Abstract
It has been shown that the structure of subjective representation space for visual form can conform to the structure of the featural dimensional space from which those forms are derived. (Cutzu & Edelman, 1996; Palomo & Watanabe, 1999) Here, the dynamics of subjective representation for visual form were assessed using parameterized morph shapes. Previous experiments (Palomo & Watanabe, 1998; 1999) demonstrated that expansion of representational space could not only account for categorical perception but could also explain alternative findings of representational space contraction (Livingston et al., 1998). The present experiments extend those findings by showing that stimulus distribution weighting alone appears to drive the distortion of representational space, and that increased discriminability within shape space subregions follows from representational space expansion. Multidimensional scaling was used to construct maps based on subjective proximity ratings for morphs before and after training on experimenter-determined categories in which one category was over-sampled relative to others while a third was neglected altogether. The over-sampled region acquired expanded representation in the subjective map, at the expense of the neglected (untrained) region. In a second experiment, we tested the discriminability of shapes within discrete category regions before and after category training and found discriminability improvements consistent with subjective space distortions. These results in subjective representation space concur with physiological findings of cortical recruitment in inferotemporal cortex and other sensory areas (Tanaka, 1996; Vogels, 1999), and suggest that the dynamics of shape representation follow a competitive process in which units vying to code for environmental stimuli gain strength and neighborhood alliances through repeated exposure to the stimuli for which they code.
Supported by NSF Grant(BCS-9905194) to TW.