LICENSING CONDITIONS IN ELLIPSIS
Transcription
LICENSING CONDITIONS IN ELLIPSIS
LICENSING CONDITIONS IN ELLIPSIS Jun Abe Tohoku Gakuin University February, 2006 Department of Literature Tohoku Gakuin University Tsuchitoi 1-3-1, Aoba-ku Sendai-shi 980-8511 Japan [email protected] Abstract In this paper, I argue that a licensing condition on ellipsis which is characterized in terms of agreement in functional projections is not at work in such a language as Japanese that lacks agreement. Instead, apparent counterparts of such elliptic constructions that are subject to Lobeck's licensing condition are licensed in terms of what is called here lexical licensing in this language. Lexical licensing is characterized in terms of not only predicate-argument but also predicate-modifier relationship. After having demonstrated that apparent cases of Sluicing and VP Ellipsis in Japanese differ from typical cases of these constructions, I claim that they involve licensing of null arguments by predicates. Furthermore, I show that there are cases of what is called here Predicate Ellipsis in Japanese, which involve null predicates that are licensed by their arguments or modifiers. I also argue that what appears to be a case of the N'-deletion construction in Japanese can be regarded as simply a special case of Predicate Ellipsis. All these arguments will lead to the conclusion that ellipsis is among the many other phenomena that are affected by the presence of agreement in a given language. Licensing Conditions in Ellipsis Jun Abe Tohoku Gakuin University 1. Introduction This paper addresses the question of how to license ellipsis, based upon a series of works done by Lobeck (1990, 1995, 1999). Lobeck (1995) proposes what she calls "Licensing and Identification of pro", which is a general condition not only on pro in its usual sense but also on a subtype of ellipsis, exemplified in (1): (1) a. John talked to Bill but Mary didn't [e]. (VP Ellipsis) b. We want to invite someone, but we don't know who [e]. (Sluicing) c. Although John's friends were late to the rally, Mary's [e] arrived on time. (Ellipsis in NP) Lobeck assumes that the elliptic sites indicated in each sentence of (1) are in fact pro of various categories; namely, VP pro in (1a), TP pro in (1b), and NP pro in (1c). Though Lobeck's condition works in a fairly complicated manner, it suffices, for the purposes of this paper, to state the way this condition works for those ellipses given in (1) as follows: (2) Pro must be the complement of a functional head that agrees with its specifier. Thus, in (1a), the VP pro is licensed by the T which is in agreement with Mary, and in (1b), the TP pro is licensed by the C which is in agreement with who, and in (1c), the NP pro is licensed by the D which is in agreement with Mary's. Lobeck (1995) also argues that not all kinds of ellipsis are subject to this condition. She argues that Gapping and Stripping, illustrated below, are of a different type: (3) a. Mary met Bill at Berkeley and Sue [e] at Harvard. b. Jane gave presents to John, but not [e] to Geoff. (Gapping) (Stripping) That there are mainly two types of ellipsis constructions are already noted by Jackendoff (1971) and Williams (1977), among others. For instance, these two types of ellipsis constructions differ in whether they can appear in subordination, as shown below: (4) a. Charlie will leave town if his mother-in-law doesn't [e]. (VP Ellipsis) b. John takes LSD, although I don't know why [e]. (Sluicing) c. Lincoln's portrait didn't please me as much as Wilson's [e]. (5) a. *Sam played tuba whenever Max sax. b. *Felex sneezed since Max (too). (Ellipsis in NP) (Gapping) (Stripping) Furthermore, these two types of ellipsis differ in whether they are sensitive to Lobeck's condition given in (2). Thus, in Gapping and Stripping, there do not seem to be any functional heads that could license the elliptic sites of these constructions. Notice also that in these constructions, the elliptic sites do not appear to make a constituent, as is clear from the examples given in (3). Based upon these observations, Abe (1996) claims that this type of elliptic constructions does not involve a null category such as pro but rather involves deletion. Thus, the sentences in (3) have the underlying structures given in (6): (6) a. Mary met Bill at Berkeley and [Sue met Bill at Harvard] b. Jane gave presents to John, but not [Jane gave parents to Geoff] Then, following Abe and Hoshi (1997), at Harvard in (6a) undergoes rightward movement and is adjoined to T', as shown in (7a), and to Geoff in (6b) undergoes leftward movement and is adjoined to TP as shown in (7b); see Abe and Hoshi (1997) for the question why rightward movement is involved in (7a) and leftward movement in (7b). (7) a. Mary met Bill at Berkeley and [Sue [T' [T' Past [VP meet Bill ti]] at Harvardi]] b. Jane gave presents to John, but not [TP to Geoffi [TP Jane gave parents ti]] The surface structures of the sentences in (6) are derived by deleting the bold-faced parts in (7) under identity with the corresponding parts of the first conjuncts if it is assumed that identity should be defined in terms of non-distinctness. With this much in background, this paper addresses the following question: does the same picture emerge if we look at Japanese ellipsis constructions? In particular, is Lobeck's licensing and identification condition on pro also applicable to such languages as Japanese that appear to lack agreement? I claim that the answer to this question is negative and that her condition should be parameterized in terms of functional licensing vs. lexical licensing; that is, since Japanese does not have an agreement system, or only has a defective system of it, as claimed by Fukui (1986) and Kuroda (1988), the licensing of ellipsis by functional categories does not exist in this language and instead the Japanese apparent counterparts of English ellipsis constructions that involve licensing by functional categories involve licensing by lexical categories. It will be claimed that lexical licensing should be characterized in terms of the relationship between predicates on the one hand and arguments and modifiers on the other. This hypothesis has some plausibility, to begin with, when we consider the difference between Japanese, Korean and Chinese on the one hand and some Romance languages such as Italian and Spanish on the other with respect to the distribution of pro in its usual sense; that is, pro is largely restricted to the subject of a tensed clause in the latter type of languages, which exhibit rich agreement systems to license it, whereas pro appears fairly freely in argument positions in the former type of languages, which appear to lack agreement systems. From this point of view, it is natural to claim, following Kim (1999) in essence, that in the Romance type of languages, pro is licensed and identified by -features of T, hence its distribution being restricted to the subject of a tensed clause, whereas in the Japanese type of languages, pro is licensed by virtue of its argumenthood and its value is determined, according to Kim (1999), by way of copying of its antecedent or reconstruction in the sense of Fiengo and May (1994); I will return to this issue in Section 2.2. This paper is organized as follows: In Section 2, I demonstrate, following the arguments given by Nishiyama et al. (1996) and Kuwabara (1996), that apparent cases of Sluicing are in fact ones of Reduced Clefts and can be characterized as involving licensing of null arguments by copulas, and also, following the arguments given by Kim (1999) and Hoji (1998), that apparent cases of VP Ellipsis simply involve null arguments that are licensed by predicates in general. In Section 3, I demonstrate that there are cases of what is called here Predicate Ellipsis in Japanese, which involve null predicates that are licensed by their arguments or modifiers. In Section 4, I argue that what appears to be a case of Ellipsis in NP in Japanese can be regarded as simply a special case of Predicate Ellipsis. 2. No licensing by functional categories: cases of Sluicing and VP Ellipsis Suppose that Fukui (1986) is right in claiming that Japanese does not have any functional category or, if any, only has a defective category in not inducing Spec/head agreement. Then, given Lobeck's licensing condition, we will predict that Japanese does not have Sluicing, VP ellipsis and Ellipsis in NP in the sense that all these constructions involve licensing by functional categories. I would like to examine in this section whether this prediction is borne out. I will first examine Sluicing in some details, regarding which Takahashi (1994) claims that there is an instance of such a construction in Japanese. I will present objections to this claim by Nishiyama et al. (1996) and Kuwabara (1996), who argue that an apparent instance of Sluicing is in fact an instance of Clefting. As for VP ellipsis, Otani and Whitman (1991) claim that there is such a construction in Japanese but I will present Kim's (1999) argument that such a construction should be reanalyzed as NP ellipsis. As for Ellipsis in NP, Saito and Murasugi (1990) establish that there is a case of such a construction in Japanese. I will claim in Section 4 that there is a plausible alternative analysis to such a construction that involves licensing by lexical categories rather than by functional categories. 2.1. Apparent cases of Sluicing Takahashi (1994) provides an example such as (8) as an instance of Japanese Sluicing: (8) Minna-wa [John-ga dareka-o aisiteiru to] everyone-Top -Nom someone-Acc love itta ga, Comp said but boku-wa [dare-o [e] ka] wakaranai. I -Top who-Acc Q know-not 'Everyone said John loves someone, but I don't know who.' Objecting to this claim, however, Nishiyama et al. (1996) observe that the copula da can be inserted after the wh-phrase in (8), as shown below: (9) Minna-wa [John-ga dareka-o aisiteiru to] everyone-Top -Nom someone-Acc love boku-wa [dare-o da ka] wakaranai. I -Top who-Acc be Q know-not itta ga, Comp said but Notice that the copula da cannot be inserted in a normal question such as (10), which corresponds to the second clause of (8) in which the content of the elliptic site is spelled-out: (10) Boku-wa dare-o [John-ga aisiteiru] (*da) ka wakaranai. I -Top who-Acc -Nom love be Q know-not 'I don't know who John loves.' Thus, it is doubtful, Nishiyama et al. claim, that (8) is an instance of Sluicing. Rather, based upon this observation, they claim that it is an instance of Clefting, so that the underlying structure of the relevant part of (8) corresponds to the following sentence: (11) Boku-wa (John-ga aisiteiru no-wa) I -Top -Nom love dare-o (da) ka wakaranai. NO-Top who-Acc be Q know-not 'I don't know who it is that John loves.' where NO in the gloss stands for a nominalizer. What appears to be an instance of Sluicing in (8) is then derived from (11) by deleting the presuppositional part the one who John loves and the copula da. Further confirmation of this claim comes from the fact that in (8) the accusative case marker -o can be dropped, as shown below: (12) Minna-wa [John-ga dareka-o aisiteiru to] everyone-Top -Nom someone-Acc love itta ga, Comp said but boku-wa dare(-o) (da) ka] wakaranai. I -Top who(-Acc) be Q know-not This fact is straightforwardly captured under the Cleft approach since it is well-known that there are two types of Clefting in Japanese, one with a case marker attached to a focused phrase and the other without such a case marker, as demonstrated by Hoji (1987); a typical instance of Japanese Clefting is given below: (13) John-ga sita no-wa tennis(-o) da. -Nom did NO-Top tennis(-Acc) be 'It is tennis that John played.' Kuwabara (1996) gives further evidence for the claim that a Sluicing-like construction such as (8) is in fact a Reduced Cleft, that is, a Cleft with its presuppositional part omitted. This is concerned with examples of embedded elliptical clauses that do not seem to be licensed by agreeing complementizers, as illustrated below: (14) a. Taroo-wa [pro Naomi-ni -Top hanataba-o ageta to] itteita ga, -Dat bouquet-Acc gave Comp said but Ziroo-wa [daiamondo-no yubiwa-o -Top diamond -Gen ring (da) to] itteita. -Acc be Comp said lit. 'Taro said that he gave a bouquet to Naomi, but Ziro said that [a diamond ring].' b. Tuma-wa [watasi-ga itinen-mae wife -Top I to] soko-de Suzuki-ni atteiru -Nom one year-before there-at -Dat saw iu ga, watasi-wa [Suzuki-ni (da) kadooka] yoku oboeteinai. Comp say but I -Top -Dat be whether well remember-not lit. 'My wife says that I saw Suzuki there a year ago, but I don't really remember whether [Suzuki].' These examples look like cases of Stripping that occur in embedded clauses. Thus, in (14a), [he gave t to Naomi] appears to be elided and a diamond ring is left behind in the complement clause of say. Likewise, in (14b), [I saw t there a year ago] appears to be elided and Suzuki is left behind in the complement clause of remember. Recall, however, that Stripping does not occur in the subordinate clause, as illustrated in (5b), so it is unlikely that these examples are cases of Stripping. Notice that in these examples, copular insertion is possible after the remnant phrases, as shown in the same examples above, which is reminiscent of a reduced Cleft analysis of Sluicing-like examples, advocated by Nishiyama et al. (1996). Based upon these facts as well as many others, Kuwabara (1996) claims that both Sluicing-like examples and those examples in (14) should be given a unified analysis; that is, they should be treated as cases of Reduced Clefts. A further argument for the Reduced Cleft approach can be provided with island sensitivity. It has been widely observed that Sluicing show island insensitivity in many cases (see Chung et al. (1995) and Merchant (2001), among others). Chung et al. (1995), for instance, provide the following data: (15) a. Sandy was trying to work out which students would be able to solve a certain problem, but she wouldn't tell us which one. b. That certain countries would vote against the resolution has been widely reported, but I'm not sure which ones. c. The administration has issued a statement that it is willing to meet with one of the student groups, but I'm not sure which one. These data show that Sluicing is insensitive to the wh-island (cf. (15a)), the subject island (cf. (15b)) and the complex NP island (cf. (15c)). Contrary to these facts, Japanese Sluicing-like cases consistently show island sensitivity. In fact, Takahashi (1994) observes that what he takes as instances of Sluicing exhibits island effects, as shown below (the data are slightly modified): (16) a.??Mary-wa [John-ga nanika-o -Top uwasa-o kiita -Nom something-Acc bought Comp-Gen rumor-Acc heard sooda ga, boku-wa [nani-o I-heard but I katta to]-no ka] siritai. -Top what-Acc Q want-to-know 'I heard that Mary heard the rumor that John had bought something, and I want to know what (Mary heard the rumor that John had bought t).' b.?