As noted in Section 2.3, Gärdenfors ’ conceptual spaces theory of concepts is an example of a similarity space theory, which on the face of things puts it in company with prototype and exemplar theories of concepts, in the empirical tradition, and in contrast with e.g. theory theory and informational atomism, in the rationalist tradition – though that, I will argue, would at best be an over-simplification, and indeed, as is clear from passages in his book, Gärdenfors sees his theory as being compatible with and complementary to these other approaches. 1 Specifically, Gärdenfors sees his approach as a bridging account between different levels of explanation of cognition more generally, and different accounts of concepts more specifically. Indeed, I believe his theory is well placed to support the “toggling effect ” thesis I proposed in the last chapter. Sometimes with similarity-space-based theories, concepts that are more similar are grouped closer together (more literally or more metaphorically, depending on the theory), while those that are more dissimilar are grouped further apart. Fodor has argued, quite convincingly I think, that such an approach is doomed to failure, for invariably the measures of similarity that are being assumed depend upon an underlying layer of strict identity.(1998, p. 32) For example, if two prototypes are more similar or less similar depending on how many features they share, then those features must,