , sentences like (14) below, if feature checking is applied, would be wrongly predicted to observe the block-ing effect, since the local subject and the matrix subject do not agree in person features. (14) Ni i shuo-guo naben shu hai-le ziji i ma?

you say-GUO that-CL book hurt-Perf self Q ‘Did youi say that that book hurt youi?’

Third, sub-commanding cases are also problematic for feature compatibili ty checking. According to feature percolation (under the assumption that inanimate subjects are not possible antecedents of ziji ), the sub-commanding possessor will percolate its person feature to the embedded subject and ziji will get this feature. Since the person feature of the possessors in (10a), (10b), (11b), and (11c) is not compatible with that of the matrix subject, the blocking effect should be ob-served for all the sentences. This prediction is only compatible with (10a) and (11b) but not with (10b) or (11c). Hence, sentences like (10b) and (11c) are not expected for the feature agreement checking approach. Furthermore, since the basic assumption for feature percolation (inanimate NPs cannot be antecedents of ziji ) is in doubt, the feature compatibilit y checking approach is in trouble.

Besides, first and second person non-subjects cause a serious problem for this approach, since feature checking (Spec-Head agreement) only applies to sub-jects or NPs contained in a subject. Thus, the approach in question makes either no prediction or wrong predictions to the sentences in (4).6

Finally, sentences like (12a) is even more problematic for the approach in question. Since the local subject is animate, feature percolation cannot apply. As the local subject and the matrix subject have compatible person feature, (12a) is predicted to allow long distance binding, which contradicts the fact.

4.2 Closeness and Potential Binders

Huang and Tang (1991) differentiate sub-commanding cases from the normal cases of ziji binding. Since their treatment of the latter also uses feature com-patibilit y checking to explain the blocking effect, and thus will have similar problems that the feature compatibilit y checking approach has, as pointed out in the previous subsection, I will focus on their treatment of the former which is dealt with by the following condition. (15) A reflexive α may take a NP β as its binder if (a) β sub-commands α; and

(b) There is no NP γ, γ a potential binder for α, such that γ is closer to α

Haihua Pan

288

than β is. (16) β sub-commands α iff β is contained in an NP that c-commands α or that

sub-commands α, and any argument containing β is in subject position. (17) A potential binder is any c-commanding animate subject in the governing

category of α (whether or not it is co-indexed with α).7 Notice that the proposed condition requires that the antecedents of ziji sub-command ziji , and long distance binding is possible only if there is no closer potential binder. Since a potential binder must be a c-commanding animate sub-ject, the analysis of ziji by Huang and Tang (1991) can avoid the problems caused by intervening inanimate subjects without possessors, as shown in (18). (18) Ni/Lisi i de xin biaoming naben shu hai-le ziji i.

You/Lisi DE letter indicate that-CL book hurt-Perf self ‘Your/Lisi’si letter indicates that book hurt you/himi?’

However, their analysis has the problems listed in the following. First, the con-dition defined in (15) above is an alternative way of extending the governing category of the reflexive, though it is not incorporated into the definition given in Binding Condition A. This suggests that there exists some redundancy in Huang and Tang’s account, since, with LF movement, there is no need to extend the governing category with additional rules, and LF movement provides it for free.

Second, the account of Huang and Tang (1991) is too strong in ruling out a remote human subject from binding the reflexive if closer subjects are animate but non-human NPs, as shown in (19) below. For these cases, Huang and Tang (1991) would predict that long distance binding of ziji is not possible, since the closer animate non-human subjects are potential binders, and thus should block the human sub-commander in the matrix subjects from binding ziji . However, this prediction is not consistent with the facts. In (19) following, the subjects in the embedded clause have the feature [+animate, –human], but they do not block long distance binding of ziji by the possessors in the matrix subjects. (19) a. Zhangsani de xin biaoming [NP Lisi j de gou ] k zai ziji i/?j/k de

Zhangsan DE letter indicate Lisi DE dog at self DE fangjian jiao-guo. room bark -GUO

‘Zhangsan’s letter indicates that Lisi’s dog barked in his room.’ b. Zhangsani de hua biaoming [NP yitiao gou ] j zai ziji i/?j de

Zhangsan DE speech indicate one -CL dog at self DE fangjian shuiji ao.

Why the Blocking Effect?

289

room sleep ‘Zhangsan’s words indicate that a dog was sleeping in his room.’

Similarly, other sub-commanding cases are also problematic for Huang and Tang’s analysis. Although they can handle the problematic cases for the feature agreement checking approach such as (10b) and (11c), they have problems with (10a) and (11b) that are not problematic for the feature agreement checking ap-proach. Since the inanimate local subjects (with or without possessors) are not potential binders, no blocking is expected under Huang and Tang’s analysis, which is at odds with examples like (10a) and (11b).

Although we can change the potential binder definition to c-commanding human subject to cover sentences like (19), the following sentences are still problematic for the new definition of potential binders. This is because under the new definition the local human subject should block the NPs farther away from binding the reflexive ziji , which contradicts the fact, as exempli fied in (20). (20) a. Ni i de xin anshi Lisik hai-le ziji i/k.

you DE letter hint Lisi harm-Perf self ‘Youri letter hinted that Lisi harmed youi/himself.’

b. Zhangsani de xin anshi Lisik hai-le ziji i/k. Zhangsan DE letter hint Lisi harm-Perf self

‘Zhangsan’si letter hinted that Lisi harmed himi/himself.’ Finally, examples like (12) given earlier are also problematic for Huang & Tang (1991). Neither the feature compatibil ity checking nor the condition defined in (15) can correctly predict the binding patterns in (12).

Since Huang & Tang only consider the closer c-commanding animate subject NPs and do not differentiate the person feature of the potential candidate and the competing closer NPs, their account of ziji fails to deal with the blocking effect, though they employ concepts like closeness and potential binders.

4.3 Logophoricity and Perspectivity

Some researchers argue that long distance binding should be accounted for by discourse factors such as logophoricity (Maling 1984, Zribi-Hertz 1989, Reinhart and Reuland 1991, 1993, Yu 1991, Yan Huang 1994, etc.), perspectiv-ity (Kuno 1987, Sells 1987, Zubin et al. 1990, N.-C. Li 1991, Iida 1992, etc.), emphasis, and intensive pronouns (Baker 1995). Others argue that thematic prominence or a combination of it with subject is the right condition to explain the long distance binding property of reflexives (Chou 1992, Xu 1993, 1994).

Haihua Pan

290

Clements (1975) uses the notion of logophoricity, which was originally intro-duced by Hagège (1974), to differentiate morphologically different pronouns in the studies of African languages. A logophor refers to an entity “whose speech, thoughts, feeling, or general state of consciousness are reported” (Clements, 1975: p. 141), and typically appears with predicates of communication and men-tal experience. Maling (1984) argues that Icelandic long distance bound reflex-ive sig is logophoric. Sells (1987) tries to reduce the notion of logophoricity into three more primitive ones: the SOURCE of the report, the person whose mental state or attitude is described by the content of the report (SELF), and the person from whose point of view the report is made (PIVOT). He claims that Japanese zibun is constrained by PIVOT. Zribi-Hertz (1989) argues that an English reflexive can be long distance bound if it refers to the minimal subject of consciousness (SC), which is very similar to the logophoricity account. Reinhart and Reuland (1991) also treat the exceptional cases of long-distance binding of reflexives as something like logophoricity, claiming that the deictic center plays a crucial role in the interpretation of long distance bound reflexives. Based on Kuno’s (1987) empathy concept, Iida (1992) argues that it is the speaker’s perspective that makes long distance binding possible for Japanese zibun. After discussing Sells’ approach to logophoricity, Stirling (1993) argues that it is possible to unify the three notions of Sells’ by the notion of validator, which has the feature [−first person] and is responsible for the truth of the sen-tence involved. Following Sells (1987), Yu (1991) and Y. Huang (1994) claim that Chinese ziji is logophoric, and thus must be constrained by logophoricity. Besides, Chen (1992) uses topicality and Sells’ PIVOT notion to account for the long distance binding properties of ziji , while Li (1991) argues that perspective-taking is the crucial factor for the interpretation of unbound and long distance bound ziji . In this subsection I will concentrate only on logophoricity and Li’s account, a representative of the perspectivity approach, and briefly discuss their limitations. 4.3.1 LONG DISTANCE BOUND ZIJI IS NOT A LOGOPHORIC PRONOUN

According to Clements (1975), Sells (1987), and Stirling (1993), a logophoric pronoun has the following three properties: (a) it does not exhibit the blocking effect; (b) it can always have the source or the subject of consciousness as its antecedent, though it is not compatible with perception verbs; and (c) it cannot have the first person pronoun as its antecedent. However, ziji does not have any of the three properties. First, ziji exhibits the blocking effect, as pointed out in Section 1, namely that an intervening first/second person pronoun blocks all the remote third person non-co-argument subjects from binding ziji , but a third per-son NP does not necessarily block first/second person pronouns from binding

Why the Blocking Effect?