*Mary-wa [John-ni nanika-o -Top atta sooda ga, -Dat something-Acc gave woman-Dat met I-heard but boku-wa [nani-o I ageta] onna-ni ka] siritai. -Top what-Acc Q want-to-know 'I heard that Mary met a woman who had given something to John, and I want to know who (Mary met a woman who had given t to John).' c.?*Mary-wa [dareka-ga kubi-ni natta kara] okotteru sooda ga, -Top someone-Nom was-fired because be-angry I-heard but boku-wa [dare-ga ka] siritai. I -Top who-Nom Q want-to-know 'I heard that Mary is angry because someone was fired, and I want to know who (Mary is angry because t was fired).' These data show that what Takahashi takes as instances of Sluicing is sensitive to the Complex NP island (cf. (16a,b)) and the adjunct island (cf. (16c)).1 I will not go into details about the question why genuine cases of Sluicing are insensitive to such islands as illustrated in (15), but if we take those Japanese constructions given in (16) to belong to the same class of construction as those given in (15), the differences between (15) and (16) with respect to island sensitivity are unexpected. On the other hand, they are compatible with what the Reduced Cleft analysis will predict, since Hoji (1987) observes independently that Japanese Clefts show island sensitivity, as illustrated below: (17) a.??Mary-ga [John-ga t -Nom kono hon-o katta to]-no uwasa-o kiita no-wa -Nom bought Comp-Gen rumor-Acc heard NO-Top da. this book-Acc be 'It is this book that Mary heard the rumor that John had bought t.' b.?*Mary-ga [John-ni t ageta] onna-ni -Nom atta no-wa kono hon-o da. -Dat gave woman-Dat met NO-Top this book-Acc be 'It is this book that Mary met a woman who had given t to John.' c.?*Mary-ga [t kubi-ni natta kara] -Nom was-fired okotteru no-wa kono otoko-ga da. because be-angry NO-Top this man-Nom be 'It is this man that Mary is angry because t was fired.' (17a,b,c) correspond in relevant respects to (16a,b,c), respectively, and the correspondence also seems to hold in their acceptability; that is, the pair of (16a) and (17a) is not as degraded as those pairs of (16b)/(17b) and (16c)/(17c). As noted with the example given in (13), the case marker of a focused phrase can be omitted in Japanese Cleft constructions, and Hoji (1987) observes that when it is, the island effects go away. Thus, when the case markers are omitted from the focused phrases in (17), all the three examples show dramatic improvement in acceptability, though (17b,c) are still slightly degraded. Interestingly, much the same situation emerges with the examples given in (16); that is, when the case markers are omitted from the wh-phrases, the resulting sentences show dramatic improvement and their acceptability seems to correspond to that of those sentences in (17) that involve case drop. These facts are straightforwardly accommodated under the Reduced Cleft analysis for what Takahashi regards as instances of Sluicing. To conclude this section, I believe that the above arguments are so powerful that it will be fair to reject the claim that such a Japanese example as given in (8) is a case of Sluicing. 2.2 Apparent cases of VP Ellipsis Otani and Whitman (1991) claim that such a construction as exemplified below may involve VP Ellipsis: (18) John-wa [zibun-no tegami]-o suteta. Mary-mo [e] suteta. -Top self -Gen letter -Acc discarded 'John threw out his letter. -also discarded Mary did too.' At first sight, the second sentence of this example appears to involve a null object, but according to Otani and Whitman's analysis, the second sentence of this example can have roughly the following structure: (19) [IP NPsubj [I' [VP [NP e] tverb] V-Infl]] Here, the verb is raised into Infl, so that the elliptic site can be the whole VP containing the null object and the trace of the verb. Given that the verb of the first conjunct is also raised into the above Infl, then the VP of the first conjunct corresponds to zibun-no tegami 'self's letter' and the trace of the verb. And after this VP is copied onto that of the second conjunct, we will successfully get the appropriate meaning. The strongest motivation for this VP Ellipsis analysis, according to Otani and Whitman, is that it can properly capture the availability of a sloppy reading in such a construction; to put it in other words, the availability of a sloppy reading correlates with that of the VP Ellipsis analysis in question. Thus, the following example contrasts with (18) in that it does not allow a sloppy reading: (20) Zibun-no hatake-no ninzin-ga McGregor ozisan-no daikoobutu desita. self -Gen garden-Gen carrot-Nom Peter-mo [e] daisuki Mr. -Gen big-favorite was desita. -also very-fond-of was 'The carrots from self's garden were Mr. McGregor's big favorite. Peter was also fond of [e].' The second sentence of this example can only mean that Peter was also very fond of the carrots from McGregor's garden. This will follow straightforwardly from Otani and Whitman's analysis, since there is no way of analyzing the second sentence of this example as involving VP Ellipsis, lacking the antecedent VP to be copied. Given this argument, the validity of Otani and Whitman's claim that such an example as (18) may involve VP Ellipsis should be examined on the basis of whether the correlation between the availability of a sloppy reading and that of VP Ellipsis analysis is real. Though it is true that such an example as (20) lacks a sloppy reading, there is another account for this fact that is no less plausible, as Otani and Whitman mention briefly; that is, the availability of a sloppy reading requires "some type of parallelism constraint." It would not be unreasonable to claim that unlike (18), (20) does not satisfy the parallelism constraint in question. Otani and Whitman would claim that their VP Ellipsis analysis ensures how such a parallelism constraint is satisfied in order to get a sloppy reading. The question, then, comes down to whether the parallelism constraint in question is exactly what the VP Ellipsis analysis demands. A good case to examine this question is one in which the coordinated sentences appear to meet such a parallelism requirement and their corresponding subjects are involved in producing a sloppy reading. (21) a. Relevant examples are such as the following: Zibun-no hatake-no ninzin-ga McGregor ozisan-no daikoobutu desita. self -Gen garden-Gen carrot-Nom Mr. -Gen big-favorite was [e] Peter-no daikoobutu de-mo arimasita. -Gen big-favorite -also was lit. 'The carrots from self's garden were Mr. McGregor's big favorite, and [e] was Peter's big favorite, too.' b. John-wa zibun-no tuma-ga kiree da to omotteiru. -Top self -Gen wife-Nom beautiful be Comp think Bill-mo [e] kiree -also da to omotteiru. beautiful be Comp think lit. 'John thinks self's wife is beautiful. Bill also thinks [e] is beautiful.' It seems to me that in both these examples, the second sentences allow sloppy readings.2 According to Otani and Whitman's analysis, they should not allow such readings, since they do not involve configurations that can be analyzed as VP Ellipsis constructions. Thus, to the extent that the facts exemplified in (21) are real, they undermine the validity of Otani and Whitman's analysis considerably. Otani and Whitman (1991) give another piece of evidence to support their analysis, which is concerned with what they call locality of the sloppy identity interpretation. (22) As is well-known, in the following example: Johni thinks that Bill likes himi. (a) and Mary does too. (b) and Mary thinks that Bill does too. the continuation given in (22a) has the sloppy reading according to which Mary thinks that Bill likes her, whereas the continuation given in (22b) does not have such an interpretation. Otani and Whitman, then, claim that "the same restriction applies to the sloppy identity interpretation of null objects in Japanese examples like the following:" (23) a. John-wa [NY Times-ga zibun-no kizi-o -Top inyoositeiru to] kiita. -Nom self -Gen article-Acc be-quoting Comp heard 'Johni heard that the NY Times is quoting selfi's article.' b. Bill-mo [NY Times-ga [e] inyoositeiru to] -also kiita. -Nom be-quoting Comp heard lit. 'Bill also heard that the NY Times is quoting [e].' Though Otani and Whitman claim that (23b) does not allow the sloppy interpretation according to which Bill also heard that the NY Times is quoting Bill's article, it seems to me that such an interpretation is not inconceivable with this sentence. I believe that we can get such an interpretation more clearly in such an example as the following, in which the preference to the referential reading is suppressed (see note 2): (24) Hotondo subete-no hito-ga [NY Times-ga zibun-no kizi-o almost to every -Gen person-Nom -Nom self -Gen article-Acc be-quoting kiita ga, Bill-wa [NY Times-ga [e] inyoositeiru to] Comp heard but -Top inyoositeiru kiiteinakatta. -Nom be-quoting Comp didn't-hear lit. 'While almost everyonei heard that the NY Times is quoting selfi's article, Bill didn't hear that the NY Times is quoting [e].' Thus, Otani and Whitman's argument for their VP analysis based upon locality of the sloppy identity interpretation is not well-founded.3 Kim (1999) also argues against Otani and Whitman's (1991) VP Ellipsis analysis for null object constructions, based upon a variety of facts, including those related to locality of the sloppy identity interpretation, mentioned just above. He proposes instead that such null object constructions as considered by Otani and Whitman simply constitute part of the more general phenomena of null arguments seen in such languages as Chinese, Japanese and Korean, and that they all involve NP Ellipsis. He shares with Otani and Whitman the idea that null arguments of these languages are not simply empty pronouns. This point is most clearly illustrated by such a Japanese example as the following, which is adapted from a Japanese translation of Xu's (1986) Chinese example: (25) Hotondo subete-no hito-ga almost zibun-o hihansita ga, John-dake-wa [e] every -Gen person-Nom self -Acc criticized but only-Top hihansinakatta. didn't-criticize lit. 'Almost everyone criticized self, but only John didn't criticize [e].' The most natural interpretation of the second sentence of this example is the sloppy reading according to which only John criticized himself. If the null object of the second sentence were simply an empty pronoun, then it would violate Condition B of the binding theory. (26) Let us compare this example with the following: John-wa [e] hihansita. -Top criticized i) *John criticized himself. ii) John criticized someone else. Interestingly, this sentence does not have the reading where John criticized himself when uttered out of the blue. The contrast between (25) and (26) with respect to the availability of the self reading suggests that only when zibun 'self' is supplied in a context such as (25) is the self-reading available. Kim demonstrates that the NP Ellipsis analysis gives a straightforward account to this contrast. In (26), the only antecedent for the null object is John and hence after it is copied onto the null object, it induces a Condition C violation.4 On the other hand, in (25), zibun can be copied onto the null object of the second sentence, thereby giving rise to the self reading. This NP Ellipsis analysis should require the kind of special mechanism that is independently needed in VP Ellipsis to accommodate a strict reading as well as a sloppy reading. (27) Consider the following example: Mike-ga zibun-no kodomo-o butta. -Nom self -Gen child -Acc hit Sosite Jeanne-mo [e] butta. then -also hit lit. 'Mike hit his child, and then Jeanne also hit [e].' The second sentence of this example allows the sloppy reading where Jeanne hit her child and also allows the strict reading in which Jeanne hit Mike's child. To accommodate this fact, Kim basically follows Fiengo and May's (1994) mechanism of copying or reconstruction. According to this mechanism, the strict reading of the second sentence in (27) is obtained when zibun 'self' is taken as an -occurrence of this pronoun, that is, when it is taken in a sense as a referring expression, and his (=Mike's) child is copied onto the null category of the second sentence. On the other hand, the sloppy reading of this sentence is obtained when zibun is taken as a -occurrence of this pronoun, that is, when the literal content of self is copied.5 Alternatively, we may rely on the null topic analysis proposed by Huang (1982, 1984) to derive the strict reading of the second sentence in (27). Suppose that the NP Ellipsis analysis allows only literal copying of the material of the antecedent of a null argument, so that copying of zibun-no kodomo 'self's child' onto the null object of the second sentence in (27) induces only its sloppy reading. Suppose further that a null argument has another way of obtaining its semantic value; that is, acting as a free variable and being bound by a null topic in order to obtain its value. Given this option, the second sentence of (27) can have the following representation: [Øtopic(i) [Jeanne-mo [e](i) butta]] (28) Given that a null topic picks up an entity as its value which is salient in the discourse, it is natural to consider that the null topic in (28) refers to Mike's child, as is verified by the naturalness of the following example, in which an overt topic phrase referring to Mike's child is supplied to (27):6 (29) Mike-ga zibun-no kodomo-o butta. -Nom self -Gen child -Acc hit Sosite Mike-no kodomo-wa then -Gen child -Top Jeanne-mo [e] butta. -also hit 'Mike hit his child, and then as for Mike's child, Jeanne also hit him/her.' There is independent evidence for this null topic analysis to derive the strict reading of the NP Ellipsis in question. Discussing what he calls quantificational null objects in Japanese, Takahashi (1995) provides the following examples: (30) a. Hitori-no zyosi-gakusei-ga taitei-no sensei-o sonkei-siteiru. one -Gen female student-Nom most-Gen teacher-Acc respect ‘One female student respects most teachers.’ b. Hitori-no dansi-gakusei-mo [e] sonkei-siteiru. one -Gen male student -also respect lit. ‘One male student also respects [e].’ As is well-known, the scope interaction in Japanese respects the rigidity condition, according to which a structurally higher QP necessarily takes scope over a lower QP. Thus, in (30a), hitori-no zyosi-gakusei ‘one female student’ must take scope over taiteino sensei ‘most teachers’. As for (30b), Takahashi (1995) observes that it can be ambiguous depending upon what is supplied to the content of the null object. According to the present analysis, one reading is derived from literal copying of taiteino sensei and the other is derived by means of a null topic, which amounts to the reading in which the null topic functions as an E-type pronoun in the sense of Evans (1980), referring to the set denoted by the domain of most teachers. Takahashi further observes that in (30b) the subject must take scope over the null object on either reading, as is expected from the rigidity condition. With this much in background, Takahashi provides the following data, which contrast in an interesting way with those given in (30): (31) a. Taitei-no sensei-o hitori-no zyosi-gakusei-ga sonkei-siteiru. most-Gen teacher-Acc one -Gen female student-Nom respect ‘One female student respects most teachers.’ b. Hitori-no dansi-gakusei-mo [e] sonkei-siteiru. one -Gen male student -also respect lit. ‘One male student also respects [e].’ These data minimally differ from those given in (30) in that (31a) involves scrambling of the object to the top of the sentence. As has been well-known since Kuroda (1971), when the object QP is scrambled over the subject QP, it makes the sentence scopally ambiguous. Thus, (31a), unlike (30a), can have the reading in which the object taitei- no sensei ‘most teachers’ takes scope over the subject hitori-no zyosi-gakusei ‘one female student’. As for (31b), Takahashi makes the interesting observation according to which this sentence is also scopally ambiguous; that is, it allows the reading in which the null object takes scope over the subject QP. This is unexpected if the null object remains in its original position. A further interesting fact about (31b), which seems unnoticed by Takahashi (1995), is that when the null object takes scope over the subject, it necessarily functions like an E-type pronoun. This follows immediately under the present analysis: The reading derived from literal copying of taitei-no sensei has the LF representation of S-O-V order, thereby giving rise to only the interpretation in which the subject takes scope over the object. The E-type reading, on the other hand, involves a null topic in the top of the sentence, hence allowing this topic phrase taking scope over the subject QP, just as in (31a). Thus, to the extent that we take the relevant facts correctly, they lend strong support to the null topic analysis to derive the “strict” reading of the NP Ellipsis in question. Now turning back to cases that involve literally copying antecedent NPs onto null arguments, Hoji (1998) makes interesting observations with respect to what portion of the antecedent of an NP is in fact copied. (32) Consider one of his examples, given below: A: John-ga zibun-no kuruma-o aratta. -Nom self -Gen car -Acc washed 'John washed his car.' B: Bill-mo [e] aratta. -also washed lit. 'Bill also washed [e].' Hoji observes that in this dialogue, B's utterance can be interpreted in many ways, as given below: (33) Bill also washed a car/John's car/Bill's car. Based upon such facts, Hoji suggests that what is copied onto a null argument may be simply the N head of its antecedent and that "a nominal projection whose sole content is its head N can be interpreted in various ways as just indicated," (p. 142) for example in (33). It seems, however, that these facts simply show the possibility of copying only the N head of the antecedent of a null argument, but not the obligatoriness of such a way of copying, since when only the head of zibun-no kuruma 'self's car' in (32A), namely kuruma 'car' is supplied to the null object in (32B), the resulting sentence does not have all the readings given in (33), as shown below: (34) A: John-ga zibun-no kuruma-o aratta. -Nom self -Gen car -Acc washed 'John washed his car.' B: Bill-mo kuruma-o -also car aratta. -Acc washed 'Bill also washed a car.' (34B) allows only the indefinite reading of a car. Thus, the indefinite reading of a car in (32B) is derived by copying the N head of the antecedent of the null object, but the other two readings should be derived in other ways; the sloppy reading is derived by copying the whole NP of the antecedent of the null object and the strict reading is derived when the null object functions as a variable bound by a null topic operator and this operator picks up John's car as its value. To summarize, it should be clear from the above discussions that null arguments in Japanese are not simply referring expressions such as pro used in Romance languages, and that we need such a mechanism as provided by the NP Ellipsis analysis and reinforced by the option of null topics binding null arguments, to capture their available interpretations properly. Accepting such a revised version of Kim's (1999) NP Ellipsis analysis as a means of identifying the content of a null argument, we can then raise the question what licenses a null argument in such a language as Japanese, Korean, and Chinese. It is standardly claimed that so-called pro-drop phenomena observed in Romance languages involve agreement; that is, in such languages, pro is licensed by way of Spec/head agreement between pro and Infl/Tense, which usually accompanies an overt manifestation of it. By contrast, in such languages as Japanese, Korean and Chinese, null arguments do not seem to be licensed by way of Spec/head agreement and their distribution is more liberal than that of pro observed in Romance languages. Given that null arguments in these languages are identified in a way quite different from the way pro is identified, it is natural to expect that the licensing of those null arguments involve a different mechanism. by way of lexical selection. The most natural candidate would be to license them Accordingly, I propose that in a language that lacks agreement, Lobeck's licensing condition is off, and instead what we may call lexical licensing is at work in such a language. Thus, the difference with respect to the existence of agreement gives rise to such a parametric variation: (35) Functional vs. Lexical Licensing Parameter If a language L exhibits agreement, then a null category is constrained by Lobeck's (1995) licensing condition. If L lacks agreement, then a null category is licensed by selection of a lexical head.7 Now going back to Japanese Reduced Cleft constructions discussed in Section 2.1., they have the following schematic structure: [CP e] XP-case (da) (36) Here, the empty CP corresponds to the presuppositional part of Clefting, including the nominalizer -no plus the topic marker -wa. We can claim that this empty CP is licensed as the subject of the copular da. Therefore, we can reduce both apparent cases of Sluicing and VP Ellipsis to those involving null arguments and the latter are licensed by way of selection of lexical heads. 3. Predicate Ellipsis In the preceding section, I have argued that a language such as Japanese that lacks agreement involves lexical licensing for a null category. The cases illustrated there are ones in which null arguments are licensed by their predicates. In this section, I demonstrate another type of ellipsis in Japanese that involves lexical licensing, namely one in which a null predicate is licensed by its argument(s) or modifier(s), which I call Predicate Ellipsis. (37) Cases in point are illustrated below: Bill-ga Susan-o -Nom sikatta ra, John-mo Mary-o [Pred e]. -Acc scolded and -also -Acc 'Bill scolded Susan and also John Mary.' (38) a. Bill-ga Susan-o -Nom b. sikatta ra, John-ga Mary-o [Pred e]. -Acc scolded and -Nom -Acc John-ga Mary-o [Pred e]. -Nom -Acc (37) appears to be a case of Gapping in Japanese, where the missing predicate corresponds to that of the first conjunct, hence interpreted as sikatta 'scolded'. Interestingly, however, it is crucial in this sentence for the subject of the elided clause, namely John, to bear the marker -mo, meaning 'also'. If we change this marker into the nominative marker -ga, as given in (38a), the interpretation available to (37) does not obtain any more. Rather, the elided clause is interpreted as an unidentified event or state in which John and Mary both participate; hence it basically has the same meaning as (38b), the difference being whether the unidentified event or state is qualified by the preceding event expressed by the first conjunct of (38a). Let us then call the type of Predicate Ellipsis constructions that involves -mo in the elided clause the PEC(-mo) and the type of PEC that involves -ga in the elided clause the PEC(-ga). At first sight, it appears that the PEC(-ga), illustrated in (38), does not complete sentences and hence might be regarded as ungrammatical as they stand. Thus, the grammatical status of these sentences might be said to correspond to that of an English sentence like I saw, I kissed, etc. with missing objects. Nonetheless, it is not so hard for Japanese native speakers to imagine that these sentences are uttered out of the blue in a situation where the speaker tried to say, for example, John killed Mary, but he/she could not complete the sentences because he/she was so shocked or perplexed. Since these fragments describe events or states in which John and Mary are both involved, a natural response to these fragments is, what happened to John and Mary or what did John do to Mary, etc. Thus, sentence (38b), for instance, can be a fragment of any sentence given in (39) below, but of no sentence given in (40). (39) a. John-ga Mary-o -Nom korositesimatta. -Acc killed 'John killed Mary.' b. John-ga Mary-o -Nom aisiteiru. -Acc love 'John loves Mary.' c. John-ga Mary-o steru tumori-da. -Nom -Acc dump intend 'John is going to dump Mary.' (40) a. John-ga [Mary-o Bill-ga -Nom korositesimatta to] -Acc -Nom killed omotta. Comp thought 'John thought that Bill killed Mary.' b. John-ga [Mary-o aisiteiru] hito-o -Nom -Acc love nikundeiru. person-Acc hate 'John hates the person who loves Mary.' c. John-ga [Mary-o steta] -Nom koto-o kookai-siteiru. -Acc dumped fact-Acc regret 'John regrets that he dumped Mary.' In (39), John and Mary serve as arguments of the same predicates whereas in (40) they belong to different predicates although they are linearly adjacent and hence what follows could be a possible continuation of (38b). From these observations, it will not be unreasonable to hypothesize that the PEC(-ga) involves a null category of a predicate type and that there is a constraint at work between such a null predicate and the remnant phrases. Turning back to the PEC(-mo), illustrated in (37), this sentence, at first blush, appears to be an instance of Gapping, but it turns out that it is simply a special case of the PEC. The strongest reason for this is that the elliptic site in such a construction is not forced to be strictly identical to the antecedent predicate. Consider the following examples: (41) a. Bill-ga Susan-ni 1000-en ageta ra, John-mo Mary-ni [Pred e]. -Nom -Dat -yen gave and -also -Dat 'Bill gave Susan 1000 yen and also John Mary.' b. Bill-ga Susan-o -Nom syokuzi-ni tureteitta ra, John-mo Mary-o [Pred e]. -Acc dinner-to took and -also -Acc 'Bill took Susan for dinner and also John Mary.' The most natural interpretations of the elided predicates in (41a) and (41b) are the ones derived by copying the antecedent predicates; namely, "gave 1000 yen" in (41a) and "took for dinner" in (41b). Nonetheless, they are not the only interpretations available to these sentences but rather a variety of interpretations seem to be available depending upon the contexts. Thus, the elided predicate in (41a) can mean, for instance, "gave money" or "gave a present", etc. and that in (41b) can mean, for instance, "took to a bar for drink" or "took to a nice place for walking", etc. These facts indicate that the PEC(-mo) is simply a special case of the PEC in general, hence involving a null category of a predicate type and that the content of this null category is supplied by whatever event or state is consistent with some sort of parallelism requirement imposed by the lexical morpheme -mo 'also'. This hypothesis is further motivated by the fact that this type of construction also obeys the co-argumenthood constraint observed in the PEC(-ga); that is, the remnant arguments of a null predicate must be ones of the same predicate. Thus, consider the following sentences: (42) a.?*Bill-ga [Susan-o -Nom sono sensei-ga sikatta to] itta ra, -Acc that teacher-Nom scolded Comp said and John-mo Mary-o [Pred e]. -also -Acc 'Bill said that that teacher scolded Susan, and also John Mary.' b. *Bill-ga [Susan-ni attakoto-ga aru] hito-o -Nom -Dat have-seen sikatta ra, person-Acc scolded and John-mo Mary-ni [Pred e]. -also -Dat 'Bill scolded a person who has seen Susan, and also John Mary.' A parallelism requirement imposed by the morpheme -mo 'also' attached to John and the case particles -o and -ni attached to Mary in the elided clauses in (42a,b) demands that John function as something equivalent to Bill and Mary to Susan, which then establishes that these two arguments do not belong to the same predicate. Thus, the unacceptability of the sentences given in (42) clearly indicates that the PEC(-mo) also obeys the co-argumenthood constraint. 