Word Order, Structure and Relativization

1 Word Order, Structure and Relativization* YEN-HUI AUDREY LI University of Southern California 1. Introduction Nouns in languages like English appear to differ greatly from those in languages like Chinese1. Prominent differences between them include the fact that nouns in English generally do not occur in the "bare" form but nouns in Chinese can.2 Moreover, a bare noun in Chinese can be interpreted as definite, indefinite or generic; these different interpretations of an English nominal phrase rely on the occurrence of the definite article the, indefinite article a or some other device. Take a definite expression for instance. Chinese is happy with a bare noun [N], such as shu 'book'; whereas English requires the definite article [the N] the book(s).3 A very interesting question arises from such a difference is the structure of nominal phrases in these languages. An article in English, for instance, plays a key role in the structural representation: a nominal phrase is represented as a Determiner Phrase (DP), headed by a Determiner (D) (see Abney 1987). The definite article the can be in the D position (see Longobardi 1994, Chierchia 1998, among others). In other words, the book can have the structure in (1), headed by the:4 (1) DP / \ D NP / \ the book What, then, is the structure of a definite expression such as shu in Chinese? Is it like (1) with a null D head (where N might move to)? Or, does it have a simpler structure, such as (2)? *Many thanks to Tom Ernst, Jim Huang, Yafei Li , Bingfu Lu, Sze-Wing Tang and Zoe Wu for their help with this work. 1 Chinese in this work refers to Mandarin Chinese. 2 Mass nouns in English can occur in the bare form. 3 Demonstratives, possessives etc.which also make a nominal phrase definite can be analyzed in the same way, which we disregard here. 4 In this work, we do not concern ourselves with the functional projections between D and N; see Carstens (1991), Ritter (1991, 1995), Valois (1991), for instance, for discussions on some of such projections. See Tang (1990), Li (1998, 1999), Cheng and Sybesma (to appear) among others for the functional projections within nominal phrases in Chinese. 2 (2) NP | shu There are theoretical and empirical concerns in answering these questions. Theoretically, if shu is interpreted in the same way as the book , should they have the same structure? Both the positive and negative answers to this question have been explored. On the one hand, the assumption is that an N is an N and a D is a D semantically; that is, an N is a property-denoting expression (predicate-type) and a D is an individual-denoting expression (argument-type). They remain constant. A bare noun in languages like Chinese therefore needs to have a more complex structure than a simple N(P) when it is an argument (see Li 1997, 1998, 1999). In other words, a definite bare noun shu in an argument position should be projected as a DP instead of a NP. On the other hand, the assumption is that type-shifting can take place in some languages (see Chierchia 1998). A bare noun interpreted as definite in an argument position, for instance, has undergone type-shifting and takes on the argument role, rather than the predicate role. Since type-shifting has taken place, (2) can be a representation for a nominal expression in an argument position in Chinese. Theoretically, then, the choice between (1) and (2) relies on the adoption or rejection of type-shifting as an option in such cases. The empirical question is therefore more interesting. Is there empirical evidence helping us to determine if a DP structure is needed in languages like Chinese? Does the availability of a DP projection, in addition to an NP, provide opportunities to derive more empirical generalizations than an analysis that does without a DP projection? Li (1997, 1998, 1999) argues favorably for a positive answer, based on the facts concerning the word order and constituent structures within nominal expressions, the distribution of nominal phrases within a sentence, and the distribution of the plural marker men in Mandarin Chinese. Indeed, many interesting facts are derived by the presence of D and an N-to-D movement process. On the other hand, Fukui and Takano (1999), based on a theory of universal phrase structure proposed by Takano (1996) and Fukui and Takano (1998) and a prohibition against right-adjunction structures (Kayne 1994), argue that the word order between a relative clause and the head it modifies (the relative head) in different types of languages can be derived from a contrast in nominal structures. They argue that there are Japanese type and English type