Results are reported from a neuropsychological investigation of a large group of schizophrenic patients (n = 141) and normal control subjects (n = 59) who performed the Wisconsin Card Sorting Test (WCST). The schizophrenic patients were divided into two DSM-III-R diagnostic subgroups: paranoid and nonparanoid schizophrenia. The performance of the WCST by schizophrenic patients was poor compared with that by normal control subjects; furthermore, paranoid patients made a higher number of perseverative errors than did nonparanoid patients. The deficit in WCST performance, which has been widely reported to characterize patients with schizophrenia, appears to be related to the clinical profile and neuropathological differences that contribute to the heterogeneity of the disorder.