HOW TO PROBE EXPLETIVES Jun Abe Tohoku Gakuin University January 2009 1-3-1 Tsuchitoi, Aoba-ku, Sendai 980-8511 JAPAN [email protected] TEL&FAX: 81-22-721-3239 HOW TO PROBE EXPLETIVES Abstract. This article aims to argue for the Move-approach to expletive constructions, according to which expletives satisfy the EPP by moving to a designated position rather than by being simply externally merged into it. In particular, it defends Sato’s (2008) analysis of the there-construction under Chomsky’s (2008) probe-goal system, which crucially assumes that the expletive and its associate make a constituent underlyingly and that the associate carries partitive Case which needs to be checked in accordance with the probe-goal mechanism. It is demonstrated that this analysis is both conceptually and empirically superior to any known version of the External Merge-approach to such a construction. Further, I propose a revision of Sato’s system according to which expletives are characterized differently from other ordinary DPs in terms of formal features: they carry unvalued -features. This makes it possible to accommodate cases in which expletives appear to be involved in default agreement due to the intervention effects induced by the experiencer phrases of raising predicates. Key Words: expletives, partitive Case, default agreement 1. Introduction It is no doubt that the investigation of how expletive constructions are derived has given a significant impact on existing movement theories at various stages of the development of generative grammar. The reason for this is that most cases of the expletive constructions have corresponding movement cases that involve raising of DPs to the 1 positions occupied by expletives. Thus, such an instance of the English there-construction as (1a), for example, has the corresponding raising case shown in (1b). (1) a. There is someone in the garden. b. Someone is in the garden. The standard claim with such a pair of sentences is that they share an underlying structure such as (2) below and that (1a) is derived from this structure by inserting there in the subject position indicated by e, whereas (1b) is derived from it by moving someone to that position. (2) [ e is someone in the garden] Since the advent of Case theory originally proposed by Vergnaud and extensively developed by Chomsky (1981), it has been a central issue how the associate of there (i.e., someone in (1a)) obtains Case. Concretely, given that in the derivation of (1b), someone raises to the subject position to obtain nominative Case, it raises the question how someone in (1a) obtains Case. Relating to this issue is the peculiar agreement pattern observed with such a construction. As witnessed by comparing (1a) with the following example, the matrix T agrees not with its subject there but rather with its associate. (3) There are some men in the garden. Obviously, this raises the question how it is possible to establish an agreement relation between T and the associate of there, given that the latter does not occupy the normal position with which T agrees. These are basic issues of expletive constructions that 2 have been tackled since the LGB era. Since Chomsky (1986) proposed a LF movement theory of the there-construction, according to which the associate of there undergoes LF movement to the position occupied by there, he has taken a consistent stance in approaching these issues; that is, he takes it that there has no direct relevance for the way its associate obtains Case from and agrees with T. Rather, there is taken as just a place holder for subject position to satisfy the Extended Projection Principle (henceforth, EPP). This basic idea is carried over to the most recent system of feature checking theory under the Minimalist Program, often called probe-goal system (cf. Chomsky (2000)), in which T probes with respect to its -set and finds the associate of there as a goal, checking the -set against that of the latter, and also the nominative Case of the associate is checked by T as a reflex, and finally there is inserted into the Spec-TP to satisfy the EPP feature. This article aims to argue against this position and to support the following claims: (4) a. Expletives and their associates form constituents underlyingly, following Sabel (2000), Hornstein and Witkos (2003) and Sato (2008), among others. b. Expletives have unvalued -features that may or may not be valued by their associates. Under this alternative system, such a there-construction as (1a) does not share with the corresponding raising case (1b) the underlying structure given (2), but rather has roughly the following underlying structure: (5) [ e is [there someone] in the garden] 3 (1a) is derived from this structure by raising there to the subject position marked by e, in much the same way as (1b) is derived from (2) by raising someone. Nonetheless, there are two crucial differences between the two cases: (i) the associate of there receives Case other than nominative and (ii) even though there is directly involved in checking the -features of T, the features it bears originally are unvalued, as indicated in (4b). I argue that this approach to expletive constructions, which is named as Move-approach in contrast with the standardly assumed External Merge-approach, is superior to the latter in both conceptual and empirical considerations. Especially, I demonstrate, along the lines of Sato’s (2008) approach, that it fits neatly into Chomsky’s (2008) probe-goal system and that it properly captures the core data of expletive constructions that have been extensively discussed in the minimalist literature. Furthermore, I argue that the apparent empirical problems that will be raised against the Move-approach with respect to the intervention effects caused by experiencer phrases in the seem-type construction are naturally solved by attributing the unvalued status to the -features of expletives, as characterized in (4b). The paper is organized as follows: Section 2 outlines Sato’s (2008) analysis of the there-construction under Chomsky’s (2008) probe-goal system and provides further arguments for this analysis. Moreover, I try to extend Sato’s analysis to the expletive it-construction, pointing out that the Move-approach is better off than the External Merge-approach in this extended domain as well. Section 3 raises empirical problems against the Move-approach with respect to different agreement patterns shown between the cases involving the expletive there and the corresponding raising cases in the 4 seem-type construction with an experiencer phrase. In Section 4, I provide solutions to these problems, making crucial use of the unvalued status of the -features of expletives, which makes it possible for them to be involved in default agreement. Further, I provide a consequence of this revised Move-approach concerning the there-construction embedded in the ECM, which contrasts with that involving a raising predicate in the availability of default agreement. 2. Arguments for the Move-Approach to the There-Construction 2.1 An Outline of Sato’s (2008) Analysis As a starting point, I first outline Sato’s (2008) analysis of the there-construction, which instantiates the Move-approach to this construction under Chomsky’s (2008) probe-goal system. One of the conceptual problems Sato raises against the External Merge-analysis of the there-construction is concerned with the singularity of the way the expletive there is involved in the probe-goal system. Typically, External Merge is related to a D-structure property, i.e., building up structures for selectional relations, whereas Move is involved in deriving surface structures from these basic structures through feature-checking under the probe-goal system. Given this situation, the role the expletive there plays in this system is peculiar in that it is involved in feature checking by way of External Merge. Thus, consider how (1a) is derived under the probe-goal system assumed by Chomsky (2000). The relevant stage of derivation of this sentence 5 is given below:1 (6) [T [vP be [SC someone in the garden]]] At this stage, T probes with respect to its -set and finds someone as its goal, checking these features as well as the nominative Case feature of the latter. Then, the expletive there is externally merged with the structure in (6) to satisfy the EPP feature of T. Under this system, expletives are only elements that satisfy EPP features by way of External Merge.2 This anomaly of expletives induces an undesirable complication to Chomsky’s (2000) probe-goal system. Let us consider the derivation of the following example: (7) There is likely to be someone in the garden. Since Chomsky assumes that all types of T, whether they are finite or infinitival, carry EPP-features, the embedded infinitival T in (7) needs an element to satisfy its EPP feature, and in this case, the expletive there serves to this end. Hence, it is inserted into the Spec of the embedded TP, as shown in (8a), and then is raised into the Spec of the matrix TP to satisfy an EPP feature again, which gives rise to (8b): (8) a. [TP there to [vP be [SC someone in the garden]]] b. [TP there T [vP be [AP likely [TP t to [vP be [SC someone in the garden]]]]]] This shows that the expletive there can undergo Move under Chomsky’s system of 1 In this structure, SC stands for small clause. In this paper, I will not be committed to the exact structure of the small clause given in (6). 2 See Groat (1999) for a similar reasoning with respect to the singularity of the way there satisfies the EPP, which leads him to adopt the Move-approach. 6 feature checking, and this is exactly the cause of an undesirable complication. Chomsky (2000) assumes that in the probe-goal system, a goal must be active to serve as such and that in order to be active, it must carry an uninterpretive feature; this is called Activation Condition. Thus, to guarantee that the expletive there can undergo Move, which requires a preceding application of the operation Agree in terms of a probe-goal relation, it needs to carry an uninterpretable feature. At the same time, however, it must be defective, in contrast with other full-fledged DPs, in the sense that it cannot enter an ordinary Agree relation with T. Notice that in (7), the matrix T agrees not with there but rather with its associate someone, as is clear from the comparison of (7) with the following example: (9) There are likely to be some men in the garden. To capture this fact, Chomsky proposes that the expletive there does not carry a full -set but only carries an uninterpretable person feature (henceforth abbreviated as [uPerson]). Thus, at the stage just before there undergoes Move to the matrix TP to yield (8b), T probes with respect to its -set and hits there first as a goal, since the latter is active and carries part of a full -set. This nonetheless does not lead to checking of the -set of T due to the defective -features of there, which then allows T to probe again. It then hits someone and its -set is properly checked against that borne by someone, as well as the Case-feature of the latter as a reflex. This explains why T agrees with the associate of there rather than there itself. Finally, as there has served as a goal for T even though it does no establish an Agree relation, it undergoes Move to the matrix Spec-TP to satisfy the EPP-feature of T. 7 Though the characterization of there as carrying [uPerson] may work well in deriving such a there-construction as (7) and (9) as well as accounting for the agreement pattern observed in such a construction, it nonetheless causes an apparently unnecessary complication in deriving a simple there-construction such as (1a), as pointed out by Nomura (2004) and Richards and Biberauer (2005). Let us consider again (6), repeated below, which represents the stage of derivation of (1a) just before there is inserted. (10) [T [vP be [SC someone in the garden]]] As mentioned above, from this stage of derivation, there is inserted into the Spec-TP to satisfy the EPP-feature of T. Note that there carries [uPerson] that needs to be checked, which then forces there to act as a probe. Chomsky (2000) claims that this uninterpretive feature is checked against the -set of T, but as pointed out correctly by Richards and Biberauer (2005), the latter ought to be inactive, as its -set is already checked, hence not being able to be involved in checking the [uPerson] of there. 