Abstract: In his 1993 Nicod Lectures, Jerry Fodor proposed a solution to a certain version of the problem of‘inscrutability of reference’, which problem poses a challenge to a certain naturalistic, computational approach to cognition which Fodor has favoured. The problem is that purely informational accounts of an agent's mental contents cannot discriminate meanings finely enough. Fodor proposes a strategy of solution which appeals to the inferential dispositions of agents to discriminate contents more finely. After a brief exposition of the problem and Fodor's bid for solution, I employ three counterexamples to argue that Fodor's proposal cannot succeed.