This paper seeks to clarify whether zibun, a Japanese pronoun, obligatorily gives rise to a de se interpretation in the complements of certain control predicates, particularly those that represent reported speech, belief, or desire. The central question is whether zibun behaves more like the control PRO or the English pronoun he in such environments. It is shown that the interpretation of zibun varies depending on the complementizers it co-occurs with, namely to and koto. The complementizer to requires a de se reading of zibun, whereas this requirement does not seem to hold for koto. This contrast appears to be correlated with a structural difference between to-headed and koto-headed complements: only the former projects a speech act layer, while the latter lacks such a projection.
Asako Matsuda (Sat,) studied this question.