3.1 How to license null predicates Having established that the PEC(-ga) and PEC(-mo) share the same syntactic structure for the elided clauses in which a null predicate is involved, let us first consider how this null predicate is licensed. Notice that it is unlikely that this null predicate is licensed by Lobeck's licensing condition, since it is fairly obvious that there is no functional element that could enter into a Spec/head relation in order to license this null predicate. Rather, I propose that this null predicate is lexically licensed. A natural hypothesis in this case will be that a null predicate is licensed by the very existence of its arguments. Let us then hypothesize, as a first approximation, that lexical licensing is modified as in (43): (43) A null category NC is lexically licensed by either (i) a predicate P if NC is P's argument, or (ii) an argument A if NC is A's predicate. There are a number of questions to raise here. For one thing, one will raise whether it is necessary that two arguments be involved in licensing a null predicate. The answer is negative, as shown below: (44) a. Bill-ga Susan-o -Nom sikatta ra, John-mo [Pred e]. -Acc scolded and -also 'Bill scolded Susan and also John.' b. (Bill-ga Susan-o -Nom sikatta ra,) John-ga [Pred e]. -Acc scolded and -Nom These sentences are just as good as those in (37) and (38). One may also raise the question whether it is really crucial for a null predicate to be licensed by its argument. It would be predicted that if arguments are replaced by modifiers in those constructions under consideration, the sentences will become unacceptable. that things are more complicated to test this prediction. It will turn out, however, If we confine ourselves to considering the PEC(-ga), it seems that a stronger restriction is operative. Let us first consider the following examples: sikatta ra,) Mary-o [Pred e]. (45)?*(Susan-o Bill-ga -Acc -Nom scolded and (46) a. (Bill-ga Susan-ni -Nom -Dat atta ra,) John-ga Mary-ni [Pred e]. -Dat saw and b.?*(Susan-ni Bill-ga -Acc -Nom -Dat atta ra,) Mary-ni [Pred e]. -Nom saw and -Dat The generalization that will emerge from these observations is that when only one argument appears in the construction in question, only a subject or a phrase marked with the nominative case marker -ga is allowed.8 (47) illustrates cases in which only a modifier appears in such a construction, and these sentences are as bad as (45) and (46b). (47) a.?*(Oka-no ue-de John-ga asondeita ra,) ki-no sita-de [Pred e]. hill -Gen surface-on -Nom was-playing and tree-Gen under '(John was playing on a hill and) under a tree ...' b.?*(Katana-de John-ga asondeita ra,) hanmaa-de [Pred e]. sword-with -Nom was-playing and hammer-with '(John was playing with a sword and) with a hammer ...' When two or more arguments appear, one of them must be a subject, so that the examples given below are not as good as those given in (38): (48) a.?*(Bill-ni Susan-o -Dat syookaisita ra,) John-ni Mary-o [Pred e]. sono sensei-ga -Acc that teacher-Nom introduced and -Dat -Acc lit. '(That teacher introduced Susan to Bill and) John-Dat Mary-Acc ...' b.?*(Bill-ni hon-o sono sensei-ga ageta ra,) John-ni zassi-o [Pred e]. -Dat book-Acc that teacher-Nom gave and -Dat magazine-Acc lit. '(That teacher gave a book to Bill and) John-Dat a magazine-Acc ...' (49) a.?*(Susan-o oka-no ue-de Bill-ga mituketa ra,) -Acc hill-Gen surface-on -Nom found Mary-o ki-no and sita-de [Pred e]. -Acc tree-Gen under lit. '(Bill found Susan on a hill and) Mary-Acc under a tree ...' b.?*(Susan-o katana-de Bill-ga tataiteita ra,) Mary-o -Acc sword-with -Nom was-hitting and hanmaa-de [Pred e]. -Acc hammer-with lit. '(Bill was hitting Susan with a sword and) Mary-Acc with a hammer ...' Furthermore, as long as one remnant is a subject, the other can be a modifier, as shown below: (50) (Bill-ga oka-no ue-de/ katana-de asondeita ra,) -Nom hill -Gen surface-on/sword-with was-playing and John-ga ki-no sita-de/hanmaa-de -Nom tree-Gen under hammer-with lit. '(Bill was playing on a hill/with a sword and) John-Nom under a tree/with a hammer ...' Thus, what we have found out so far about what licenses a null predicate seems to be that a subject or a phrase marked with the nominative marker -ga licenses such a null predicate. Leaving aside the question why this is so for the moment, notice that the lexical licensing given in (43) is not incompatible with the data we have so far observed, since a subject is an argument of a null predicate. To consider next the PEC(-mo), it turns out that all the instances considered above become acceptable when -mo 'also' is attached to the remnant when there is one and to any remnant when more than one exits, as shown below:9 (51) a. Susan-o Bill-ga sikatta ra, Mary-mo [Pred e]. -Acc -Nom scolded and -also 'Bill scolded Susan and also Mary (Bill scolded).' b. Susan-ni Bill-ga atta ra, Mary-ni-mo [Pred e]. -Dat -Nom saw and -Dat-also 'Bill saw Susan and also Mary (Bill saw).' (52) a. Oka-no ue-de John-ga asondeita ra, ki-no sita-de-mo [Pred e]. hill -Gen surface-on -Nom was-playing and tree-Gen under -also 'John was playing on a hill and also under a tree (someone was playing).' b. ra, hanmaa-de-mo [Pred e]. Katana-de John-ga asondeita sword-with -Nom was-playing and hammer-with-also 'John was playing with a sword and also with a hammer (someone was playing.)' (53) a. Bill-ni Susan-o -Dat sono sensei-ga syookaisita ra, -Acc that teacher-Nom introduced and John-ni-mo Mary-o [Pred e]/ John-ni Mary-mo [Pred e]. -Dat-also -Acc -Dat -also lit. 'That teacher introduced Susan to Bill and also Mary to John (that teacher introduced).' b. Bill-ni hon-o sono sensei-ga ageta ra, -Dat book-Acc that teacher-Nom gave and John-ni-mo zassi-o [Pred e]/ John-ni zassi-mo [Pred e]. -Dat-also magazine-Acc -Dat magazine-also lit. 'That teacher gave a book to Bill and also a magazine to John (that teacher gave).' (54) a. Susan-o oka-no ue-de Bill-ga mituketa ra, -Acc hill-Gen surface-on -Nom found and Mary-mo ki-no sita-de [Pred e]/ Mary-o -also tree-Gen under ki-no sita-de-mo [Pred e]. -Acc tree-Gen under-also lit. 'Bill found Susan on a hill and also Mary under a tree (Bill found).' b. Susan-o katana-de Bill-ga tataiteita -Acc sword-with ra, -Nom was-hitting and Mary-mo hanmaa-de [Pred e]/ Mary-o hanmaa-de-mo [Pred e]. -also hammer-with -Acc hammer-with-also lit. 'Bill was hitting Susan with a sword and also Mary with a hammer (Bill was hitting).' Notice that the acceptability of the sentences given in (52) indicate that a modifier can license a null predicate in the PEC(-mo). Given that this ellipsis construction belongs to the same category as the PEC(-ga), we need to modify the licensing condition given in (43) in such a way that not only an argument but also a modifier can serve as a licenser for a null predicate, as given below: (55) A null category NC is lexically licensed by either (i) a predicate P if NC is P's argument, or (ii) an argument A or a modifier M if NC is A or M's predicate. This characterization is less attractive than the original licensing condition given in (43) in that it loses the symmetrical relation between a predicate and an argument, that is, the relation in which one licenses the other, whichever occurs as a null category. I will not dwell on this point any further here, leaving the formulation of lexical licensing of null categories in Japanese as in (55), but simply point out that the gist of lexical licensing of a null category is maintained in that licensing of a null category in Japanese is not mediated by agreement in functional categories, in accordance with the parametrization given in (35).10 One may raise the question why tense is also suppressed in the PEC. I have been discussing this construction as if only a predicate were suppressed, but the fact is that a predicate as well as a tense is suppressed. Notice that it is impossible to suppress only a predicate, leaving a tense, as shown below: (56) *John-ga [V e]-ta. -Nom Past This is because a tense morpheme in Japanese is an affix, so that it needs a predicate to support it. Thus, in all instances of the PEC, what is missing should be a Tense projection as well as a verb projection. I have been discussing how a null predicate is licensed by the lexical licensing in question, ignoring a tense projection. The question to raise now is how a null category that includes a tense is licensed. The lexical licensing in question does not appear to give a sufficient answer, since a tense is usually claimed to be not directly related to an argument or a modifier. I suggest that Saito and Hoshi's (2000) mechanism of -marking provides a way to solve this problem. According to their mechanism, a subject can be base-generated in the Spec of TP and a theta-role is assigned to it by a verb after this verb is raised to T, as schematized below: (57) TP 2 NP-Nom T' VP 2 V+T 1 tV Given this, it will be reasonable to claim that T and V constitute a complex predicate in a sense to be made precise and they as a whole participate in assigning their theta-roles to their arguments. Following this intuition, let us define a COMPLEX PREDICATE as follows: 11 (58) Given a sequence of X1+X2 ... Xn where Xi is either a predicate or a tense, the whole sequence constitutes a COMPLEX PREDICATE iff X1 licenses an argument or a modifier in the adjoined position of Xn. Here, that a predicate licenses an argument means that the former assigns one of its theta roles to the latter, and that a predicate licenses a modifier means that the two phrases are in a predicate-modifier relationship. Further, let us define an argument or a modifier of a COMPLEX PREDICATE as follows: (59) An XP is an argument or a modifier of a COMPLEX PREDICATE X1+X2 ... Xn iff it is an argument or a modifier of any Xi and is within the projection of Xn. Given Saito and Hoshi's mechanism of -marking, V+T in (57) constitutes a COMPLEX PREDICATE, since V assigns its subject theta-role to NP-Nom in Spec of TP after it raises to adjoin to T, and this subject is an argument of the COMPLEX PREDICATE, according to (59). We can now modify the licensing condition given in (55) into the one given below: (60) A null category NC is lexically licensed by either (i) a predicate P if NC is P's argument, or (ii) an argument A or a modifier M if NC is A or M's COMPLEX PREDICATE. With this modified licensing condition in mind, let us first consider a case of the PEC(-ga) in which a subject or a phrase with the nominative marker -ga licenses a null predicate, as illustrated in (44b). Let us assume that this fragment has the following underlying structure: (61) TP 2 NP-Nom T' ! Ø According to (60), the null category that corresponds to T' in (61) must be taken as a COMPLEX PREDICATE of the subject; otherwise, it will not be licensed by this subject. This in effect functions to restrict the set of possible interpretations that can be given to this null COMPLEX PREDICATE. In the present case, the null COMPLEX PREDICATE will receive an appropriate interpretation as long as whatever main verb is supplied to this null category takes the subject in Spec-TP as its argument, since such a main verb and the tense above it can constitute a COMPLEX PREDICATE, according to the above mechanism. Thus, the sentences given in (62) are possible continuations of (44b): (62) a. John-ga Mary-o sikatta. -Nom -Acc scolded 'John scolded Mary.' b. John-ga Bill-ni [yameru yooni] itta. -Nom -Dat stop said 'John told Bill to stop it.' Let us next consider a case of the PEC(-mo), as illustrated in (44a). In this case, it is first necessary to determine where the phrase accompanied by -mo is basegenerated. There is a good reason to assume that such a phrase is base-generated in a functional category above TP. (63) a. Consider the following examples: Mary-moi John-ga kinoo ei sikatta. -also -Nom yesterday scolded 'Also Mary, John scolded yesterday.' b. Mary-moi John-ga [Bill-ga -also kinoo ei sikatta to] omotteiru. -Nom -Nom yesterday scolded Comp think 'Also Mary, John thinks that Bill scolded yesterday.' c. Mary-moi John-ga [Bill-ga ej ei syookaisita] hitoj-ni -also -Nom -Nom atta. introduced person-Dat saw 'Also Maryi, John saw a person whoj Bill introduced heri to ej.' These examples show that non-subjects accompanied by -mo can appear to the left of subjects. Furthermore, the grammaticality of (63c) shows that such a phrase can be associated with a gap within an island, which then suggests that it is base-generated in its surface position. Let us call the functional category that accommodates such a phrase Top(icalization)P. (64) The underlying structure of (44a) will then be as follows: TopP 2 XP-also Top' ! Ø Again, according to (60), the null category that corresponds to Top' in (64) must be taken as a COMPLEX PREDICATE of the XP in its Spec; otherwise, it fails to satisfy this lexical licensing condition. Hence, the null predicate will receive an appropriate interpretation as long as whatever main verb is supplied to this null category takes the phrase in Spec-TopP as either its argument or its modifier. (44a) is an instance of the PEC(-mo) in which a phrase accompanied by -mo is an argument of a null COMPLEX PREDICATE, and the sentences given in (52) are instances in which such phrases are modifiers of null COMPLEX PREDICATEs. Thus, the sentences given in (65) and (66) are possible continuations of (44a) and those given in (52), respectively. (65) a. John-mo Mary-o -also sikatta. -Acc scolded 'John also scolded Mary.' b. John-mo Bill-ga -also sikatta. -Nom scolded 'Also John, Bill scolded.' (66) a. Ki-no sita-de-mo dareka-ga asondeita. tree-Gen under-also someone-Nom was-playing 'Also under a tree, someone was playing.' b. Hanmaa-de-mo dareka-ga asondeita. hammer -with-also someone-Nom was-playing 'Also with a hammer, someone was playing.' However, there appear to be counterexamples to the present analysis of the PEC(mo). Consider the following examples: (67) a. John-moi Mary-ga [Bill-ga ei kinoo -also sikatta to] itta. -Nom -Nom yesterday scolded Comp said 'Also John, Mary said that Bill scolded yesterday.' b. Ki-no sita-de-moi Mary-ga [John-ga ei asobu no]-o tree-Gen under-also -Nom negatteita. -Nom play NO-Acc was-hoping 'Also under a tree, Mary was hoping that John would play.' c. Hanmaa-de-moi Mary-ga [John-ga ei asob-eru yooni naru hammer -with-also -Nom no]-o -Nom play-can toward become NO-Acc negatteita. was-hoping 'Also with a hammer, Mary was hoping that John would be able to play.' It seems that these sentences are possible continuations of (44a), (52a) and (52b), respectively. Note that given the present analysis, these facts are unexpected, since the phrases accompanied by -mo in sentence-initial position do not seem to serve as arguments of COMPLEX PREDICATES which should be made out of the rest of the sentences. Note further that such exceptional behaviors do not extend to cases of the PEC(-ga). Let us consider the following examples: (68) a. [Bill-ga Susan-o -Nom sikatta to] Mary-ga itta ra, John-mo [Pred e]. -Acc scolded Comp -Nom said and -also lit. 