languages. In the English type, a DP is projected and N-to-D raising must take place. Such a raising process would place a relative clause, which is base-generated as left- adjunction to N, after the relative head N.5 By contrast, the type of languages like Japanese, which uses bare nouns as definite, indefinite and generic expressions, does not have a D projection. Because of the lack of a D projection, a N-to-D raising process cannot take place. A relative clause therefore maintains to be left-adjoined to N and, accordingly, surfaces on the left of the head N. Such a parametric difference between these two types of languages, according to Fukui and Takano, is not only manifested in the contrast in ordering between a relative clause and its head but also reflected in how relative clauses are formed and licensed in these two different types of languages. If Fukui and Takano are correct, their analysis would provide strong empirical support for the absence of a DP projection and N-to-D raising in languages like Japanese. It would also pose interesting challenges to Li's claim that Chinese does project a DP and that N-to-D movement does take place in this language. After all, Chinese and Japanese have identical word order with respect to a relative clause and the relative head. Moreover, Chinese appears to have similar properties concerning relativization as Japanese, in addition to the fact that both languages allow a bare noun in an argument position. 5 Kayne (1994), following Vergnaud (1974), argues that English relativization is a promotion process which raises the relativized nominal to the D position, the relative clause being a complement of D. 3 In brief, Chinese seems to belong to the Japanese type of languages, rather than the English type, at least with respect to the ordering of a relative clause and its head and the form/interpretation of bare nouns. This raises the question of whether there exists empirical evidence against the projection of a DP and the presence of N-to-D raising in Chinese, contrary to what Li has claimed. In this work, I will show that the various empirical generalizations claimed to be derivable from the parametric differences in the structure of nominals between the Japanese and English types of languages discussed in Fukui and Takano cannot be correct considering the Chinese relativization structures. The conclusion has to be drawn that Fukui and Takano's alleged correlation between N-to-D raising and the ordering between a relative clause and its head cannot be upheld. None of the properties of relativization that are claimed to be related to word order and N-to-D raising are true in Chinese. There is, thus, no evidence against the existence of a DP projection and N-to-D raising in Chinese. 2. Word order and the structure of relative clauses A major claim of Fukui and Takano's work is that the ordering between a relative clause and the head modified by the relative clause is the result of different structure and derivation. Adopting Kayne's proposal that only left-adjunction is allowed in the grammar, Fukui and Takano claim that when N is raised to D, a relative clause, being left-adjoined to N, would surface in a position following the relative head. This is illustrated by the English example in (3) and the structure in (4): (3) a picture which John saw yesterday.6 (4) DP / \ a D' /\ picture D' /\ N D /\ CP N=tpicture /\ which On the other hand, if N does not undergo raising, a relative clause stays left-adjoined to N. It surfaces at a position preceding the head noun, according to the spell-out rule proposed by Fukui and Takano (1998). This is illustrated by the Japanese example in (5), with the structure in (6): (5) a. John-ga kinoo mita syasin John-Nom yesterday saw picture 'the/a picture that John saw yesterday' b. John-ga kinoo atta gakusei John-Nom yesterday met student 'the/a student who(m) John met yesterday' 6 Fukui and Takano (1998) assume with Takano 1996 that a complement is generated on the left of a head. 4 (6) N1 / \ CP N1 =syasin The generalization thus emerges: the word order [relative clause + relative head] indicates N is not raised to D and the reverse order [relative head + relative clause] indicates N is raised to D. The former is illustrated by Japanese and the latter by English. Fukui and Takano further note that there is a variety of differences between the English type and the Japanese type languages which can "fall out in a simple and elegant fashion, based solely on the single parametric difference between the languages: English exhibits N-to-D raising, while Japanese does not (simply because the latter language lacks the category D)."