interg2.gif (2223 bytes)


Verbs and Governing 
Category Determination

Horng-Ming Wu


      In this paper we will observe some binding phenomena concerning the pronoun ta 'he' in embedded subject position. We will see that some verbs in matrix position can affect the reference determination of the embedded subject pronoun. Verbs that can change the coreferential relation between the matrix subject quantifier and the pronoun pose problems for the binding theory. The main question to be raised in this investigation is why and how certain elements can work on the binding relation between NPs in a sentence. This paper will present a syntactic explanation of the changing binding effect associated with the relation between the quantificational subject and the embedded subject pronoun ta 'he'.

Key words: Governing Category, Pronoun, Binding




     In this paper I will examine certain exceptional binding phenomena that pose interesting questions for binding theory. The main question to be raised in this investigation is whether verbs can affect the binding relation between NPs in a sentence by changing the governing category of the binding domain. If so, one might ask what mechanism allows this and under what circumstances. It is my intention in this paper to try to answer these questions.




     A binding problem concerning the Specified Subject Condition (SSC) was pointed out in Chomsky (1986a), as shown in (1):


a. Theyi told [my stories about themi]

b. Theyi heard [my stories about themi]

c. * Theyi told [stories about themi]

d. Theyi heard [stories about themi]



     Since SSC requires that pronouns not be bound and anaphors not be free in the domain of the most local subject, the local domain for sentences (1a-b) is the embedded NP because it contains the nearest subject my. The pronouns are free in this domain. SSC applies correctly. However the embedded NP in sentences (1c-d) has no subject, the nearest subject for the pronoun in (1c) is the matrix subject They. The local domain is the entire sentence in which the pronoun is not free; (1c), therefore, violates Binding Condition B and is correctly predicted ungrammatical. Nevertheless, sentence (1d), having the same structure except for having a different verb, is not as expected. Though the pronoun is not free in its local domain, the sentence is grammatical. To explain this problem, Chomsky suggests the following structures for (1c-d):


a.*Theyi hold [ PROi stories about themi ]

b. Theyi heard [ PROjstories about themi ]



     The indexing tells us that in (2a) the stories are by the subject they; in (2b) the stories are by someone outside the sentence. Under this explanation, SSC is then applicable as the embedded NP now has a subject PRO. Binding principle B is therefore consistent. Although Chomsky paid no attention to the role of the verb in binding (as the only difference between the two sentences is the matrix verb), sentences in (1c-d) raise the interesting question whether verbs play some sort of syntactic role in determining the binding relation; more precisely, whether the choice of the verb affects the pronoun's antecedent. In theory, such a possibility seems quite reasonable. As is well known, in certain cases the choice of one constituent may have some sort of effect on the choice of another. In theta theory, for example, the thematic role of the subject in a sentence is determined by the choice of the object. Compare, for instance:


a. Bill broke a hand yesterday.

b. Bill broke a window yesterday.


5.      We note that sentences (3a) and (3b) share the same structure, except that the verb in (3b) takes the NP window as its object; the subject Bill is then assigned the role AGENT, since Bill is the one who initiates the action¾ breaking a window. However, Bill in (3a) cannot be considered AGENT because the verb takes the NP hand as its complement. Bill in this case turns out to be the one who undergoes the action pointed out by the predicate. The subject is therefore assigned the role THEME (or PATIENT). The theta role of a subject may sometimes depend on the semantics of verb and its complement compositionally. This fact in turn suggests that it is not implausible to expect a close correlation between verbs and binding. More straightforwardly, the idea that verbs may affect the binding relation may be another possible account in addition to Chomsky's ‘PRO as subject in NPS’ for the problem shown in (1c-d).