291

ziji long distance. Second, ziji cannot have the NP carrying the role source as its antecedent, as shown in (21). (21) Zhangsani cong Lisi j nar tingshuo naben shu hai-le ziji i/* j.

Zhangsan from Lisi there hear that-CL book hurt-Perf self ‘Zhangsan heard from Lisi that that book hurt himself.’

Also, ziji cannot have the subject of consciousness as its antecedent as long as there is an intervening first/second person pronoun, cf. (10a) and (12a), though this would be possible if ziji were a logophoric pronoun.

Third, although the logophoricity theory can account for long distance bind-ing in belief contexts or sentences with embedded complement clauses when there is no first/second person pronoun intervening, it fails to account for similar sentences which contain an intervening first/second person pronoun, e.g. (10a) and (12a). The theory also fails to explain why long distance binding is possible in the following sentences (Carl Lee Baker’s lecture notes): (22) a. Johni minglin Bill j [S PRO gei ziji i/j guahuzi] .

John order Bill to self shave ‘John ordered Bill t o shave him/himself.’

b. Johni bi Bill j [S PRO gei ziji i/j guahuzi] . John force Bill to self shave

‘John forced Bill t o shave him/himself.’ In (22) the local domain is the embedded infinitive clause indicated by S. Since the matrix object Bill controls PRO, Binding Condition A can only predict that Bill is the antecedent of ziji . That is, it fails to predict that the matrix subject can also be the antecedent. Since John in (22) is hardly a source, nor are sentences in (22) reports about John’s feeling, thoughts, etc., the binding pattern observed in (22) is problematic for the logophoricity account.

Besides, ziji can appear in the perception verb context, and can be long dis-tance bound, as long as there is no intervening first/second person pronoun, as exempli fied in (23). (23) Zhangsani tingshuo [NP Lisi j de gou ] k zai ziji i/j/k de fangjian jiao.

Zhangsan hear Lisi DE dog at self DE room bark ‘Zhangsan heard that Lisi’s dog is barking in his own room.’

Finally, logophoric pronouns are said to be in complementary distribution

with first person pronouns, so they cannot have them as their antecedents, i.e. the (c) property of logophoric pronouns. However, long distance bound ziji does

Haihua Pan

292

not have the (c) property of a logophoric pronoun. It can readily have the first person pronoun wo ‘ I’ as its antecedent, as shown in sentences like (6a, c), (7a) and (8a).

Since none of the three properties of logophoric pronouns are possessed by ziji , it is very diff icult to argue that ziji is a logophoric pronoun.8 4.3.2 THE PERSPECTIVE-TAKING ACCOUNT IS NOT ADEQUATE

N.-C. Li (1991) argues that Chinese long distance bound ziji is constrained by

perspective-taking, and it is compatible only with reflective mental states. She differentiates unbound ziji and long distance bound ziji . Although both have the experiencer as antecedents, they differ in the sense that the former appears in the expressive framing which does not contain the coding of the experiencer, while the latter occurs in the reportive framing with the coding of the experiencer. She suggests that the former is related to the erperiencer’ s private experience, and the latter must refer to the experiencer who is in a mental state of self-awareness. The prediction of her analysis of long distance bound ziji is that, in statements about facts and pure emotion or perception, ziji cannot be long distance bound. Li’s account can provide an explanation to why first/second person pronouns cannot always bind ziji long distance when there is a third person human NP intervening, since according to her, only verbs that are related to the experi-encer’s private experience, or to the reflective consciousness of the experiencer allow the long distance binding of ziji . Although I agree with her on the role played by verbs and on the claim that ziji can have the projected ego as its ante-cedent, Li’s analysis is not adequate in the following aspects.

First, the sentence below contains a factive verb zhidao ‘ know’ , so it should not allow ziji to appear in it according to Li’s account. This is because it is not guaranteed that Zhangsan is always self-conscious, as the belief can be attrib-uted to him by the speaker without his awareness. However, ziji can appear in sentences like this, which suggests that self-consciousness is not a necessary condition for ziji to have a long distance antecedent. (24) Zhangsani zhidao naben shu dui ziji i buli .

Zhangsan know that-CL book to self not-good ‘Zhangsan knows that book is not good for him.’

Second, sentences containing non-reflective reports also allow long distance

bound ziji to appear in them, which is not compatible with Li’s account, because there should be no coding of the experiencer in non-reflective reports, but long distance bound ziji must point to the projected ego of the experiencer, which requires the coding of the experiencer.

Why the Blocking Effect?

293

(25) Zhangsani tingshuo Lisi qiangzou-le ziji i de shu.

Zhangsan hear Lisi rob-Perf self DE book ‘Zhangsan heard that Lisi took away his (own) book.’

Also perception verbs do allow long distance bound ziji , as exempli fied in (23) and (25) given in the previous subsection.

Finally, Li does not provide an account to explain the blocking effects. Even with the highly reflective verb yiwei ‘ think’ , ziji cannot be long distance bound in the relevant sentences, as exempli fied in (26). (26) Zhangsani yiwei wo/ni mei nazou ziji * i de shu.

Zhangsan think I/you not take self DE book ‘Zhangsan thought I/you did not take away his book.’

From the discussions in this section we can see that previous approaches do

not provide a satisfactory account for the blocking effect of long distance bound ziji . Thus, an alternative account is called for. 5. SELF-ASCRIPTION AND LONG DISTANCE BOUND ZIJI Although I agree with the view that the properties of Mandarin reflexive zij i cannot be properly accounted for by pure syntactic conditions like Binding Con-dition A, and discourse and pragmatic principles are necessary to account for its properties, I think that the correct condition for long distance bound ziji is not one that utili zes concepts like logophoricity or perspectivity, but one that em-ploys self-ascription, as will be argued in the rest of this chapter. Note that my account assumes that the locally bound ziji is dealt with by the Anaphor Condi-tion as proposed in Pan (1998), which roughly corresponds to the category---syntactic reflexives---in Pollard and Xue (1998).

5.1 My View of Mandarin Reflexives

In Pan (1995, 1997) I divide Chinese reflexives into contrastive and non-contrastive reflexives.9 The contrastive reflexive category also includes mor-phological reflexives like benren ‘proper-person’ , benshen ‘proper-body’ , zishen ‘ self-body’ , and their compound forms. The non-contrastive reflexive category includes locality and self-ascription reflexives. Figure 1 below represents my current view on Chinese reflexives. Notice that the category self-ascription is now under the new category long distance bound reflexives that also includes

Haihua Pan

294

other long distance bound reflexives such as the ones that are constrained by logophoricity. For example, Mandarin compound reflexive taziji can be used logophorically. See the relevant discussions in Pan (1998) and Section 6 follow-ing for the differences between logophoricity and self-ascription.

Morphological Reflexives

Non-contrastive Contrastive (Local or LD bound)

Locality LD Bound Reflexives ziji , ta-ziji , wo-ziji ,

benren, benshen, etc.

ziji , ta-ziji Self-Ascription Others (including logophoricity)

ziji ta-ziji

Figure 1. My Current View of Mandarin Reflexives In Pan (1997) I argue that the three different categories of reflexives are sub-

ject to different conditions. The contrastive ones, besides the contrastive condi-tion, are also subject to a prominence condition like centrality, as suggested in Baker (1995), namely that a contrastive reflexive must have the central character in the discourse as its antecedent.10 The locality reflexive is constrained by a revised condition on the basis of Chomsky’s (1981) Binding Condition A. The condition utili zes the concepts of closeness and prominence (see Pan (1998) for relevant discussion). Since only long distance bound reflexives including self-ascription ziji observe the blocking effect, I will not discuss locality reflexives and contrastive reflexives any further in this chapter.

5.2 Beliefs de re, de se, and de dicto

Before getting to my analysis of self-ascription ziji , let me first introduce some necessary concepts for subsequent discussion. There are three kinds of belief. Belief de re is a belief about an entity. Belief de se, introduced in Lewis (1979), is a belief about the believer him/her-self, which corresponds to self-

Why the Blocking Effect?

