Aspectual Shift via Supervaluation

Aspectual shifts in connection with temporal adverbial modification usually are treated by close analogy with type-coercion in programming languages. The paper shows that by taking a good metaphor too literally, semantic research so far did not do full justice to the kind of flexibility observable in the natural language examples. A new proposal is developed within the framework of finite-state temporal semantics combined with a supervaluational concept of underspecification. The simple shifting algorithm used in the present approach generates the correct set of possible readings on the basis of linguistic input only and, furthermore, may claim cognitive plausibility.