     There are two reasons for preferring the verb-based alternative. First, the following sentences in (4) suggest that the verb hear in English may play a part in binding:


a.   *Theyi discussed stories about them i

b.  *Theyi knew stories about themi

c.   *Theyi wanted stories about themi

d.   *Theyi liked stories about them i

e.    *Theyi read stories about themi

f.   *Theyi wrote stories about them i

g. *Theyi published stories about them i

h. *Theyi told stories about them i

i.   Theyi heard stories about them i



     Second, the idea that PRO in [Spec NP] position may serve as subject creates problems. For example, by adopting this proposal, (5a) below would have an incorrect syntactic representation something like (5b):      


a. Theyi heard [NP stories about each other i]

b.*Theyi heard [NP PROj stories about each other i]



     Clearly, (5b) is not consistent with Condition A. The reciprocal anaphor is not bound in the embedded NP (= its governing category), so the sentence is starred. Chomsky realizes this problem, and so suggests that the presence of PRO as subject inside NPS is optional. When PRO is not present, the governing category for the reciprocal anaphor in (5a) is the entire sentence because it contains the most local subject they, and this is the domain where the anaphor is bound. The sentence is thus correctly predicted, as (6) shown:


 [GC Theyi heard stories about each other i]

     Superficially, the problem seems to be plausibly resolved. However, Kuno (1987) points out certain problems concerning Chomsky's PRO suggestion and argues that the problem should be dealt with semantically rather than syntactically. For instance, Kuno (1987, pp. 171-172) indicates that Chomsky's  ‘PRO as subject in NPS’ assumption fails to explain why the sentences such as (7a) is unacceptable, whereas (7b) is nearly acceptable or acceptable:


a. *John i wrote for Mary a profile of him i
          which was ten pages long.

b. \/ /?John i wrote for Mary i a profile of herself i  
          for the application.

According to Chomsky, (7) would have the following syntactic representation in (8):


a.*John i wrote for Mary [PRO i's profile of him i], 
     which was ten pages long.

b.\/ /?John i wrote for Mary [PRO i's profile of herself i
     for the application.



     That is, just like the verb tell, write must also assign subject control to the PRO so that the unacceptability of (7a) can be correctly explained, as shown in (8a). Yet this is not the case in (7b), which is not an unacceptable sentence, even though the binding between John, PRO, and herself clearly shows that the sentence should not be acceptable. Kuno, therefore, concludes that this suggests that the phenomenon requires a semantic explanation rather than a syntax one.



     Note now what would happen in (1c-d) if we do not adopt Chomsky's ‘PRO as subject in NPS’ explanation. Suppose we assume Chomsky's (1996) binding theory, repeated here:


A.    An anaphor must be bound in a local domain.

B.    A pronoun must be free in a local domain.

C.    An r-expression must be free.

Recall that governing category in this version is defined roughly as in (10) below.


 The GC for a is the minimal CFC that contains 
 a and a governor   of a  and in which a's binding 
 condition could, in principle, be satisfied.

Following (10), sentences (1c-d) would have the following interpretation instead of (2).


a. *[They i told stories about them i]

b. [They i heard stories about them i]



     That is, the governing category for the pronoun them is the entire sentence, since it contains the pronoun itself, its governor about, which is the head of the PP about them, and the most local subject they serving as the binder. In (11a) the sentence's ungrammaticality is precisely predicted by Condition B because the pronoun is not free in this domain. But the grammaticality of (11b) cannot be explained. Note that, however, the sentences shown in (4) suggest that the verb hear may be important in deciding the binding relation. A natural question to ask here then is: how can the verb affect the binding? A possible explanation may be that the sentences in (4) may have the following syntactic representations.


a. *[GC Theyi discussed stories about themi]

b. *[GC Theyi knew stories about themi]

c. *[GC Theyi wanted stories about themi]

d. *[GC Theyi liked stories about them i]

e. *[GC Theyi read stories about them]

f .*[GC Theyi wrote stories about them i]

g. *[GC Theyi published stories about them i]

h. *[GC Theyi told stories about them i]

i.*[Theyi heard [GC stories about them ]

(12) suggests that sentences (a-h) involve unmarked structure, in which the governing category for each sentence is the root sentence; while sentence ( i ) has a marked structure in which the governing category for the pronoun them is the embedded NP. In other words, the determination of the governing category in these sentences is somehow determined by the verb. Based on this assumption, sentences (1c-d) may now be alternatively considered as having the following structures instead of the (2) or (11).


a. *[ GC They i told stories about them i]

b.   They i heard [GC stories about them i]


12.      Thus, while the pronoun in (13a) is bound, it is free in (13b). Their grammaticality is therefore accounted for by Condition B of the binding theory. However, there is one question that arises further in connection with this explanation. That is, how do verbs determine the governing category? We will return to this question in the later discussion.



