ABSTRACT
This chapter has reviewed the literature on visual object recognition impairments attributable to a presemantic deficit and that are specific to biological object categories. Several authors concur that such impairments are not a function of a categorically organised visual object recognition system but rather that they reflect the greater visual similarity of objects within biological than man-made categories. In particular, it has often been assumed that the greater within-category similarity for biological objects renders them more difficult to dissociate from each other and, therefore, more susceptible to the effects of brain damage. The studies of patient ELM that are reviewed above have provided the first controlled experimental demonstration that visual similarity and semantic proximity do indeed jointly determine visual object recognition performance in category-specific visual agnosia. These investigations have implications that extend beyond the particular case of ELM. Notably, results have argued for distributed representations of object shapes that are made of collections of discrete features. Support for this assumption has been found in patient IL (who is another case suffering from CSVA), DAT patients, and neurologically intact observers. Finally, we report computational models that are capable of replicating the interactive effects of visual similarity and semantic proximity documented in patients ELM and IL while implementing a distributed code for shape representation.
