Abstract
A requirement on any theory of vagueness is that it solve the sorites paradox. It is generally agreed that there are two aspects to such a solution: one task is to locate the error in the sorites argument; the second task is to explain why the sorites reasoning is a paradox rather than a simple mistake. I argue for a further constraint on approaches to the second task: they should conform to the standard modus operandi in formal semantics, in which the semantic theory one develops is taken to be implicit in the ordinary usage of competent speakers. Thus it should not turn out that one's explanation of why ordinary speakers react to the sorites reasoning in the way they do depends on speakers not thinking that the semantics of vague language is governed by the theory one is advocating. I then argue that, out of the current main contenders for a theory of vagueness, only theories that posit degrees of truth can meet this further constraint.