The psychometric properties of a clinician-administered, DSM-IV-based, structured clinical interview for pathological gambling (SCI-PG) were examined. Seventy-two consecutive subjects requesting treatment for gambling problems were administered the SCI-PG. Reliability and validity were determined. Classification accuracy was examined using longitudinal course of illness. The SCI-PG demonstrated excellent inter-rater and test-retest reliability. Concurrent validity was observed with the South Oaks Gambling Screen (SOGS). Discriminant validity was observed with measures of anxiety and depression. The SCI-PG demonstrated both high sensitivity and specificity based on longitudinal assessment. The SCI-PG demonstrated excellent reliability and validity in diagnosing PG in subjects presenting with gambling problems. These findings require replication in other groups to examine their generalizability.