Two Epstein-type overinclusion tasks and two Password-type editing tasks were given to 80 schizophrenics and 20 normal controls. For each of the two types of test, one test was in a free-response format and one was in a choice format. Test format did not prove to be an important variable in accounting for group differences. Schizophrenics rated as severely disordered and delusional showed a nonspecific deficit on all four tests. Less disordered and nondelusional subjects performed in the normal range on both overinclusion tests but showed a deficit on both editing tasks. This suggests that although overinclusion is a correlate of degree of pathology, role-taking deficit is a correlate of the schizophrenia-nonschizophrenia dichotomy.