Abstract
In Japanese, case marking is generally correlated to a category of a case-assigner and its projection. A verbal case (i.e. Nom, Acc, Dat) is assigned by a verb in the verbal projection and a nominal case (i.e. Gen) by a noun in the nominal projection. Case-marking patterns associated with temporal affix constructions in Japanese pose a problem on the general case-marking pattern, since they allow 1) a verbal case marking, which involves only verbal cases, 2) a nominal case marking, which involves only nominal cases, and 3) a mixed case marking, which involves both cases. This paper attempts to solve the problem within a LFG framework. In particular, we argue that Bresnan (1997)'s head sharing analysis, which allows a verbal head to be mapped to the same f-structure as its sister NP as an extended head (or a f-structure head) of the sister NP, is empirically preferred to solve the problem of mixed case marking, since it captures the property of temporal affix constructions that allow mixed case marking, that is, the property such that the mixed case marking is allowed if the temporal affix constructions are verbal projections headed by a single verb. We propose a case licensing system based on phrase structure rules and the constructive case theory without assuming a case-assigning head at c-structure. In particular, the phrase structure rules reflect the fact that a verbal case is licensed under VP and a nominal case under NP (i.e. the general case-marking pattern) and that a VP allows a nominal case marking to a non-external argument NP (i.e. mixed case marking).