295

ascription. It involves self-ascribers, and implies that the believer (consciously) ascribes a property to him/her-self. Belief de dicto is a belief about a proposi-tion and quantifying-in is not allowed, i.e., the elements in the belief cannot be accessed externally, i.e. outside the scope of the believer. Lewis (1979) suggests that belief de dicto is subsumed under belief de se, since, if a person believes a proposition, then we can say s/he has a belief about her/him-self that s/he lives in the world where the relevant proposition is true, and thus this person has a de se belief about him/her-self, which involves the relevant proposition. Because the de se belief in the de dicto case is attributed to him/her by the speaker, the person in question is not necessarily aware of it, and thus s/he is not necessarily self-conscious, cf. endnote 12.

According to Lewis (1979), self-ascription can be applied to any attitude re-lated to belief, knowledge, and desire. It also applies to any attitude that amounts to lack of belief, knowledge or desire. This latter case is possible, because, as noted before for de dicto beliefs, the speaker can attribute the relevant de se be-lief to the person in question without his/her awareness.

Connected with self-ascription are the concepts of self-ascriber and property: a self-ascriber ascribes a property to him/her-self. First/second person pronouns are obligatory self-ascribers, while third person human NPs are optional self-ascribers, as shown in (27). (27) a. I think I am smart.

b. You think you are smart. c. John thinks he is smart.

Unlike the referent of John in (27c) who can have either a de re or de se belief about himself, the referents of I and you in (27) can only have a de se belief about themselves. The above difference indicates that, whenever the speaker or addressee ascribes some property to him/her-self, it must be a de se belief, or a self-ascription. Note that the ambiguity in (27c) comes from the third person pronoun he. It can be interpreted as a referential pronoun or a bound variable pronoun, whereas the first occurrence of first and second persons in (27a) and (27b) is not a bound variable pronoun, and the second occurrences of them have to be bound variable pronouns.11

Furthermore, although third person NPs, e.g. John, can be referred to as the believer/carrier of belief, or the entity named John, first/second person pronouns can be referred to as the believer/carrier of belief only. Since first and second person pronouns have obligatory de se beliefs, they are obligatory self-ascribers in the context of belief de se, whereas third person NPs are optional self-ascribers in the de se context.

Haihua Pan

296

In the discussion following we assume that all the sentences in the language represent a belief, knowledge, or desire of the carrier. Even for simple sentences like Zhangsan xihuan Lisi ‘Zhangsan likes Lisi’ ---a statement made by the speaker---is a belief of the speaker. Note that the sentence like Zhangsan zhidao Lisi xihuan Wangwu ‘Zhangsan knows that Lisi li kes Wangwu’ can also be a belief of the speaker. Besides, there are two other situations (readings) for this sentence: (a) Zhangsan self-ascribes the belief ‘ Lisi li kes Wangwu’ ; and (b) the speaker attributes to Zhangsan the de se belief---a belief about himself---that he lives in the world where the proposition Lisi li kes Wangwu is true. In both situa-tions, Zhangsan has the knowledge that he lives in the world where the proposi-tion Lisi li kes Wangwu is true, though in the (b) reading Zhangsan is not self-conscious, as he may not know that he was attributed a de se belief, and the speaker can stand at his place, viewing the object in question, in order to satisfy the requirement of self-ascription (See more relevant discussion in Section 6 in this chapter.). Also note that, for the latter sentence Zhangsan zhidao Lisi xi-huan Wangwu, we have one self-ascription embedded in another. The speaker self-ascribes the following belief: that Zhangsan self-ascribes the belief that Lisi likes Wangwu.

5.3 Self-Ascription Ziji Points to the Carrier of Belief

N.-C. Li (1991) points out that there is an unbound ziji which refers to the speaker, as exempli fied here:

(28) Ziji neng qu nar ma?

self can go there Q ‘Can self (I) go there?’

She claims that this ziji is referential, which is why it can be used alone. How-ever, I would rather say that ziji is bound in a discourse in which (the projected ego of) the speaker is its antecedent. Besides the speaker, an addressee can also be the antecedent of the so-called unbound ziji when the speaker addresses the addressee using (29). Note that (29) can also be used to talk about a third party that is salient in the discourse. (29) Ziji weishenme bu qu ne?

self why not go Q ‘Why didn’ t self (you) go?’

Furthermore, a sentence with an object ziji can also have the speaker as ante-

cedent if the speaker asks him/her-self the question as in (30a).

Why the Blocking Effect?

297

(30) a. Zhangsan xihuan ziji ma?

Zhangsan like self Q ‘Does Zhangsan like (my)self?’

b. Zhangsan zhidao Lisi xihuan ziji ma? Zhangsan know Lisi li ke self Q

‘Does Zhangsan know that Lisi li kes myself/him/himself?’ This is true even for a complex sentence like (30b). Besides the normal readings with Zhangsan and Lisi as the antecedents of ziji , the speaker can also be the antecedent of ziji in (30b), though I admit that this reading is more difficult to get than the ones with Zhangsan and Lisi. Note that this reading is highly reflec-tive; actually all the similar readings are presented in the form of a question, which implies that the speaker has an image of himself who has the property denoted by the sentence in question. By contrast, one of the readings in (30b), i.e. with Zhangsan as antecedent, need not be reflective, since the speaker can attribute Zhangsan a de se belief λx[Lisi xihuan x] ; in this case Zhangsan is not necessarily self-conscious, as he may not know it.12 This reading suggests that Li’s requirement that the antecedent of long distance bound ziji must be the ex-periencer of a reflective mental state is not the correct characterization of long distance bound ziji for third person believers, though she may be right with re-spect to first/second person pronouns. We can say that Zhangsan may self-ascribe the knowledge about the fact that Lisi li kes him. Even his lack of self-ascribing the knowledge also falls in the domain of self-ascription, as pointed out by Lewis (1979), and repeated in the previous subsection. This is because the speaker can attribute Zhangsan a de se belief without his awareness, and it is the speaker who makes the connection between Zhangsan and ziji . Note that in this case Zhangsan may be conscious, but he is not self-conscious. What is cru-cial here is that Zhangsan is not necessarily self-conscious, and whether he is conscious or not is not relevant to our discussion of self-ascription.

From the discussion above we can see that long distance bound ziji points to the carrier of belief, knowledge or desire, and it is thus constrained by self-ascription (see more relevant discussions in Pan 1995, 1997). 5.4 The Condition for Self-Ascription Ziji

Assuming the discussion above, I propose the following condition for self-ascription ziji . (31) The Condition for Self-Ascription Ziji

Haihua Pan

298

Ziji can be bound to the carrier of belief, the most prominent self-ascriber, in a linguistic domain γ iff there is no blocker in the believed proposition contained in γ.

(32) The Prominence Condition α is the most prominent self-ascriber in γ iff there is no β in γ such that β appears higher in one of the following hierarchies than α.

a. SUBJ > OBJ or OBLIQUE b. Dominating NPs > Dominated NPs

(33) α is a blocker for β if α is a self-ascriber such that (a) α precedes ziji ; and (b) neither α nor the NP controlled by it is an argument of an irreflexive predicate containing ziji . In a de se belief situation, self-ascribers include all the referents of the animate subjects of attitudinal verbs, and first and second person pronouns. The hierar-chy in (32b) only compares NPs that have dominance relationships with one another. The domain γ includes the clause that contains at least one attitudinal verb, and one sentence may have more than one domain if it contains more than one attitudinal verb. The believed proposition is the complement clause of the attitudinal verb under consideration.

Now I will show that the condition in (31) can handle all the long distance bound cases of ziji , and can provide a natural account to the blocking effect. (34) a. Johni yiwei Bill j xihuan ziji i/j.

John think Bill li ke self ‘ Johni thinks Bill j likes himi/himself j.’ b. Johni yiwei [naben shu] j hai-le ziji i/* j.

John think that-CL book hurt-Perf self ‘ Johni thinks that book hurt himi.’ c. Johni yiwei wo/ni j xihuan ziji * i/j.

John think I/you like self ‘ Johni thinks I/youj likes himself* i/my/your-self j.’ d. Woi yizhi yiwei Bill j xihuan ziji i/j, keshi wo zuo le.

I always think Bill like self but I wrong Prt ‘I i always thinks Bill j likes mei/himself j, but I was wrong.’

(35) a. [S1 Johni zhidao [S2 Bill j juede Markk xihuan ziji i/j/k]] . John know Bill think Mark like self

‘ John knows that Bill t hinks that Mark likes himself/him.’ b. [S1 Johni juede [S2 wo/ni j zhidao Markk xihuan ziji * i/j/k]] .

