On the Extractability of Adjuncts in Chinese Relative Clauses* Yiu Yuk Man, Shi Dingxu, Jiang Yan Department of Chinese & Bilingual Studies The Hong Kong Polytechnic University This study attempts to explain the inconsistent acceptability judgments resulting from the extraction of the object of adjunct PPs in Chinese relative clauses. The facts observed is that the extraction of arguments and the object of obligatory adjunct PPs is always judged to be acceptable, whereas the extraction of the object of optional adjunct PPs is sometimes judged to be acceptable and sometimes unacceptable. Such an inconsistency is in contrast to the uniform pattern of acceptability judgment predicted by a syntactic analysis. In this study, a syntactic-pragmatic account is proposed which can better explain the observed facts and is compatible with the Minimalist Program (MP). This study shows that extractions of the object of optional adjunct PPs are syntactically legitimate and the related inconsistent judgments can be resolved by pragmatics. Particularly, a successful enrichment of a structure ensures a correct interpretation of an utterance which in turn entails an acceptable judgment of the utterance. 1. Extracting the Object of Adjunct PPs Within the framework of Generative Grammar, asymmetry in extractability between different categories has often been observed (Ross 1967, Chomsky 1977, Kayne 1981, Rizzi 1982). For example, the subject/object superiority and the adjunct/complement superiority have been observed in many languages. With respect to Chinese, Huang (1982) observes another asymmetry which is between argument and adjunct. The above asymmetries usually display an across-the-board pattern of extractability and are believed to be caused by syntactic factors. Therefore, syntactic explanations such as the Empty Category Principle (ECP) are often used to explain such asymmetries (Huang 1982, Chomsky 1986, Rizzi 1990, Lasnik & Saito 1990). Following Chomsky (1977) and Huang (1982), the present study assumes that relativization is a kind of extraction as it exhibits the characteristics of wh-movement which is subject to certain locality constraints (Chomsky 1977, Huang 1982, Aoun et al. 1987). In this study, the pattern observed is that in Chinese, extractions of arguments and the object of obligatory adjuncts are always judged to be acceptable, whereas extractions of the object of optional adjuncts are judged to be acceptable in some cases and unacceptable in other cases. Extractions of the object of optional adjuncts lack a uniform pattern of judgments which is different from the consistent pattern of judgments observed in the asymmetries discussed above. Therefore, structural explanation alone is not likely to be adequate in explaining the observed facts. The fact that in relative clauses, extractions of an argument are always judged to be acceptable is illustrated by (1)-(4): (1) ec kai che de ren he person who drives’ operate car DE person (2) ec da yu de ren he person who catches catch fish DE person fishes’ (3) wo mai ec de shu he book which I I buy DE book bought’ (4) ta xihuan ec de shuiguo he fruit which he likes’ he like DE fruit Examples (1) and (2) both involve extracting the subject, i.e., ren erson’ whereas (3) and (4) involve extracting the object, i.e., shu ooks’ and shuiguo ruit’. In all these cases, an acceptable judgment results and an EC appears in the position where the extraction has taken place. In Chinese, adjuncts are not exclusively NPs but can as well be PPs (Chao 1968). The present study will only focus on extractions of the object of adjunct PPs. These adjunct PPs are adverbials referring to time, location, manner, instrument, goal, comparison etc. Time and location adverbials are two kinds of adjuncts which can appear preverbally as NPs as in (5) and (6), whereas other adjuncts must appear as preverbal PPs as in (7)-(9): (5) ta zuotian zou le (Time) e left yesterday.’ he yesterday leave ASP (6) women wu-li qu (Location) e go inside of the we house-inside go house.’ (7) wo *(chao) na bian zou (Goal) went that way.’ I toward that side go (8) ta *(bi) na ge ren gao (Comparison) e is taller than that he than that CL person tall person.’ (9) ta *(yong) qianzi xiu che (Instrument) e repairs cars with he with plier repair car pliers.’ In relative clauses, after the object of an adjunct PP is extracted, the adjunct PP position can be either filled by the same preposition plus a resumptive pronoun or it can simply be occupied by an EC. For example: (10) Zhangsan gen ta tiaowu de huoban Zhangsan with him/her dance DE partner he partner Zhangsan danced with’ (11) wo zai nali qing ke de jiujia I at there invite guest DE restaurant he restaurant where I held the banquet’ In (10), after the object of the adjunct PP is extracted, i.e., huoban artner’, the resumptive pronoun ta /he’ is inserted to the object position of the preposition gen ith’. Similarly, in (11) after the object of the adjunct PP is extracted, i.e., jiujia estaurant’, the resumptive pronoun nali here’ is inserted to the object position of zai t’. In both examples, the resumptive pronouns are coreferential with the head nouns. If the preposition is not present, the adjunct PP position will only be occupied by an EC. (12) Zhangsan ec tiaowu de huoban Zhangsan dance DE partner he partner Zhangsan danced with’ (13) wo ec qing ke de jiujia I invite guest DE restaurant he restaurant I held the banquet’ In (12) and (13), after the object of the adjunct PP is extracted, i.e., huoban artners’ and jiujia estaurant’, the preposition is not present and only an EC appears in the adjunct PP position. Examples (12) and (13) have the same interpretation as their respective counterparts in (10) and (11). Specifically, the head noun huoban artners’ in (10) and (12) is related to the adjunct position inside the relative which bears a theta-role of comitative, whereas the head noun jiujia estaurant’ in (11) and (13) is related to the adjunct position inside the relative which bears a theta-role of locative. (10) and (11) differ from (12) and (13) in that at PF the former has the adjunct position filled by a full-fledged preposition and a resumptive pronoun, whereas the latter has the adjunct position left empty. In Chinese, there are adjunct PPs which are so closely related to the verb that they and the verb together are assumed to form a discontinuous constituent with the adjunct functioning like an object of the verb. Structurally, these adjuncts occur preverbally as PPs. Semantically, they are indispensable to render a complete sentence meaning. This study assumes that these preverbal adjunct PPs are obligatory adjuncts in the sense of Jackendoff (1990) and are subcategorized in the lexical entry of the verb in contrast to optional adjuncts which are not subcategorized by the verb. For example: (14)*wo bu xie yi gu na ge ren I not disdain one look that CL person look down upon that person and do not want to cast a single sight on him.’ (15) wo dui na ge ren bu xie yi gu I toward that CL person not disdain one look (16)??wo bu xie yi gu I not disdain one care In (15), the PP dui na ge ren oward that person’ illustrates the so-called obligatory adjuncts. The idiom chunk bu xie yi gu isdain (sth.) so much that one does not even want to cast a single sight on (it)’ requires an object NP which bears a patient theta-role. However, the object NP can not immediately follow the idiom chunk. Otherwise, the sentence will be ruled out as ill-formed as in (14). As a result, the object NP, i.e., na ge ren hat person’ has to occur preverbally as the object of the preposition dui oward’ as in (15) and the whole PP dui na ge ren oward that person’ functions as an obligatory adjunct of the idiom chunk. If (15) is uttered without the obligatory adjunct dui na ge ren oward that person’, the sentence sounds incomplete as in (16). (17)*wo gan xingqu manhua shu I feel interest comic book (18) wo dui manhua shu gan xingqu I toward comic book feel interest am interested at comic books.’ (19)??wo gan xingqu I feel interest In (18), the verb gan eel’ is a transitive verb which takes xingqu nterest’ as its object and the V-O in turn requires an object NP which bears a patient theta-role, an entity affected by the action of the verb. Again, if such an object NP occurs immediately after the V-O chunk, the sentence will be ruled out as ill-formed as in (17). As a result, the object NP manhuashu omic books’ has to occur preverbally as the object of an obligatory adjunct PP as in (18). If (18) is uttered in isolation without the obligatory adjunct dui manhua shu oward comic books’, it sounds incomplete as shown in (19). When the object of obligatory adjuncts is extracted and the adjunct position filled by an EC, the resulting relatives are always judged to be acceptable as in (20) and (21): (20) wo ec gan xingqu de shu I feel interest DE book he book which I am interested at’ (21) wo ec ti le henduo yijian de dianshi jiemu I raise Asp much criticism DE TV program he TV program on which I have much criticism’ In (20) and (21), when the object of the obligatory adjuncts is extracted, i.e., shu ooks’ and dianshi jiemu V programs’, the resulting relatives are both judged to be acceptable. On the other hand, extractions of the object of optional adjunct PPs vary in their judgments. In some cases, an extraction of the object of an optional adjunct PP is judged acceptable, as in (22): (22) ta ec sha ren de dongji he kill people DE motive he motive he killed people’ In other cases, an unacceptable judgment arises when the object of an optional adjunct PP is extracted, as in (23): (23)*wo huxi de chuang I breathe DE bed he bed I breathe’ With regard to the inconsistent judgments resulting from the extraction of the object of optional adjunct PPs, a syntactic account has been proposed by Ning (1993) and it will be briefly outlined in the next section. 