'That Bill scolded Susan, Mary said and also John ...' b. [Bill-ga Susan-o -Nom sikatta to] Mary-ga -Acc scolded Comp itta ra, John-ga [Pred e]. -Nom said and -Nom lit. 'That Bill scolded Susan, Mary said and John ...' In accordance with the facts shown in (67), John-mo 'John-also' in (68a) can be taken to be in a correspondent relation with the embedded subject of the first conjunct, namely Bill. Thus, the following sentence can be a possible continuation of (68a). (69) John-mo Susan-o -also sikatta to Mary-ga itta. -Acc scolded Comp -Nom said lit. 'Also John, that he scolded Susan Mary said.' In (68b), by contrast, the corresponding interpretation is impossible, and hence the following sentence cannot be a possible continuation of (68b): (70) John-ga Lucy-o sikatta to Nancy-ga itta. -Nom -Acc scolded Comp -Nom said 'That John scolded Lucy Nancy said.' Rather, John-ga must be taken as the matrix subject of the null COMPLEX PREDICATE, and hence the following sentence can be a possible continuation of (68b): (71) John-ga [sore-wa uso da to] itta. -Nom that -Top lie be Comp said 'John said that that was a lie.' Why is it the case, then, that only the PEC(-mo) allows a looser relationship between the phrases marked with -mo and the rest of the sentences? There is a natural answer to this question. Saito (1985) suggests that in Japanese topic constructions, the topic "is licensed by some sort of 'aboutness relation' holding between the topic and the rest of the sentence." (p. 287) Typical examples that motivate this claim is ones in which the topic is not associated with any argument position inside the rest of the sentence, as observed by Kuno (1973) with such examples as the following: (72) a. Sakana-wa [tai-ga fish oisii]. -Top snapper-Nom tasty 'Speaking of fish, snapper is tasty.' b. Hana-wa [sakura-ga ii]. flower-Top cherry-Nom good 'Speaking of flowers, cherry blossoms are good.' Admitting that such an aboutness relation holds between the topic and the rest of the sentence, it is possible to claim that this relation suffices to license a null COMPLEX PREDICATE in the configuration given in (64), since a phrase accompanied by -mo can also function in the same way as a topic, as witnessed by the following example: (73) Sake-wa [karakuti-ga oisii] ga, wain-mo [karakuti-ga oisii]. alcohol-Top dry one-Nom tasty and wine-also dry one-Nom tasty 'Speaking of Japanese sake, dry sake is tasty, and as for wine as well, dry wine is tasty. Further evidence for the present claim comes from the fact that the topic phrase marked with -wa can also participate in the PEC, as illustrated below: (74) [[Bill-ga Susan-o -Nom sikatta to] Mary-ga itta] no-ni taisite, John-wa [Pred e]. -Acc scolded Comp -Nom said NO-to contrary -Top lit. 'While that Bill scolded Susan, Mary said, John ...' As expected under the present claim, not only a sentence such as (75) below in which John-wa clearly functions as an argument of the main verb but also one such as (76) in which an aboutness relation holds between John-wa and the rest of the sentence can be possible continuations of (74). (75) John-wa [sore-wa uso da to] itta. -Top that -Top lie be Comp said 'John said that that was a lie.' (76) John-wa Lucy-o -Top sikatta to Nancy-ga itta. -Acc scolded Comp -Nom said 'As for John, that he scolded Lucy Nancy said.' Thus all these facts lead to the conclusion that the licensing condition given in (60) must be supplemented by the following statement: (77) An XP in Spec-TopP can license a null COMPLEX PREDICATE by way of an aboutness relation. Let us now consider such cases as given in (37) and (38) where more than one argument is involved in licensing a null COMPLEX PREDICATE. Recall that I have shown above that there is some locality condition operative to such a construction, as exemplified by the fact that the sentences in (39) are possible continuations of (38b), but not those in (40) and also that the sentences given in (42) are unacceptable. These facts indicate that the two remnants in such constructions must be co-arguments of the null predicates. This will follow immediately under the present analysis. Consider the underlying structures of the elided parts in (37) and (38), which are shown below: (78) a. TopP b. 2 John-also 2 Top' 2 Mary-Acc TP John-Nom T' 2 Top' Mary-Acc ! T' ! Ø Ø Here each occurrence of Mary-Acc is temporarily put in the adjoined positions of Top' and T', respectively; the exact structures of (78a,b) will be presented in the next subsection. Notice that in these structures, the empty predicates are licensed by either John-also/-Nom or Mary-Acc as their argument, according to the licensing condition (60), but strictly we have not specified yet which phrase should serve as a licenser of the null predicates. Whichever serves as such, it establishes that it functions as an argument of the null predicates. This, in turn, establishes that the other argument also functions as an argument of the same null predicates; otherwise, it will not be given any appropriate interpretation. Note that a unique property of the PEC is that the content of the null predicate may remain unfilled even when it is interpreted in the semantic component, and hence that the remnants of this construction are required to have some semantically justifiable relationship with the null predicate in order to be properly interpreted; a predicate-argument relationship in the cases under consideration. The same holds true for cases such as those given in (50) and (54), in which one of the remnants is a modifier of some sort. Thus, such cases should also respect the same locality condition as observed with cases such as those given in (37) and (38) in which the remnants are both taken as arguments. That this is the case is shown below; (79a) and (79b) are possible continuations of (50) and (54), respectively, whereas none of those given in (80) and (81) can be: (79) a. John-ga ki-no sita-de/hanmaa-de hito-o korosita. -Nom tree-Gen under hammer-with person-Acc killed 'John killed a person under a tree/with a hammer.' b. Mary-mo ki-no sita-de/hanmaa-de dareka-ga mituketa/tataiteita. -also tree-Gen under hammer-with someone-Nom found /was-hitting 'Also Mary, someone found under a tree/was hitting with a hammer.' (80) a. John-ga [ki-no sita-de/hanmaa-de Bill-ga hito-o korosita to] -Nom tree-Gen under hammer-with -Nom person-Acc killed Comp omotta. thought 'John thought that Bill killed a person under a tree/with a hammer.' b. John-ga [ki-no sita-de/hanmaa-de hito-o korosita] otoko-o -Nom tree-Gen under hammer-with person-Acc killed man -Acc mokugekisita. witnessed 'John witnessed the man who killed a person under a tree/with a hammer.' (81) a. Mary-mo [ki-no sita-de asondeita] hito-ga mituketa. -also tree-Gen under was-playing person-Nom found 'Also Mary, a person who is playing under a tree found.' b. Mary-mo [hanmaa-de asondeita] hito-ga tataita. -also hammer-with was-playing person-Nom hit 'Also Mary, a person who is playing with a hammer hit.' We have not considered those cases of the PEC which have the combination of XP-also and NP-Nom; a relevant example is something like the following: (82) Bill-ga Susan-o -Nom sikatta ra, Mary-mo John-ga [Pred e]. -Acc scolded and -also -Nom lit. 'Bill scolded Susan and also Mary, John-Nom ...' The most natural interpretation of the second elided clause is that John also scolded Mary, though other interpretations are possible if provided with appropriate contexts. Under the present analysis, the elided clause in (82) has the following structure: (83) TopP 2 Mary -also Top' 2 Top TP 2 John-Nom T' ! Ø Recall that it has been shown above that a phrase in Spec-TopP can be in an aboutness relationship with the rest of the sentence. Thus, in (83), Mary-also can be interpreted as such and hence the null predicate can be solely licensed by John-Nom. This leads to the prediction that the combination of XP-also and NP-Nom does not require such a coargumenthood constraint as observed above as long as XP-also functions as the topic of the whole sentence and NP-Nom functions as the subject of the null predicate. This is in fact borne out; consider the following examples: (84) a. Bill-ga [sensei-ga Susan-o tataita to] -Nom teacher-Nom -Acc hit itta ra, Mary-mo John-ga [Pred e]. Comp said and -also -Nom lit. 'Bill said that the teacher hit Susan and also Mary John-Nom ...' b. Bill-ga [Susan-o -Nom tataiteita] hito-o sikatta ra, Mary-mo -Acc was-hitting person-Acc scolded and -also John-ga [Pred e]. -Nom lit. 'Bill scolded a person who was hitting Susan and also Mary John-Nom ...' The elided clause of (84a) can be interpreted as 'speaking of Mary, John said that the teacher also hit her' and that of (84b) can be interpreted as 'speaking of Mary, John also scolded a person who was hitting her.' In contrast, John-Nom in the elided clause of (84a) cannot be taken in such a way as to correspond to the embedded subject of the first conjunct, namely, sensei-ga 'teacher-Nom', thereby the elided clause being interpreted as 'speaking of Mary, Bill said that John also hit her.' This is exactly what we predict, since in this case, John-ga must license the null predicate as its subject. The same holds true for (84b). Finally, recall that the null category of the PEC is in fact a COMPLEX PREDICATE that consists of V and T and that a licenser of this null predicate can be an argument or a modifier of V according to the mechanism proposed by Saito and Hoshi (2000). Given this, it is predicted that a null predicate can be extended not only to a V+T projection but also to what is standardly called a complex predicate+T projection. Saito and Hoshi (2000) argue that a combination of more than one predicate can be treated as what is called here a COMPLEX PREDICATE in the sense defined in (58). Thus, in the trees given below: (85) a. XP ... 2 b. A X' ... 2 B XP 2 X' 2 A X' 2 YP 1 X' 2 X' B 2 X 2 X Y X 2 tY Y X suppose that Y is a two-place predicate and A and B are its arguments, and that X is a higher predicate that has its own theta roles to assign. According to Saito and Hoshi's mechanism of -marking, Y in (85a) can raise to adjoin to X and in that position, Y can assign its theta-roles to A and B, which are located in a X' projection. There is another derivation possible under their theory of -marking, as shown in (85b), in which Y is directly merged with X, without having its own projection, and assigns its theta-roles to A and B. According to Hoshi (2001), this way of -marking is possible at least in Japanese unless any functional category such as T or C intervenes between Y and X. It does not matter for the present purpose which option of -marking is taken, but notice that under the definitions given in (58) and (59), X and Y constitute a COMPLEX PREDICATE and A and B are arguments of this complex predicate. Given this, it is predicted that any complex predicate that allows such -marking as given in (85) can constitute the content of a null predicate of the PEC. This is in fact borne out. Let us again consider a fragment such as (38b). The sentences given below can be possible continuations of this fragment: (86) a. John-ga Mary-o izime-hazimeta. -Nom -Acc bully started 'John started to bully Mary.' b. John-ga Mary-o -Nom sin-aseta. -Acc die-let 'John let Mary die.' According to Saito and Hoshi's analysis, izime 'bully' in (86a) can raise and adjoin to hazime 'start' and further the amalgamated predicate can further raise and adjoin to T. Suppose that John-ga and Mary-o are located in the T projection. Then, they can receive its -roles from start and bully, respectively, since the amalgamated predicate is in the adjoined position of T. In this case, izime-hazimeta 'started to bully' as a whole constitutes a COMPLEX PREDICATE and hence can be the content of the null category of (38b). Likewise, in (86b), sin 'die' can be raised to ase 'let' and the whole V complex can then be raised to T. Suppose again that John-ga and Mary-o are located in the T projection. In that case, these arguments can receive their -roles from the V complex in the adjoined position of T. Then, sin-aseta 'let die' as a whole constitutes a COMPLEX PREDICATE and hence can be the content of the null category of (38b). The following examples show that the same holds true for the PEC(-mo): (87) a. Bill-ga Susan-o -Nom izime-hazimeta ra, John-mo Mary-o [Pred e]. -Acc bully started and -also -Acc lit. 'Bill started to bully Susan and also John Mary-Acc ...' b. Bill-ga Susan-o sin-aseta ra, John-mo Mary-o [Pred e]. -Nom -Acc die-let and -also -Acc lit. 