3 On top of this, this checking is apparently unnecessary in deriving such a simple there-construction as (1a) in a proper way, which then strongly suggests that the [uPerson] that is attributed to there is just a technical artifact that is posited only to 3 A reviewer kindly reminds me of another significant problem to Chomsky’s (2000) system, pointed out by Bokovi (2007): under this system, nothing guarantees that probing by there with respect to its [uPerson] feature cannot take place when it is inserted into the Spec-TP at the stage given in (8a). If there established an Agree relation with someone at this stage, it could not move to the matrix Spec-TP due to the Activation Condition. 8 capture the fact that there undergoes Move in the derivation of such a there-construction as (7) and (9). This undesirable complication seems to occur due to the anomaly of there as being involved in checking an EPP-feature by way of External Merge. Sato (2008) also raises an empirical problem with the External Merge-approach to the there-construction, which is concerned with one of the data regarding such a construction that have attracted much attention under the Minimalist Program: (11) a. There is likely to be someone in the garden. (=(7)) b. *There is likely someone to be in the garden. Chomsky (2000) basically follows Chomsky (1995) in accounting for the ungrammaticality of (11b) by the Merge-over-Move Principle, which dictates that Merge must be chosen over Move at a given stage of derivation if both operations are available. This principle is conceptually motivated by the economy consideration according to which Move is a costly operation, contrary to External Merge, in that it applies to a syntactic object that is already integrated into a structure. To see how this principle works, let us consider the stage of derivation of (11a, b) where the embedded infinitival T is merged with the following vP, as shown below: (12) [to [vP be someone in the garden]] At this stage, the EPP-feature of the infinitival T needs to be satisfied and two options are available for this purpose: either externally merge there or move someone into the Spec-TP. The Merge-over-Move principle dictates that the former option be chosen, as shown below: (13) [TP there to [vP be someone in the garden]] 9 (= (8a)) Then the derivation proceeds in the way indicated in (8), giving rise to sentence (11a). Since the Merge-over-Move Principle forces there-insertion at the stage given in (12), there is no legitimate way to produce sentence (11b), which could be derived if someone was moved to the embedded Spec-TP at this stage. Sato (2008) points out correctly that such an analysis in terms of the Merge-over-Move Principle does not work any more under the current conception of the Minimalist Program, since Move is now characterized as a cost-free operation. Thus, Chomsky (2004) claims that Move (or Internal Merge according to his terminology) is “an operation that is freely available,” and “accordingly, displacement is not an ‘imperfection’ of language; its absence would be an imperfection.” (p. 8) Given this conception of Move, the Merge-over-Move Principle will lose its basis in terms of economy considerations, and hence we need a different account for the ungrammaticality of such a case as (11b). 4 In order to solve the problems pointed out above, Sato (2008) adopts a Move-approach to the there-construction, such as the one proposed by Sabel (2000) and Hornstein and Witkos (2003), in which the expletive there forms a constituent with its associate underlyingly, as given below: 4 See Epstein and Seely (1999, 2006) and Bokovi (2002, 2007) for additional arguments against the Merge-over-Move Principle. 10 (14) P 3 there the associate [+][-Case] [+][-Case(Part)] Sato does not commit herself to the identification of the category that constitutes there and its associate (that is why she exploits P) nor to the internal structure of this category. There carries interpretable -features and an uninterpretable Case-feature, just like other ordinary DPs, and it shares the values of its -features with its associate, as indicate by [+]. Even though Sato does not address the question of what internal structure guarantees the fact that there and its associate share the values of -features, her analysis will be quite compatible with Sabel (2000) and Hornstein and Witkos (2003), who claim that their internal structure is something like the following: (15) DP 3 D 1 there[+][-Case] NP the associate[+][-Case(Part)] In this structure, the agreement relation between there and its associate is captured in the structural relation of the D head to its NP complement, as is the case with French articles such as le, la, les, for instance, as claimed by Hornstein and Witkos (2003). The associate carries not only -features but also an uninterpretable Case-feature. 11 Sato assumes, following Belletti (1988), that this Case-feature must be valued as partitive due to the selectional requirement imposed by there upon its associate. This captures so-called definiteness effects exhibited by the associate; that is, the fact that it must be an indefinite.5 Notice that there and its associate act as independent nominal elements in the sense that each carries its own - and Case-features. In this respect, Sato’s analysis will be more accurately couched with the internal structure given in (15) than with other alternatives under which there and its associate constitute a subject-predicate relation.6 Furthermore, it will be argued in the next subsection that the idea that there and its associate make a constituent underlyingly should be extended to the pair of the expletive it and its associate clause, and that the reason why the latter pair does not share -features unlike the former is attributed to the selectional relation of 5 See Chomsky (1995) and Bokovi (2007) for a claim to the effect that the expletive there carries a D-feature that requires NP as its associate at LF and hence the latter needs to be an indefinite, on the assumption that definiteness resides in D. A reviewer points out that since partitive Case has semantic import, it will not be natural to assume that it is uninterpretable/unvalued. To resolve this inadequacy, we may adopt Pesetsky and Torrego’s (2007) feature checking mechanism, according to which formal features are characterized in terms of both interpretability and valuation. Given this mechanism, it will not be unnatural to assume that the associate of there carries an interpretable and unvalued Case feature, whose value is supplied as partitive by a relevant verb. For expository purposes, I keep to outlining Sato’s (2008) original mechanism in the text. See Sections 3 and 4 for relevant discussions regarding interpretability and valuation. 6 See Hoekstra and Mulder (1990) and Moro (1997), among others, for the claim that there is a predicate and its associate serves as its subject, and Williams (1994) and Hazout (2004) for the opposite view. 12 D to its complement CP; see (35). To the extent that the availability of -feature sharing is naturally captured by the selectional relation of D to NP vs. CP, the internal structure given in (15) will be more appropriate under Sato’s theory than the subject-predicate structure, advocated by the scholars mentioned in fn. 6. Belletti claims that “[b]ecause partitive is an inherent Case assigned by verbs, it will typically be assigned to the direct object of both transitive and unaccusative verbs,” (p. 3) which is clearly indicated by such a language as Finnish that has an overt manifestation of partitive Case, as illustrated below: (16) a. Hän pani kirjoja pöydälle. he put bookpart, pl on the table ‘He put some books on the table.’ b. Helsingistä tulee kirjeitä. from Helsinki comes letterpart, pl ‘There come some letters from Helsinki.’ c. *Miehiä on tavannut Pekan manpart, pl aux3sg met kadulla. Pekkagen in the street ‘Some men met Peter in the street.’ The ungrammaticality of (16c) follows from the fact that partitive Case is not assigned by T. Thus, the assumption that the associate of there must have its uninterpretable Case feature valued as partitive serves to capture its limited distribution. For instance, it accounts for the contrast in grammaticality between the sentences in (1a), (3), (7) and (9) on the one hand and the following sentences on the other: 13 (17) a. *There seem that a lot of people are intelligent. b. *There seem to a lot of people that it is raining outside. In these sentences, the associate of there, i.e., a lot of people appears in a position where it is unable to get its uninterpretable Case feature valued as partitive, and hence the derivations do not converge or if they do, they violate the selectional requirement imposed by there upon its associate. Sato assumes, along the lines of Lasnik’s (1995) theory of partitive Case, that this Case is involved in the probe-goal system exactly in the same way as other structural Cases. This assumption leads us to consider an idiosyncratic property of the there-construction that appears to make partitive Case differ from other structural Cases at least in a certain variety of languages, including English: the type of verbs which are able to appear in this construction is limited to unaccusative verbs, including be, as noted by Hoekstra and Mulder (1990):7 (18) a. There just may not exist a solution (to this problem). b. There occurred a catastrophe (in that country). (19) a. *There walked a man with a dog. b. *There ate a man a pudding. The grammaticality contrast between the sentences in (18) and those in (19) will be 7 The types of the expletive construction illustrated in (19) are in fact grammatical in such languages as Dutch and Icelandic with the word orders indicated in the ungrammatical English examples. The ungrammaticality of these examples has nothing to do with their word orders but simply indicates that intransitives and transitives are incompatible with the there-construction. 14 accounted for if we stipulate that those verbs that are able to value the uninterpretable Case feature of the associate of there as partitive are confined to unaccusative verbs. Given Belletti’s (1988) claim that partitive Case is in principle assigned by any type of verbs, however, it is desirable to dispense with this stipulation. I will demonstrate directly how this can be done under Sato’s approach to the there-construction. Given that partitive Case is valued by verbs, a question arises as to whether the vP that is involved in checking this Case constitutes a strong phase, since the usual assumption with this latter notion is that when v is involved not only in Case checking but also in licensing an external argument, its maximal projection constitutes a strong phase. Sato assumes that the answer is positive; that is, Case checking suffices for a given vP to constitute a strong phase. This is crucial for the probe-goal system she assumes in which the search domain of a probe is regulated by the Phase-Impenetrability Condition (henceforth, PIC), as stated below:8 (20) Phase-Impenetrability Condition In Phase with head H, the domain of H is not accessible to operations outside , only H and its edge are accessible to such operations. (Chomsky 2000, p. 108) Thus, according to this condition, once a partitive Case is checked in vP, the domain of v, namely VP, is not accessible to any operation at the next phase level. 8 Sato (2008) adopts the version of the PIC proposed by Chomsky (2000), but it would not matter, as far as I can see, if she adopted the different version of the PIC assumed by Chomsky (2001). This is because under the probe-goal system invented by Chomsky (2008), which Sato adopts, only a phrase head is eligible for a probe, which neutralizes a crucial difference between the two different versions of the PIC. 15 Let us now see how there-constructions are derived under the probe-goal system of Chomsky (2008) with the assumptions made by Sato (2008). To take (1a) for illustration, the first stage where a phase head probes is the following: (21) [v’ v* [V’ be [DP there someone] in the garden]]] At this stage, be inherits -features from the phase head v* and probes for them. There are two options that can serve as goals for this probing: there and someone. Sato claims that in this situation, the associate someone should be chosen as a goal, since be is able to provide a partitive Case as a reflex of the checking of -features.9 After this Agree relation is established, someone is raised into the Spec of V. At the same time, the expletive there is raised into the Spec of v* thanks to the edge feature of the latter; otherwise, its Case feature would remain unvalued and hence the derivation would crash.10 This completes the derivation of the lower v*P, as given below:11 9 Or the expletive there will not be an appropriate receiver of partitive Case, given that it occupies the D head, as assumed in the structure given in (15), and that partitive Case is provided only to NP. 10 A reviewer raises the question whether there prevents be from targeting the associate as its goal because there is closer to be than its associate. There are at least two possibilities to address this question. One is, as the reviewer suggests, to rely on the notion of equidistance, so that there may not block be from targeting the associate because both there and its associate are equidistant from the target. The other possibility, which is more interesting under the probe-goal system assumed here, is to exploit the notion of parallel operations in such a way that in one span of such operations, their targets do no interfere in each other’s operations with respect to minimality. Thus, in the v*-V level of operations given in the text, (22) is derived from (21) at once, all operations involved taking place in parallel. 16 Hence, there and its (22) [v*P therei [v*’ v*+be [VP someonej [V’ tbe [DP ti tj] in the garden]]]] The next phase level that can undergo probing is the matrix C, as shown below: (23) [C’ C [T’ Pres [v*P therei [v*’ v*+be [VP someonej [V’ tbe [DP ti tj] in the garden]]]]]] At this stage, T inherits -features from the phase head C and probes for them. In this case, there is the only candidate for a goal, since any element inside VP is not accessible for this probing, according to the PIC. Then the checking of -features between T and there leads to providing nominative Case feature to the latter as a reflex, which establishes an Agree relation and brings there into the Spec of T. This completes the derivation of sentence (1a), as shown below: (24) [CP [TP therei [T’ Pres [v*P t’i [v*’ v*+be [VP someonej [V’ tbe [DP ti tj] in the garden]]]]]]] Notice that since the expletive there shares the values of its -features with its associate, the fact that the -features of there is checked with those of T results in the required agreement pattern in which T agrees with the associate of there. This way of deriving a there-construction is immune from the conceptual problems pointed out above. For one thing, there is no singularity that applies only to associate, which are both targets of separate operations, do not interfere in each other’s operations. 11 In this derivation, be is raised into the above v*, as is the case with other main verbs. Given that it behaves like an auxiliary with respect to SAI and the relative order with not, etc., it should raise further into T when no other auxiliary occupies that position. For an expository purpose, however, we will dispense with this further movement to T in the representations given in the text, since nothing hinges on whether it really takes place or not. 17 the expletive there in the way the - and Case-features are checked in the present probe-goal system: after an Agree relation is established between a probe and a goal in terms of the - and Case-features, the goal itself or a phrase including it is pied-piped to the Spec of the probe, irrespective of whether the targeted phrase is an ordinary DP or the expletive there. Furthermore, there will arise no complication that has arisen under the External Merge-approach when we consider a more complex case of the there-construction. Let us consider the derivation of (7) under the present system. The first stage at which a phase head has done probing is the same as that for (1a), namely the one given in (22), repeated below: (25) [v*P therei [v*’ v*+be [VP someonej [V’ tbe [DP ti tj] in the garden]]]] The next phase level at which probing takes place is the matrix CP; the intermediate TP is not a phase by definition nor the matrix vP due to its lack of Case checking. Thus we reach the following stage: (26) [C’ C [T’ Pres [vP [v’ v+be [AP likely [T’ to [v*P therei [v*’ v*+be [VP someonej [V’ tbe [DP ti tj] in the garden]]]]]]]]]] At this stage, T inherits -features from the phase head C and probes for them. In this case again, there is the only candidate for a goal, since any element inside VP is not accessible for this probing, according to the PIC, as indicated by striking out the relevant part. Then the checking of -features between T and there leads to providing nominative Case feature to the latter as a reflex, which establishes an Agree relation and brings there into the Spec of T. Since the intermediate infinitival T has an EPP-feature, the movement of there must be successive-cyclic, which leads to the final representation 18 of sentence (7), as shown below: (27) [CP [TP therei [T’ Pres [vP [v’ v+be [AP likely [TP t”i [T’ to [v*P t’i [v*’ v*+be [VP someonej [V’ tbe [DP ti tj] in the garden]]]]]]]]]]]] Notice that nothing special is required in deriving such a complex there-construction as (7). This situation sharply contrasts with that found under the External Merge-approach, in which positing a [uPerson] feature to capture the fact that there can undergo Move in such a case as (7) leads to an unnecessary complication in such a simple case as (1a), that is, the need for there to serve as a probe with respect to the [uPerson] feature. On top of that, the present system dispenses with re-probing by T, arguably a desired consequence, which is necessary under the External Merge-approach. Recall that in deriving such a there-construction as (7) under the External Merge-approach, the matrix T that probes for its -set first hits there as a goal, but that due to the defective -features of there, T needs to probe again and hits someone as an appropriate checker of the -set. This is apparently not an efficient way of checking relevant features, and hence it would be better to dispense with it. Thus the present system is advantageous in this respect as well. As for the empirical problem pointed out above regarding the pair of there-constructions given in (11), Sato (2008) claims that the ungrammatical case (11b) is straightforwardly excluded under the present system. Let us consider again the stage of derivation given in (26), which is shared by (11a) and (11b). In order to derive sentence (11b) from this stage, someone needs to raise to the intermediate Spec-TP, but as indicated by striking out the relevant part, the PIC prohibits it from 19 doing so. 12 Hence, there is no way of deriving (11b) under the present system. Therefore, the ungrammaticality of this sentence follows automatically without any further stipulation. Let us finally address the question of how the present system derives the fact, noted above, that only unaccusative verbs appear to be able to license the associate of the there-construction. Let us first consider the case of the there-construction that involves an intransitive verb. Suppose that the complex form consisting of there and its associate occupies the subject position of such a verb, as schematized below: (28) C [T’ T [v*P [DP there associate] v* [VP Vintr]]] Given this structure, there is no way for the associate to obtain partitive Case, under the assumption that it must be obtained by a verb. Hence, such a case is excluded for exactly the same reason as the cases in (17).13 Let us next consider the case that 12 As a reviewer points out, raising of someone into the intermediate Spec-TP would induce a minimality violation, since there is closer to that position than someone. Furthermore, even if someone were able to raise somehow into the intermediate Spec-TP, that would then block raising of there into the matrix Spec-TP due to a minimality violation. framework. See Groat (1999) for such an account under Chomsky’s (1995) Notice that neither account in terms of minimality will be maintained if we adopt the second possibility suggested in fn. 10, according to which the targets of parallel operations do not interfere in each other’s operations regarding minimality. In that case, we need to rely on the PIC to exclude the derivation of (11b). 13 We have not considered the possibility that in (28), the expletive is involved in checking of the -features of T and is raised into the Spec-TP, whereas the associate is raised into the Spec-CP thanks to the edge feature of C, hence accessible to the next v*-V phase level. I assume that the movement of the associate across the expletive is 20 involves a transitive verb. Suppose that the complex form consisting of there and its associate occupies the object position of such a verb, as schematized below: (29) [v*P DP v* [VP Vtr [DP there associate]]] Notice that we are assuming that any type of verbs can provide partitive Case in principle. Hence, in (29), V and the associate can enter into an Agree relation, checking both -and Case-features against each other, and then the associate raises into the Spec-VP. After there raises into the Spec-v*P thanks to the edge feature of v*, we reach the following stage of derivation: (30) [C’ C [T’ T [v*P therei [v*P DP [v*’ v*+V [VP associatej [V’ tV [DP ti tj]]]]]]]] At this stage, T necessarily picks up there as a goal, since the latter is the closest to T, and hence the external argument DP is left unchecked, leading the derivation to crash. This is why transitive verbs cannot be involved in the there-construction. As a result, only unaccusative verbs can lead instances of the there-construction to convergent derivations. 2.2 An Alternative of the External Merge-Approach There is one competing alternative of the External Merge-approach, proposed by Abe (1997) and Bokovi (2002), which is worth considering here. Abe and Bokovi claim, contrary to Chomsky (2000), that no EPP requirement is imposed upon raising infinitival clauses. Under this assumption, such an instance of the there-construction banned by a general condition to be specified, and hence leave such a possibility out of consideration. 