Wisconsin Card Sorting Test (WCST) performance was investigated in patients with complex partial seizures (CPS) of dominant (n = 16) or nondominant (n = 19) temporal-lobe origin as well as in an epilepsy control group (n = 6) which consisted largely of patients with primary generalized epilepsies. Fifty seven percent of the CPS group (39% of dominant and 74% of nondominant temporal-lobe patients) performed in a manner suggestive of frontal-lobe pathology compared to 17% of the epilepsy controls. There were no differences among the groups in the number of categories achieved, but there were substantial qualitative differences in problem-solving efficiency. Nondominant temporal patients manifested more total errors and perseverative errors relative to both dominant temporal and epilepsy controls, and more perseverative responses relative to epilepsy controls. Dominant temporal-lobe patients showed more perseverative errors than epilepsy controls. Finally, a consecutive series of patients who underwent partial temporal-lobe resection were examined 6 months post-surgically and they were found to manifest significantly fewer perseverative responses. The possible reasons for "frontal-like" performance in CPS patients with documented temporal-lobe pathology are discussed and a specific hypothesis is suggested.