John think I/you know Mark like self ‘ John thinks that I/you know that Mark likes himself/me/you.’

Why the Blocking Effect?

299

In (34a) the domain γ is the whole sentence, and the believed proposition is the embedded clause. Since the local subject is not a first or second person pronoun, and neither is it the subject of an attitudinal predicate, it is not a self-ascriber. Because there is no self-ascriber in the believed proposition contained in γ and thus no blocker, the only self-ascriber, the matrix subject, will be assigned the antecedent of ziji . Hence, (31) predicts that long distance binding is possible in (34a), which is consistent with the fact. (The coindexation with j is due to the locality ziji, which is not the concern of this chapter.) In (34b), like (34a), the domain γ for long distance bound ziji is the matrix clause, and the local subject is not a self-ascriber. Since only the matrix subject is a self-ascriber and no other NPs are more prominent, it can be the antecedent of ziji as predicted by (31). (The locality ziji in (34b) is excluded independently, because ziji carries the experiencer role, but naben shu ‘ that book’ has the inanimate feature, and thus they are incompatible with each other.) In (34c), the first/second person pronoun wo/ni in the embedded clause, a self-ascriber, precedes ziji and is not an argument of an irreflexive predicate containing ziji , so it is a blocker in the be-lieved proposition, according to (33). Thus, (31) would predict that long-distance binding of ziji is not possible, which is consistent with the fact. In (34d) the matrix subject in the first conjunct can be the antecedent of ziji just like in (34a), since the local subject is not a self-ascriber, and is thus not a blocker.

In (35a) there are two domains for the long distance bound ziji : the matrix clause (S1) and the intermediate clause (S2). For domain S1, since there is a self-ascriber in the believed proposition S2, i.e. the subject of the attitudinal predicate juede ‘ think’ in S2, this self-ascriber may function as a blocker for the matrix subject if it is an obligatory self-ascriber. But as pointed out in the previous sub-section, third person NPs are optional self-ascribers, so when the intermediate subject does not function as a self-ascriber, and is thus not a blocker, the matrix subject can be the antecedent of ziji . As for the intermediate subject in domain S2, similar to (34a), (31) allows it to be the antecedent of ziji . Thus (31) cor-rectly predicts the possible readings in (35a).

In (35b), the first/second person pronouns in the believed proposition S2 pre-cede ziji and are not an argument of an irreflexive predicate containing ziji , so they function as blockers for the matrix subject, just like in (34c). Since they are obligatory self-ascribers, the intervening first/second person pronouns will block the matrix subject from binding ziji . Note that, unlike the matrix subject, first/second person pronouns in (35b) can be the antecedents of ziji , as they are also the most prominent NPs in the domain (S2), and there are no (other) self-ascribers in the believed proposition in this domain.

Since all the obligatory self-ascribers discussed so far are subjects, let us ex-amine cases in which the obligatory self-ascribers are non-subjects, to see if (31) also makes correct predictions.

Haihua Pan

300

(36) a. Johni shuo Bill k gei wo/ni j kan-guo ziji * i/j/k de shu.

Jogn say Bill to I/you see-GUO self DE book ‘John said that Bill showed me/you his/my/your book.’ b. Johni shuo Bill j gaosu-guo ni/wok Markn da-le ziji * i/?j/*k/n yixiar.

John say Bill tell -Guo you/I Mark hit-Perf self once ‘John said that Bill t old you/me that Mark hit himself/him once.’

In (36a) the obligatory self-ascriber wo/ni ‘ I/you’ are non-subjects; they are obliques indicated by preposition gei ‘ to.’ Since there is no requirement of sub-jecthood in the definition of blockers in (33), (31) would predict that the obliga-tory self-ascriber in (36a), though not a subject, can block the long distance binding of ziji by the matrix subject, because the obligatory self-ascribers wo/ni in the believed proposition---the intermediate clause---are blockers for the ma-trix subject, according to (33); note that they precede ziji and are not an argu-ment of an irreflexive predicate containing ziji . This prediction is borne out by the impossibilit y of long distance binding of ziji observed in (36a). In (36b) the obligatory self-ascribers wo/ni are not subjects, either; they are objects. Just as in (36a) the object wo/ni in (36b) blocks the long distance binding of ziji by the matrix subject, as predicted by (31). This is because the obligatory self-ascriber first/second person pronouns in the intermediate clause also function as blockers for the matrix subject.

Note that there is a difference between (36a) and (36b): the first/second per-son pronouns do not block the local subject from binding ziji, though they do for non-local subjects except for the co-arguments of f irst and second person pro-nouns.13 This is a difference between locality ziji and long distance bound ziji : only the latter observe the blocking effect. Also note that the fact that the local subject can be antecedent of ziji with or without the first/second person pro-nouns intervening is accounted for by the condition on locality ziji.14 Hence, it is not a problem for our account of long distance bound self-ascription ziji .

Condition (31) also predicts that long distance binding of ziji is possible if there is an obligatory self-ascriber that does not intervene between the potential antecedent and ziji . This prediction seems to be correct, as the contrast between the sentences in (37) exempli fies. (37) a. Johni shuo Bill j ba ziji i/j/ *k de shu songgei-le wo/nik.

John say Bill BA self DE book give-Perf I/you ‘ John said that Bill gave his own books to me/you.’

b. Johni shuo Bill j songgei-le wo/nik yiben ziji * i/j/?k de shu John say Bill give-Perf I/you one-CL self DE book

‘ John said that Bill gave me/you one of his own books.’

Why the Blocking Effect?

301

The obligatory self-ascribers wo/ni are obliques in the embedded clause, but they do not intervene between the matrix subject and ziji in (37a), though they do in (37b). Since the obligatory self-ascribers wo/ni do not block long distance binding only in sentences like (37a), we can say that obligatory self-ascribers can block long distance binding of ziji only if they intervene between the two NPs under discussion. This is why we require the self-ascriber to precede the reflexive in the definition of blockers in (33a).

Sentences like (37a) suggest that an account stipulating that ziji must take the perspective of the sentence in question is not adequate, as also argued in the previous section. Under the perspective-based account the first or second person NP in the embedded clause would be the deictic center; note that word order should be irrelevant in an account based on perspectivity. Thus the first or sec-ond person NP would have to be the perspective-taker. That is, an account em-ploying perspectivity would predict that the long distance binding of ziji in (37a) is blocked, since a first/second person pronoun---a perspective-taker---appears in it. This account would also predict that the referent of the first or second per-son NP can always be the antecedent of ziji , a prediction that is at odds with the binding pattern observed in (37a). Hence, the facts here further show that the account proposed in this section is superior to an account based on perspectivity.

Another property of (31) is that it requires the antecedent of ziji to be the most prominent self-ascriber. Let us see if this requirement is fulfill ed in (38) below. (38) a. Bill i cong wo/ni j nar tingshuo Suek piping-le ziji i/* j/k.

Bill from I/you there hear Sue criticize-Perf self ‘Bill heard from me/you that Sue criticized herself/him.’ b. Wuqing de shishi i gaosu Markj Suek pian-le ziji * i/j/k.

cruel DE fact tell Mark Sue cheat-Perf self ‘The cruel fact told Mark that Sue cheated herself/him.’

In (38a) there is no intervening self-ascriber in the believed proposition, the em-bedded clause, so the NPs in the matrix clause are possible antecedents of ziji . If there is no requirement of the most prominent self-ascriber, then both the ma-trix subject and the oblique first/second person pronouns, obligatory self-ascribers, should be possible antecedents of ziji . However, the fact that only the subject can be the antecedent suggests that some kind of prominence relation is necessary. Such a requirement is specified in (31). Note that sentences like (38b) are not counterexamples to our account. This is because the inanimate subject is not a self-ascriber in nature, and thus the matrix object can be the an-tecedent of ziji in (38b).

Haihua Pan

302

Sentence (39), similar to (12a), suggests that the prominence requirement should not be imposed on the blockers (self-ascribers), and our account correctly predicts the long distance binding of ziji is not possible in (39). This is because the first person pronoun in the believed proposition is a blocker for the matrix subject, as it precedes ziji and is not an argument of an irreflexive predicate containing ziji , according to (33). (39) Baoyüi yiwei wodej xueshengk bu xihuan ziji * i/* j/k. Baoyü think my student not like self

‘Baoyü thinks that my student does not like himself.’ Our account can also cover cases like (40) and (41) which are claimed to be evidence for the influence of pragmatic factors (world knowledge) in the deter-mination of the antecedents of ziji .15 (40) Zongtongi qing woj [S PROj zuo zai ziji i de shenbian].

president ask I sit at self DE side ‘The president asked me to sit beside him.’