21 as (7) is derived without applying Move to there; in other words, by simply inserting there into the matrix Spec-TP, as shown below: (31) [TP there is likely [TP to be someone in the garden]] Under this analysis, the ungrammaticality of (11b) immediately follows from the Last Resort Principle, since to move someone to the intermediate Spec-TP would have no triggering. On the other hand, Abe and Bokovi claim that in such a raising case as the following: (32) Someone is likely to be in the garden. someone undergoes successive-cyclic movement, passing through the intermediate Spec-TP. They follow Chomsky (1993) in that successive-cyclic movement is not a result of satisfying the EPP-features of intermediate TPs but rather follows from the Shortest Move (henceforth, SM, which is phrased as Minimize Chain Links in Chomsky and Lasnik (1993)), which dictates that each application of Move cannot pass a possible landing site. Though this alternative of the External Merge-approach has many interesting empirical consequences other than providing a very neat explanation to the pair given in (11) (see Section 3), it gives rise to a couple of both conceptual and empirical problems, when couched under Chomsky’s (2008) probe-goal system. Let us begin with conceptual problems. Abe (1997) and Bokovi (2002) assume that the expletive there is inserted by External Merge into a Case position to satisfy a Case requirement of a head, i.e., the Inverse Case Filter. Given the assumption that intermediate infinitival Ts do not carry any feature, including an EPP-feature, it follows that there does not 22 move under the system invented by Abe and Bokovi; in this sense, their approach is viewed as a more radical External Merge-approach than that advocated by Chomsky (2000). For one thing, this system suffers the singularity of the expletive there in that it is the only element that is involved in feature checking by way of External Merge. Since Abe and Bokovi assume that there carries a Case-feature, it will have to function as a probe after it is externally merged into a given structure under the probe-goal system. Thus, to take (1a), repeated below as (33) with a relevant structure, for illustration, (33) [TP there T [vP v+is someone in the garden]] after there is externally merged into the Spec-TP, it should probe for its Case-feature and finds T as its goal, checking this feature.14 A crucial question to raise here is how the -features of T are checked. If we follow the way an Agree relation is established under Chomsky’s (2008) probe-goal system, it will be most natural to claim that the -features of T are valued as a reflex of Case-feature checking. However, this alone does not capture the fact that in the there-construction, T agrees with the associate of there. In order to capture this, we need some extra operation. We might say that before there is inserted into the Spec-TP, T has already established an Agree relation with its associate with respect to their -features. 14 Since the Case-feature of the One may notice that this way of feature checking is anomalous, given the standard assumption that Agree precedes External Merge or Move. But in Section 3, I adopt the claim, made by Goto (2008), that the EPP can be satisfied by Move before Agree takes place in English. Hence, this anomaly carries over to our system. 23 associate is taken care of by be, T will still have the ability of checking the Case-feature of there. Though this works technically, it is not as efficient as the Move-approach adopted here in that T establishes more than one Agree-relation, one involving -features and the other involving Case-feature, which happens only when T has the expletive there in its Spec position. This inefficiency has the same flavor as re-probing necessary for Chomsky’s (2000) analysis of the there-construction.15 Let us now move on to empirical problems. It is well-known that there is another type of expletives that do not induce agreement between T and their associates, but rather are directly involved in an agreement with T. The expletive it in English is such a case, as illustrated in the following paradigm: (34) a. That he’ll resign and that he’ll stay in office seem at this point equally possible. b. *It seem at this point equally possible that he’ll resign and that he’ll stay in office. c. It seems at this point equally possible that he’ll resign and that he’ll stay in office. (McCloskey 1991, p. 564-565) (34a) shows that when that-clauses are conjoined in subject position, they exhibit a plural agreement with T, as witnessed by the morphological form of seem. 15 Under the Another way of capturing the agreement pattern in question is to relate there and its associate by way of covert movement; either by adjoining the associate to there, as in Lasnik (1992, 1995) and Abe (1997) or the other way around, as in Bokovi (1997). Since such an option does not seem to fit well into Chomsky’s (2008) probe-goal system, I will leave it aside. 24 probe-goal system adopted here, this agreement fact will follow if we assume that the conjoined that-clauses are base-generated after the adjective possible and that they have [+plural] value for their number feature. Thus, when the matrix T probes for its -features, it finds the conjoined that-clauses as its goal, hence its -features valued by those of the latter.16 The contrast between (34b) and (34c), on the other hand, shows that the presence of the expletive it in subject position forces a singular agreement with T. How can the radical External Merge-approach account for this agreement pattern? Nomura (2004) points out correctly that under the system in which the expletive it is directly merged into Spec-TP, it is predicted that (34b) rather than (34c) should be grammatical, since T searches for an element within its c-command domain, hence choosing the conjoined that-clauses as its goal in exactly the same way as in (34a). Thus, given that the Spec/head relation is no longer relevant for feature checking, any version of External Merge-approach to expletive constructions will be confronted with the problem of how to capture the fact that such an expletive as the English it exhibits an agreement with T.17 16 Given the Activation Condition, the conjoined that-clauses must carry an uninterpretable feature so as to undergo Move after an Agree relation is established between them and T. this checking. I leave open exactly what uninterpretable feature is involved in See Bokovi (1995) for the claim that a clause optionally carries a Case-feature and that when it undergo A-movement, it is required to carry this feature. 17 Bokovi (1997) argues that (34b) does not necessarily show the agreement between the expletive it and T, as McCloskey argues, but rather that this agreement pattern can be taken as a result of the first conjunct agreement with the associate of it. The latter agreement pattern is independently observed with the expletive there, as he provides a 25 The Move-approach adopted here, on the other hand, is free from such a problem if we make the natural assumption that the expletive it and its associated clause make a constituent underlyingly, just like the there-construction. The expletive it crucially differs, however, from the expletive there in that it does not share the values of its -features with its associate. We can capture this fact by positing the underlying structure of it and its associate as something like the following:18 relevant example such as the following: (i) There is/*are a man and five women in the garden. Though this analysis correctly captures the agreement pattern exhibited by (34b), it raises a problem with that exhibited by (34a). Under the present probe-goal system, it will be predicted that it should show a singular agreement for exactly the same reason as (34b) does. Thus, a question still remains as to how to capture the whole agreement pattern given in (34) under the External Merge-approach. I thank a reviewer for bringing Bokovi’s (1997) analysis to my attention. 18 See Bokovi’s (1997) for the claim that “the associate of expletive it must be a clause specified as [+tense],” (p. 90) which captures the fact that it can host infinitival control clauses whereas it cannot host infinitives containing NP-traces, as illustrated below: (i) a. It was arranged [PRO to leave]. b. *Iti was believed [ti to be someone in the garden]. We might capture this fact on the assumption that only [+tense] infinitival clauses are CPs and that the expletive it selects CP as its complement. 26 (35) DP 3 D 1 it[+][-Case] CP the associate[+] It is not unnatural to assume that unlike the D-N association, which induces -feature sharing, the D-C association does not induce such an effect. As a result, the expletive it has its own values of -features, namely third person singular. In deriving such a sentence as (34c), the matrix T probes for -features and finds it as its goal. Notice that the associate that-clause is invisible to such probing due to lack of uninterpretable features. Then, the -features of T are checked against those of it and the latter obtains nominative Case as a reflex. This correctly captures the fact that the expletive it agrees with T. Notice furthermore that this construction differs from the there-construction in whether the associate carries a Case such as partitive. Recall that we have discussed toward the end of the preceding subsection how the present system can derive the fact that only unaccusative verbs can be involved in the there-construction. In this account, partitive Case plays a crucial role: since this Case must be obtained by a verb, the associate cannot function as the external argument of an intransitive verb (and a transitive verb for that matter), and when the associate appears in the internal argument of a transitive verb, partitive Case must be checked by this verb, which thus makes it necessary for the expletive there to be checked by T and hence leaves the external 27 argument unchecked. It is then predicted under the present system that there will be no such restriction with respect to the type of verbs in the it-construction, and this is apparently the case. Furthermore, it is predicted that the expletive it can not only be checked by a finite T with nominative Case but also by a transitive verb with accusative Case. That this is in fact the case is demonstrated in detail by Postal and Pullum (1988); some examples that show that the expletive it can appear in object position are reproduced below: (36) a. I dislike it that he is so cruel. b. I didn’t suspect it for a moment that you would fail. c. I regret it very much that would could not hire Mosconi. These facts thus lend support to the position according to which the there-construction and the it-construction are treated on a par, that is, the expletives and their associates form a constituent underlyingly, and the different distributions exhibited by the two constructions are ascribed to the presence of partitive Case in the associate of there. A second problem that will arise with the radical External Merge-approach has to do with the following pair of sentences discussed by Lasnik (1995): (37) a. There is likely to be a building demolished. b. There is a building likely to be demolished. In (37a), a building functions as the internal argument of demolished, and hence it must have moved from the complement position of the verb to whatever position precedes it. Lasnik claims that the movement is to Spec-Agro to check the partitive Case of a building. Given this, it will be expected that (37b) is derived by moving a building 28 from the complement position of demolished all the way to the surface position in a successive-cyclic fashion, passing though the intermediate Spec-TP. However, this expectation is not fulfilled, as witnessed by the fact that extraction of such an adjunct as how is possible with (37a), but not with (37b): (38) a. Howi is there likely to be [a building demolished ti]? b.?