     In sum, the English data illustrated in this section proposes that verbs may be responsible for the determination of governing category. In that regard, I will present evidence found in Chinese in the following sections.




     Aoun and Li (1988) claim that bound pronouns in Chinese can not appear in such a context as shown in (14).


             Context                      Referential Pronoun

a.  NPi   V  [s' NPi  V   NP] 

b.  Zhangsani  shuo  tai    de   le            jiang 

                   say   he   get   ASP     prize 

    "Zhangsan said that he got the prize." 

                                  Context                  Bound Pronoun

c.  NPi   V  [s' NPi  V   NP]                   *

d. *Meigereni  dou  shuo  tai de       le       jiang

                 everyone    all     say   he   get     ASP  prize

                "Everyone said that he got the prize."



     In (14b) ta ‘him’ is a referential pronoun since it is coindexed with a name called ‘Zhangsan’, while in (14d) ta is a bound pronoun for its referent is ‘Meigeren’, a quantifier. As they point out, only referential pronouns, but not bound pronouns, can appear in the subject position of an embedded clause. Nevertheless, let's consider the following sentences that have the similar structures (i.e. they all comply with the context (14c)) except for having different verbs in place of shuo ‘say’.


a. Meigereni   dou   ziren         tai   hen     you    danliang

    everyone     all   self-know   he  very   have     courage

    "Everyone considered himself a courageous person."

b. Meigereni   dou    jiazhuang  tai    bu    zhidao  zhejian      shi

    everyone      all      pretend    he not      know   this          matter

    "Everyone pretended he didn't know the matter."

c. Meigereni   dou    fashi   tai    zaiyebu   zuo    huaishi          le

    everyone      all     swear he    no more do      bad thing    ASP

    "Everyone swore that he no more did the bad things."

d. Meigereni  dou   mengxiang   tai     jiushi    neige   baiwanfuweng

    everyone       all       dream        he   is exactly  that      millionaire

    "Everyone dreamed of being that millionaire."



     Contrary to the prediction, the sentences in (15) are all perfectly acceptable. The bound pronoun ta in each sentence is in the subject position of the embedded clause, and is naturally interpreted as referring to the quantifier Meigeren ‘everyone’. To see if verbs do possibly play a crucial role in the interpretation of pronoun reading, let's now replace the matrix verb in each sentence to see the result.


 a. *Meigereni  dou    shuo     tai   hen   you   danliang

 everyone    all       say     he     very  have  courage

 "Everyone said that he was a courageous person."

 b. *Meigereni   dou    zhidao  tai   bu   zhidao   zhejian     shi

 everyone     all        know  he  not    know     this          matter

 "Everyone knew that he didn't know the matter."

           c. *Meigereni   dou    yiwei    tai     zaiyebu   zuo     huaishi        le

            everyone     all       think       he     no more  do   bad thing      ASP

            "Everyone thought that he no more did the bad things."

            d. *Meigereni    dou    zhichu   tai   jiushi        neige   baiwanfuweng

            everyone       all     indicate  he  is exactly    that        millionaire

         "Everyone pointed out that he definitely was the millionaire."


17.      The resultant structures clearly suggest that the sentences’ acceptability is dependent on the particular verbs. This observation can be further supported by comparing (17) with (18).


Zhangsani    chengren  zijii     shuoguo     tai you   nü    pengyou

                     admit     self      have said     he   have  girl   friend

"Zhangsan admitted that he himself had said that he had a girl friend."


