The purpose of this dissertation is to reconsider reflexives in Japanese through the following three steps: (a) separation of genuine reflexive elements from elements that are confounded as reflexives, (b) classification of reflexive anaphors into subtypes based on their semantic difference, and (c) classification of predicates that occur with anaphors.

Many researchers have worked on the reflexive element zibun (self), but Japanese has other reflexive elements as well. These elements including zibun have not only the reflexive anaphor usage but also other ones. All the instances are, however, often lumped together under one category: reflexive. I distinguish genuine reflexive anaphors in Japanese from elements that are confounded as reflexive elements, by scrutinizing their syntactic and semantic properties and behavioral differences.

Further, I claim that reflexive anaphors are classified into two subtypes as Pure reflexive anaphors and Near reflexive anaphors (Lidz, 1996, 2001a,b) based on their semantic property. Observing several languages from different language families, I propose that there is a parametric variation with respect to the two-type distinction of reflexive anaphors among languages. In languages like Japanese, anaphors in the form of affix are Pure reflexive anaphors, while non-affixal anaphors are Near reflexive anaphors. On the other hand, in languages like Dutch, the morphological composition (complexity) of anaphor corresponds to the two-type anaphor distinction. What yields this variation is also discussed.

In considering reflexives, it is important to know the nature of reflexive anaphors, but it is also essential to understand the nature of predicates that occur with an anaphor. One of the unsolved questions in the research of reflexives in Japanese is that the anaphor zibun cannot take a local antecedent when it occurs with a certain type of verb, although anaphors should be locally bound. Several studies have demonstrated that the availability of local binding of an anaphor depends on the property of its cooccuring predicate (Reinhart and Reuland, 1993, Bergeton, 2004, among others). Discussing how the type of reflexive and the type of predicate relate, I propose a way to categorize predicates in Japanese into subtypes based on the analysis in Bergeton (2004). By going through the three steps, I give an answer to the unsolved question.