We examine the influence of contrast categories on the internal graded membership structure of natural language categories. To this end we contrast two exemplar models in their account of typicality: According to the GCM, typicality is the summed similarity towards all category members. According to the SD-GCM, typicality is determined by both the summed similarity towards all category members and the summed dissimilarity towards members of contrast categories. For five animal categories, we contrast the SD-GCM and the GCM in their account of typicality. Results indicate that the internal category structure can indeed be co-determined by dissimilarity towards potential contrast categories.