Proceedings of the Annual Meeting of the Cognitive Science Society

The categorization literature often considers two types of cat-egories as equivalent: (a) standard categories and (b) negationcategories. For example, category learning studies typicallyconflate learning categories A and B with learning categoriesA and NOT A. This study represents the first attempt at de-lineating these two separate types of generated categories. Wespecifically test for differences in the distributional structure ofgenerated categories, demonstrating that categories identifiedas not what was known are larger and wider-spread comparedto categories that were identified with a specific label. We alsoobserve consistency in distributional structure across multiplegenerated categories, replicating and extending previous find-ings. These results are discussed in the context of providing afoundation for future modeling work.