Passive, Causative, and Light Verbs: a Study on Theta Role Assignment

Abstract

In the enterprise of generative grammar, it has been considered desirable to minimize assumptions and eliminate redundancy. Following this tradition, Chomsky (1992) proposed that the internal interface levels, D-structure and S-structure, should be eliminated from the grammatical model, because there is no conceptual necessity for them.^ In this thesis, I show that to analyze Japanese passives, Romance causatives, and light verb constructions, we must abandon D-structure as a pure representation of GF-$\\theta$ relations. Thereby, I propose analyses which are compatible with Chomsky's new grammatical model, and explore their theoretical and empirical consequences.^ In chapter 2, I point out that the ni direct passive verb in Japanese has dual characteristics, the combined function of the passive morpheme -en and the verb get of the get passive. To explain these properties, I propose that this passive verb triggers Passivization at the initial point of the derivation, but assigns $\\theta$-roles at a later point of the derivation (Hoshi 1994b, cf. Washio 1989-90).^ I argue in chapter 3 that like the ni direct passive verb, the Italian causative verb and the Japanese benefactive verb assign $\\theta$-roles not at the initial point of the derivation, but later in the course of the derivation. However, as opposed to the ni direct passive verb, the benefactive verb excorporates from the embedded verb for tense feature checking in syntax as does the Italian causative verb (Guasti 1992). Consequently, the proposal suggests that Japanese is an overt V-to-I language (Tada 1990, among others).^ In chapter 4, I demonstrate that there is an instance in which $\\theta$-marking takes place in LF. Based on our previous work (Hoshi and Saito (1993) and Saito and Hoshi (1994)), I propose that in the light verb construction, the nominal $\\theta$-marker incorporates to the light verb, and assigns $\\theta$-roles at the clausal level in LF. ^

Cite this work

No public copy is stored for this record, so only the metadata is cited.

Stored only when both papers are in the corpus — 21 citations exist upstream.