2. Primary and Secondary Adjuncts Ning (1993) classifies adjuncts into two groups, i.e., primary and secondary adjuncts. Primary adjuncts include those with one of the four features below: PLACE, TIME, MANNER/INSTRUMENT and REASON and are extractable as in (24)-(27): (24) ta ec xiu che de cheku he fix car DE garage he garage where he fixed cars’ (Place) (25) ta ec xiu che de na-ge wanshang he fix car DE that evening he evening when he fixed cars’ (Time) (26)a.ta ec xiu che de fangfa he fix car DE way he way he fixed cars’ (Manner) b.ta ec xiu che de qianzi he fix car DE pliers he pliers with which he fixed cars’(Instrument) (27) ta ec xiu che de yuanyin he fix car DE reason he reason why he fixed cars’ (Reason) (Ning 1993:95) On the other hand, secondary adjuncts have features other than those of primary adjuncts and are not extractable, as shown in (28)-(32): (28)*ta ec tiaowu de guniang he dance DE girl he girl he danced with’ (Comitative) (29)*ta ec xiao-le-xiao de na-ge ren he smile-Asp-smile DE that man he man he cast a smile’ (Goal) (30)*ta song-le yiben shu gei ec de na-ge ren he give-Asp one book to DE that man he man he gave a book to’ (Dative) (31)*ta lai ec de na-ge xiaozhen he come DE that town he town where he came’ (Ablative) (32)*ta ec gao de na-ge ren he tall DE that man he man whom he is taller than’ (Comparative) (Ning 1993: 96) To account for the asymmetry in extractability between primary and secondary adjuncts, Ning proposes a syntactic analysis based on the Minimalist Program. He assumes that only adjuncts with single-word adjunct operators can be extracted. Such a single-word adjunct operator is only available to primary adjuncts but not to secondary adjuncts. In particular, Ning analysis is that primary adjuncts with the single-word adjunct operators can move to [Spec, CP], checking off its feature against the C head. As a result, the derivation converges. On the other hand, secondary adjuncts without the single-word adjunct operators cannot move to [Spec, CP], failing to check off its feature against the C head, hence the derivation crashes. Under Ning analysis, extractions of each group of adjuncts are expected to have a consistent judgment. Particularly, it is expected that extractions of primary adjuncts are always judged acceptable, whereas extractions of secondary adjuncts are always judged unacceptable. However, it is observed that extractions of primary and secondary adjuncts do not display a consistent pattern in judgment. For example: (33)*wo chuan yifu de xiaozhen I wear clothes DE town he town where I wore clothes’ (Place) (34)*ta shenghuo de na ge xingqitian he live DE that CL week he week when he lived (his life)’ (Time) (35)*ta xiu che de huaban he repair car DE skateboard he skateboard with which he repairs cars’ (Instrument) (36)*ta huxi de yuanyin ta breathe DE reason he reason he breathes’ (Reason) Examples (33)-(36) involve extracting the so-called primary adjuncts and such extractions are expected to be acceptable when in fact they are not. (37) ta tiaowu de huoban he dance DE partner he partners he danced with’ (Comitative) (38) wo ca pixie de guke I shine shoes DE guest he customers I shone shoes for’ (Goal) (39) ni song li de dui xiang you send gift DE targeted person he targeted person to send gifts to’(Dative) (40) xian xue liu chulai de shangkou fresh blood come out DE wound he wound which the blood came from’ (Ablative) In contrast, examples (37)-(40) involve extracting the so-called secondary adjuncts and such extractions are expected to be unacceptable but they turn out as acceptable. The lack of an across-the-board pattern of judgment within each group of adjuncts suggests that Ning generalization and analysis are not adequate and the fact observed in the present study needs to be otherwise explained 3. A Syntactic-Pragmatic Account Taking note of the inconsistent judgments resulting from extractions of the object of optional adjuncts, contradicting what a syntactic analysis will predict, we assume that the object of optional adjuncts is always extractable involving the same structure and we suggest that the related inconsistency be explained by pragmatic factors. Following Chomsky (1977), we take relativization to be a kind of extraction. Particularly, it involves the extraction of an empty operator (Huang 1982, Li 1990, Ning 1993). Such an extraction creates a chain between the moved operator and the trace left inside the relative clause and the chain is subject to certain locality constraints. The head noun is then coindexed with the moved operator and is in turn indirectly coindexed with the trace.