The syntax of Japanese tokoro-clauses: Against control analyses

The Japanese tokoro-clause construction has long been an intriguing topic in Japanese linguistics because of its mismatch between syntax and semantics. Syntactically, the tokoro-clause itself seems like an object of the matrix verb through being assigned an accusative or a dative particle, while semantically the subject of the tokoro-clause is construed as the thematic object of the matrix verb. It has been argued that this mismatch can be reconciled by assuming backward object control in which the unpronounced matrix object is coreferential with the downstairs tokoro-clause subject: NP eci [NPi V tokoro]-acc/dat V. While acknowledging four different types of tokoro-clauses according to the particle and the matrix verb, it is argued that none of them involve obligatory control, let alone backward control. Specifically, two of them are complements on their own, and the other two are vP-adjuncts. Although there is a null matrix object in the latter case, it does not have a c-command relation with the tokoro-clause subject, and its referential relation is pronominal coreference rather than obligatory control.