This view of Goal constructions contrasts with a standard analysis proposed by Hoji and others, which suggests that the Goal asymmetrically instructs the Theme in both D and S structure. The ungrammaticality of (14b) follows from the fact that the R-expression is bound by the pronoun at LF. This contrast cannot be captured if one assumes that the Objective c commands the Theme at D structure, as suggested by Hoji.
We will show that the structural prominence of the theme over the target is attested in some other constructions which are related to target constructions. This is confirmed by the ungrammaticality of (25b), which contains a transitive verb, showing that the subject containing an anaphor is not bound by the object. These results indicate that the theme projects higher than the target in non-accusative target constructions as in (22).
The UTAH implies that the structural prominence of the Theme over the Goal must also apply in transitive Goal constructions. As we pointed out above, we assume that the Goal NP is governed by an empty P, as illustrated in (27). He also argues that the trace is assigned as an anaphor Case so that it is licensed at LF.9.
We suggest that the PP in（27） moves out of the D-structure position, leaves a trace, and that the trace is assigned by the verb Case to be licensed at LF.
Source Constructions
Although (34) is superficially similar to Goal constructions such as John-ga Mary-ni hon-o ageta 'John gave Mary the book', we will argue that they are syntactically distinct. In particular, we will propose that the source argument asymmetrically c-commands the topic argument at each level. John-ga Mary-{ni/kara} hon-o kari-ta John-NOM Mary-{DAT/from} book-ACC borrow-PAST.
Juan-ken Maria-{isu/isu} husto-husto isuna John-NOM Maria-{DAT/manipud} dayta a damag-ACC mangngeg-NAPALABAS. Kas pagarigan, ti taudan nga argumento ket mabalin laeng a mamarkaan kadagiti karakter kadagiti sumaganad a pagarigan. Juan-dayta Maria-{*dayta/dayta} dayta-wenno sungbatan-nga Juan-NOM Maria-{DAT/manipud} idea-ACC take-PAST.
Shibatani (1978) points out that the NP appeared in Burimi constructions is semantically similar to the NP (from the phrase) in passives such as John-ga Mary-ni butareta 'John was hit by Mary', in a sense that two NPs represent the Action Agent. The above claim that the dative-marked source serves as a secondary agent intuitively motivates us to propose that the Source is projected higher than the Subject. These results suggest that the Source c-commands the Topic asymmetrically and that the sequence NP NP-Q does not derive from composition.
If the Source construction is structurally parallel to the Target construction with the structure（41b）, the ungrammaticality of. If the Source construction is derived from (41b) by Dative Shift, it is expected that the Source can remain in situ, separated from a predetermined VP. On the other hand, (41a) predicts that Source cannot be separated by VP preposition, as it is strictly within VP.
The entire VP can be preposed in the Source construction, as illustrated in (43a), while Source cannot be separated from VP-. The structural relationship between the Source and the Theme in the D-structure is preserved during the course of the derivation. The source argument created in internal subject position is compositionally assigned an agent (secondary) role by V' which consists of the verb and its subject argument.
Benefactive Constructions and the Theory of Control In this section, we will deal with benefactive constructions that
John's Mary's hon-o cat's mosha John-NOM Mary-DAT libër-ACC blej jap-PAST. Shoku i Gjonit Mary-hon-i John-NOM Mary-DAT-libri-ACC marrë-PAST. John's Mary-his hon-o age-this John-NOM Mary-DAT book-ACC give-PAST.
Shibatani（1978） claims that useful constructions contain an embedded clause and derive from the Equi-NP-Deletion rule. First, we check whether the benefactive construction contains an embedded clause by examining the scope interpretation of the quantifiers. Assume that the dative phrase is a quantifier and is shuffled over another quantifier in subject position.
Daremo-ni dareka-ga akusyu-site age-ta everyone-DAT someone-NOM shake=hand give-PAST This contrast indicates that the dative phrase in the te-ager construction is also an element of the matrix clause. The Japanese reflexive zibunsisin is assumed to be a local anaphora that obeys condition A of the binding theory. John-ga Maryi-ni kanozyo?＊i-no syasin-o totte morat-ta John-NOM Mary-DAT she-GEN picture-ACC take receive-PAST.
John-ga Maryi-ni kanozyoi-no syasin-o totte age-ta John-NOM Mary-DAT she-GEN picture-ACC take give-PAST. Taking this thematic similarity as a clue, it is reasonable to argue that the te-ager construction is structurally parallel to the Goal construction, while the te-moraw construction to the Source construction. Mary-ni hana-o katte age］-saei John-ga ti sita Mary-DAT flower-ACC buy give-even John-NOM did Lit.
Mary-ni cake-o yaite morai(saei John-ga ti cita Mary-DAT cake-ACC bake receive-and John-NOM did Lit. Daytoy ti mangtignay kadatayo a mangisingasing a dagiti te-ager a konstruksion, ngem saan a te-moraw a konstruksion, ti nagtaud manipud iti Dative Shift Ni John ni Mary's hon-o's girl John-NOM Mary-DAT libro-ACC gumatang get-PAST.
In his discussions of the constructions for promise and obey, Larson argues that the former is a two-object verb, but not the latter. John-ga Mary-{ni/kara} ［PRO homete］morat-ta John-NOM Mary-{DAT/from} praise-PAST.
Conclusions
Impersonal passives and the unaccusative hypothesis, Proceedings of the Fourth Annual Meeting of the Berkeley Linguistic Society, 157─189. 1) Later we will propose a slightly modified hierarchy, where a living Source fits between the Experiencer and the Topic in the hierarchy. In both cases, the subject 'someone' is moved to subject position by A-movement, which is caused by the passive morphology in (ia) and the unaccusative verb in (1b).
However, the following example seems to suggest that the hierarchical order between the theme and the location is reversed. If this is correct, it appears that the theme is not moved across the location in (ib). 7）Note that the ni in transitive Objective constructions does not block mutual c-command relation between a floated quantifier and its host NP.
8)Kayne (1983) also assumes that an empty P blocks case assignment of its complement NP by a verb and claims that the verb assigns Case to the PP and it is transmitted to that NP via the empty P. According to his analysis of dative displacement as P-incorporation in a verb, the verb can direct an NP complement to the. 10) Following Baker (1988, 1991b) and Raposo and Uriagereka (1990), we assume that both the visibility condition in the LF and the case filter in the S (or PF) structure are necessary in the grammar.
11) We suggest that the reason that PP cannot stand in the internal subject position stems from the general inability of PPs to involve predicative relations. 15) Hoji et al. (1989) point out that example (ia) is ungrammatical due to the violation of the proper conjunction Condition. We will develop a slightly different account, due to the following fact that the VP containing a trace can be preposed in the passive,.
John-ga Mary-ni ［ kanozyo-ga itte］morat-ta John-NOM Mary-DAT she-NOM will receive-PAST 'John received the favor of Mary's passing'. John-ga Mary-ni ［ zibun-ga utatte］age-ta John-NOM Mary-DAT self-NOM sing give-PAST 'John gave Mary the favor of singing'. 21）Note that the bare VP analysis can still be available if one accepts Williams (1980) predicate condition in place of the SSC.
