A new problem for the linguistic doctrine of necessary truth∗

My target in this paper is a view that has sometimes been called the ‘Linguistic Doctrine of Necessary Truth’ (L-DONT) and sometimes ‘Conventionalism about Necessity’. It is the view that necessity is grounded in the meanings of our expressions—meanings which are sometimes identified with the conventions governing those expressions—and that our knowledge of that necessity is based on our knowledge of those meanings or conventions. In its simplest form the view states that a truth, if it is necessary, is necessary (and knowably necessary) because it is analytic. It is widely recognized that this simple version of the view faces a prima facie problem with the existence of the necessary a posteriori. Assuming that all analytic truths are a priori, if there are necessary a posteriori truths then there are necessary synthetic truths—contradicting the view’s claim that all necessary truths are analytic. Contemporary L-DONTers have things to say about the problem, but in this paper I want to suggest that there is a different, more serious, problem which arises from the phenomenon of indexicality, which L-DONTers have not taken account of. Though there are many versions of the problem, a simple one is this. Consider Kaplan’s celebrated sentence: