意味論研究会2016年5月13日

The central question in the research on ellipsis phenomena in natural language is whether abstract structural representation needs to be posited to account for the various properties that such phenomena exhibit. Based on a reconsideration of the major arguments put forward in Kennedy and Merchant (2000) and Kennedy (2003), we answer negatively to this question. While we agree with Kennedy and Merchant that purely interpretive approaches to ellipsis have difficulties in accounting for certain combinatorial properties that some ellipsis phenomena in English exhibit, this does not immediately lead to the conclusion that a full syntactic representation (in the form of trees) needs to be represented in the ellipsis site. To demonstrate this point, we offer an alternative analysis, couched in Hybrid Type-Logical Categorial Grammar, which captures the same range of facts by merely making the anaphoric process that underlies ellipsis sensitive to the combinatorial properties of the antecedent expression. The general conclusion we arrive at is consonant with the conclusion independently arrived at by various authors in the recent literature (Barker 2013, Chung 2013, Yoshida et al. 2016) that ellipsis exhibits only 'partial' syntactic sensitivity.

(Joint work with Bob Levine.)