Proceedings of the Annual Meeting of the Cognitive Science Society

The Wisconsin Card Sorting Test (WCST) is commonly used to assess executive (dys-)function, particularly in neuropsychological patients. Performance on the test typically yields two types of error: perseverative errors, where participants persist in applying an inferred rule despite negative feedback, and set-loss errors, were participants cease applying an inferred rule despite positive feedback. The two types of error are known to dissociate. In this paper we apply an existing model of the WCST -- the model of Bishara et al. (2010) -- to a novel dataset, focussing specifically on the distribution of the two types of error over the duration of the task. Using Maximum Likelihood Estimation to fit the model to the data, we argue that the model provides a good account of the performance of some participants, but a poor account of individual differences. It is argued that this is because the model is essentially a competence model which fails to incorporate performance factors, and that accounting for the different types of error, and in particular the error distribution during the task, requires incorporating performance factors into the model. Some consequences of this for the broader enterprise of developing normative competence models are discussed.