Schizophrenic patients were carefully diagnosed and screened for a history of neurological disorders. Diagnosis and subtyping was based on DSM-III-R criteria, using the Structured Clinical Interview for DSM-III-R, which was administered by trained interviewers and confirmed by a research psychiatrist. The schizophrenic patients were compared with an age-matched control group on an extensive battery of neuropsychological measures. The undifferentiated/disorganized schizophrenic patients were consistently the most impaired on a broad range of tasks. When the effect of symptom severity and drug level were statistically controlled (analysis of covariance), however, the magnitude and number of differences were substantially reduced. The perseverative error score from the Wisconsin Card Sort Test showed the greatest difference between the groups. However, the strongest and most consistent effects were observed in relation to symptom ratings. These data indicate the importance of controlling for medication and symptom severity, and suggest that current diagnostic classifications may not be the most useful factors for studies of the cognitive correlates of schizophrenia.