Age-related differences in the stability of categorization performance and prototype versus exemplar strategy use

Category learning is the ability to link items to a shared label based on their common features. Despite being important for individuals of all ages, there is little research on how older age affects category learning and generalization. Moreover, there is no research, to date, on the extent to which individuals tend to adopt the same categorization strategy across different to-be-learned categories. In this study, healthy young (aged 18-30, n = 77) and older adults (aged 60+, n = 74) completed two categorization tasks in separate sessions approximately 48 hours apart, with the category structures remaining constant across sessions but the stimuli differing. We compared categorization accuracy and prototype versus exemplar model fits across age groups and across sessions. In both sessions, there was an age deficit specific to learning items near the category boundary. Despite differences during learning, there were no significant age differences in generalization performance. Fitting formal prototype and exemplar models revealed a comparable proportion of young and older adults best fit by the prototype model overall, but a continuous measure of prototype model fit advantage showed higher correlation across sessions in older compared to young adults. Together, these findings show that older adults consistently rely on typical training items to learn categories and are more stable in their prototype strategy use than young adults, which can be effective in supporting generalization performance at young-adult levels.