The semantics of vagueness: supertruth, subtruth, and the cooperative principle

Experimental results are used to assess the predictions of supervaluation, subvaluation, and many-valued logic with regards to the problem of vagueness. Supervaluation predicts a truth-value gap in the borderline range of a predicate, thus assigning the predicate neither true nor false as value, subvaluation predicts a truth-value glut, where the predicate is both true and false, and many-valued logics assign at least three truth-values to propositions in their domain. The results of the experiment are analyzed and shown to oppose each of these frameworks, and instead to favor an approach in which the predicate and its negation are false in the borderline range, but where their conjunction is true.