Abstract
This study focusses on a set of syntactic constructions, commonly referred to as Specificity Effects, in which the possibility of extracting from within or binding into a Noun Phrase depends on non-syntactic properties of the NP. In contrast to a number of studies of this phenomenon in the syntactic literature (Fiengo and Higginbotham (1981), Bower (1987), Diesing (1990)), it is argued that such cases of opacity should not be defined in syntactic terms. The non-syntactic status of the constraints underlying Specificity Effects follows from a number of empirical and theoretical considerations. Empirically, the received grammaticality judgments on unacceptable cases of extraction from NPs are reconsidered. In particular, data that have been used in syntactic definitions of the class of opaque NPs are shown to vary according to, or to be mitigated by, purely contextual factors. Theoretically, it is argued that independently motivated mechanisms required to explain certain pragmatic properties of NPs suffice to account for Specificity Effects. Thus, syntactic accounts of this phenomenon are superfluous.
The account of Specificity Effects argued for in this work utilizes the representation of the beliefs underlying the communicative import of propositions proposed in Ludlow and Neale (1991). Based on their approach to specificity, the inability to quantify into specific NPs follows from a pragmatic equivalent to structurally determined wide scope readings. The class of NPs which induce such wide scope readings is defined in terms of two distinct pragmatic properties. The first is referentiality as defined in Burge (1974), where the interpretations of referential NPs are relativized to acts of reference performed on particular occasion of utterance. The second property is based on the approach to anaphora with (in)definite NPs in Hawkins (1991) and involves the implicatures speakers utilize in conveying whether or not a referent is identifiable. These pragmatically defined properties are shown to correlate with the opacity of NPs more closely than syntactic or semantic properties.
(To return to the main page, click here)
(For ordering information, click here)