Indefinites as Epsilon Terms: A Labelled Deduction Account

Abstract

It is a well known problem of on-line parsing that quantified NPs, in particular indefinites, may be interpreted as giving rise to logical forms in which they have a scope very different from that indicated by the surface sequence of expressions in which they occur — a problem which cannot be reduced to positing referential uses of indefinites; cf. (Farkas, 1981; Abusch, 1994). Various computational and formal systems have tackled this ambiguity problem, either, as is familiar, by positing processes of restructuring or storage (Montague, 1974; Cooper, 1983; May, 1985; Morrill, 1994; Pereira, 1990), or by building underspecified structures including unscoped or partially scoped representations of quantified NPs (Alshawi & Crouch, 1992; Reyle, 1993; Pereira & Pollack, 1991). In all of these cases, there is implicit recognition that the determination of scope choice is not incremental, but can only be defined once the total structure is complete. In this chapter we propose that the interpretation of indefinite NPs involves an anaphoric-like dependency, in which the indefinite is lexically projected as involving a dependent name for which the anchor of the dependency has to be chosen online. The dependency is represented by an indexing on the name, the index indicating the expression to which the dependent element is to be anchored.