*Howi is there a building likely to be [demolished ti]? Lasnik asserts that if such a raising derivation as described just above were available to (37b), then we would not expect that (38b) were ungrammatical, and hence that this sentence should be analyzed differently. He suggests that it involves a reduced relative clause that modifies the associate of there; that is, likely to be demolished is a reduced relative modifying a building. (38b) is then explained as a violation of the complex NP island condition. Lasnik prevents the raising derivation in question for (37b) by assuming, following Belletti (1988), that partitive Case is inherent. In (37a), the partitive Case of a building is properly licensed, since its -role assigner, i.e., demolished, is covertly raised into Agro to license it. In (37b), by contrast, the partitive Case of a building is not properly licensed, since its licenser demolished is located too far down in the structure to raise into the matrix Agro. Although the characterization of partitive Case as inherent has an independent motivation, as convincingly demonstrated by Bellletti (1988) with Italian cases, it is not at all obvious whether such a characterization has any good motivation in English. Under the standard small clause analysis of there-constructions involving be, a sentence such as (39a) below has roughly the 29 underlying structure given in (39b): (39) a. There will be a man available. b. e will be [SC a man available] In this case, the licenser of partitive Case for a man is be, which does not have any -relation with this NP, whereas its -licenser is available. contradicts with the claim that partitive Case is inherent. Thus, this apparently To solve this problem, Lasnik (1995) claims that be here acts as a light verb which has no -role of its own and hence that this allows the predicate of the complement clause of be to raise covertly into the latter verb. Given the present framework of the probe-goal system invented by Chomsky (2008), however, such a covert head movement will not be warranted. Furthermore, it is not at all clear under this proposal how to deal with such a there-construction in which the predicate of the associate of there constitutes PP, as illustrated in (1a), whose underlying structure will be the following: (40) e will be [SC someone in the garden] In this case, it is rather obvious that someone and in the garden enters a subject-predicate relation, but it is not at all obvious whether this relation is something that has to be captured as a -relation. This leads further into the question of what raises covertly into be in this case. Thus, I take it that these problems give us a good motivation to look for a new explanation for why the raising derivation is impossible for (37b). Notice that as far as I can see, the radical External Merge-approach will not have anything to contribute to answering this question. Under the Move-approach, on the other hand, we can offer a fairly natural 30 explanation of how to prevent the raising derivation for (37b). There are actually two such derivations we need to consider, depending upon which occurrence of be is involved in checking partitive Case. Let us first consider the one in which the lower be enters into an Agree relation with a building with respect to - and partitive Case-features. This derivation is excluded for basically the same reason as (11b), repeated below: (41) *There is likely someone to be in the garden. Recall that this sentence is ruled out because after someone enters into an Agree relation with the lower be and then is moved to the embedded Spec-VP, the PIC makes it invisible from the next higher phase level, so that it ought not to undergo further movement. Accordingly, if a building in (37b) enters an Agree relation with the lower be, then it should not undergo movement across the v*-V phase level in which the Agee relation is established, due to the PIC. The more interesting derivation to consider is the one in which the upper be is involved in checking partitive Case. Thus, the crucial stage of this derivation is the following: (42) [v*’ v* [VP be [AP likely [TP [T’ to [vP [v’ v+be [VP demolished [DP there a building]]]]]]]]] At this stage, the upper verb be, which inherits -features from v*, probes for these features and then finds a building as its goal (here there cannot be a goal, either because otherwise a building could not obtain partitive Case or because be will not be able to value the Case-feature of the expletive on the assumption that partitive Case is assigned only to NP; see fn. 9). After an Agree relation is established between be and a 31 building, the latter phrase must undergo Move to the matrix Spec-VP. On the other hand, there also needs to undergo Move to the matrix Spec-v*P to be visible from the next higher phase level at which it needs to obtain nominative Case. If these applications of Move were successful, then the derivation would converge, contrary to our expectation. There is a natural way, however, to exclude such applications of Move. The key point is that each application of Move must take place in a successive-cyclic fashion passing through the intermediate Spec-TP. Suppose, following Chomsky (1993), that successive-cyclic A-movement is forced by the SM, as assumed by the radical External Merge-approach. The application of Move to a building into the matrix VP is obviously an instance of A-movement, and hence is required to pass through the intermediate Spec-TP in accordance with the SM. The other application of Move, i.e., moving there into the matrix v*P is a little bit tricky, but since the final destination of there is the matrix Spec-TP, it is natural to regard such movement as an instance of A-movement as well, and hence it also needs to pass through the intermediate Spec-TP. Then, given that these two applications of Move take place in parallel under the probe-goal system of Chomsky (2008), we can attribute the illegitimacy of such applications to a restriction to the effect that the intermediate Spec-TP cannot be used by more than one element at a time.19 In this way, we can 19 Notice that this does not exclude the possibility of having multiple Specs in the projection of a phase head such as v* and C, which could be obtained when the phase head probes more than once. Hence, if the claim in the text is correct, it will be predicted that multiple Specs are allowed only for phase heads. right prediction or not needs further research, however. 32 Whether this is the I thank a reviewer for pointing correctly exclude the raising derivations as illegitimate for (37b) under the Move-approach, which then counts as a good advantage over the radical External-Merge approach.20 3. Empirical Problems I have argued that the Move-approach to the there-construction proposed by Sato (2008) is both conceptually and empirically superior to the External Merge-approach under the probe-goal system of Chomsky (2008). However, this approach is also not free from any problem. In this section, I point out a couple of empirical problems with this approach and provide a solution to them in the next section. Bokovi (2002) provides a fair amount of evidence for his claim that expletives do not move. What he regards as “probably the strongest piece of evidence” (p. 196) is provided by the experiencer blocking effect in French. Whereas English allows a raising construction with an experiencer phrase, as shown below: (43) John seems to Mary to be smart. French does not allow such a construction, as illustrated below (the examples in (44) and (45) are taken from Bokovi (2002)): out the importance of this consequence. 20 It is beyond the scope of this paper to discuss other variants of the Move-approach to expletive constructions and to compare them with the one adopted here. See Bowers (2002), Nomura (2003) and Richards and Biberauer (2005) for the claim that expletives are base-generated in Spec-vP or Spec-VP and are moved to the above Spec-TP. 33 (44) a. *Deux soldats semblent au two soldiers seem général manquer (tre manquants) la to-the general to-miss to-be missing at the caserne. barracks ‘Two soldiers seem to the general to be missing from the barracks.’ b. * Deux soldats semblent au two soldiers seem général tre arrivés en ville. to-the general to-be arrived in town ‘Two soldiers seem to the general to have arrived in town.’ However, the counterparts of these sentences that involve expletives in the matrix subject position are acceptable, as illustrated below: (45) a. Il semble au général y avoir deux soldats manquants la caserne. there seems to-the general to-have two soldiers missing at the barracks ‘There seem to the general to be two soldiers missing from the barracks.’ b. Il semble au général tre arrivés deux soldats en ville. there seems to-the general to-be arrived two soldiers in town ‘There seem to the general to have arrived two soldiers in town.’ Bokovi takes the grammaticality of these sentences involving the expletive il, contrary to the corresponding raising sentences given in (44), as indicating that the expletive does not undergo Move but rather is directly inserted into the matrix TP. This evades a violation of relativized minimality that is induced by the intervening experiencer phrase in such raising cases as in (44). On the other hand, the paradigm presented in (44) and (45) is a challenging problem to the version of the Move-approach 34 to the there-construction proposed by Sato (2008), according to which the cases involving the expletive il, as presented in (45), ought to be ungrammatical due to a violation of relativized minimality induced by the movement of the expletive across the experiencer phrase. A second problem with this version of the Move-approach is concerned with a peculiar paradigm of agreement that emerges in the English sentences corresponding to those given in (44) and (45). As noted above with example (43), English allows a raising construction with an experiencer phrase and this construction exhibits the normal pattern of agreement between the matrix subject and verb, as shown below: (46) a. b. A man seems to Mary to be in the room. Some men seem to Mary to be in the room. In contrast with this pattern, when the expletive there is involved in this construction, an unexpected pattern manifests itself, as pointed out by Boeckx (2000); compare the sentences in (47) with those in (48): (47) a. There seems to Mary to be a man in the room. b. There seems/?*seem to Mary to be men in the room. (Boeckx 2000) (48) a. There seems to be a man in the room. b. There seem to be men in the room. When an experiencer phrase is omitted, as in (48), the main verb agrees in number with the associate of there, as expected. On the other hand, when an experiencer phrase exists, as in (47), such a normal agreement pattern collapses and the main verb exhibits its singular form, irrespective of whether the associate of there is singular or plural. 