Underspecification and the Semantics / Pragmatics Interface

The notion of UNDERSPECIFICATION [17, 2, 21, 25] has been proposed as a possible solution to the problem of specifying a theory of the semantics / pragmatics interface—more specifically, of the process of identifying an interpretation in context—that is consistent with psychological, as well as linguistic evidence. The assumptions about the respective roles of semantics and pragmatics adopted in standard semantic theories such as [14] are illustrated in Fig. 1. Grammar (syntax and semantics) specifies all the distinct disambiguated interpretations of an ambiguous sentence like Kermit croaked,1 typically characterized by means of a DERIVATION TREE annotated with expressions in a DISAMBIGUATED LANGUAGE. The role of pragmatics is to filter those which are implausible and / or choose the preferred one. Theories of utterance processing that directly implement this view of the respective roles of semantics and pragmatics are confronted with the problem that the number of distinct disambiguated interpretations of any given sentence is in general very large—scopally ambiguous sentences such as (1), for example, might have n! readings, where n is the number of scope-taking operators. It's not clear how humans could be able to generate, let alone choose from, all these interpretations. 1I assume that the lexical item croak is semantically ambiguous between a property which is true of objects that utter frog-like sounds, and a property which is true of objects which died. (1) In most democratic countries most politicians can fool most of the people on almost every issue most of the time. [Lincoln; reported in [11].]