[1] According to Ning (1993), Chinese relatives have the structure in (41): (41) [NP [CP [Spec Opi ] [C’ [IP ...ti...] [C de]]] [NP NPi]] As previously mentioned, after the object of adjunct PPs is extracted, the adjunct position can be either filled by the same preposition and a resumptive pronoun or by an EC. In the former, the preposition and the resumptive pronoun must co-occur as in (42): (42) wo yong tai qie rou de daoi I with it cut meat DE knife cut meat with a knife.’ (42) shows that after the operator is extracted, if the preposition yong ith’ is to be kept, the resumptive pronoun ta t’ need to be inserted to the object position of yong ith’, coindexing with the moved operator. The head noun dao knife’ is then coindexed with the operator and is in turn indirectly coindexed with the resumptive pronoun. On the other hand, if the preposition is not present, only an EC will appear in the adjunct PP position as in (43): (43) wo ti kan shu de fangjiani I see book DE room he room where I read’ In (43), after the operator is extracted, a trace is left inside the relative clause coindexing with the moved operator. The head noun fangjian he room’ is then coindexed with the moved operator and is in turn indirectly coindexed with the trace. A related fact observed from the extraction of the object of optional adjuncts is that when the object of optional adjunct PPs is extracted, the resulting relative is always judged acceptable if the adjunct position is filled by the same preposition and a resumptive pronoun. On the other hand, if the preposition is not present and only a trace appears in the adjunct position, the acceptability judgment of the resulting relative is inconsistent. It is commonly observed that if structural constraints are imposed on an extraction, then all structures resulting from such extractions should be consistently ruled out as ill-formed. This line of reasoning prompted Ning (1993) to suggest that there is a structural difference between the so-called primary adjuncts and secondary adjuncts. Extractions of the former is not constrained whereas extractions of the latter is. However, we have observed that extractions of the object of adjuncts within each group give rise to inconsistent judgments. This in turn suggests that the inconsistency in judgments resulting from the extractions is not likely to be caused by structural constraints. We are of the opinion that both relatives with the adjunct PP position filled by a trace and relatives with the adjunct PP position filled by a preposition and a resumptive pronoun have the same structure. Precisely, relativized adjunct PPs are generated with a preposition and an NP operator.[2] During the computation, the NP operator is extracted leaving a trace inside the relative clause. When the computation splits into LF and PF, on the one hand, at LF the preposition remains intact. On the other hand, at PF the preposition is either deleted or kept. In the latter case, if the preposition is deleted, the trace will not be realized as a resumptive pronoun. If the preposition is kept, the trace has to be realized as a resumptive pronoun as it is a general constraint in Chinese that preposition stranding is not allowed. In other words, at PF the preposition can be overtly present or not present at all whereas at LF the preposition is assumed to be always present as it has content contributing to the sentence meaning.[3] As a result, the obligatory presence of the preposition at LF will ensure the correct interpretation of the relatives. Under the assumption that all extractions are legitimate involving an identical structure, we will argue that the observed unacceptability of relatives can be explained by pragmatics. When a well-formed relative is taken out of a context for a native speaker to make a judgment on its acceptability, the native speaker is likely to judge the relative acceptable when he can relate the head noun to an empty position inside the relative and unacceptable when he cannot relate the head noun to an empty position inside the relative. To establish such a relationship, the native speaker will use clues that are available to him such as structural information, lexical information, encyclopedia knowledge and context (Sperber & Wilson 1995). However, according to relevance-theoretic pragmatics, establishing a relationship between the head noun and an empty position takes time and effort. The native speaker will do it only if he can balance the processing effort involved and the contextual effects yielded (Sperber & Wilson 1995). Otherwise, he