35 Under the Move-approach to the there-construction, the contrast in agreement pattern between the raising cases, illustrated in (46), and those involving the expletive there, illustrated in (47), is unexpected, since both cases involve raising of a nomimal element out of the embedded infinitival clause across the experiencer phrase. This calls for an explanation. Before moving on to providing a solution to these problems in the next section, I present here Goto’s (2008) work to address the question of why English and French are different regarding the availability of the raising construction with an experiencer phrase, as shown in (43) and (44). First of all, Goto follows Chomsky (1995), Bokovi (2002) and others in that such French cases as in (44) are a violation of relativized minimality; more precisely a violation of the Minimal Link Condition (henceforth, MLC) in Chomsky’s terms. Thus, in both cases of (44), the matrix T cannot probe into an element inside the embedded infinitival clause, since the experiencer phrase is the closest to the probe, and hence the embedded subject cannot be moved up to the matrix Spec-TP. Goto then considers how such an English counterpart as (43) can evade a violation of the MLC. I summarize Goto’s proposals as follows: (49) a. Following Chomsky (2000), the EPP is a selectional requirement rather than a feature to be checked. b. The EPP can be satisfied by Move before Agree takes place in English. c. Following Abe (2002), Probe is subject to the MLC while Move is subject to the SM. 36 Abe (2002) proposes that so-called movement involves two steps, i.e., Probe (or Select in his terms) and Move, and that each operation is subject to a minimality condition, as indicated in (49c). Thus, to take (50) below for illustration, (50) John seems to be smart. let us consider the stage of its derivation just before John undergoes movement: (51) [C’ C [T’ T [vP v [VP seem [T’ to be John smart]]]]] At this stage, the matrix T probes for its -features and hits John without violating the MLC. After an Agree relation is established, John undergoes Move, which is subject to the SM. Since the latter condition demands that Move not skip a possible landing site, John undergoes successive-cyclic movement, passing through the intermediate Spec-TP, thereby the derivation reaching the following stage:21 (52) [C’ C [TP Johni T [vP v [VP seem [TP ti’ to be ti smart]]]]] Goto’s core proposal is given in (49b), which is flatly opposed to Chomsky’s (2000) assumption that Move must take place after an Agree relation is established. He then assumes that such an application of Move is subject to the SM, along the lines of Abe (2002)’s minimality conditions. Let us now consider how this mechanism properly generates such a raising case as (43); the first stage where relevant operations take place in this case is the matrix C-T 21 This account of how Abe’s (2002) system of movement works misses the full definition of “possible landing site”, which needs to be worked out, partly because his system is not fully compatible with the present probe-goal system, especially regarding whether the notion of phase plays a crucial role. For the present purpose, I simply assume that Spec-TP counts as a possible landing site for so-called A-movement. 37 level, as shown below: (53) [C’ C [T’ T [vP v [VP seem to Mary [T’ to be John smart]]]]] At this stage, John can move up to the matrix Spec-TP before Agree takes place and this movement can take place successive-cyclically by making John pass through the embedded Spec-TP, hence satisfying the SM; crucially, crossing the experiencer phrase does not cause a SM violation, since it is not a possible landing site for the movement in question. The derivation then proceeds to the following stage: (54) [C’ C [TP Johni T [vP v [VP seem to Mary [TP ti’ to be ti smart]]]]] Goto claims that at this stage, John probes for its uninterpretable Case-feature and finds T as a value assigner of this feature, and further that it values the -features of T in return. This completes the derivation of (43).22 In order to accommodate the ungrammaticality of such a French counterpart as in (44) under Goto’s mechanism, it will be necessary to assume that the option given in (49b) is unavailable for such a case. Then the ungrammaticality of the sentences given in (44) immediately follows exactly in the way explained above, since the unavailability of this option makes Agree a prerequisite for Move. Though Goto tries to derive the availability of the option (49b) from a language parametrization, it will be more 22 See Bokovi’s (2007) for a similar mechanism of the way feature checking is implemented, though it remains to be seen whether his whole system is in fact compatible with the present probe-goal system; especially, his theory aims to address such issues as the look-ahead problem of computation, the elimination of the EPP, the Activation Condition and the Inverse Case Filter, which I am unable to discuss in this paper. 38 desirable, as a reviewer points out, to attribute it to an independently motivated lexical parametrization. Here I suggest one possible way of doing so in terms of Pesetsky and Torrego’s (2007) feature characterization. They assume that formal features are characterized with both interpretability and valuation. While in most cases, these two notions coincide in such a way that being interpretable entails being valued and being uninterpretable entails being unvalued, I propose that this correlation breaks up in the characterization of Case-feature in the following way:23 (55) a. English D-N Case-feature: uninterpretable, unvalued b. French D-N Case feature: uninterpretable, valued I speculate that the difference between English and French nominals with respect to Case valuation may be attributed to the morphological property of “rich declension”; i.e., those nominals which show rich declension have valued Case-features whereas those which are poor in declension have unvalued Case-features. Given this characterization, the availability of (49b) follows from Pesetsky and Torrego’s (2007) assumption given below, which I assume in what follows: (56) Only unvalued features can act as probes. The derivation of such an English raising case as (43) given above in (53) and (54) still holds with the feature characterization given in (55a). Crucially, John is able to act as a probe for its Case-feature after it moves up to the matrix Spec-TP, thanks to its 23 See Bokovi (2008) for the claim that gender-feature is characterized as uninterpretable and yet valued, though the conceptual motivation for such a characterization is different from the one given in the text. 39 unvalued status. In French cases such as given in (44), on the other hand, the option (49b) is unavailable since if we followed this option, the DP moved up to the matrix subject position would have to be able to act as a probe, but it would not, due to lack of any unvalued features. The remaining option is to make the matrix T probe into the infinitival clause to hit the subject DP as its goal, but this inevitably violates the MLC due to the presence of the experiencer phrase. Goto (2008) also provides an account for Boeckx’s (2000) agreement paradigm given in (47) under his mechanism, but he crucially assumes the External Merge-approach to the there-construction: Since there is externally merged into the matrix Spec-TP in such cases as in (47), the availability of the option given in (49b) is irrelevant here. Thus, the matrix T cannot probe into the embedded infinitival clause due to the intervention effect caused by the experiencer phrase, hence manifesting a default value with respect to its -features. The same explanation would hold for the grammaticality of the French counterparts of expletive constructions exemplified in (45), but crucially under the External Merge-approach. In the next section, I provide a solution to Boeckx’s (2000) agreement paradigm given in (47) as well as the grammaticality of the French expletive constructions given in (45) under the Move-approach advocated by Sato (2008). 4. A Revision of Sato’s (2008) Mechanism In order to solve the problems raised above against the Move-approach, I examine how expletives are characterized in terms of formal features, revising Sato’s (2008) original 40 assumptions. She treats the expletive there on a par with other DPs in carrying full-fledged Case- and -features. I oppose this characterization, rather following Chomsky (2000) in that expletives should be characterized differently. Let us first suppose, following Chomsky, that expletives simply lack Case-features. This may be supported by the fact that they do not exhibit any case declension, as far as I know, which is most notable with such an expletive as English there that has a locative form in its origin. As for the -features of expletives, I propose the following, adopting Pesetsky and Torrego’s (2007) characterization of formal features: (57) a. The expletive there carries uninterpretable and unvalued -features. b. The values of the -features may be supplied by its associate. Given that expletives lack semantic content, it is most natural to assume that the -features of expletives are uninterpretable. Recall Sato’s assumption that the expletive there shares the values of its -features with its associate. Departing from this assumption, I assume that the -feature sharing is optional, as indicated in (57b). Given the internal structure of the expletive there and its associate, as shown in (15), which is repeated below, (58) DP 3 D 1 there[+][-Case] NP the associate[+][-Case(Part)] 41 this means that the D-N agreement is optional here, so that the value of the -features of there can be [+] or [-].24 Concentrating first on the case where the -features of the expletive there are valued by those of its associate, so that it carries uninterpretable valued -features, note that this new characterization of the expletive there in terms of formal features nonetheless keeps Sato’s analysis of the there-construction almost intact, since the uninterpretability of the -features serves to the same end as the uninterpretability of the Case-feature borne by there in her original analysis, especially to satisfy the Activation Condition.25 To take (1a) again for illustration, (21), reproduced below, represents the first stage where a phase head probes: 24 See Hornstein and Witkos (2003) for relevant discussions. 25 This new characterization of there will nonetheless yield some theoretical consequences that cannot be fleshed out in detail in the present paper, especially, relating the issue of whether it should carry Case. For one thing, it argues against the existence of the Inverse Case Filter, since under the present system, there is able to check the relevant features of T and V (in ECM cases) even though it lacks Case-feature. A reviewer has brought to my attention such a case that demonstrates that the expletive there is subject to so-called Case adjacency ((i) is taken from Hazout (2004)): (i) *For unexpectedly there to be a unicorn in the garden is unlikely. In light of the present system of expletives, the apparent adjacency requirement holding between for and there in (i) cannot be attributed to a requirement on Case checking per se. Rather, it might be attributed to a constraint on the Agree relation established between the complementizer for and its goal. Though this idea may be extendable to such a case where an Agree relation is established in the v*-V phase level, it is not clear whether such an adjacency requirement equally holds for the C-T phase level in a tensed clause. I must leave this issue for future research. 