The CIDI-SAM is a fully-structured interview that ascertains DSM-III, DSM-III-R, Feighner, RDC and ICD-10 diagnoses for alcohol, tobacco and nine classes of psychoactive drugs. It was designed at the request of the WHO/ADAMHA Task Force on Psychiatric Assessment instruments to expand the substance abuse sections of the CIDI. Using a test-retest design, the diagnostic and item reliabilities of this instrument were tested in a sample of 39 patients in substance abuse treatment at three St. Louis treatment facilities. Kappa values and their 95% confidence intervals, and Yule's Y values are reported. The average kappa for DSM-III substance disorders was 0.84, for DSM-III-R it was 0.82. We report, on the average, excellent kappa values for individual alcohol and drug symptoms. We also ascertained from the respondent's themselves the reasons why answers to specific questions might have been discordant. The findings from this unique 'discrepancy interview' are reported.