*Zhangsani     shuo   zijii    shuoguo   tai      you    nü     pengyou

                         say    self     have said  he    have   girl       friend

"Zhangsan said that he himself had said that he had a girl friend."


18.      Obviously some verbs in these instances do play a part in binding. How they are able to do so will be the focus of the next section.





     To explain why some verbs have an effect on changing the binding relation between NPs in certain structures, two types of verbs in Chinese must be first identified. Consider the following two non-exhaustive lists:


                   a. Strong Subject Orientation (SSO)

zi-ren           ‘self-know’

jiazhuang      ‘pretend’

fashi              ‘swear’

chengren        ‘admit’

xiang-zhdao   ‘want-know’

mengxiang       ‘dream’

zhaoren           ‘confess’


          b. Weak Subject Orientation (WSO)

shuo               ‘say’

renwei           ‘think’

xihuan           ‘like’

xiangxin        ‘believe’

zhidao           ‘know’

juede             ‘feel’




     How might we distinguish these two types of verbs semantically? One fundamental feature that SSO verbs have in common is that they seem to always report on the subject's actions with respect to a mental state, whereas WSO verbs seem to merely report on the mental state itself (however, the verb shuo ‘say’ is an exception to this generalization). In general, whether a verb has strong or weak subject orientation property is not evident if the embedded subject pronoun is referential. For instance:


a. Zhangsani shuo    tai   de    le              jiang (WSO)

                     say       he   get    ASP          prize

  "Zhangsan said that he got the prize."

b. Zhangsani    jiazhuang   tai      de    le        jiang (SSO)

    Zhangsan      pretend    he    get    ASP     prize

   "Zhangsan pretended that he got the prize."



     As we can see, both types of verbs in (20) are possible. Nevertheless, this is not the case when a bound pronoun appears in the embedded subject position. More clearly, we find only SSO verbs may occur. Compare:


a. *Meigereni     dou    shuo    tai     de   le           jiang

     everyone      all     say        he   get   ASP          prize

     "Everyone said that he got the prize."

b. Meigereni    dou    jiazhuang      tai   de         le         jiang

    everyone    all      pretend         he   get     ASP      prize

   "Everyone pretended that he got the prize."


22.      When a matrix verb is a WSO verb, it would behave like a barrier for coreference between the bound pronoun and the matrix subject, as (21a) shows. On the other hand, when the verb is SSO, this barrier disappears, as illustrated in (21b).



     Based on the above observation, we seem to be able to give a schema to conclude the finding:


a. NP + V + [R-pronoun + V + NP]

                          Subject Orientation of the R-pronoun

SSO verbs ((20a))                        Yes

WSO verbs ((20b))                      Yes

b. NP + V + [B-pronoun + V + NP]

                        Subject Orientation of the B-pronoun

SSO verbs ((21a))                        Yes

WSO verbs ((21b))                       No


24.      (22) clearly illustrates why the two groups of subject orientation verbs listed in (19) can be defined in terms of the notion ‘strong’ and ‘weak’, as the former shows consistent subject orientation no matter which type of pronominal occupies the embedded subject position, whereas the latter does not.



     There seems to exist another difference between SSO and WSO verbs. Topicalization of the embedded subject bound pronoun is allowed by weak subject orientation verbs, but not SSO verbs. Compare, for example:


a.  taj,     Meigereni dou    shuo tj     de     le          jiang                 

     he    everyone       all  say            get   ASP          prize

     "He, everyone said that e got the prize."

b. * tai/j,   Meigereni   dou   jiazhuang  ti/j    de   le     jiang

     he      everyone       all        pretend e    get   ASP    prize

    "He, everyone pretended that he got the prize."