(abstract)7 The differences discussed are briefly sketched in the following subsections. 2.1. Relative pronoun According to the authors, the presence or absence of a relative pronoun (such as which in (3)) can be derived from the structural contrast between (4) and (6). Assuming that a relative pronoun must be bound by the relative head, they claim that a relative pronoun in a structure like (4) in English is legitimate. In this structure, the relative head (the raised N) c-commands the relative pronoun which. The relative pronoun which is properly bound and therefore English allows a relative pronoun in a relative clause. By contrast, the relative head N1 (either the lower N1 or the two segmented category [N1, N1]) in (6) does not c-command CP because N1 has two segments, if the definition of c-command in Chomsky (1986) and Kayne (1994) is adopted which incorporates the notion of segments and exclusion. 8 When the relative head does not c-command the relative clause CP, it does not c-command a relative pronoun in the CP. For (6) to be well-formed, the relative clause cannot contain a relative pronoun; otherwise it would not be c-commanded and properly bound by the relative head. Consequently, Japanese does not have a relative pronoun in its relative clause. The contrast between the structures in (4) and (6) captures the contrast between the existence of a relative pronoun in English and the lack of one in Japanese. 2.2. Operator movement The lack of a relative pronoun, according to Fukui and Takano, indicates that a Japanese relative clause is not "operator-oriented." It is not licensed as modifier of the relative head through the mediation of a relative pronoun functioning as an operator creating an open position within the relative clause. Instead, it is licensed by an "aboutness" relation between the relative clause and the relative head. They claim that such an aboutness condition is not peculiar to the licensing of the 7 Fukui and Takano's proposal requires N-to-D movement to apply in all cases in English. This contrasts with, for instance, the analysis of Chierchia (1998) or Longobardi (1994) which does not raise N to D when the definite article the occurs. In such an analysis, the occupies the D position and N stays in the N position, N being a property-denoting, predicate type of expressions. Fukui and Takano place the in the Spec of D position and requires N to move to D. 8 The definition adopted in Fukui and Takano is this: (i) X c-commands Y iff X excludes Y and every element that dominates X dominates Y. (ii) X excludes Y iff no segment of X dominates Y. The two segmented category [N1, N1] is the relative head. Because it does not exclude CP with the upper N1, a segment of [N1, N1], dominating CP. Even if the lower N1 is the relative head, it does not c-command CP either, given that the upper N1, which is a segment of the category [N1, N1], dominates CP, thereby failing to satisfy the exclusion condition in (ii). 5 Japanese relative clause. It can also be seen in the licensing of certain topic constructions in Japanese and English: (7) sakana-wa tai-ga ii fish-Top red-snapper-Nom good 'As for fish, a red-snapper is the best.' (8) As for sports, I like baseball best. Attributing to Kuno (1973) and Murasugi (1991), they argue that this aboutness condition is also true of Japanese relative clauses. Thus, in the following example, the relative clause can be interpreted as being about a picture. (9) John-ga kinoo mita syasin John-Nom yesterday saw picture 'the/a picture John saw yesterday' Being in an aboutness relation with the relative head, the relative clause is properly licensed. Furthermore, because there is no operator movement in Japanese relativization and an aboutness condition is sufficient to license a relative clause, it follows that gapless relative clauses are possible in this language, such as (10a) and relativization is not subject to the island conditions. (10) a. [syuusyoku-ga taihen na] buturigaku employment-Nom difficult is physics 'physics (that) finding a job is difficult' b. [pro kiteiru yoohuku-ga yogoreteiru]sinsi is.wearing suit-Nom is.dirty gentleman 'the/a gentleman who the suit that is wearing is dirty' 2.3. Relative complementizer Another contrast that Fukui and Takano claim to follow from the said structural and movement differences is that a relative clause in Japanese, unlike an English one, has a TP structure, rather than a CP structure. Japanese relative clauses do not need CP: because of the lack of N-to-D raising, Japanese relative clauses cannot be operator-oriented. The fact that an operator is not needed makes the Spec of CP unnecessary. Following Diesing(1990)'s