may give up at any stage during the interpretation process. In other words, the balance between the effort involved in processing a relative and the reward obtained from correctly interpreting such a relative determines the extent to which the native speaker will try to use clues available to him to interpret the relative. For example: (44) wo xiu che de qianzi he plier I repair cars with’ I repair car DE plier (45) wo xiu che de zhi he paper I repair cars with’ I repair car DE paper (44) and (45) are almost identical except that they have a different head noun. The theta-grid of xiu epair’ has two arguments which are filled in both examples. Therefore, the native speaker is likely to relate the head noun in (44) and (45) to an adjunct position. In the terminology of relevance theory, one needs to enrich the underdetermined information provided by the incomplete syntactic string, supplying it with the explicatures of the utterance, so as to obtain the complete logical form, which will then serve as the basis for semantic interpretation. In (44) and (45), to correctly relate the head noun to the adjunct position involves many factors such as the native speaker experience, knowledge of the world etc. Imagining that a man in prehistoric times, who has never seen che ars’ or qianzi liers’ and does not have the knowledge of what che ars’ and qianzi liers’ are, is asked to judge whether (44) and (45) are acceptable, he is likely to judge both as unacceptable. This is because he fails to correlate xiu che ars’ with qianzi liers’ are and how they are related to xiu epair’ or zhi aper’. On the other hand, if a mechanic is asked to judge whether (44) and (45) are acceptable, he is likely to judge (44) acceptable as it is part of his knowledge that mechanics often use qianzi liers’ to repair cars. As a result, he can readily establish a relationship between xiu che epairing cars’ and qianzi liers’. In contrast, in (45) the relationship between xiu che epairing cars’ and zhi aper’ does not yield immediate contextual effects and has to be ruled out. Nevertheless, if the mechanic believes that it is still worth his effort to establish a relationship between xiu che epairing cars’ and zhi aper’ will he use other clues to help him to interpret (45). In this case, the mechanic may try to define an appropriate context which will help him to relate zhi aper’ to xiu che epairing cars’.[4] For example, if the mechanic had seen some people putting a piece of paper on the floor on which they lied when repairing cars and if he is able to retrieve this piece of information from his memory to interpret (45), then he is again likely to judge the relative acceptable. On the other hand, if the mechanic does not have such a piece of information or if he does but is unable to retrieve it to interpret (45), he is likely to judge (45) unacceptable. However, when interacting with others in spontaneous verbal communication, a native speaker is rarely asked to give judgment on a sentence or on an utterance that is taken out of a context. Instead, a sentence or an utterance is often interpreted within a particular context. This means that when interpreting an utterance, the native speaker will seldom make choices among possible interpretations. Instead, he will hit on the one that yields enough contextual effects without unjustifying processing effort and take it to be the optimally relevant interpretation that can be derived from the utterance. It goes without saying that the selection of the correct context by the native speaker depends on many factors. In verbal communication, the speaker has to decide, even if unconsciously on most occasions, how much information he should state explicitly in his utterance which is based on his assessment of the hearer knowledge (Sperber & Wilson 1995). On the other hand, when a hearer tries to interpret an utterance, he makes all sorts of inferences based on clues that are available to him, again as a usually unconscious process. The success of communication depends on the accuracy of the inferences made by the speaker and the hearer (Sperber & Wilson 1995). If either side makes a wrong judgment or inference, then the communication may break down or misunderstanding may arise. As previously mentioned, after the object of an adjunct PP is extracted, the speaker can either delete the preposition or keep it at some stage during the computation. Between the two choices, the tendency in Chinese is to delete the preposition and to avoid using resumptive pronouns.