     In (23a), the pronoun in the structure can only refer to someone outside the sentence. Therefore, topicalization of the pronoun from the embedded subject position to the highest C-specifier position is fine, as this does not violate Koopman and Sportiche's (1983) Bijection Principle, which requires that every variable be bound by exactly one operator, and every operator bind exactly one variable. Thus, the trace left by the pronoun is a variable, because the pronoun was moved to an A’-position. The trace is bound by one operator only. Consider now the LF representation of (23b) in (24):


*[tai/j,   Meigereni   [ti/j dou jiazhuang [ ti/j de le jiang ]]]

   he     everyone             all   pretend           get ASP prize


27.      As shown in (24), each variable is bound by two operators, i.e., the pronoun and the quantifier; and in a like manner, each operator binds two variables. As the pronoun has been moved to the highest C-specifier position, where the quantifier is unable to c-command it, according to the schema (22a), then, the binding between the pronoun and the quantifier can not be licensed, for the pronoun can only be bound by a c-commanding subject in a higher position with the presence of an SSO verb. This may explain why topicalization of the embedded subject pronoun is allowed with WSO verbs only: since when a WSO verb shows up, the bound pronoun in the embedded subject position always takes a different index from the matrix subject, which will ensure that the pronoun does not violate the Bijiection Principle.




     Thus far we have observed that the non-matrix subject pronoun reading may be changed when an SSO verb is involved. We then must ask how a verb or verb class can affect a binding relation. This section aims to suggest an explanation to this question.



     The suggestion presented in this paper involves the idea that governing category is sensitive to verb class.


If a verb is SSO, the governing category for the 
pronoun is the smallest S containing the pronoun.



     Now let's see how this assumption works on the problems. Consider first the following sentence:


 [GC Meigereni dou fashi     tai  de    le         jiang]

       everyone       all   swear he  get    ASP     prize

       "Everyone swore that he got the prize."



     According to Aoun and Li (1988), (26) must obey the A’-disjointness requirement, which requires that the pronoun have the entire sentence as its governing category.


The A’-disjointness Requirement
A pronoun must be A’-free in the least CFC 
containing a SUBJECT  and the pronoun



     But since (26) involves an SSO verb, the pronoun ta ‘he’ must then be free in the smallest clause. This requirement entails that the governing category of the pronoun in (26) must be drawn back from the root sentence to the least clause domain, as illustrated in (28):


[ Meigereni dou fashi      [GC tai de         le   jiang]

        | GC ------contract-------à | GC

 everyone all          swear he    get      ASP     prize

 "Everyone swore that he got the prize."

Turn now to a sentence that involves no SSO, i.e. (29):


*[GC Meigereni dou    yiwei   tai      de      le      jiang]

         everyone all   think    he       get     ASP      prize

 "Everyone thought that he got the prize."



     Since sentence (29) has no an SSO verb in the matrix position, but rather a WSO, it obeys the A’-disjointness requirement, which will predict that the entire sentence is the governing category of the pronoun. The ungrammatical reading of the pronoun is thus correctly predicted. Now consider the following sentence.


Zhangsani chengren  zijii     shuoguo   tai you   nü       pengyou

                   admit    self   have said   he  have      girl   friend

"Zhangsan admitted that he himself had said that he had a girl friend."



     The pronoun ta ‘he’ in (30) is A-free in its smallest CFC domain, in accordance with the folllowing A-disjointness requirement suggested by Aoun and Li (1988):


The A-disjointness Requirement

 A pronoun must be A-free in the least Complete 
 Functional Complex (CFC) in which it occurs



     However, according to the A’-disjointness requirement, the pronoun should be A’-bound by ziji at LF in its binding domain, which, nevertheless, would bring an ungrammatical reading. To be in accord with (25), the governing category for the pronoun should be the smallest sentence as this sentence contains an SSO verb chengren ‘admit’. We therefore obtain the following correct result by retracting the governing category from IP2 to the smallest IP.


[IP1 Zhangsani chengren [IP2 zijii shuoguo [IP3 tai you nü pengyou]]]

      | GC --contract--à | GC

               admit   self  have said   he  have girl   friend

"Zhangsan admitted that he himself had said that he had a girl friend."