suggestion that a functional category is present in the structure only when it is necessary, Fukui and Takano claim that CP is not necessary in a Japanese relative clause and is not projected. The lack of a CP projection means the lack of a C which hosts a complementizer. This captures the fact that Japanese does not have the counterpart of the complementizer that in the English relative clause a picture that John saw yesterday.9 9 The argument is not easy to follow here. Even though an operator in the Spec of CP is not needed, it does not mean that the head is not needed and therefore it does not mean that CP is not needed. Consider the English example (i): (i) That he is here is important. In this sentence, there is no requirement of the presence of an operator in Spec of CP. However, the complementizer that must be present. The fact that a Spec of CP is not needed does not mean that the head C is also not needed. 6 2.4. Internally headed relative clause Finally, Fukui and Takano note that only languages of the Japanese type allow internally headed relative clause. Following Cole(1987)'s claim that an internally headed relative clause has a head that is a null pronominal correferential with the internal head, they note that the English structure (4) violates Binding Principle C because the head pro c-commands the internal head. By contrast, a pro in the relative head position does not c-command the internal head in (6). Consequently, only Japanese allows internally headed relative clauses. Briefly summing up, the contrast between the English structure (4) and the Japanese structure (6), as manifested in the different ordering between a relative clause and a relative head, derives the different properties of the English and Japanese relative clauses with respect to the possibility of a relative pronoun, operator movement, a complementizer and an internally headed relative clause. The lack of an operator movement is related to the availability of an abountess condition licensing a relative clause which does not obey island conditions. The lack of operator movement and the availability of licensing by an aboutness condition also make a gapless relative clause possible. When the relevant Chinese facts are considered, however, the correlated properties just mentioned do not fall out as predicted. The word order facts suggest that Chinese should be like Japanese. On the other hand, most of the other related properties suggest that Chinese should be more like English. The mixture of the clustering of properties raises doubt on the correctness of the typological claims we have just seen. 3. Correlated properties? A first look at the Chinese relative clauses may suggest that they are just like Japanese relative clauses. Like Japanese that only allows a relative clause to precede the relative head, Chinese also requires a relative clause to precede the relative head. In other words, Chinese should belong to the Japanese type of languages and have the structure in (6). If Chinese has the structure in (6), should we, then, expect Chinese relative clauses to have the properties in 2.2.-2.5? Unfortunately, if we examine the facts concerning Chinese relative clauses more carefully, we find that almost all the predictions are not true. 3.1. Relativization processes Before showing how the predictions in 2.1.-2.4. are not born out with respect to the Chinese relative clauses, we begin with the discussion on how relative clauses are formed in Chinese. 3.1.1. Movement There is evidence that relativization in Chinese need to be derived from movement of the relative head directly (see Vergnaud, 1974, Kayne 1994). For instance, part of an idiomatic expression can become a relative head: (11) a. [[ta kai ti de] daoi] dou hen chenggong (kai-dao 'open knife he open De knife all very successful =operate') 'All the operations he performed were successful.' b. [[ta chi ti de] cui ] bi bieren duo. (chi-cu 'eat vinegar he eat De vinegar compare others more = be jealous) 'He is more jealous than the others.' c. [[ta you ti de] moi] meiren ting-de-dong. (you-mo '(transliteration of) humor') he hu- De mor nobody listen-able-understand 'Nobody can understand his humor.' 7 Long distance movement is possible and the movement is sensitive to island conditions: (12) a. [[ta renwei [wo yinggai kai ti ] de] daoi ] dou hen nan. he think I should open De knife all very difficult 'The operations that he thinks I should perform are all difficult.' a'. *[[ta kai e ] hen zhongyao de] dao] ---subject island he open very important De knife 'the operation that it is important that he performs.' b. [[ta renwei [wo keyi chi ti ] de] cui ] hai bu zhi zhexie.