[5] [6] As a result, when the object of an adjunct PP is extracted, the speaker will always choose to delete the preposition if he assumes that the hearer is able to access the correct piece of information which can help him to recover the deleted preposition. On the other hand, the speaker will choose to keep the preposition if he assumes that the hearer may not be able to recover the preposition if it is deleted. In that case, keeping the preposition will allow the hearer to correctly interpret his utterance. For example: (46) wo ec hen bu gaoxing de fangjian I very not happy DE room he room where I had some very bad experience’ (46) shows that when the speaker thinks that the hearer is able to recover the deleted preposition, he will always choose to delete the preposition to avoid being repetitive. With the deletion of the preposition, no resumptive pronoun needs to be inserted to the object position of the adjunct PP. As a result, no overt clue is given in (46) to relate the head noun to the empty adjunct position. In this case, the speaker expects the hearer to recover the deleted preposition. On the other hand, when the speaker assumes that deleting the preposition may cause misunderstanding as the hearer is not able to recover the deleted preposition, then the speaker will choose to keep the preposition as in (47): (47) wo zai nali hen bu gaoxing de fangjian I at there very not happy DE room he room where I had some very bad experience’ (47) shows that when the speaker thinks that the hearer is not able to correctly interpret his utterance when the preposition is deleted, he will choose to keep the preposition. The resumptive pronoun nali here’ is inserted to the object position of zai t’ so that preposition-stranding can be avoided. In cases such as this, the preposition zai t’ and the resumptive pronoun nali here’ provides an overt clue showing how to relate the head noun to the adjunct PP position. Therefore, the utterance itself provides enough syntactic clues to the hearer to relate the head noun to the adjunct position. From the speaker point of view, (46) and (47) carry the same amount of information. They only differ in whether an overt clue, i.e., the preposition, is given in the utterance to show the hearer how to relate the head noun to the adjunct position. The speaker decision of whether to delete or to keep the preposition depends on his judgment of the hearer knowledge. When the speaker makes a wrong choice miscalculating the hearer knowledge, he will either be considered repetitive or uninformative (Grice 1989). That is equal to saying that his utterance is not optimally relevant (Sperber & Wilson 1995). For example, when the speaker chooses to keep the preposition, assuming that the hearer cannot recover the deleted preposition when in fact the hearer can, the speaker will be considered repetitive, providing too much information. On the other hand, when the speaker chooses to delete the preposition, assuming that the hearer can recover the deleted preposition when in fact the hearer cannot, the speaker will be considered uninformative, providing too little information. On the other hand, the speaker choices will also induce two consequences on the hearer side. First, if the preposition is kept, the hearer is given overt syntactic and structural clues to relate the head noun to the adjunct position. As a result, the hearer can always correctly interpret the speaker utterance and need not accommodate. Second, if the preposition is deleted, the hearer is given no overt syntactic and structural clue to relate the head noun to the adjunct position. As a result, the hearer need to accommodate, drawing different clues from his resources in order to correctly interpret the speaker utterance. Given a relative, the hearer has to relate the head noun to the empty position inside the relative in order to correctly interpret the speaker utterance. If there is an empty argument position inside the relative, the head noun is likely to be related to that position. For example: (48) mai shu de ren he person who bought books’ buy book DE person (49) wo mai de shu he book which I bought’ I buy DE book The theta-grid of mai uy’ has two arguments, external and internal argument. In (48), the object position of mai uy’ is filled by shu ooks’ whereas the subject position is not filled. As a result, the hearer is likely to relate the head noun ren eople’ to the empty subject position. Similarly, in (49) the subject position of mai uy’ is filled by wo ’ whereas the object position is not filled. As a result, the hearer is likely to relate the head noun shu ooks’ to the empty object position. In cases in which all argument positions inside the relative are filled, the head noun is likely to be related to an adjunct position. For example: (50) wo ec qie rou de dao he knife I cut meat with’ I cut meat DE knife In (50), the theta-grid of qie ut’ has two arguments, external and internal argument, which are already filled by wo ’ and rou eat’. Therefore, the empty position inside the relative is likely to be an adjunct position and the hearer is likely to relate the head noun dao knife’ to that adjunct position. If the head noun is related to the position of an obligatory adjunct PP, the hearer can usually establish a correct relationship between the head noun and the empty adjunct position. Hence, correctly interpreting the speaker utterance. (51) wo dui shu gan xingqu am interested at books.’ I toward book feel interest (52) wo ec gan xingqu de shu he books which I am I feel interest DE book interested in’ In (51), the PP dui shu oward books’ is an obligatory adjunct of the predicate gan xingqu e interested in’. In (52), after the object of the adjunct PP, i.e., shu ooks’ is extracted, the preposition dui oward’ is deleted. Since an obligatory adjunct is part of a fixed expression and is subcategorized by the verb, lexical information can always be used to recover the deleted preposition. It thus explains why the hearer can easily relate the head noun to the obligatory adjunct position, correctly interpreting the speaker’ utterance. On the other hand, when the head noun is related to the position of an optional adjunct and no overt syntactic or structural clue is given, the hearer can correctly interpret the relative only if he can establish a correct relationship between the head noun and the optional adjunct position with clues that are available to him such as lexical information, encyclopedia knowledge, immediate physical environment, preceding discourse, etc. (Sperber & Wilson 1995). In cases in which the hearer tries to establish an appropriate context which can help him to relate the head noun to the optional adjunct position, if the hearer shares the same knowledge with the speaker, it is easier for him to establish such a relationship. In contrast, if the hearer does not share the same knowledge with the speaker, it is likely to be more difficult for him to establish such a relationship. For example: (53) wo hen bu gaoxing de na liang chezi hat car in which I had some I very not happy DE that CL car very bad experience’ Assuming that the speaker utters (53) when he means that he had some very bad experience when he was inside that car. In other words, the optional adjunct PP is used to indicate location. If the hearer does share such a piece of knowledge with the speaker, he can easily establish the relationship between the head noun na liang chezi hat car’ and the optional adjunct position, recovering the information carried by na liang chezi hat car’, i.e., location. As a result, the hearer can correctly interpret the speaker utterance and communication can go on. On the other hand, if the hearer does not share such a piece of knowledge with the speaker, the hearer may establish an incorrect relationship between the head noun na liang chezi hat car’ and the optional adjunct position. For example, the hearer may interpret the head noun na liang chezi hat car’ as an instrument with which when the speaker used it, he had some very bad experience. As a result, the hearer cannot recover the information carried by the optional adjunct PP failing to correctly interpret the relative and communication breaks down. Likewise, the hearer may simply have no way to relate the head noun with the remaining string. Hence he cannot obtain the logical form of the sentence and no interpretation is assignable as a result. To summarize, the correct interpretation of an adjunct relative depends on the correctness of judgments and inferences made by the speaker and the hearer. The speaker has to decide how much information he should state explicitly in his utterance so that the hearer can correctly interpret his utterance. If the speaker makes a wrong decision, he will be either repetitive, providing too much information or uninformative providing too little information to the hearer. On the other hand, given no overt syntactic clue, i.e., when the preposition is deleted, the hearer may try to use other clues that are available to him to interpret an optional adjunct relative. In cases in which the hearer tries to establish an appropriate context to interpret an optional adjunct relative, it is easier for him to establish such a context if he shares the same knowledge with the speaker. Otherwise, the hearer is likely to make more effort to establish such a context and he may give up at any stage during the conversation when he thinks it is not worth his effort to establish such a context. 