42 (59) [v’ v* [V’ be [P there someone] in the garden]]] At this stage, be inherits -features from the phase head v* and probes for them. They are uninterpretable unvalued features. For this probing, the associate someone needs to be chosen as a goal, since be is able to provide partitive Case as a reflex of the checking of -features; if there were chosen as a goal, then there would be no way for the associate to obtain partitive Case, or be will not be able to value the Case-feature of the expletive, to begin with, on the assumption that partitive Case is assigned only to NP. After someone and there are raised into the Spec of V and of v*, respectively, the derivation reaches the next phase level, represented in (23), reproduced below: (60) [C’ C [T’ Pres [v*P therei [v*’ v*+be [VP someonej [V’ tbe [P ti tj] in the garden]]]]]] At this stage, T inherits uninterpretble unvaluded -features from the phase head C and probes for them. In this case, there is the only candidate for a goal, since any element inside VP is not accessible for this probing, according to the PIC. As there carries valued -features, it supplies those values to T, thereby establishing an Agree relation and bringing there into the Spec of T, as shown in (24), reproduced below: (61) [CP [TP therei [T’ Pres [v*P t’i [v*’ v*+be [VP someonej [V’ tbe [P ti tj] in the garden]]]]]]] Since the expletive there shares the values of its -features with its associate, the fact that the -features of there have valued those of T results in the required agreement pattern in which T agrees with the associate of there. There is one crucial respect in which the characterization of there as carrying uninterpretable valued -features differs from Sato’s original characterization: while in the latter, there carries an unvalued feature, that is, one with respect to Case, it does not 43 carry any such feature in the former. This is of great significance, given Pesetsky and Torrego’s (2007) assumption regarding the eligibility of probing, given in (56), reproduced below: (62) Only unvalued features can act as probes. Recall that under Goto’s (2008) approach, English raising cases have access to the option given in (49b) but that once this option is chosen, the moved element must act as a probe to check its unvalued/uninterpretable feature, as illustrated with the derivation of (43) in (53)-(54). Thus, given (62), it follows that the expletive there, when it has its -features valued by its associate, has no access to the option given in (49b), contrary to normal DPs, since it does not have any unvalued feature, hence not acting as a probe. This in turn leads us to consider how such there-constructions as in (47), repeated below, are derived. (63) a. There seems to Mary to be a man in the room. b. There seems/?*seem to Mary to be men in the room. Notice that unless we rely on the option given in (49b), as we did in deriving such corresponding raising cases as given in (46), there is no way to derive these sentences properly under the present assumptions. Notice also that the main verb seem in (63) takes a third person singular form, arguably a default one, no matter whether the associate of there is singular or plural. Thus, the correct derivations of (63a, b) should capture this agreement pattern as well. Schütze (1999) observes many cases of the there-construction in which T fails to agree with a plural associate DP and instead manifests a singular form; some relevant 44 examples are given below: (64) There is/’s/are lots of cookies on the table. (65) a. There appear/?appears to be cookies on the table. b. There tend/?tends to be cookies on the table when Johnnie comes home. For such cases, Schütze adopts the view that “I [=T] cannot enter any agreement relationship at all in the presence of such subjects [as there] and therefore shows up as a default form.” (p. 477) Following this idea, I propose that the default agreement manifested in (64) and (65) is induced when the -features of there are not valued; to put it in other words, I put forth the following: (66) Default agreement occurs if two occurrences of unvalued -features establish an Agree relation. To illustrate how this works, let us consider the following schematic configuration: (67) …T… … [ there unvalued/uninterpretable associate] … unvalued/uninterpretable Here T probes for the uninterpretable/unvalued -features that are inherited from the above C and finds there as its goal, which also has uninterpretable/unvalued -features. I am assuming that feature matching as a result of probing deprives the feature(s) involved of the label [uninterpretable], so as not to violate Full Interpretation at the interface, and that the two occurrences of unvalued -features remain unchanged after an Agree relation is established between them. And thanks to this establishment, there is raised into the Spec-TP. Finally, I assume that a last resort strategy of the following sort for unvalued -features works just before the Transfer operation applies: 45 (68) Unvalued -features may be supplied with default values that are specified in a given language. Since English has a third person singular form as default, the two occurrences of the unvalued -features in T and there bear this form. This is what happens when T fails to agree with a plural associate DP and instead manifests a singular form in some instances of the there-construction. Given the option (49b), we have another there-constructions that involve default agreement. way of deriving those Suppose that in (67), the expletive there undergoes Move before an Agree relation is established, as shown below: (69) therei T… unvalued/uninterpretable … [ ti associate] … unvalued/uninterpretable Since there carries unvalued -features, it is able to function as a probe for these features, and then finds T as its goal, establishing an Agree relation. Since this matching of -features involves unvalued ones, this gives rise to default agreement, according to (66). Given this option of derivation, we can not only provide proper derivations to such sentences as in (63), repeated below, but also explain why they exhibit default agreement. (70) a. There seems to Mary to be a man in the room. b. There seems/?*seem to Mary to be men in the room. In order to derive these sentences properly, we need to rely on the option given in (49b), since otherwise the matrix T could not probe into the embedded infinitival clause due to the intervention effect caused by the experiencer phrase. 46 Furthermore, the expletive there cannot share the values of -features with its associate in these cases, since if it did, then it would not be able to act as a probe after it is raised into the matrix Spec-TP, due to lack of unvalued features. Hence, the only way to rely on the option (49b) is to keep the -features of there unvalued, so that it can properly act as a probe after it is raised into the matrix Spec-TP. In this way, there establishes an Agree relation with the matrix T in terms of unvalued -features, which are then supplied with default values; hence the matrix verb takes the third person singular form, i.e., seems. Much the same explanation holds for the cases of the French il-construction that involve raising verbs, as illustrated in (45a, b), reproduced below, which are grammatical, unlike the corresponding raising cases illustrated in (44): (71) a. Il semble au général y avoir deux soldats manquants la caserne. ‘There seem to the general to be two soldiers missing from the barracks.’ b. Il semble au général tre arrivés deux soldats en ville. ‘There seem to the general to have arrived two soldiers in town.’ The French expletive il is more like the English expletive it rather than there in that it always induces a third person singular form on the finite verb with which it enters into a subject-predicate relation, even if its associate is plural, as illustrated below: (72) a. Il a été mangé des pommes. ‘It has been eaten apples.’ b. Il est venu quelques hommes. ‘It has come some men.’ (Hoekstra and Mulder 1990, p. 47) Given the present way of characterizing expletives in terms of formal features, it is most 47 natural to hypothesize that the expletives il and it are minimally different from the expletive there in that their -features are not valued by their associates.26 Thus, we can classify expletives into at least two kinds in terms of the availability of this valuation, as summarized below: (73) a. b. Expletives carry uninterpretable and unvalued -features. The values of the -features may or may not be supplied by their associates, depending upon the inherent nature of the expletives. According to this characterization, the expletives il and it differ from their pronominal counterparts in that while the latter take the third person singular forms as a reflex of their internal valued -features, the former happen to take the same form due to the default agreement strategy stated in (68).27 With this much in background, the grammaticality of the French impersonal constructions given in (71) immediately follows, given the above characterization of the 26 See the discussion around (35) for the expletive it and its that-clause associate. 27 Bokovi (2002, 2007) observes that quasi-argument expletives show a blocking effect caused by the experiencer phrase, as shown below: (i)?*Il semble au général avoir plu. there seems to-the general to-have rained ‘It seems to the general to have rained.’ He attributes the ungrammaticality of this sentence, contrary to such cases as in (71), to the fact that “il is actually -marked by plu in (i).” Under the present theory of expletives, this may be taken as indicating that the quasi-argument il has an uninterpretable, valued Case-feature as well as interpretable, valued -features, just like ordinary DPs in French. 48 expletive il. In each case, il first undergoes Move to raise into the matrix Spec-TP. Notice that this option is available to il, unlike other normal DPs in French, since it carries unvalued -features. Then, it probes for its unvalued -features and finds the matrix T as its goal. This feature matching makes the -features borne by the matrix T and il both get rid of the label [uninterpretable] and establishes an Agree relation between them. Since both elements have unvalued -features, the establishment of an Agree relation between them makes these features obtain default values due to the strategy stated in (68), thereby leading the derivation to converge. The English expletive it shows the same property as the French il in relevant respects. (74) a. Consider (74a) below, whose rough underlying structure is given in (74b). It seems to Mary that John is honest. b. e T seem to Mary [it [that John is honest]] In (74b), the expletive it first moves to the matrix Spec-TP and then probes for its unvalued -features and finds the matrix T as its goal. After these two elements have established an Agree relation, the unvalued -features of the two elements are supplied with default values. This completes the derivation with the correct outcome. Finally, let us consider the case where the v*-V phase level probes and finds an expletive as a goal to see whether any interesting result is obtained. Such a case is instantiated in the ECM construction, as exemplified below: (75) We proved there to be a thief among us. Under the present assumptions, (75) has the following stage of derivation, at which the matrix v*-V is to probe for -features: 49 (76) At [v*’ v* [V’ prove [T’ to [v*P therei [v*’ v*+be [VP a thiefj [V’ tbe [ ti tj] among us]]]]]]] this stage, the verb prove finds there as its goal and checks its uninterpretable/unvalued -features against those of the expletive, whether the latter are valued or not, establishing an Agree relation, and then there is raised into the Spec-VP. 28 After prove is raised into the above v*, we can derive the correct word order. With this much bearing in mind, let us consider cases of the ECM construction that appear to induce intervention effects due to the presence of PP phrases between the matrix verbs and the following infinitival clauses (the examples are cited from Takano (1998)): (77) a. *We proved to the authorities Smith to be the thief. b. We proved Smith to the authorities to be the thief. (78) a. *We proved to the authorities there to be a thief among us. b. *We proved there to the authorities to be a thief among us. The ungrammaticality of (77a) and (78a) will be immediately accounted for under the present probe-goal system, according to which Smith and there must be raised into the matrix Spec-VP, thereby resulting in preceding any phrase inside the VP; hence they necessarily precede the PP to the authorities, as in (77b) and (78b). 