Volume 2002 (2002) Issue 122 Pages 163-180
This paper adds some support to the claim that in Japanese a floating quantifier (FQ) and its host-NP can form a single constituent (more specifically, can form an NP). The claim itself is found in Kamio (1977, 1983) and Kawashima (1998) among others, but Koizumi (2000) presents an alternative analysis of the crucial example and raises a question on the validity of the argument for this claim. The aim of this paper is to argue that the alternative analysis proposed in Koizumi (2000) is not adequate and that the original claim should still be regarded as tenable.
The key construction is a coordination in terms of -to, which is often assumed to be a conjunction of two constituents of the same kind. This paper describes how the to-coordination involving an FQ is restricted, and points out that there are numerous instances of to-coordination which cannot be accounted for under the analysis proposed in Koizumi (2000). This paper also shows that a sequence made up of an FQ and its host-NP should be clearly distinguished from other sequences containing an FQ. It is shown that the former can be conjoined with an ordinary NP while the latter cannot. I leave open the question why the latter can sometimes be conjoined, but the fact that the former can be conjoined with an ordinary NP indicates that it is an NP, a conclusion which supports the claim in Kamio (1977, 1983) and Kawashima (1998).