     In this paper I focus on the relation between verbal class and binding, and discover that certain verbal classes play a role in determining the binding relation, particularly associated with pronoun, in some sentences. The determination of the binding relation can be theoretically achieved by assuming that the verbal classes, when involved, are able to contract the governing category to the smallest clause containing the pronoun, conforming the coreferential relation to the binding theory.



Aoun, Joseph (1985) A Grammar of Anaphora, Cambridge: MIT Press.

Aoun, Joseph (1986) “Bound Pronouns in Chinese,” in S. Berman, J. W. Choe, and J. McDonough, eds., Proceedings of the 16th Annual Meeting of NELS, GLSA, University of Massachusetts, Amherst.

Aoun, Joseph and Yen-hui Audrey Li (1988) “Minimal disjointness,” paper from the Proceedings of the Seventh West Coast Conference on Formal Linguistics, pp. 29-39.

Battistella, E. (1989) “On the Distribution of PRO in Chinese,” Natural Language and Linguistics Theory 3, 317-340.

Bickerton, D. (1987) “He Himself: Anaphor, Pronoun, or…?” Linguistic  Inquiry 18, 345-348.

Bouchard, D. (1985) “The Binding Theory and the Notion of Accessible SUBJECT,”Linguistic Inquiry 16, 117-133.

Chao, Yuen-ren (1968) A Grammar of Spoken Chinese, Berkeley and Los Angeles: University of California Press.

Chomsky, Noam (1980) “On Binding,” Linguistic Inquiry 11: 1-46.

Chomsky, Noam (1981) Lectures on Government and Binding, Dordrecht: Foris. Chomsky, Noam (1982) Some Concepts and Consequences of the Theory of   Government and Binding, Cambridge, Mass.: MIT Press.

Chomsky, Noam (1986a) Barriers, Cambridge, Mass.: MIT Press.

Chomsky, Noam (1986b) Knowledge of Language: Its Nature, Origin, and Use, New York: Praeger.

Chomsky, Noam (1996) The Minimalist Program, Cambridge, Mass.: MIT Press. Cole, P., and Sung Li-May (1990) “Non-C-commanding Antecedents of Bound Anaphors and Feature Percolation in Government and Binding

Theory,” Paper from Proceeding of the 9th WCCFL, 523-533.

Cook V.J. (1988) Chomsky’s Universal Grammar: An Introduction, Basil Blackwell Inc. New York.

Eng, M. (1989) “Pronouns, Licensing, and Binding,” Natural Language and Liguistic Theory 7, 51-92.

Giorgi, A. (1987) “The Notion of Complete Functional Complex: Some Evidence from Italian,” Linguistic Inquiry 18, 511-518.

Haegeman, L. (1991) Introduction to Government and Binding Theory, Basil
Blackwell. Haik, I. (1984)
“Indirect Binding,” Linguistic Inquiry 15, 185-223.

Higginbotham, J. (1980) “Pronouns and Bound Variables,” Linguistic Inquiry 11, 679-708.

Huang, C-T. James (1982) Logical Relations in Chinese and the Theory of Grammar, Ph.D. dissertation, MIT.

Huang, C-T. James (1983) “A Note on the Binding Theory,” Linguistic Inquiry 14.3.

Katada, Fusa (1988) “LF-binding of anaphors,” WCCFL 7, 171-186.

Koopman, H. and D. Sportiche (1983) “Variables and the Bijection Principle,” The Linguistic Review 2, 139-160.

Koster, J. (1984) “On Binding and Control,” Linguistic Inquiry 15.3, 417-459.

Kuno, Susumu. 1987 “Functional Syntax” The University of Chicago Press.

Lebeaux, David (1985) “Locality and Anaphoric Binding,” The Linguistic Review 4, 343-363.

Rizzi, L. (1990) Relativized Minimality, Cambridge, Mass: MIT Press.

Webelhuth, Gert (1995) Government and Binding Theory and the MinimalistProgram, Basil Blackwell Inc. Cambridge, Mass.




摘 要