Proceedings of the Annual Meeting of the Cognitive Science Society

In this paper we explore the effect of normal aging on executive function and present a computational account of the effectof aging in a standard executive task. We tested 25 younger adults and 25 older adults (both with no known neurologicalcondition) on the Wisconsin Card Sorting Task (WCST), a classic test of executive function. The test produces multiplemeasures related to the types of error made by participants, the rate of learning, and so on. As hypothesised, results showno difference between the groups in the number of perseverative errors (i.e., in continuing with a previously successful rulein the presence of negative feedback), but a significantly increased tendency for older adults relative to younger adults tocommit set loss errors (i.e., to switch away from a rule despite positive feedback). We fit an existing neurocomputationalmodel of the task to the experimental data by searching through the models parameter space in order to find the best set ofparameter values for the two different age groups. This leads to a proposition regarding the effect of aging on the value ofthe epsilon ctx parameter, which we argue elsewhere reflects cortical dopamine concentration. We further reanalyse thedata by clustering participants by performance (rather than by age) and show that there are multiple points in parameterspace that fit each cluster of participants. We argue on the basis of this and the behavioural data, that different parametervalues reflect different solutions to optimizing task performance, and that older participants may compensate for changesin epsilon ctx (reflecting dopamine concentration) by effortful changes in other parameters (specifically, by increasingattentional focus).