OBJECTIVE: The purpose of this study was to investigate the reliability of a new semistructured diagnostic interview, the Psychiatric Research Interview for Substance and Mental Disorders (PRISM), for substance- abusing patients. The reliability of psychiatric diagnoses for individuals who drink heavily or use drugs has been shown to be problematic. The PRISM was designed to improve the reliability for such individuals. METHOD: A test-retest reliability study of the PRISM was conducted with 172 patients being treated in dual-diagnosis or substance abuse settings. RESULTS: Good to excellent reliability was shown for many diagnoses, including affective disorders, substance use disorders, eating disorders, some anxiety disorders, and psychotic symptoms. The interview has recently been updated for DSM-IV diagnoses. CONCLUSIONS: The PRISM offers a method of producing psychiatric diagnoses with improved reliability for patients and other research subjects who have problems with alcohol or drugs.