(41) a. Zhangsan pa [S wo/ni hui chaoguo ziji ]. Zhangsan fear I /you will surpass self

‘Zhangsan fears that I/you will surpass him.’ b. Ni bu pa [S ta chaoguo ziji ma]?

you not fear he surpass self Q ‘Don’ t you fear that he surpasses you?’

Although the believed proposition, the embedded clause in (41a), contains a first/second person pronoun, we do not take the first/second person pronoun as a blocker for the matrix subject, since the predicate in the embedded clause is irre-flexive, and thus precludes co-reference of its arguments: wo/ni and ziji . This is specified in our definition of blockers in (33). Since the first/second person pro-noun is not a blocker for it, the matrix subject in (41a) can be the antecedent of ziji , according to (31). Example (40) can be dealt with similarly. The irreflexive predicate is x zuo zai y de shenbian ‘ x sits beside y,’ where x is PRO, and y is ziji . The only self-ascriber wo in (40) cannot be a blocker, since it controls PRO which is an argument of the irreflexive predicate in question. Thus, there is no blocker for the matrix subject Zongtong ‘president,’ according to (33). Hence, the matrix subject Zongtong ‘president’ can be the antecedent of ziji , as ex-pected.16, 17

Condition (31) also allows the co-arguments of first/second person pronouns to be the antecedents of long distance bound ziji , as shown in (36b) and the sen-tence following.18

Why the Blocking Effect?

303

(42) Zhangsani gaoshu woj [S ziji i/* j mei bei dahui xuanshang].

Zhangsan tell I self not by conference select ‘Zhangsan told me that he was not selected by the conference.’ According to (31), the believed proposition, the embedded clause S, does not contain a self-ascriber, so there is no blocker for the matrix subject, Zhangsan, to bind ziji . Hence long distance binding is possible in (42). Similarly, the fact that the intermediate subject, Bill , in (36b) can be an antecedent of ziji is also predicted by our condition (31).19

Note that, although the existence of first and second person pronouns is cru-cial for the blocking effect, there is no need to stipulate in condition (31) the obligatoriness of self-ascribers, as shown by the following example (43) (Carl Lee Baker, p.c.).20

(43) John renwei Bill zhidao Mark ba ziji de shu jie-gei-le John think Bill know Mark BA self DE book loan-to-Perf

ziji de pengyou. self DE friend ‘John thinks Bill knows that Mark has loaned his book to his friend.’

(44) a. Johni renwei Bill j zhidao Markk ba ziji i de shu jie-gei-le John think Bill know Mark BA self DE book loan-to-Perf ziji i de pengyou. self DE friend ‘John thinks Bill knows that Mark has loaned his book to his friend.’

b. Johni renwei Bill j zhidao Markk ba ziji j de shu jie-gei-le John think Bill know Mark BA self DE book loan-to-Perf ziji j de pengyou. self DE friend ‘John thinks Bill knows that Mark has loaned his book to his friend.’

c. *Johni renwei Bill j zhidao Markk ba ziji i de shu jie-gei-le John think Bill know Mark BA self DE book loan-to-Perf

ziji j de pengyou. self DE friend ‘John thinks Bill knows that Mark has loaned his book to his friend.’

d. *Johni renwei Bill j zhidao Markk ba ziji j de shu jie-gei-le John think Bill know Mark BA self DE book loan-to-Perf ziji i de pengyou. self DE friend ‘John thinks Bill knows that Mark has loaned his book to his friend.’

Haihua Pan

304

For ease of exposition, I have represented the long distance bound readings of sentence (43) individually in (44). The possible readings for the two occur-rences of ziji are that either John or Bill can bind both occurrences of ziji . Note that the mixed readings are not possible as shown in (44c) and (44d). Thus, sen-tence (43) suggests that, like first/second person pronouns, third person NPs, when functioning as self-ascribers, will block other NPs outside the domain from binding ziji . So we do not need to specify obligatory self-ascriber in the if-clause in condition (33). Note that sentence (43) does have mixed readings with the local subject, as exempli fied here: (45) a. Johni renwei Bill j zhidao Markk ba ziji i de shu jie-gei-le

John think Bill know Mark BA self DE book loan-to-Perf ziji k de pengyou. self DE friend ‘John thinks Bill knows that Mark has loaned his book to his friend.’

b. Johni renwei Bill j zhidao Markk ba ziji j de shu jie-gei-le John think Bill know Mark BA self DE book loan-to-Perf ziji k de pengyou. self DE friend ‘John thinks Bill knows that Mark has loaned his book to his friend.’

c. Johni renwei Bill j zhidao Markk ba ziji k de shu jie-gei-le John think Bill know Mark BA self DE book loan-to-Perf ziji i de pengyou. self DE friend ‘John thinks Bill knows that Mark has loaned his book to his friend.’

d. Johni renwei Bill j zhidao Markk ba ziji k de shu jie-gei-le John think Bill know Mark BA self DE book loan-to-Perf ziji j de pengyou. self DE friend ‘John thinks Bill knows that Mark has loaned his book to his friend.’

In (45) ziji k is locally bound, while ziji i/j is long distance bound. The sentences in (45) show that the local subject and the remote subjects can have mixed read-ings. The binding patterns in (44) and (45) indicate that local subjects function differently from long distance ones. They also suggest that locally bound zij i and long distance bound ziji are constrained by different conditions, as sug-gested in Pan (1995, 1997).

From the discussion above we can see that our account of long distance bound ziji fares better than do the previous analyses presented in this chapter. Hence, the condition in (31) is an appropriate condition for self-ascription ziji .

Why the Blocking Effect?

305

Now we can answer the question: why is there a blocking effect for long dis-tance bound ziji? It is because the difference between the first/second person pronouns and third person NPs. The blocking effect is observed because of the obligatoriness of f irst/second person pronouns being a self-ascriber if they do not agree in person features with the carrier of belief, knowledge, or desire in-volved. Since ziji points to the carrier of belief, a self-ascriber, the intervening obligatory self-ascribers will prevent it from being bound by farther-away self-ascribers. Hence, the blocking effect. 6. DISCUSSION

Like Huang and Tang (1991) and Cole and Wang (1996), we recognize the need of separating blockers from antecedents. As one can see from our condi-tion on self-ascription ziji given in (31), our concept of blockers include only self-ascribers that are (a) first/second person pronouns that precede the reflexive ziji and are not an arguments of an irreflexive predicate that contains ziji ; and (b) animate NPs that are subjects of a belief, knowledge or desire. There is no prominence restriction for being a blocker for first/second person pronouns, though there is one on being the antecedent of long distance bound ziji . Unlike Huang and Tang (1991) but in line with Cole and Wang (1996), our account also suggests that a blocker is not necessarily an antecedent of the reflexive ziji , cf. (12a) (See more discussion on the difference between blockers and antecedents in Cole & Wang (1996) and Pan (1998)).

Another issue is the relationship between the concepts of logophoricity and self-ascription. Since these concepts are closely related, it is very easy to con-fuse them. In the following I will briefly discuss their differences.

Logophoricity and self-ascription differ from each other in the following as-pects. First, according to Sells (1987), the SOURCE of the report is one primi-tive of the concept of logophoricity. However, self-ascription can never be at-tributed to the individual who carries the source role if it is not also the carrier of belief, which is why a source cannot be the antecedent of long distance bound ziji , as shown in (21). Second, the subject of consciousness is another primitive in Sells’ concept of logophoricity (SELF). By contrast, the concept of self-ascription does not necessarily require self-consciousness, for example, (46) would present a problem for the requirement of self-consciousness of the subject, whereas, in the self-ascription concept, self-consciousness is not necessary. This is because, in addition to the reading in which Zhangsan is self-conscious, i.e., that he self-ascribes the relevant belief, the de se belief can also be attrib-uted to Zhangsan by the speaker without his awareness. In this latter reading, Zhangsan is not aware of the fact that the speaker attributed him a de se belief.

Haihua Pan

306

This is what I meant earlier that in self-ascription, the person, denoted by a third person NP, who has the de se belief, need not be self-conscious, as the speaker can attribute him the relevant de se belief. Notice that the consciousness of Zhangsan is not relevant here, and what is crucial is the self-consciousness con-cept, cf. Endnote 12. Even for the positive version of (46), self-consciousness is not necessary for it to be true.21 Just like the negative version, Zhangsan can be attributed a de se belief by the speaker without his awareness. The clearer ex-ample is the question version of (46) given in (30b), since there is no guarantee that the answer to the question is positive. (46) Zhangsan bu zhidao Lisi xihuan ziji .