'Bill let Susan die and also John Mary-Acc ...' The elided clause of (87a) can be interpreted as 'John also started to bully Mary' and that of (87b) as 'John also let Mary die.' 3.2. How to license the remnants of null predicates Let us now turn to the question why only a subject or a phrase marked with the nominative marker -ga appears to be able to license a null predicate in the PEC(-ga). claim that this question is directly related to how the remnants are licensed. I Recall that those types of ellipsis that are subject to Lobeck's licensing condition, namely, VP Ellipsis, Sluicing and Ellipsis in NP, illustrated in (1), all involve Spec/head agreement. Hence, it is reasonable to assume that in these cases, the remnants are licensed precisely because they enter into Spec/head agreement. Further, in the cases of Gapping and Stripping, I have been assuming that they do not involve a null category but rather involve coordination of full-fledged clauses in their underlying structures with deletion afterwards. Hence, what appears to be remnants in these constructions are not really remnants, so the issue in question will not arise in this type of construction. What about the Japanese constructions under consideration? Let us consider the structure given in (78b) again. I have so far been discussing how the null predicate is licensed, but I have so far mentioned nothing about how to license John-Nom and Mary-Acc, and it is not obvious how they are licensed. This might suggest that there is no licensing condition on remnants of ellipsis, but we can find a variety of facts that suggest that the remnants must be somehow licensed. I will provide here a fact of Sluicing in Bavarian German, reported by Lobeck (1995), which is a good case to suggest remnant licensing. Lobeck reports that according to Bayer (1984), "in Bavarian German, COMP[+WH] can realize agreement features of Person and Number, features which 'copy' those of embedded INFL," (p. 58) as shown below (here only an English gloss of the original sentence is provided): (88) you wanted come, but we know not when-2sg (you) come wanted-2sg 'You wanted to come, but we don't know when you wanted to come.' In this sentence, when, which is in the embedded COMP, shows an agreement in 2sg with the embedded INFL. Lobeck further notes that "although Sluicing is generally allowed in indirect questions in Bavarian German, the nominal agreement features in COMP exemplified in [88] can never be overtly realized when IP is empty" (p. 59), as shown below: (89) you wanted come, but we know not when(*-2sg) [e]. What she concludes from these facts as well as others is that "the appearance of morphological agreement features in COMP in Bavarian appears to be contingent on the occurrence of the same features in INFL." (p. 59) A natural explanation of this fact is that in a Sluicing case such as (89), if the 2sg morpheme shows up in the wh-phrase when, there is nothing that can license this morphology. If this sort of account is on the right track, then it suggests that a licensing condition is operative in ellipsis not only to a null category but also to remnants. Lobeck notes crucially that this account, if correct, lends strong support to the type of analysis that assumes base-generated empty categories in such a construction as Sluicing. Assuming it to be true that remnants of an elliptic construction need to be licensed, let us turn to the structure given in (78b) and raise the question how the subject and the object are licensed. I propose that argument remnants of the PEC are subject to the following condition: (90) Licensing of Argument Remnants in the PEC Argument remnants are licensed in terms of their case features. Furthermore, I follow Saito (1982, 1985) in assuming that the accusative case -o and the dative case -ni are assigned or checked by a lexical predicate whereas the nominative case -ga is a default case and is assigned to or checked with a phrase that is located in a T projection. (91) a. Thus, in the sentences given below, John-ga Mary-o -Nom nagutta. -Acc hit 'John hit Mary.' b. John-ga Mary-ni atta. -Nom -Dat saw 'John saw Mary.' Mary-o and Mary-ni are licensed in the sense relevant to (90) by the verbs nagutta 'hit' and atta 'saw', respectively, by receiving or checking an accusative case in the former and a dative case in the latter. On the other hand, John-ga is licensed when it is located in the T projection and receives or checks a nominative case. The claim that the nominative case -ga is a default case is supported by so-called multiple subject constructions such as given below: (92) a. Zoo-no hana-ga nagai. elephant-Gen nose-Nom long 'Elephants' noses are long.' b. Zoo-ga hana-ga nagai. elephant-Nom nose-Nom long (93) a. Bunmeekoku-no dansee-no heekinzyumyoo-ga mizikai. civilized country-Gen man -Gen average life span-Nom short 'Civilized countries' men's average life spans are short.' b. Bunmeekoku-ga dansee-ga heekinzyumyoo-ga mizikai. civilized country-Nom man -Nom average life span-Nom short In (92), zoo 'elephant' serves as the possessor of hana 'nose' and it is realized with the genitive -no in (92a), as expected. ga, as shown in (92b). It can also be realized with the nominative marker - In this case, it is very unlikely that ga-marking is done by a particular lexical head but rather it is more reasonable to claim that elephant in (92b) is located in the Spec-TP in surface structure and hence is marked with -ga. Likewise, in (93), bunmeekoku 'civilized country' is the possessor of dansee 'man' and civilized countries' men in turn is the possessor of heekinzyumyoo 'average life span'. these possessors are marked with -no, as in (93a), as expected. marked with -ga, as shown in (93b). Hence, They can also be Again, this fact suggests that they can be marked with -ga when they are located in the T-projection, irrespective of their lexical selection. With this much in background, let us turn back to the structure given in (78b), reproduced below: (94) TP 2 John-Nom T' 2 Mary-Acc T' ! Ø In this structure, John-ga is licensed according to (90) since it is located in the Spec-TP, hence being able to receive a default nominative case, whereas Mary-o is not licensed since there is no verb that could assign or check its accusative case. Then, how is this phrase licensed? I suggest that it is licensed by adjoining to a phrase that is independently licensed, that is, adjoining to John-Nom in (94). Thus, a more accurate structure is something like the following: (95) TP 2 [[John-Nom] [Mary-Acc]] T' ! Ø This way of licensing is modeled on the way of licensing a wh-adjunct, proposed by Saito (1994), according to which a wh-adjunct can be licensed by way of adjoining to a wh-phrase that is independently licensed. He observes that in Japanese an adjunct wh- phrase such as naze 'why' can be saved by a higher wh-argument when it appears in a typical island violation configuration, as shown below: (96) a. *John-wa [sono hon-o naze katta] hito-o sagasiteiru no. -Top that book-Acc why bought person-Acc be-looking-for Q lit. 'Q John is looking for the person [that bought that book why]?' b.??John-wa [nani-o naze katta] hito-o sagasiteiru no. -Top what-Acc why bought person-Acc be-looking-for Q lit. 'Q John is looking for the person [that bought what why]?' (96a) is a typical instance of island effects, because the adjunct wh-phrase naze appears within a complex NP island. This is expected under the assumption, made by Abe (1993) and Tsai (1994), that adjunct wh-phrases in situ must move to Spec-CP to be licensed, unlike argument wh-phrases, which can be licensed by way of binding, hence not exhibiting island effects. In (96b), the object within the relative clause is changed to a wh-phrase, and the sentence shows dramatic improvement. To account for this contrast, Saito gives the characterization given below for the way wh-adjuncts are licensed: (97) A wh-adjunct is licensed by means of either (a) or (b): a. it moves to Spec-CP to agree with a [+WH] Comp b. it is adjoined to a higher wh-argument According to this characterization, naze in (96b) can be licensed by way of adjoining to nani-o 'what' and it does not have to move out of the complex NP island; hence this sentence improves in its acceptability. Notice that the saving effects of wh-adjuncts arise when they are adjoined to wharguments at LF; that is, the adjunction operation involved is a covert one. Extending Saito's idea, Sohn (1994) and Takano (2002) argue that such an adjunction operation can take place overtly. What I would like to suggest here is that such a saving effect by way of one phrase adjoining to another can take place by Direct Merge, to use Hoshi's (2001) terminology. merged. Notice that in (95), John-Nom and Mary-Acc are directly This is because there is no other position for Mary-Acc that can be base- generated and properly licensed; hence Mary-Acc is licensed by being directly merged with John-Nom.12 It should be clear now why only a subject or a phrase marked with the nominative marker -ga appears to be able to license a null predicate of the PEC(-ga), as far as argument remnants are concerned. The answer is that only a phrase marked with the nominative marker -ga can license itself on its own in the construction in question; that is, by being located in a T-projection. Thus, while (44b) is acceptable, (45) is not, since Mary-o needs something it can adjoin to. The same explanation holds for the contrast between (46a) and (46b), since not only a phrase with the accusative marker -o but also one with the dative marker -ni is not licensed on its own and hence needs something it can adjoin to. Let us now turn to cases of the PEC(-ga) that involve modifiers. The most natural assumption on the licensing of modifier remnants of this construction will be something like the following: (98) Licensing of Modifier Remnants in the PEC Modifier remnants are licensed by the presence of their modifiees. Given this licensing condition, the fragments given in (47) are ungrammatical since there is no lexical verb that could license the modifiers under a tree/with a hammer.13 The licensing condition in question is also needed to explain the ungrammaticality of the fragments given in (49); otherwise, they would be grammatical in the configuration in which Mary-o is left-adjoined to the modifiers. The grammaticality of (50) indicates that a modifier can also be saved by adjoining to a phrase with the nominative marker -ga. It is predicted under the present analysis that a phrase that modifies TP can be an independent remnant of the PEC. This is in fact borne out with a tense modifier such as kinoo 'yesterday', as illustrated below: (99) (Ototoi Susan-o Bill-ga sikatta ra,) kinoo Mary-o/Mary-o kinoo before yesterday -Acc -Nom scolded and yesterday -Acc -Acc yesterday lit. '(The day before yesterday, Bill scolded Susan and) yesterday Mary-Acc/ Mary-Acc yesterday ...' It seems that this fragment is much better than those in (49). Thus, the following sentence seems to be a possible continuation of this fragment: (100) Kinoo Mary-o/Mary-o kinoo John-ga sikatta. yesterday -Acc -Acc yesterday -Nom scolded 'Yesterday John scolded Mary.' The reason why the fragment in (99) is grammatical is that kinoo 'yesterday', unlike under a tree and with a hammer, is licensed in the same way as a phrase with the nominative marker -ga; that is, by a T-projection. Given this, Mary-o in (99) is licensed by either left- or right-adjoining to kinoo (see note 12). With this tense modifier, we can further demonstrate that not only arguments but also modifiers are able to license a null predicate under consideration, since even if Mary-o is replaced by a modifier of the main predicate in (99), the resulting sentence is still grammatical, as shown below: (101) (Ototoi oka-no ue-de usi-o mikaketa ra,) before yesterday hill-Gen surface-on cow-Acc found and kinoo ki-no sita-de kinoo [Pred e]. sita-de/ki-no yesterday tree-Gen under tree-Gen under yesterday lit. '(The day before yesterday, I found a cow on a hill and) yesterday under a tree/under a tree yesterday ...' Thus, the following sentence can be a possible continuation of this fragment: (102) Kinoo ki-no sita-de/Ki-no sita-de kinoo uma-o mikaketa. yesterday tree-Gen under tree-Gen under yesterday horse-Acc found 'Yesterday, I found a horse under a tree.' Let us now turn to cases of the PEC(-mo), dealing with the fact that all the cases of the PEC that only involve non-subjects or modifiers of the main predicates become grammatical when -mo is added to one of the remnants of this construction, as illustrated in (51)-(54). Recall that the underlying structure of a case of the PEC(-mo) that involves one remnant is given in (64), repeated below: (103) TopP 2 XP-also Top' ! Ø Given this structure, the fact mentioned above is accounted for rather straightforwardly under the assumption that a phrase accompanied with -mo (or the topic marker -wa for that matter) is licensed in the Spec-TopP in the sense relevant to the licensing condition on remnants of the PEC. Furthermore, in those cases of the PEC(-mo) involving a remnant that lacks an independent licenser, we can simply follow the analysis of the PEC(-ga) in claiming that such a remnant is licensed by adjoining to another remnant which is independently licensed in the Spec-TopP. Thus, the underlying structure of (37) is not that given in (78a) but rather the following: (104) TopP 2 [[John-also] [Mary-Acc]] Top' ! Ø Here, Mary-Acc, which lacks its inherent licenser, namely its main predicate, is licensed by way of adjoining to John-also, which is independently licensed in the Spec-TopP.14 Finally, let us briefly discuss some consequences of the present analysis of the PEC concerning multiple subject constructions. Consider the following examples: (105) a. Zoo-ga [Pred e]. -Nom 'Elephants ...' b. Zoo-ga hana-ga mizikaku natteiru. -Nom nose-Nom short becoming 'As for elephants, their noses are becoming short.' It is natural to take (105b) as a possible continuation of (105a). arise as to how the null predicate in (105a) is licensed. A question will, then, Notice that in (105b), zoo-ga 'elephant-Nom' functions thematically as the possessor of hana 'nose' and the subject of the predicate mizikaku natteiru 'becoming short' is noses, not elephants. Thus, if the speaker intended to say (105b) when he/she uttered (105a), the null predicate would not be lexically licensed by zoo-ga 'elephant-Nom', since the latter does not serve as an argument of the predicate becoming short, still less its modifier. Nonetheless, the fact that (105b) is a possible continuation of (105a) suggests that zoo-ga 'elephant-Nom' does function as an argument of the rest of the sentence in an extended sense. It is not unreasonable to claim, following the Aristotelian idea that a proposition consists of a subject and a predicate, zoo-ga 'elephant-Nom' in (105b) functions as the subject of hana-ga mizikaku natteiru 'their noses are becoming short'; hence it can license the null predicate that corresponds to their noses are becoming short. Thus, this indicates that the licensing condition given in (60) must be supplemented by way of predication in the sense of Aristotle as well as by way of aboutness relation stated in (77). There is a further piece of evidence that supports the claim just made above. Let us consider the following examples: (106) a.?*Zoo-ga hana-ga [Pred e]. (cf. Zoo-no hana-ga [Pred e].) elephant-Nom nose-Nom b.?