Fallibilism and Consequence

Abstract

Alex Worsnip argues in favor of what he describes as a particularly robust version of fallibilism: subjects can sometimes know things that are, for them, possibly false (in the epistemic sense of ‘possible’). My aim in this paper is to show that Worsnip’s argument is inconclusive for a surprising reason: the existence of possibly false knowledge turns on how we ought to model entailment or consequence relations among sentences in natural language. Since it is an open question how we ought to think about consequence in natural language, it is an open question whether there is possibly false knowledge. I close with some reflections on the relation between possibly false knowledge and fallibilism. I argue that there is no straightforward way to use linguistic data about natural language epistemic modals to either verify or refute the fallibilist thesis.