Zhangsan not know Lisi like self ‘Zhangsan does not know that Lisi li kes him.’

Self-ascription is also different from the third primitive in Sells’ logophoricity

concept, point of view (PIVOT). Self-ascription is person neutral, namely that it can be combined with all three persons, whereas point of view has separate per-sons: first, second or third person point of view, depending on which NP is the point of reference (Mitchell 1986). The closest one among the three points of view to self-ascription is the first person point of view. However, there are at least two differences between them, as far as I can see. (47) a. Zuozi zai wo de zuobian.

table at I DE left ‘The table is to my left.’

b. Zuozi zai ziji de zuobian. table at self DE left ‘The table is to self’ s left.’

The first difference between wo ‘ I’ and ziji is that the interpretation for (47a)

is fixed to the speaker, but that for (47b) is not. In (47b), ziji can refer to the speaker, the addressee, or a third party individual. The last reading is possible if the speaker and addressee are talking about a particular third person. This dif-ference becomes clearer if we embed (47) in a sentence. With the different per-son NPs as the matrix subject, the interpretation is very different for (47b), though for (47a) it always has the reading that the point of reference is the speaker. The second difference is that ziji can refer to the projected ego, while wo cannot. It seems that wo is external in nature, but ziji is internal in nature. In other words, wo is existentially bound, i.e. identified outside the scope of the carrier of belief, while ziji is not. If we really want to name self-ascription ziji in

Why the Blocking Effect?

307

terms of the concept of point of view, we can view it as presenting the self point of view, which is the essential property of self-ascription.

Another difference between logophoric pronouns and self-ascription ziji is that the former cannot have a first person pronoun as antecedents, whereas the latter can easily do so, as pointed out in Section 4.3.1.

Maybe one can see the difference between self-ascription and the concept of point of view (and logophoricity) more clearly by using the following analogy. Suppose that a viewer (the speaker) is seeing things from a point of reference, and the reference can be first, second, and third person NPs. If the viewer takes the perspective of a third person NP, then we call this a third person point of view. Note that the viewer can have his/her independence from the referent of the third person NP, namely that they stand at different places. This is the so-called logophoricity use of reflexives. However, if the viewer becomes one with the referent of the third person NP, namely that they at least stand at the same place, then we have the self point of view, or it involves self-ascription. Hence, the difference between logophoricity and self-ascription is that the former allows the independence of the viewer from the point of reference, while self-ascription does not. That is, logophoricity does not necessarily require the two (the re-viewer and the point of reference) become one, though self-ascription does re-quire it. Note that the self-ascription requirement is satisfied only if the speaker stands at Zhangsan’ s place, and views the object in question from that place.

Similarly, if the second person pronoun is the point of reference, and the viewer takes the perspective of the referent of the second person pronoun, but still keeping his/her independence, then we have a second person point of view. However, if the two become one, i.e. that the viewer loses his/her independence, then we have the self point of view again. Note that the independence of the viewer from the point of reference and the self-consciousness of the individual used as the point of reference are independent concepts. Although the viewer or speaker is always self-conscious, the individual as the point of reference is not necessarily so, especially when the point of reference is a third person NP, though s/he may be conscious that is not at issue here, cf. endnote 12. Self-ascription does not require the individual---the point of reference---to be self-conscious, as pointed out previously and in Section 4.3, though it does require that the viewer and the point of reference become one, i.e. at least standing at the same place.22

An anonymous reviewer asks why can’ t the speaker take the point of view of Zhangsan in (46). Actually, the speaker can take Zhangsan’s point of view. The issue is the relationship between the speaker and Zhangsan. There are two pos-sible ways for the speaker to take Zhangsan’ s point of view: (a) the speaker is independent of Zhangsan, e.g. they stand at different places, and the speaker views the object from Zhangsan’ s perspective; and (b) the speaker and Zhang-

Haihua Pan

308

san stand at the same place, or in my terminology, they become one. The first case is the so-called third person point of view, while the second case is the self point of view. I think that sentences like (46) only allow the second case that corresponds to self-ascription, though logophoricity may allow both cases. Hence, only self-ascription corresponds exactly to the use of long distance bound ziji .

In summary, self-ascription and logophoricity are two different concepts, though they overlap in meaning. I think the concept of self-ascription is more primitive, and covers more cases than logophoricity.23

7. CONCLUSION

In this chapter I have presented systematic facts on the blocking effect of long-distance bound bare reflexive ziji in Mandarin Chinese. I have argued for the non-symmetrical nature of the blocking effect and against the symmetrical unlike-person blocking view. I have shown that not only subjects or NPs con-tained in subjects but also non-subjects can block long distance binding of ziji as long as they are first or second person pronouns, or contain first or second per-son pronouns. Hence the blocking effect is more sensitive to the person feature than to the local subject. I claim that long distance bound ziji is constrained by self-ascription, and have proposed a condition in (31) to account for the binding properties of ziji . I have explained the blocking effect by appealing to the spe-cial role played by the first and second person pronouns in self-ascription: since first and second person pronouns are obligatory self-ascribers, they will block long distance binding of ziji by third person NPs. Since sometimes third person NPs can indeed block the long distance binding of ziji by first/second person pronouns, though they are not as strong as the cases with first/second person pronouns as blockers, we need to investigate the reason why. One possible reason is that, following Li (1991), we may say that it is re-quired to have highly reflective contexts for long distance bound ziji , though unlike Li (1991), we think the relevant constraint only applies to first/second person pronouns, not to third person NPs, when they function as long distance antecedents. Hence, there is another asymmetry between third person NPs and first/second person pronouns in the sense that highly reflective contexts are re-quired only when the latter function as the antecedents of long distance bound ziji . We leave this issue open for future research.

Why the Blocking Effect?

309

ACKNOWLEDGEMENTS Parts of this chapter were presented at the 1997 LSA Workshop on Long Dis-tance Reflexives in Ithaca, New York. The author would like to thank the par-ticipants for interesting questions. He would also like to thank Peter Cole, James C.-T. Huang, Manfred Krifka, Carlota Smith, Jonathan Webster, and Liejiong Xu for helpful comments and suggestions on an earlier version of this chapter. Thanks also go to the two anonymous reviewers for the helpful comments and suggestions. As usual, the author is responsible for all the potential errors. REFERENCES Baker, Carl Lee (1995) Contrast, discourse prominence, and intensification, with

special reference to locally free reflexives in British English. Language 71(1), 63--101.

Battistella, Edwin. (1989) Chinese reflexivization: A movement to INFL ap-proach. Linguistics 27, 987-1012.

Chen, Ping (1992) The reflexive ziji in Chinese: Functional vs. formalist ap-proaches, In Thomas H.-T. Lee (ed.), Research on Chinese Linguistics in Hong Kong, pp. 1-36, Linguistics Society of Hong Kong.

Chierchia, Gennaro (1989) Anaphora and Attitude de se, In R. Bartsch, et al. (eds.) Semantics and Contextual Expression, pp. 1--31, Foris Publications, Dordrecht, Holland.

Chomsky, Noam (1981) Lectures on Government and Binding. Foris Publica-tions, Dordrech, Holland

Chou, Xiao-Li. (1992) An alternative approach to Chinese reflexives. MA report, the University of Texas at Austin, USA.

Clements, George N. (1975) The logophoric pronoun in Ewe: Its role in dis-course. Journal of West African Languages 10, 141--177.

Cole, Peter, G. Hermon & L.-M. Sung. (1990) Principles and parameters of long distance reflexives. Linguistic Inquiry 21, 1--22.

Cole, Peter and L.-M. Sung. (1993) Feature percolation and Mandarin reflexives. Jorunal of East Asian Linguistics 2, 91--118.

Cole, Peter and L.-M. Sung. (1994) Head movement and long distance reflex-ives. Linguistic Inquiry 25, 239-62

Cole, Peter and C.-Z. Wang. (1996) Antecedents and blockers of long distance reflexives: The case of Chinese ziji . Linguistic Inquiry 27, 357-90

Hagège, Claude. (1974) Les pronoms logophorique. BSLP, 69, 287--310.

Haihua Pan

310

Huang, C.-C. James and C.-C. Jane Tang. (1991) The local nature of long dis-tance reflexive in Chinese. In Jan Koster and Eric Reuland, editors, Long dis-tance Anaphora. Cambridge University Press, Cambridge.

Huang, Yan. (1994) The Syntax and Pragmatics of Anaphora: A Study with Spe-cial Reference to Chinese. Cambridge University Press, New York.

Huang, Yun-Hua. (1984) Reflexives in Chinese. Studies in English Literature and Linguistics 10.

Iida, M. (1992) Context and Binding in Japanese. Ph.D. thesis, Stanford Univer-sity, Stanford, CA.

Kuno, Susumu. (1987) Functional Syntax: Anaphora, discourse, and empathy. University of Chicago Press.

Lewis, David. (1979) Attitudes de dicto and de se. The Philosophical Review 88, 29-80.

Li, F.-X. (1990) The effect of person hierarchy on the blocking of the long dis-tance binding of the Chinese pronoun ziji , In: Proceedings Of WECOL 19.

Li, N.-C. (1991) Perspective-Taking in Mandarin Discourse. Unpublished Ph.D. dissertation, State University of New York at Buffalo.

Li, YaFei (1993) What makes long distance reflexives possible? Journal of East Asian Linguistics 2, 132--66.

Maling, Joan (1984) Non-clause-bounded reflexives in modern Icelandic. Lin-guistics and Philosophy 7, 211-241.

Mitchell , Jonathan E. (1986) The Formal Semantics of Point of View. Unpub-lished Ph.D. dissertation, University of Massachusetts, Amherst, MA.

Pan, Haihua (1991) Pro and long distance reflexive ziji in Chinese. paper pre-sented at the 3rd North-American Conference on Chinese Linguistics, Cor-nell University, Ithaca, NY.

Pan, Haihua (1995) Locality, Self-Ascription, Discourse Prominence, and Man-darin Reflexives. Ph.D. thesis, Dept. of Linguistics, the University of Texas at Austin, Austin, TX.

Pan, Haihua (1997) Constraints on Reflexivization in Mandarin Chinese. Gar-land Publishing, Inc., New York.

Pan, Haihua (1998) Closeness, prominence and binding theory. Natural Lan-guage & Linguistic Theory 16, 771-815.

Pollard, Carl and Ivan A. Sag. (1994) Head-Driven Phrase Structure Grammar. The University of Chicago Press and CLSI Publication.

Pollard, Carl and Ping Xue. (1998) Chinese reflexive Ziji : Syntactic reflexives vs. nonsyntactic reflexives. Journal of East Asian Linguistics 7, 287-318.

Progovaç, Ljilj ana. (1992) Relativized SUBJECT: Long distance reflexives without movement. Linguistic Inquiry 23, 671--680.

Progovaç, Ljilj ana. (1993) Long distance reflexives: Movement-to-Infl vs. rela-tivized SUBJECT. Linguistic Inquiry 24, 755--772.

Why the Blocking Effect?

311

Reinhart, Tanya and Eric Reuland. (1991) Anaphors and logophors: An argu-ment structure perspective. In Jan Koster and Eric Reuland, editors, Long distance Anaphora. Cambridge University Press, Cambridge.

Reinhart, Tanya and Eric Reuland. (1993) Reflexivity. Linguistic Inquiry 24, 657--720

Sells, Peter. (1987) Aspects of logophoricity. Linguistic Inquiry 18, 445--479. Stirling, Lesley (1993) Switch Reference and Discourse Representation. Cam-

bridge University Press, Cambridge. Tang, C.-C. Jane (1985) A Study of Reflexives in Chinese. Unpublished M.A.

thesis, National Taiwan Normal University. Tang, C.-C. Jane (1989) Chinese reflexives. Natural Language and Linguistic

Theory 7(4), 93--121. Xu, LieJiong. (1993) The long distance binding of ziji . Journal of Chinese Lin-

guistics 21, 123--141. Xu, LieJiong. (1994) The antecedent of ziji . Journal of Chinese Linguistics, 22,

115--137. Xue, Ping, Carl Pollard, and Ivan Sag (1994) A new perspective on Chinese ziji .

In Proceedings of the 13th West Coast Conference on Formal Linguistics, pp. 432-447. The University of Chicago Press and CSLI Publication.

Yu, Willi am. (1991) Logophoricity in Chinese. paper presented at the 3rd North-American Conference on Chinese Linguistics, Cornell University, Ithaca, NY.

Zribi-Hertz, Anne (1989) Anaphor binding and narrative point of view: English reflexive pronouns in sentences and discourse. Language 65(4), 695--727.

Zubin, David A., Soonae Chun, and Naicong Li. (1990) Misbehaving reflexives in Korean and Mandarin. In Kira Hall and Jean-Pierre Koenig, eds., Pro-ceedings of the 16th Annual Meeting of the Berkeley Linguistics Society, Vol. 16: General Session and Parasession on the Legacy of Grice. Berkeley: Berkeley Linguistics Society.

NOTES 1The following abbreviations are used in this chapter: CL (classifier), Perf (perfective marker), DE (possessive/modifier/complementizer marker), ZAI (progressive marker), ZHE (imperfective non-progressive marker), GUO (the experiential marker), Q (question morpheme), Prt (sentence-final particle), Conj (sentence connective), etc. Also note that the Chinese names Zhangsan, Lisi, and Wangwu given in the chapter are generic names rather than real names.

Haihua Pan

312

2F.-X. Li (1990) also notices the asymmetrical nature of long distance binding in Chinese. I differ from him, though, in the sense that I group first/second per-son pronouns together in contrast with third person NPs, but he differentiates all three person forms. I think that first person pronouns can bind ziji with a second person pronoun intervening only if ziji is contrastive. Since it belongs to the contrastive reflexive, and is thus subject to different conditions, this usage of ziji should not interfere with the characterization of non-contrastive reflexives (see Pan (1997) for relevant discussion). 3Xue, Pollard, and Sag (1994) propose a new category Z-pronoun for ziji be-sides locality ziji subject to local o-binding and discourse ziji , and suggest that it must be o-bound (see Pollard and Sag 1994 for the concepts of (local) o-binding). Although they suggest differentiating two types of blocking effect: unlike-person blocking for ziji in complement clauses and animate blocking for ziji in sub-commanding cases, Xue, Sag, and Pollard (1994) do not provide a clear account to explain the blocking effect for ziji except pragmatic factors. Hence I will not discuss their analysis further in the rest of this chapter.

4The reader can find a detailed critique of all the previous analyses mentioned in this chapter and other analyses of Chinese reflexives in Pan (1995, 1997). 5Note that it is not adequate to claim that cases like (13) are metaphoric exten-sion (or personification), as suggested in Tang (1989). One weakness of the personification claim is that there is no clear condition to determine when per-sonification takes place, and when it does not. We think that ziji , in general, can take either an animate or inanimate NP as its antecedent, although we under-stand that, for a specific case, it depends on the semantics of the verb to deter-mine the animacy feature of ziji , and whether the NP in question can be an ante-cedent of ziji . For instance, in the case following, the inanimate subject cannot be the antecedent of ziji because of the animate feature ziji has that is determined by the experiencer role ziji carries.

i. *Na-bian wenzhang hai -le ziji. That-CL article hurt-ASP self Lit: That article hurt self.

What we are saying basically is that the animacy feature of ziji is not determined lexically or inherently, rather it is determined by context.

6We notice that in Cole and Wang (1996) they claim that first/second person pronouns do not block long distance binding of ziji by a third person NP if they are the BA object or an oblique introduced by the passive marker bei ‘by.’ It seems that there is some speaker judgement variation here. Although this possi-bilit y may be extended to adjuncts, it does not apply to objects as in (4a). That is, even if we agree in not counting obliques and adjuncts as possible blockers, objects and possessors in a subject are still problematic to the feature agreement

Why the Blocking Effect?

313

checking approach. The relevant examples are (4a) and (12a). Besides, all the arguments against this approach presented earlier are still valid. 7(17) is a summary of Huang and Tang’s (1991) original descriptions. 8Both anonymous reviewers point out that the notion of logophoricity has been used in varying ways by different authors, and thus my arguments against the individual notions of logophoricity proposed by Sells (1987) do not refute the general notion of logophoricity. However, we notice that the general notion of logophoricity is not well defined. One possible definition is given in Reinhart and Reuland (1993): non-argument reflexives or arguments that are focused are logophors, but this notion only applies to compound reflexives like himself, not to bare reflexives like ziji . If we use the conception that all the cases that do not obey Binding Condition A belong to the so-called logophoricity, then it is very clear to see that this general notion of logophoricity is too weak to account for the properties of ziji . Although it can predict the long-distance binding possibil -ity, the general notion fails to explain the blocking effect, and it also fails to ac-count for the subject orientation property of ziji . One of the reviewers asks what category the logophoric pronouns belong to if logophoric pronouns are not de se anaphors. As shown in Figure 1 in Section 5.1 following, I take the logophoric reflexives as belonging to a separate category that is dominated by long distance bound reflexives that also dominate the self-ascription reflexives. See also endnote 23.