*Bunmeekoku-ga -Gen nose-Nom dansee-ga [Pred e]. civilized country-Nom man -Nom (cf. Bunmeekoku-no dansee-ga [Pred e].) civilized country-Gen man -Nom These examples do not sound good. Notice that they should be legitimate in terms of licensing of remnants, since the two phrases in each fragment are marked with the nominative –ga. Recall, however, that such remnants as in (106) need to be co- arguments of a null predicate; otherwise, both remnants will not be given appropriate interpretations. Given this, the fragments in (106) cannot be properly interpreted. Suppose that (105b) was what the speaker really wanted to say when he/she uttered (106a). Then, Zoo-ga is the subject of the rest of the sentence, i.e., hana-ga mizikaku natteiru 'their noses are becoming short', whereas hana-ga 'nose-Nom' is the subject of mizikaku natteiru 'becoming short'. Hence, these two remnants are not co-arguments of the same predicate, thereby (106a) being unable to receive any appropriate interpretation. The same account also holds for (106b) (cf. (93b)). As expected under the present analysis of the PEC, if the first occurrence of the NP marked with -ga is replaced by that with the topic marker -wa, then the resulting fragments are perfectly acceptable, as shown below: (106) a. Zoo-wa hana-ga [Pred e]. elephant-Top nose-Nom b. dansee-ga [Pred e]. Bunmeekoku-wa civilized country-Nom man -Nom This is because in these cases, the phrases zoo-wa 'elephant-Top' and bunmeekoku-wa 'civilized country-Top' are situated in the Spec-TopP and can have aboutness relations with the following TPs. This in turn makes it possible for the phrases hana-ga 'nose- Nom' and dansee-ga 'man-Nom' to be the sole subjects of the null predicates, hence making the co-argumenthood constraint irrelevant. 4. Ellipsis in NP in Japanese When I argued in Section 2 that Lobeck's licensing condition in terms of Spec/head agreement is not at work for Japanese elliptic constructions, one question was left unanswered, namely the question whether there is any Japanese counterpart of Ellipsis in NP or to use a more familiar term the N'-deletion construction, exemplified in (1c). Saito and Murasugi (1990) do claim that there is such a construction in Japanese, providing examples such as the following: (107) a. Gakusei-no sensei-e-no izon-wa yuruseru ga, student -Gen teacher-on-Gen reliance-Top can-tolerate but insei-no [e]-wa yurusenai. grad. student-Gen -Top cannot-tolerate 'I can tolerate undergraduates' reliance on teachers, but I cannot tolerate grad. students' [e].' b. Taroo-no kenkyuu-ni taisuru taido-wa -Gen research-toward ii ga, attitude-Top good but Hanako-no [e]-wa yokunai. -Gen -Top not-good 'Taro's attitude toward research is good, but Hanako's [e] is not.' If these sentences are real instances of the N'-deletion construction that is licensed by way of Spec/head agreement in D projections, as Saito and Murasugi claim that they are, then they will constitute strong counterexamples to the main claim of this paper. Thus, I would like to propose an alternative analysis of these constructions; that is, I claim that the N'-deletion construction exemplified in these sentences is subject to the same way of licensing as the PEC, namely, lexical licensing by arguments or modifiers. An initial motivation to regard these two constructions alike will obtain when we observe that it is possible to utter only the second half of the sentences of (107) out of the blue, as shown below: (108) a. Insei-no [e]-wa yurusenai nee. grad. student-Gen -Top cannot-tolerate 'I cannot tolerate grad. students' [e], can I?' b. Hanako-no [e]-wa yokunai nee. -Gen -Top not-good 'Hanako's [e] is not good, is it?' In order to facilitate imaginable contexts in which these sentences may be uttered, the ending morpheme nee is added to them, which corresponds most appropriately to a tag question in its meaning. When one hears these sentences, he/she will naturally respond by saying what? or what can't you tolerate about graduate students and what is not good about Hanako?, as expected. Hence, it seems that these constructions do not necessarily require antecedents, just as in cases of the PEC. Secondly, this construction allows not only one remnant, as exemplified above, but also allows two remnants, as shown below: (109) a. ?John-no Mary-e-no izon-wa -Gen yuruseru ga, -on-Gen reliance-Top can-tolerate but Bill-no Susan-e-no [e]-wa yurusenai. -Gen -on-Gen -Top cannot-tolerate 'I can tolerate John's reliance on Mary, but I cannot tolerate Bill's [e] on Susan.' b. ?John-no Mary-e-no hihan-wa -Gen ii ga, -to-Gen criticism-Top good but Bill-no Susan-e-no [e]-wa yokunai. -Gen -to-Gen -Top not-good 'John's criticism of Mary is good, but Bill's [e] of Susan is not good.' In this respect too, this type of construction exhibits the same property as the PEC. Suppose then that the same analysis is given to these cases as to those of the PEC in terms of lexical licensing rather than Lobeck's functional licensing in terms of agreement. Then, from the licensing condition on the PEC given in (60), it is expected that the null category of the construction under consideration also needs to be licensed by either an argument or a modifier. At first blush, this expectation does not seem to be borne out, as Saito and Murasugi claim that while arguments can license the elliptic site of this construction, as exemplified above and more below in (110), adjuncts cannot, as shown in (111): (110) a. Taroo-no kenkyuu-ni taisuru taido-wa Hanako-no [e] yorimo yoi. -Gen research-toward attitude-Top -Gen than good 'Taro's attitude toward research is better than Hanako's [e].' b. Rooma-no hakai-wa Kyooto-no [e] yorimo hisan datta. -Gen destruction-Top -Gen than horrible was 'Rome's destruction was more horrible than Kyoto's [e].' (111) a. *Saikin-wa hare-no hi-ga ame-no [e] yorimo ooi. recent-Top clear-Gen day-Nom rain-Gen than plentiful 'Recently, there have been more clear days than rainy [days].' b. *Hutakire-no hamu-wa yuusyoku-ni naru two slices-Gen ham -Top supper-to ga, make-up but hitokire-no [e]-wa naranai. one slice-Gen -Top not-make-up 'Two slices of ham make up a supper, but one slice [of ham] does not.' In (110a), Hanako's serves as the subject of the null category corresponding to attitude toward research and in (110b), Kyoto's serves as the object of the null category corresponding to destruction; hence the grammaticality of these sentences. The same account will hold for the sentences in (107)-(109). On the other hand, in (111a), ame- no 'rain-Gen' is not an argument of the null category corresponding to days nor in (111b) is hitokire-no 'one slice-Gen' is an argument of the null category corresponding to hams, and hence they are excluded, Saito and Murasugi claim. However, further examination of relevant data shows that modifiers can also license the elliptic site of the N'-deletion construction. (112) a. Consider the following examples: Ano toozi-no Rooma-no hakai-wa that time -Gen hisan konniti-no [e] yorimo -Gen destruction-Top today -Gen than datta. horrible was lit. 'That time's destruction of Rome was more horrible than today's [destruction of Rome].' b. Sono basyo-de-no Yamada sensei-e-no that place -in -Gen hihan-wa yuruseru ga, teacher-to-Gen criticism-Top tolerable but kono basyo-de-no [e]-wa yurusenai. this place -in-Gen -Top intolerable lit. 'I can tolerate in that place's criticism of Prof. Yamada, but not in this place's [criticism of Prof. Yamada].' The grammaticality of these sentences clearly demonstrates that not only arguments but also modifiers can license the elliptic site of the N'-deletion construction. We can attribute the ungrammaticality of the sentences in (111) not to the fact that adjunct genitive phrases do not license N'-deletion, as Saito and Murasugi claim, but to the fact that they involve deletion of nominals that are not predicates and hence cannot be null according to the licensing condition on null categories given in (60). In this way, we can maintain this licensing condition in explaining the grammaticality of all the data of the N'-deletion construction we have considered so far.15 Let us now move on to the issue of licensing of remnants in this construction. Recall that I claimed in the last section that the nominative case marker -ga is assigned as a default case when a phrase is located in a T-projection, and hence such a phrase is licensed on its own in the PEC. How about a phrase bearing the genitive case marker - no? Interestingly, this case marker also behaves as a default case, since, as Saito and Murasugi note, any NP or PP must bear a genitive marker when it is located in an Nprojection, as shown below: (113) a. yuubokumin-no nomad tosi-no hakai -Gen city-Gen destruction 'the nomad's destruction of the city' b. Hanako-no Taroo-e-no hihan -Gen -to-Gen criticism 'Hanako's criticism of Taro' c. Taroo-no genzitu-kara-no toohi -Gen reality -from-Gen escape 'Taro's escape from the reality' d. Taroo-no Hanako-e-no izon -Gen -on-Gen reliance 'Taro's reliance on Hanako' Thus it is quite natural to regard the genitive marker -no as a default case, exactly like the nominative marker -ga. Hence, any phrase bearing the genitive marker is licensed on its own by simply being located in a N-projection involving ellipsis. Most of the examples that we have seen so far which involve only one remnant are those in which the remnant serves as the subject of the null category. We have seen one example in which the remnant serves as the object of the null category, namely (110b), and also examples in which modifiers license N' elliptic sites, namely a NP modifier in (112a) and a PP modifier in (112b). It is also predicted that an argument PP bearing the genitive marker can also be a remnant of the elliptic construction in question. This is also borne out, as shown below: (114) a. John-no Mary-e-no -Gen Susan-e-no izon-wa yuruseru ga, -on-Gen reliance-Top can-tolerate but [e]-wa yurusenai. -on-Gen -Top cannot-tolerate lit. 'I can tolerate John's reliance on Mary, but I cannot tolerate [e] on Susan.' b. John-no Mary-e-no hihan-wa -Gen ii ga, -to-Gen criticism-Top good but Susan-e-no [e]-wa yokunai. -to-Gen -Top not-good lit. 'John's criticism of Mary is good, but [e] of Susan is not good.' Further, we have seen that the Japanese N'-deletion construction can involve more than one remnant, as illustrated in (109), in which one remnant serves as subject and the other serves as PP argument. In fact, other combinations of remnants are possible, as illustrated below in (115), where (115a) involves a modifier and an object NP as remnants of the N' elliptic site and (115b) involves a modifier and a PP argument. (115) a. Ano toozi-no Rooma-no hakai-wa that time -Gen yorimo hisan than konniti-no Kyooto-no [e] -Gen destruction-Top today -Gen -Gen datta. horrible was lit. 'That time's destruction of Rome was more horrible than today's [destruction] of Kyoto.' b. Sono basyo-de-no Yamada sensei-e-no that place-in-Gen hihan-wa yuruseru ga, teacher-to-Gen criticism-Top tolerable but kono basyo-de-no insei-e-no [e]-wa yurusenai. this place-in-Gen grad. student-to-Gen -Top intolerable lit. 'I can tolerate in that place's criticism of Prof. Yamada, but not in this place's [criticism] of a grad student.' All these data are exactly what we predict under the present assumptions about the way remnants of ellipsis are licensed, since a phrase bearing the genitive case marker -no can be licensed on its own by being located in a N-projection. To sum up, I have demonstrated a plausible new analysis of what Saito and Murasugi (1990) take as an instance of the N'-deletion construction, in terms of Predicate Ellipsis and lexical licensing. Granted that I have not given a more thorough outline of Saito and Murasugi's analysis of the N'-deletion construction along the way of Lobeck's licensing condition, the above arguments should not be taken to be enough for refutation of their analysis. However, to the extent that independent motivations are provided for lexical licensing of ellipsis in Japanese, the analysis of the Japanese N'deletion construction in terms of lexical licensing is more likely to be on the right track. 5. Concluding remarks I have argued in this paper that a licensing condition on ellipsis such as Lobeck's (1995) which is characterized in terms of agreement in functional projections is not at work in such a language as Japanese which lacks agreement. Instead, apparent counterparts of such elliptic constructions that are subject to Lobeck's licensing condition are licensed in terms of what is called lexical licensing in this language. Lexical licensing is characterized in terms of not only predicate-argument but also predicate-modifier relationship. In Section 2, I demonstrated, following the arguments given by Nishiyama et al. (1996) and Kuwabara (1996), that apparent cases of Sluicing are in fact ones of Reduced Clefts and can be characterized as involving licensing of null arguments by copulas, and also, following the arguments given by Kim (1999) and Hoji (1998), that apparent cases of VP Ellipsis simply involve null arguments that are licensed by predicates in general. In Section 3, I demonstrated that there are cases of what is called here Predicate Ellipsis in Japanese, which involve null predicates that are licensed by their arguments or modifiers. In Section 4, I argued that what appears to be a case of the N'-deletion construction in Japanese can be regarded as simply a special case of Predicate Ellipsis. All these arguments lead to the conclusion that ellipsis is among the many other phenomena that are affected by the presence of agreement in a given language and that parametrization is involved accordingly in terms of functional vs. lexical licensing. Notes 1 It has been well-known since Lasnik and Saito (1992) that Japanese does not exhibit subject island sensitivity, so it is generally the case that the subject island is not exploited to test whether a particular Japanese construction exhibits island sensitivity. Takahashi (1994) provides an example that he claims to show that Japanese Sluicing is sensitive to the wh-island, as given below (again, the example is slightly modified): (i) Mary-wa Bill-ni [John-ga aru mono-o -Top -Dat boku-wa [nani-o I katta kadooka] kiita sooda ga, -Nom one thing-Acc bought whether asked I-heard but ka] siritai. -Top what-Acc Q want-to-know 'I heard that Mary asked Bill whether John bought a certain thing and I want to know what (Mary asked Bill whether John bought t).' Takahashi gives ?? to such an example as (i), that is, judges it just as degraded as the pure complex NP case given in (16a). It seems to me that (i) is better than (16a), though it is not crystal clear how significant the difference is. See Abe (2005) for the claim that those Japanese constructions which are claimed to involve overt operator movement are not sensitive to the wh-island. 2 The judgments of the sentences in (21) regarding the availability of sloppy readings may be somewhat insecure, but I believe that it is because the referential readings of the null subjects involved are somehow preferred. Thus, if we control the relevant examples in such a way as to suppress such a factor, then it seems that sloppy readings are more easily available, as shown below: (i) Zibun-no hatake-no ninzin-ga hotondo subete-no hito-no self -Gen garden-Gen carrot-Nom almost every -Gen person-Gen daikoobutu desita ga, [e] Peter-no daikoobutu de-wa arimasen desita. big-favoritewas but -Gen big-favorite not was lit. 'The carrots from self's garden were almost everyone's big favorite, but [e] wasn't Peter's big favorite.' (ii) Hotondo subete-no hito-ga almost da to every -Gen person-Nom self -Gen wife-Nom beautiful be Comp omotteiru ga, Bill-wa [e] kiree think zibun-no tuma-ga kiree but -Top da to omotteinai. beautiful be Comp think-not 'Almost everyone thinks self's wife is beautiful, but Bill doesn't think [e] is beautiful.' 3 See Hoji (1998) and Kim (1999) for further counterevidence to Otani and Whitman's claim that null objects exhibit the same locality effects as VP Ellipsis in the availability of sloppy identity interpretation. 4 This account raises the question why the full sentence which results from copying John onto the null object in (26) is almost perfect in its acceptability, as shown below: (i) John-wa John-o -Top hihansita. -Acc criticized 'John criticized John.' Here we may assume Tancredi's (1994) version of binding theory, in which focus plays a crucial role in regulating the distribution of NPs. Following his theory, we can claim that (i) is void of a "Condition C" violation, since the second occurrence of John is focused. On the other hand, (26) cannot escape from a violation of this condition, since the element copied from the subject is in no way regarded as focused. See Abe (2002) for relevant discussion. 5 The second sentence of (27) in fact allows the third reading according to which Jeanne hit Mike. Kim (1999) says that "this third reading obtains naturally when Jeanne became angry at Mike because she saw him hit her son." (p. 265) This reading obtains by simply copying Mike onto the null object in question. 6 One may wonder if the null topic analysis can be extended to derive such a sloppy reading as is available to (27). In that case, Kim's (1999) NP Ellipsis analysis could be replaced by this null topic analysis. This move seems doubtful, however, since when an overt topic referring to self's child is supplied to (27), the resulting sentence does not sound very good, as shown below: (i)??Mike-ga zibun-no kodomo-o butta. -Nom self -Gen child -Acc hit Sosite zibun-no kodomo-wa then self -Gen child -Top Jeanne-mo [e] butta. -also hit To the extent that the contrast between this example and (29) in their acceptability is significant, it is reasonable to confine the option of a null topic to deriving the coreferential reading of a null argument. 7 This does not mean that selection by a lexical head is a sufficient condition for licensing a null category, but rather it should be taken as a necessary condition. See Saito (1985) and Hoji (1985) for the claim that the distribution of so-called pro is restricted in category and the status of its antecedent. 8 There is one reading where (45) and (46b) can be uttered out of the blue, that is, when watasi-wa 'I-Top' is suppressed; otherwise, they do not sound very good. same qualification also holds true for the examples given in (47)-(49). The This point will be briefly addressed in footnote 14. 9 When mo- 'also' is attached to an NP with either the nominative marker -ga or the accusative marker -o, the latter marker must be dropped for an unknown reason. That is why the accusative marker -o is suppressed in examples (51a), (53a,b) and (54a,b). Note further that the most natural interpretations of (52a,b) are the ones indicated in the translations, in which the elliptic sites are not identical to the corresponding parts in the first conjuncts. Thus, these examples also show that the PEC(-mo) does not require strict identity for the elliptic site. 10 One might wonder if there is any possibility of restoring the symmetrical relation in question by admitting the existence of null modifiers. On this point, I am simply following, without any argument, the standard claim that "pro" is allowed only in argument positions in Japanese (see Murasugi and Saito (1992), among others). It was pointed out to me by an unidentified person that a sentence such as the following may be relevant for the issue at stake: (i) John-ga zibun-no kuruma-de kita ra, Bill-mo kita. -Nom self -Gen car -by came and -also came lit. 'John came by his own car and Bill came, too.' If the second clause of this sentence has the sloppy reading on which Bill came by Bill's own car, then this will indicate the existence of a null modifier, in this case, one that means "by his own car". I am not sure if (i) has this sloppy reading; it is true that this sentence can be uttered to describe the situation intended by this sloppy reading, but it seems that there is a possibility of explaining that the second clause simply expresses Bill's coming and the way Bill came is inferred by the context. I will leave this matter open here. 11 According to the definition given in (58), a COMPLEX PREDICATE is strictly a sequence of verbs and tenses, but I will also use this term to mean any projection of it containing its arguments and modifiers. 12 Here I am tacitly assuming that when a phrase XP precedes another phrase YP, then YP is right-adjoined to XP, and hence that in (95) Mary-Acc is right-adjoined to John-Nom. (i) Notice, however, that the opposite order is also allowed, as shown below: Mary-o John-ga ... -Acc -Nom In this case, it might be claimed that Mary-Acc is left-adjoined to John-Nom. But, if one considers it odd to claim that both left- and right-adjunction are allowed, then we can claim that in (i), John-Nom is right-adjoined to Mary-Acc, and that even in this case, the former can license the latter. Alternatively, considering the fact that Japanese is a head-final language and hence usually only left-adjunction is permitted, then it would be more reasonable to claim that in (i), Mary-Acc is left-adjoined to John-Nom, and in (95), John-Nom is left-adjoined to Mary-Acc. For only an expository purpose, I assume that adjunction in both directions is allowed, so that (95) involves rightadjunction and (i) involves left-adjunction. See Sohn (1994) for relevant discussion. 13 One might wonder whether null COMPLEX PREDICATES cannot serve as licensers of such modifier remnants as under a tree/with a hammer. Notice that they are labeled as a projection of T and as such include tense, so what is implicitly assumed here is that a part of a null COMPLEX PREDICATE does not serve as a licenser of a modifier remnant. See below in the text those cases in which null COMPLEX PREDICATES do act as such licensers. 14 It was noted in note 8 that the unacceptable examples in (45)-(49) become acceptable when they are interpreted in such a way that watasi-wa 'I-Top' is suppressed. This fact immediately follows under the assumption that in such an interpretation, a zero topic referring the speaker exists in the Spec-TopP, hence licensing other remnants. 15 Saito and Murasugi (1990) take the following example as ungrammatical: (i)(*)Taroo-ga Hanako-ni yoseru sinrai-wa Ziroo-ga Akiko-ni yoseru-no [e] -Nom -Dat have trust -Top -Nom -Dat have -Gen yorimo atui. than deep 'The trust that Taro has in Hanako is deeper than [the trust] that Ziro has in Akiko.' They attribute the ungrammaticality of this sentence ultimately to the status as a modifier of the relative clause that appears to license the N' elliptic site of this sentence. It does not seem to me, however, that this sentence is unacceptable. Saito and Murasugi show further that (i) contrasts with the following sentence in its grammaticality: (ii) Taroo-ga Hanako-ni yoseru sinrai-wa Ziroo-no [e] yorimo atui. -Nom -Dat have trust -Top -Gen than deep 'The trust that Taro has in Hanako is deeper than Ziro's [e].' (ii) differs from (i) in that the N' elliptic site of this sentence is supported by a genitive NP, namely, Ziroo-no, which serves as its subject rather than its modifier. Saito and Murasugi note that the contrast in grammaticality between (i) and (ii) is problematic to their analysis, but it seems to me that (i) is just as acceptable as (ii). References Abe, J., 1993. Binding conditions and scrambling without A/A' distinction. Ph.D. dissertation, University of Connecticut. Abe, J., 1996. Ellipsis: deletion, copying or both? in: Nishiyama, Y., Otsu, Y. (Eds.), Keio Studies in Theoretical Linguistics I. Institute of Cultural and Linguistic Studies, Keio University, pp. 49-87. Abe, J., 2002. Toward a better understanding of binding theory. English Linguistics 19, 54-80. Abe, J., 2005. Scrambling and operator movement. Unpublished manuscript, Tohoku Gakuin University. Abe, J., Hoshi, H., 1997. Gapping and P-stranding. Journal of East Asian Linguistics 6, 101-136. Bayer, J., 1984. COMP in Bavarian. Linguistic Review 3, 209-274. Chung, S., Ladusaw, W., McCloskey, J., 1995. Sluicing and logical form. Natural Language Semantics 3, 239-282. Evans, G., 1980. Pronouns. Linguistic Inquiry 11, 337-362. Fiengo, R., May, R., 1994. Indices and Identity. MIT Press, Cambridge, MA. Fukui, N., 1986. A theory of category projection and its applications. Ph.D. dissertation, MIT. Revised version published in 1995 as Theory of Projection in Syntax, CSLI Publications, Stanford, CA, distributed by University of Chicago Press. Hoji, H., 1985. Logical form constraints and configurational structures in Japanese. Ph.D. dissertation, University of Washington. Hoji, H., 1987. Japanese clefts and reconstruction/chain binding effects. Talk at WCCFL 6 at University of Arizona, March 21, 1987. Hoji, H., 1998. Null object and sloppy identity in Japanese. Linguistic Inquiry 29, 127-152. Hoshi, H., 2001. Relations between thematic structure and syntax: a study on the nature of predicates in Japanese. SOAS Working Papers in Linguistics and Phonetics in 2001. Huang, J., 1982. Logical relations in Chinese and the theory of grammar, Ph.D. dissertation, MIT. Huang, J., 1984. On the distribution and reference of empty pronouns. Linguistic Inquiry 15, 531-574. Jackendoff, R., 1971. Gapping and related rules. Linguistic Inquiry 2, 21-35. Kim, S., 1999. Sloppy/strict identity, empty objects, and NP ellipsis. Journal of East Asian Linguistics 8, 255-284. Kuno, S., 1973. The Structure of the Japanese Language, MIT Press, Cambridge, MA. Kuroda, S., 1971. Remarks on the notion of subject with reference to words like also, even or only: Part II. Annual Bulletin, vol. 4, 127-152. Research Institute of Logopedics and Phoniatrics, Tokyo University. Kuroda, S., 1988. Whether we agree or not: a comparative syntax of English and Japanese. in: Poser, W. (Ed.), Papers from the Second International Workshop on Japanese Syntax, CSLI, Stanford, CA, pp. 103-143. Kuwabara, K., 1996. Multiple wh-phrase in elliptical clauses and some aspects of clefts with multiple foci. Formal Approaches to Japanese Linguistics 2, MIT Working Papers in Linguistics 29, MIT, Cambridge, MA, pp. 97-116. Lasnik, H., Saito, M., 1992. Move : Conditions on Its Application and Output, MIT Press, Cambridge, MA. Lobeck, A., 1990. Functional heads as proper governors. Proceedings of NELS 20, Graduate Students Linguistics Association, University of Massachusetts, Amherst, MA. Lobeck, A., 1995. Ellipsis: Functional Heads, Licensing, and Identification, Oxford University Press, Oxford. Lobeck, A., 1999. VP ellipsis and the minimalist program: some speculations and proposals," in: Lappin, S., Benmamoun, E. (Eds.), Fragments: Studies in Ellipsis and Gapping, Oxford University Press, Oxford, pp. 98-123. Merchant, J., 2001. The Syntax of Silence: Sluicing, Islands, and the Theory of Ellipsis, Oxford University Press, Oxford. Murasugi, K., Saito, M., 1992. Quasi-adjuncts as sentential arguments. Proceedings of the Western Conference on Linguistics 5, 251-264. Nishiyama, K., Whitman, J., Yi, E., 1996. Syntactic movement of overt wh-phrases in Japanese and Korean. in: Akatsuka, N., Iwasaki, S., Strauss, S. (Eds.), Japanese/Korean Linguistics 5, CSLI Publications, Stanford, CA, pp. 337-351 Otani, K., Whitman, J., 1991. V-raising and VP ellipsis. Linguistic Inquiry 22, 345-358. Saito, M., 1982. Case marking in Japanese: a preliminary study. Unpublished manuscript, MIT. Saito, M., 1985 Some asymmetries in Japanese and their theoretical implications, Ph.D. dissertation, MIT. Saito, M., 1994. Additional-wh effects and the adjunction site theory. Journal of East Asian Linguistics 3, 195-240. Saito, M., Murasugi, K., 1990. N'-deletion in Japanese. University of Connecticut Working Papers in Linguistics 3, pp. 87-107. Saito, M., Hoshi, H., 2000. The Japanese light verb construction and the minimalist program. in: Martin, R., Michaels, D., Uriagereka, J. (Eds.), Step by Step: Papers in Minimalist Syntax in Honor of Howard Lasnik, MIT Press, Cambridge, MA, pp. 261-295. Sohn, K., 1994. Adjunction to argument, free ride and a minimalist program. in: Koizumi, M., Ura, H. (Eds.), Formal Approaches to Japanese Linguistics 1, MIT Working Papers in Linguistics 24, MIT, Cambridge, MA, pp. 315-334. Takahashi, D., 1994. Sluicing in Japanese. Journal of East Asian Linguistics 3, 265-300. Takahashi, D., 2005. Quantificational null objects and noun phrase ellipsis. Unpublished manuscript, Tohoku University. Takano, Y., 2002. Surprising constituents. Journal of East Asian Linguistics 11, 243-301. Tancredi, C., 1994. Toward eliminating binding theory. Unpublished manuscript, Yokohama National University. Tsai, D., 1994. On economizing the theory of A'-dependencies, Ph.D. dissertation, MIT. Williams, E., 1977. Discourse and logical form. Linguistic Inquiry 8, 101-139. Xu, L., 1986. Free empty category. Linguistic Inquiry 17, 75-93.