The Representational Theory of Mind, or RTM, is arguably the best hope for formulating a naturalistic theory of reasoning. According to RTM, thinking processes are implemented by kinds of computational processes that are purely constituted by the syntactic relations of Mentalese symbols. But findings with the Wason selection task suggest that the appeal to syntactic or formal relations alone is not sufficient for understanding logical inferences. In order to understand logical inferences, one also has to understand the content effect that facilitates the inferential processes. Thus a satisfactory theory of reasoning requires a theory of thinking as well as a theory of content. But there is a problem of putting these two theories together. The aim of this paper is to highlight this problem.

Key words: content, computation, mental representation, naturalizability, truth-preserving