REPRESENTATION IS REPRESENTATION OF SIMILARITIES (CORRECTED VERSION OF 126BQ)

Advanced perceptual systems are faced with the problem of securing a p rincipled (ideally, veridical) relationship between the world and its internal representation. I propose a unified approach to visual repres entation, addressing the need for superordinate and basic-level catego rization and for the identification of specific instances of familiar categories. According to the proposed theory, a shape is represented i nternally by the responses of a small number of tuned modules, each br oadly selective for some reference shape, whose similarity to the stim ulus it measures. This amounts to embedding the stimulus in a low-dime nsional proximal shape space spanned by the outputs of the active modu les. This shape space supports representations of distal shape similar ities that are veridical as Shepard's (1968) second-order isomorphisms (i.e., correspondence between distal and proximal similarities among shapes, rather than between distal shapes and their proximal represent ations). Representation in terms of similarities to reference shapes s upports processing (e.g., discrimination) of shapes that are radically different from the reference ones, without the need for the computati onally problematic decomposition into parts required by other theories . Furthermore, a general expression for similarity between two stimuli , based on comparisons to reference shapes, can be used to derive mode ls of perceived similarity ranging from continuous, symmetric, and hie rarchical ones, as in multidimensional scaling (Shepard 1980), to disc rete and nonhierarchical ones, as in the general contrast models (Shep ard & Arabie 1979; Tversky 1977).