28 The This implies that so-called “object raising” is obligatory in the ECM construction. This is, however, incompatible with what Lasnik (1999) observes with a variety of syntactic tests, which all lead to the conclusion that “object raising” is optional. See Bokovi (2002) for a reinterpretation of Lasnik’ relevant data in light of the obligatory object shift analysis. 50 grammaticality of (77b) is also immediately explained under the present assumptions, exactly like such a raising case as (43) that involves an experiencer phrase. Given the option stated in (49b), which states that Move can apply before Agree takes place in English, Smith in (77b) can raise into the matrix Spec-VP without inducing an intervention effect, since the relevant operation of Move is subject only to the SM. It then probes for its uninterpretable/unvalued Case-feature and finds the matrix V as its goal, obtaining the accusative value from the latter as well as checking its uninterpretable/unvalued -features as a reflex. In this way, the derivation of (77b) is led to converge. The most interesting case is (78b). Adopting the External-Merge approach to the there-construction, Takano (1998) accounts for the ungrammaticality of this sentence as a violation of the MLC, since the matrix V cannot probe into the infinitival clause to hit there as its goal due to the presence of the PP phrase. This account, however, faces a problem with respect to the grammaticality of the sentences given in (70), reproduced below: (79) a. There seems to Mary to be a man in the room. b. There seems/?*seem to Mary to be men in the room. Takano would predict that these sentences were ungrammatical exactly for the same reason as (78b) is, since the matrix T could not probe into the infinitival clause without inducing a violation of the MLC. Under the present probe-goal system, on the other hand, there is a natural way to exclude such a sentence as (78b). Recall that such cases as in (79) are derived when there keeps its -features unvalued, so that it may act as a 51 probe for these feature after it moves up to the matrix Spec-TP. Since the Agree relation established between there and the matrix T involves only unvalued -features, it induces default agreement. Suppose that the default value assignment stated in (68), reproduced below, is not applicable to Vs but only to Ts and expletives. 29 (80) Unvalued -features may be supplied with default values that are specified in a given language. Then the ungrammaticality of (78b) immediately follows, since the unavailability of this default agreement forces the derivation of this sentence to take the option of the -features of there being valued against those of its associate, which in turn makes the option given in (49b) unavailable due to the condition on the eligibility of probe, stated in (62). The only remaining way is to make the matrix V probe for its -features before there is raised into the matrix Spec-TP and yet the presence of the PP phrase prevents it from penetrating into the infinitival clause to hit there as its goal; hence the ungrammaticality of (78b). 5. Conclusion In this paper, I have defended the Move-approach to expletive constructions, based upon the work by Sato (2008). 29 Couched under the probe-goal system proposed by A reviewer raises the very interesting question of whether the availability of default agreement has anything to do with overt agreement, suggesting that there might be such a correlation. This leads to the interesting prediction that default agreement should occur in Vs in overt object agreement languages. I have not yet done enough research to determine whether it is in fact borne out. 52 Chomsky (2008), this approach has been shown to be superior both conceptually and empirically to any known theory of the External Merge-approach to these constructions. Conceptually, this approach makes the probe-goal system much simpler basically because it treats expletives on a par with other normal DPs in the way they are involved in checking relevant features, including satisfaction of the EPP. Empirically, it has been demonstrated that this approach provides natural accounts for the core data involving there and it-constructions in collaboration with the crucial ingredients of Chomsky’s probe-goal system; especially, the PIC and the parallel operations operative at one phase level. Furthermore, I have attempted a refinement of Sato’s analysis of the there-construction by proposing that expletives are characterized differently from other ordinary DPs in terms of formal features: they carry uninterpretable and unvalued -features. This enables us to solve some apparent problems with Sato’s original theory. In particular, I have proposed that when the -features of there are kept unvalued, it can establish an Agree relation with T after it moves up to its Spec position, and that matching unvalued -features with other unvalued ones induces default agreement. This takes care of the fact that those expletive constructions that involve raising predicates do not induce intervention effects despite the presence of experiencer phrases only when the matrix verbs manifest default agreement. Further, by making the natural assumption that default agreement is possible only with T but not with V, I have provided a plausible account to the ungrammaticality of such a case of the there-construction that involves a PP phrase intervening between the matrix ECM verb 53 and the associate of there in the infinitival clause. Though I believe that this paper has provided enough evidence for the Move-approach to expletive constructions, there still are several important points left open; in particular, we need a more thorough examination on the internal structure of the expletive-associate complex. Further, it will be necessary to compare the present theory with another variety of the Move-approach according to which expletives are base-generated separately from their associates. I hope that the present work constitutes a good basis upon which these further issues can be properly dealt with. References ABE, J. 1997. What triggers successive-cyclic movement. In ‘Is the logic clear?’: Papers in honor of Howard Lasnik, University of Connecticut Working Papers in Linguistics 8, ed. J.S. Kim, S. Oku & S. Stjepanovi, 1-20. Cambridge, MA: MIT Working Papers in Linguistics. ABE, J. 2002. On the displacement property of language and minimality. Ms., Tohoku Gakuin University, Sendai. BELLETTI, A. 1988. Unaccusatives as Case assigners. Linguistic Inquiry 19: 1-34. BOECKX, C. 2000. Quirky agreement. Studia Linguistica 54: 354-380. BOKOVI, . 1995. Case properties of clauses and the Greed principle. Studia Linguistica 49: 32-53. BOKOVI, . 1997. The syntax of nonfinite complementation: An economy approach. Cambridge, MA: MIT Press. 54 BOKOVI, . 2002. A-movement and the EPP. Syntax 5: 167-218. BOKOVI, . 2007. On the locality and motivation of Move and Agree: An even more minimal theory. Linguistic Inquiry 38: 589-644. BOKOVI, . 2008. Unifying first and last conjunct agreement. Ms., University of Connecticut, Storrs. BOWERS, J. 2002. Transitivity. Linguistic Inquiry 33: 183-224. CHOMSKY, N. 1981. Lectures on government and binding. Dordrecht: Foris. CHOMSKY, N. 1986. Knowledge of language: Its nature, origin, and use. New York: Praeger. CHOMSKY, N. 1993. A minimalist program for linguistic theory. In The view from Building 20: Essays in linguistics in honor of Sylvain Bromberger, ed. K. Hale & S.J. Keyser, 1-52. Cambridge, MA: MIT Press. CHOMSKY, N. 1995. Categories and transformations. In The Minimalist Program, N. Chomsky, 219-394. Cambridge, MA: MIT Press. CHOMSKY, N. 2000. Minimalist inquiries: The framework. In Step by step: Essays on minimalist syntax in honor of Howard Lasnik, ed. R. Martin, D. Michaels & J. Uriagereka, 89-155. Cambridge, MA: MIT Press. CHOMSKY, N. 2001. Derivation by phase. In Ken Hale: A life in language, ed. M. Kenstowicz, 1-52. Cambridge, MA: MIT Press. CHOMSKY, N. 2004. Beyond explanatory adequacy. In Structure and beyond: The cartography of syntactic structures, Vol.3, ed. A. Belletti, 104-131. Oxford: Oxford University Press. 55 CHOMSKY, N. 2008. On phases. In Foundational issues in linguistic theory: Essays in honor of Jean-Roger Vergnaud, ed. R. Freiden, C. Otero & M.L. Zubizarreta, 133-166. Cambridge, MA: MIT Press. CHOMSKY, N. & H. LASNIK. 1993. The theory of principles and parameters. In Syntax: An International Handbook of Contemporary Research, ed. J. Jacobs, A. von Stechow, W. Sternefeld, & T. Vennemann. Berlin: Walter de Gruyter. EPSTEIN, S.D. & T.D. SEELY1999. Spec(ifying) the GF subject. Ms., University of Michigan, Ann Arbor, & Eastern Michigan University, Ypsilanti. EPSTEIN, S.D. & T.D. SEELY2006. Derivations in minimalism: Exploring the elimination of A-chains and the EPP. Cambridge: Cambridge University Press. GOTO, N. 2008. Interplay between Agree and Move. Ms., Tohoku Gakuin University, Sendai. GROAT, E.M. 1999. Raising the case of expletives. In Working minimalism, ed. S.D. Epstein & N. Hornstein, 27-43. Cambridge, MA: MIT Press. HAZOUT, I. 2004. The syntax of existential constructions. Linguistic Inquiry 35: 393-430. HOEKSTRA, T. & R. MULDER. 1990. Unergatives as copular verbs: Locational and existential predication. The Linguistic Review 7: 1-79. HORNSTEIN, N. & J. WITKOS. 2003. Yet another approach to existential constructions. In Grammar in focus: Festschrift for Christer Platzack, ed. L.O. Delsing, C. Falk, G. Josefsson & H. Sigursson. Lund University, Department of Scandinavian. 56 LASNIK, H. 1992. Case and expletives: Notes toward a parametric account. Linguistic Inquiry 23: 381-405. LASNIK, H. 1995. Case and expletives revisited: On Greed and other human failings. Linguistic Inquiry 26: 615-633. LASNIK, H. 1995. Chains of arguments. In Working minimalism, ed. S.D. Epstein & N. Hornstein, 189-215. Cambridge, MA: MIT Press. MCCLOSKEY, J. 1991. There, it and agreement. Linguistic Inquiry 22: 563-567. MORO, A. 1997. The raising of predicates: Predicative noun phrases and the theory of clause structure. Cambridge: Cambridge University Press. NOMURA, M. 2004. Expletives Move! In Proceedings of the 32nd western conference on linguistics Vol. 15, ed. B. Agbayani, V. Samiian & B.V. Tucker, 207-220. Fresno, CA: California State University, Department of Linguistics. PESETSKY, D. & E. TORREGO. 2007. The syntax of valuation and the interpretability of features. In Phrasal and clausal architecture: Syntactic derivation and interpretation, ed. S. Karimi, V. Samiian & W. Wilkins, 262-294. Amsterdam: John Benjamins. POSTAL, P.M. & G.K. PULLUM. 1988. Expletive noun phrases in subcategorized positions. Linguistic Inquiry 19: 1988. RICHARDS, M. & T. BIBERAUER. 2005. Explaining expl. In The function of function words and functional categories, ed. M. den Dikken & C. Tortora, 115-153. Amsterdam: John Benjamins. SABEL, J. 2000. Expletives as features. In Proceedings of the 19th west coast 57 conference on formal linguistics, ed. R. Billerey & B.D. Lillehaugen, 411-424. Somerville, MA: Cascadilla Press. SATO, M. 2008. Where is there?: An approach to the expletive construction, Master’s Thesis, Tohoku Gakuin University, Sendai. SCHTZE, C.T. 1999. English expletive constructions are not infected. Linguistic Inquiry 30: 467-484. TAKANO, Y. 1998. Object shift and scrambling. Natural Language and Linguistic Theory 16: 817-889. WILLIAMS, E. 1994. Thematic structure in syntax. Cambridge, MA: MIT Press. 58
© Copyright 2024