Structured or semi-structured interviews, including the Structured Clinical Interview for DSM-III-R (SCID), are used widely to maximize the reliability and validity of psychiatric diagnoses. Although the reliability of such interviews appears adequate, there has been little effort to evaluate their validity. In a sample of 100 substance abuse patients, we evaluated the concurrent, discriminant and predictive validity of SCID substance use diagnoses, as well as co-morbid disorders that occur commonly among these patients. The validity of current and life-time substance use diagnoses obtained by a research technician using the SCID was good; it was moderate for antisocial personality disorder and major depression and poor for anxiety disorders. Although accurate diagnosis of substance use disorders in substance abuse patients can be accomplished by a research technician, the diagnosis of co-morbid psychiatric disorders requires either additional expertise or the use of a diagnostic instrument specially designed for that purpose.