Toward a computational cognitive neuropsychology of Wisconsin card sorts: a showcase study in Parkinson’s disease

Abstract

The Wisconsin Card Sorting Test (WCST) is an instrument for the clinical assessment of executive functions. Computational modeling of latent cognitive processes offers a route toward improved interpretability of performance on neuropsychological tests. We implemented a computational model by Bishara et al. (Journal of Mathematical Psychology, 54(1), 5–13, 2010) that allows for evaluating the role of feedback-based attentional learning. We investigated if the model differentiates between Parkinson’s disease (PD) patients on and off dopaminergic medication. We reanalyzed data from 32 patients with idiopathic PD and 35 matched healthy controls, which completed a computerized version of the WCST. The PD sample was divided into patients tested on (n = 18) and off (n = 14) dopaminergic medication. Model performance was assessed via posterior probabilities and simulations of WCST error scores. Individual model parameters were used for group comparisons. The best performing model configuration showed a single learning rate parameter for positive and negative feedback and recovered the observed WCST perseveration error and set-loss error rates. The occurrence of inference errors could not be accounted for by that model configuration. We did not observe evidence for differences in model parameters between the examined groups of individuals. Our results indicate that assuming distinct, feedback-specific learning processes are not superior to feedback-type unspecific processes in accounting for performance on the computerized WCST. Specifically, the studied model successfully accounted for some, but not all, aspects of test performance. The model parameters did not differentiate between PD patients on and off their dopaminergic medication, a finding that we discuss in the context of study design.