Japanese Passives as Control/Raising/ECM

Japanese has three passives, Direct (1), Niyotte (2), and Indirect (3), which show different syntactic and semantic characteristics (Hoshi 1999). In this paper, I propose an analysis of Japanese passives as verbs instantiating three types of non-finite complementation: control, raising, and ECM, and argue that the proposed analysis provides a principled account for them. Previous studies have identified differences and similarities across the three passives. First, both Direct and Niyotte Passive show characteristics of the " familiar " passive. Their target verb's object position is obligatorily null, and the object's semantic role is borne by the subject (1 and 2). Moreover, the subject appears to have derived from the object position, as it is compatible with modifiers which can only be associated with internal arguments, such as preverbal numeral quantifiers and resultatives (Miyagawa 1989). Indirect Passive, however, shows behaviors atypical of passive. It adds a new participant to an event, and it allows its target verb's object to remain in situ (3). Furthermore, Indirect Passive's by phrase acts like an embedded subject, unlike the by phrases in two other passives, as it can license a floating quantifier (Shibatani 1977) and the reflexive jibun (McCawley 1972). Niyotte Passive also differs from other two passives in nontrivial ways. It has niyotte as the by phrase marker, while other two passives have ni, and it is the only one that does not impose selectional restrictions upon its subject (Kuroda 1979). Finally, while both Niyotte and Direct Passive have derived subjects, only Niyotte Passive shows reconstruction effects (Kitagawa and Kuroda 1992). These differences, summarized in 4), make a unified account for the three passives difficult. Especially problematic is Direct Passive, which requires its subjects to be derived and thematic, a paradox under Principles and Parameters frame work. However, the three passives' prima facie unprincipled differences receive a principle account by analyzing the them as control, raising, and ECM. Under the proposed analysis, Direct Passive is Θ-feature driven A-movement, or control as movement (Hornstein 2003). In this theory of control, Θ-roles are features of verbs, and NPs undergo A-movement to check Θ-features. Consequently, the difference between control and raising is whether movement is to a thematic or non thematic position. Adopting the movement theory of control for Direct Passive eliminates its paradox: if Direct Passive is control, its subject is expected to be thematic and derived. It also accounts for the lack of reconstruction effects, …