Semantically-Based Ellipsis Resolution with Syntactic Presuppositions

A central issue since the late 1960’s has been the issue of whether ellipsis resolution processes are syntactic or semantic in nature.1 Recently, works that take a more generalized discourse view (Prust et al, 1993; Asher, 1993; Kehler, 1994) have suggested the need for an approach which eschews a purely syntactic or semantic approach, but rather is conditioned by discourse structure. Kehler, for instance, has modified the approach of Hankamer and Sag (1976, 1984) so that whether the resolution arises from syntactic copying or is semantic i.e. dermis from material already integrated in the discourse model, is determined by whether the coherence relation between clauses is parallel or not;. Thus, on the Sag/Hankamer/Kehler view, syntactic parallelism is not expected indeed not possible in resolution where the source has been integrated in the discourse model. In this chapter, I show that short answers as they occur in extended dialogue involve a resolution process that is perplexing for current models of ellipsis, both purely syntactic or semantic, and mixed ones such as Kehler’s. In particular, I demonstrate that elliptical options are possible at essentially unbounded distance from the original source, long after integration of material must have taken place in the discourse model, and yet, (partial) syntactic parallelism obtains.