OBJECTIVE: The authors' goals are to use scales from the MMPI hypothesized in their previous research to be correlates of liability to schizophrenia to differentiate DSM-III schizophrenia from bipolar and unipolar affective illness and to cross-validate these correlates in an independently ascertained sample of patients with Research Diagnostic Criteria (RDC) schizophrenia or affective disorder. METHOD: The criterion sample consisted of 83 patients consecutively admitted to a state-operated community mental health center. Diagnosis of schizophrenia; bipolar disorder, manic; and major depression were assigned by using DSM-III. The replication sample consisted of 60 adults with RDC diagnoses of schizophrenia, schizoaffective disorder, bipolar disorder, and unipolar disorder who were parents of children in two samples collected for a study of offspring at high risk for schizophrenia and other psychopathology. After the patients in the criterion sample were classified by logistic regression analysis, the results were used to classify patients in the replication sample. RESULTS: The MMPI indicators had adequate sensitivity, specificity, and predictive power for classifying schizophrenia, and there was a moderately high rate of diagnostic agreement between the MMPI and DSM- III. Cross-validation in the replication sample was successful. Overall, the MMPI index was an adequate inclusion and exclusion criterion not only for DSM-III-defined but also for RDC-defined schizophrenia. CONCLUSIONS: A psychometric index composed of the paranoid schizophrenia, psychoticism, and manifest hostility scales from the MMPI would be a cost-effective measure to increase diagnostic efficacy in future schizophrenia research and clinical practice.Abstract Teaser