LEARNING OF WELL-DEFINED AND ILL-DEFINED CATEGORIES IN AUTISM

Background: The cognitive process of categorization is fundamental to the organization and generalization of knowledge, both of which are impaired in autism. Of particular interest have been categories based on well-defined rules versus those in which category membership is not easily defined.

Objectives: We examined performance with ill-defined categories that were based on a prototype, or central tendency, and those on that were based on individual category exemplars. Also examined was performance with well-defined categories based on a rule about a single diagnostic feature. Categorization of trained category examples versus novel examples, differences in rates of learning between participant groups, and the role of recognition memory for learned exemplars was measured.

Methods: Twenty-six high-functioning adults with an autism spectrum disorder and 26 typically-developing controls participated. Categorization tasks involved alternating training and transfer phases and a recognition phase, similar to the task structure used by Kolodny (1994). Stimuli for ill-defined categories included random dot patterns (Posner, Goldsmith, & Welton, 1967), and landscape paintings of the European Volga region. Stimuli for well-defined categories included computer line drawings of imaginary bugs and animals, each containing six varying features, only one of which was diagnostic of category membership.

Results: Participants with autism performed as well as controls with the well-defined categories. Performance with the ill-defined categories varied. Participants with autism demonstrated evidence of prototype formation similar to that of controls although were less able to categorize new exemplars that were high distortions of the prototype. Overall performance with ill-defined categories based on individually learned exemplars was poorer for the autism group.

Conclusions: Individuals with autism appear to be better able to learn categories that are well-defined, and prototype formation may be preserved in autism. Possible reasons for discrepancies in the literature and finer evaluation of categorization performance will be discussed.