Although structured diagnostic interviews are increasingly being used in substance abuse treatment settings, there has been limited systematic evaluation of their ability to enhance reliability and validity of psychiatric diagnoses. The present report provides data on the concurrent, discriminant, and predictive validity of current substance use disorders and common comorbid diagnoses in a sample of 100 substance abuse patients. Diagnoses formulated primarily by master's-level clinicians in the usual course of their duties were compared with diagnoses formulated by research technicians using a semistructured interview. Results indicated that the validity of clinician diagnoses was good for substance use disorders, moderate for personality disorders, and poor for anxiety disorders and major depression. Greater validity was observed for substance abuse diagnoses formulated by research technicians using the semistructured interview. Based on these findings, we conclude that psychiatric diagnosis in substance abuse patients may be improved by adding elements of structured interviews to the clinician's usual assessment.