Semantic Type and Anaphoric Reference of Internally Headed Relative Clause
Rui Otake
Graduate School of International Cultural Studies, Tohoku University

Abstract
In the actual use of the language, we supplement an incomplete or underspecified representation of meaning with information inferred from the context and our world knowledge. In an attempt to formalize such an inferential mechanism, this paper focuses on Internally Headed Relative Clause (IHRC) construction in Japanese. An IHRC is a sentence which has no gap and fills in the argument slot where an individual denoting expression is normally expected (e.g. Taroo-ga Hanako-ga ringo-o muita-no-o tabeta “Taro ate the apple Hanako peeled.” Lit. Taro ate Hanako peeled an apple). As a result, the relative clause refers to some contextually understood individual, along with the proposition which is derived compositionally.
The first part of the paper addresses the issue of the semantic type of the IHRC. Among the several analyses that have been proposed, the E-type pronoun analysis (Shimoyama 1999) is a major candidate in that it descriptively captures the way how the interpretation of the IHRC interacts with the quantification. However, the E-type analysis runs the risk of proliferating the type ambiguity, since in the analysis if the IHRC appears in the scope of n quantified expressions, the E-type pronoun has to be decomposed into n bound variables and the n-placed predicate to which they apply. We will argue that this kind of ambiguity can be eliminated by regarding IHRCs in general as instances of the generalized Skolem term (Steedman 2005), individual denoting object whose denotation varies according to the values of bound variables introduced by scope-taking operators, but which is interpreted as a constant in the absence of such operators.
The second part of the paper addresses the anaphoric aspect of the IHRC, namely the problem with recovering the nominal property associated with the IHRC (e.g. being an apple). Some constraints on the interpretation of IHRCs are studied in (Kikuta 2000), but they cover only a particular subclass of the construction. In that class of IHRC, which I will refer to as headless IHRC, the relative clause lacks the “head” (i.e. noun phrase that corresponds to the nominal property of the IHRC). The two classes of IHRC (headed/headless) cannot be distinguished until the interpretation is determined, so they should be treated by some common mechanism. We suggest that the rhetorical relation, as exploited in Segmented Discourse Representation Theory (Asher and Lascarides 2003), might play a significant role in the inferential process of the anaphoric resolution of IHRCs.