9This classification does not include the adverbial usage of ziji , as discussed in Tang (1989), Pan (1995, 1997), etc. The following is a relevant example.

i. Zhangan ye ziji zuo fan. Zhangsan also self do rice

‘Zhangsan also cooks his meal himself.’ We can also group this usage of ziji with the locality reflexive, since they can only have clause bound antecedents.

10Note that Baker’s conditions are not fine-grained enough to differentiate the four contrastive reflexives in Chinese: ziji , benren, benshen, and zishen. For the relevant conditions to differentiate these reflexives, see the discussion in Chap-ter 7 of Pan (1997). 11The bound variable interpretation of pronouns is a necessary condition but not a sufficient condition for de se beliefs. Note that, even for the bound vari-able interpretation of pronoun he, there are also two readings: de re and de se. See Chierchia (1989) for relevant discussion. 12Note that I make a distinction between the concepts of consciousness and self-consciousness. The former does not entail the latter, though the latter does entail the former. A person can be conscious without being self-conscious. For example, when a person is awake, then he is conscious. He can read and speak. Nevertheless, he may not know what his name is if he has suffered a brain dam-

Haihua Pan

314

age, and has lost all of his previous memory. In this latter situation, the person in question is not self-conscious, though he is conscious. Hence, when I say a person is not self-conscious, I mean that he may not know something about him-self, though he is awake and conscious.

13(36b) also shows that first/second person pronouns do not block their co-arguments from long distance binding ziji . See more relevant discussions later.

14The condition on locality ziji and taziji is proposed in Pan (1998), and we repeat it here for reference.

(i) The Anaphor Condition An α can be the antecedent of an anaphor X in a linguistic domain γ if (a.) α is a noun phrase that does not dominate X; and (b.) There is no closer blocker in γ. (ii ) α is closer to X, the reflexive, than β iff the path from X to the minimal maximal projection dominating α is a subset of the path from X to the minimal maximal projection dominating β. (iii ) β is a blocker for α if (a) β is not less prominent than α; and (b) β, if not dominating α, must be a subject. (iv) α is less prominent than β if it appears lower in the animacy hierarchy than β does.

(v) Animacy Hierarchy (Chou, 1992) [+human] > [+animate, -human] > [-animate]

15Thanks to an anonymous reviewer for pointing out that sentence (40) may be a counterexample to the claim that first person pronouns always induce the blocking effect. Sentences like (40) are well -behaved examples showing that semantic constraints play a role in the determination of the antecedents of ziji . The involved predicates are irreflexive predicates, i.e., the meaning of the predi-cate precludes co-reference of its arguments. We have incorporated this con-straint in our definition of blockers in (33). 16The same reviewer points out that the following sentence may be a counter-example to the claim that third person NPs do not block the long distance bind-ing of ziji by first/second person pronouns. (Note that I only claim that third person NPs do not NECESSARILY block long distance binding of ziji by first/second person pronouns, cf. endnotes 19 & 20.)

(i) Ni i weishenme meiyou shixian gaosu dajia Lisi j bu xihuan ziji * i/j? you why didn’ t before tell all Lisi not like self

‘Why didn’ t you tell everyone in advance that Lisi didn’ t like you?’ Although I agree with the reviewer that first/second person pronouns are strong blockers, and third person NPs are weaker blockers, I think that the i reading in (i) is possible, at least marginally. It seems that for the remote fisrt/second per-son pronouns to be the long distance antecedents of ziji , we need to make extra effort to provide appropriate context. This is maybe the point made in Li (1991),

Why the Blocking Effect?

315

namely that long distance bound ziji requires highly reflective contexts, and it points to the projected ego. Although I don’ t think that her claim applies to third person NPs as antecedents, it may be true that it applies to first/second person pronouns as antecedents. Sentences like (i) seem to suggest that there may be additional constraints for the cases in which first/second person pronouns are the long distance antece-dents of ziji . I will l eave this issue open for future research. 17The reviewer also argues against the self-ascription account of ziji using (i) below, claiming that it is very diff icult to say the subject Zhangsan is the carrier of belief, and thus a syntactic account is necessary.

(i) Cong nei yihou Zhangsani zai meiyou jian-dao guo jiu le ziji i from that after Zhangsan again didn’ t see ASP save Perf self ming de nage ren. li fe De that-CL person ‘Since then Zhangsan has never seen the person again who saved his li fe.’

Actually, I agree with the reviewer that my self-ascription account cannot cover sentences like (i). However, these sentences are dealt with by the condition on locality ziji , as specified in endnote 14, so are sentences involved psych-verb object antecedents. 18Example (42) is from Luther Liu (1999), i.e. his (50c) in Chapter 1 of his Ph.D. dissertation, University of Cali fornia, Irvine. 19An anonymous reviewer suggests that the sentence following (Tang 1989) is another counterexample to my analysis, since there is no first/second person pronoun in the believed proposition, the embedded clause, but long distance binding of ziji is still not possible, contrary to the prediction made by my analy-sis.

(i.) Tameni renwei taj zai piping ziji * i/j. they think he at criticize self ‘They think that he is criticizing himself.’

However, my informants take the i reading in (i) to be possible, especially when tamen ‘ they’ is interpreted as a group. For instance, long distance binding of zij i is perfectly possible in (ii ) below.

(ii .) Like ban de xuesheng zhidao Laoshi zai zhanyang ziji, science class DE student know Teacher at praise self, gandao feichang gaoxing. feel very happy ‘The students from the science class(es) knew that Teacher is praising them, and felt very happy about it.’

Hence, the asymmetry between first/second person pronouns and third person NPs still holds, even if the long distance binding of ziji is not possible in (i). See also endnotes 16 & 20.

Haihua Pan

316

20The fact that 3rd person NPs can sometimes block the long distance binding of ziji by first/second person pronouns, as pointed out by one of the reviewers, does not falsify my claim that there is an asymmetry between first/second person pronouns and third person NPs in the blocking effect. This is because I only claim that third person NPs do not NECESSARILY block long distance binding of ziji by first/second person pronouns, which implies that third person NPs can, though they do not always do so. Similarly, sentences like (43) are not counter examples to the above claim of mine, either, as long distance binding is possible in the relevant sentences.

21Note that here I differ from what I said in Pan (1995, 1997), namely that the antecedents of long distance bound ziji must be self-conscious. It seems that this self-consciousness constraint should be replaced by the highly reflective context constraint and only applies to antecedents with first/second person features, i.e. it does not apply to third person antecedents.

22When I say that the viewer and the referent of the point of reference become one, I mean it metaphorically. It seems that it is more appropriate to say that the viewer stands at the same place as the referent of the point of reference, viewing the object in question, when we are talking about self-ascription in the sense that the viewer attributes a de se belief to the referent of the point of reference. In this situation the referent in question is not required to view the object, and the viewing action can be carried out by the speaker only. This is why the referent need not be self-conscious, though s/he may be conscious which is not relevant to the concept of self-ascription, as mentioned earlier.

Note that we do not deny the existence of a different reading in which the point of reference is the viewer, and thus s/he self-ascribes the relevant belief. Unlike the situation mentioned above, the referent of the point of reference, be-ing a self-ascriber, in this reading is self-conscious.

23The same reviewer claims that s/he could not see the differences between logophoricity and self-ascription, and suggests that, although I have refuted the individual notions (SOURCE, SELF, and PIVOT) of logophoricity for long dis-tance bound ziji , there is still t he general notion of logophoricity that is not re-futed. I believe that the general notion must be based on the individual notions. If each and every one of them is refuted, I could not see how the general notion could stand. If the general notion is a garbage can type, as suggested in the lit -erature, then it is still t oo weak to account for the properties of long distance bound ziji , as pointed out in endnote 8.

Also note that the logophoricity notion given in Reinhart and Reuland (1991), as pointed out in Section 4.3, is actually perspectivity (deictic center) which is refuted in Section 4.3. Hence, it is not appropriate to claim that the general no-tion of logophoricity can help to account for the long distance binding of ziji .