Abstract

This chapter addresses several arguments to the effect that similarity relations can be recognized only by means of concepts. First, it is argued that we can recognize that _x_ is more like _y_ than like _z_ even when this cannot be accounted for by appeal to a property that _x_ and _y_ share that _z_ lacks. Second, it argued that we do not need to think of representations as conceptual in order to define a kind of correctness. A sense in which a mark in perceptual similarity space may accurately represent an object or scenario is defined. Finally, it shown that Tversky's well-known account of similarity in terms of shared and differentiating features is not well supported by his experimental results.