The classic reference is Fine (1975). See also Keefe (2000). 2 See MacFarlane (2016) for a recent approach of this kind. We may then say that a given language is inconsistent exactly if competence with all elements of that language involves dispositions to accept things that are inconsistent with each other. Talk of language being inconsistent hardly has a pretheoretically clear meaning. I am using the locution partly stipulatively. The account of competence just sketched does not by itself say anything about the semantic values of the expressions of an inconsistent language. One perhaps natural suggestion would be that when the rules governing an expression are inconsistent, then the expression simply lacks a semantic value (or it has a trivial semantic value: for example, a predicate would be empty). I see two problems with this. First, this seems implausibly radical: it yields that nothing is old, or red, or a mountain, etc. Anyone who embraces this radical idea is apt to want to draw a distinction between assertible and non-assertible sentences of, for example, the form '___ is old', and so mitigate the problem by saying that those of these sentences we would take to be true are at least assertible. However, deciding under what conditions sentences are assertible presents much the same problems as deciding under what conditions they are true, and supposing one can come up with a satisfactory theory of assertibility, why not simply use that also as a theory of truth? Second, a different problem is that the incoherence is holistic. It is not just the rules for one expression