4. Concluding remarks The goal of this paper is to show that the acceptability of a sentence is not solely determined by its syntactical well-formedness. A sentence can be structurally well-formed but is yet judged to be unacceptable for other reasons. The fact observed is that when uttered in isolation, relatives involving the extraction of the object of optional adjuncts are sometimes judged to be acceptable and sometimes not. To account for such an inconsistency, a syntactic-pragmatic account is proposed. Under the assumption that extractions involving the object of adjunct PPs are all syntactically well-formed, involving an identical structure, it is suggested that such a well-formed relative will be judged acceptable only if the addressee can accommodate successfully recovering the addresser intended interpretation. To carry out a successful communication, both addresser and addressee have to make some correct judgments. The addresser has to decide how much information he should state explicitly in his utterance or sentence. On the other hand, the addressee has to accommodate differently in order to recover the correct interpretation. If both sides make a correct judgment, communication will go on. On the other hand, if either side makes a wrong judgment and the other side loses intention to continue communicating, then communication may break down. Putting it simply, the syntactical well-formedness of a sentence and the use of pragmatic strategies both contribute to successful communication. References: Aoun, Joseph., Norbert Hornstein, David Lightfoot and Amy Weinberg. (1987). wo Types of Locality”. Linguistic Inquiry 18:537-557. Chao, Yuen Ren. (1968). A Grammar of Spoken Chinese. University of California Press. Chomsky, Noam. (1977b). n WH-Movement”. In P. Cullicover, T. Wasow, and A. Akmajian, eds. Formal Syntax, New York: Academic Press. ----. (1981). Lectures on Government and Binding. Dordrecht: Foris. ----. (1986). Barriers. Cambridge: The MIT Press. Grice, Paul. (1989). Studies in the Way of Words. Harvard University Press. Huang, James C-T. (1982). Logical Relations in Chinese and the Theory of Grammar. Doctoral dissertation, MIT, Massachusetts. Jackendoff, Ray. (1990). Semantic Structures. Cambridge: The MIT Press. Kayne, Richard. (1981). CP Extractions”. Linguistic Inquiry 14:554-561. ----. (1984). Connectedness and Binary Branching, Dordrecht: Foris. Lasnik, Howard and Mamoru Saito. (1992). Move : Conditions on its Application and Output, Cambridge, Mass: The MIT Press. Li, Audrey. (1990). Order and Constituency in Mandarin Chinese. Dordrecht: Kluwer. Ning, Chunyan. (1993). The Overt Syntax of Relativization and Topicalization in Chinese. Doctoral dissertation, University of California, Irvine. Rizzi, Luigi. (1982). Issues in Italian Syntax. Dordrecht: Foris. ----. (1990). Relativized Minimality. Cambridge, Mass: MIT Press. Ross, Robert. (1967). Constraints on Variables in Syntax. Doctoral dissertation, MIT. Sperber, Dan. and Deirdre Wilson. (1995). Relevance: Communication and Cognition. Second Edition. Oxford: Blackwell. Wang, Peter. (1970). Transformational Approach to Chinese ba and bei. Unpublished Doctoral dissertation, University of Texas, Austin. -------------------------------------------------------------------------------- * This study is supported by the Hong Kong Polytechnic University Research Grant NO. 0351 302 A3 720 bearing the project name tudies on the Context-dependent Properties of Modern Chinese”. [1] That the head noun directly moves out of a relative clause, coindexing with the trace has also been proposed, e.g. Wang (1970). This proposal will not affect the argument that relativization is a kind of extraction as the coindexed chain between the head noun and the trace is also subject to the locality constraint. [2] Ning (1993) account has a similar proposal. His account differs from the present account in that the preposition in his account is unspecified but the preposition in the present account is specified. [3] The presence of an invisible preposition at LF is similar to Ning unspecified preposition. [4] Sperber & Wilson (1995) discussed in detail how an addressee selects an appropriate context to interpret an utterance. [5] Inferring from this fact, one possible explanation for such a tendency is from Grice (1989) maxim of Quantity which states that the speaker is expected to provide no more and no less information to the hearer. Keeping the preposition is likely to involve the possibility of providing redundant information which is already known by the hearer. Therefore to avoid being repetitive, the speaker will tend to delete the preposition. Rephrasing it in terms of relevance theory, as the speaker thinks the hearer in a given context can fully recover the explicature of the utterance containing an EC, he does not need to use the resumptive pronoun strategy to provide a more explicit string (i.e. with the explicature of the former sentence containing the EC already provided as part of the sentence meaning). [6] Cf. Chomsky (1981) Avoid Pronoun Principle.
