The reliability and validity of self-report data regarding substance abuse has often been questioned. To determine how best to enhance the veracity of self-report, three factors which might affect self-report veracity were examined: alcohol status at time of interview; level of cognitive functioning; and method of self-report data collection. Subjects were 234 admissions to an inpatient substance abuse treatment unit. Self-report data were collected via both personal interview on the day of admission and questionnaire within the first week of stay. Self-reports concerned use of alcohol, cocaine, and marijuana in the days preceding admission. Test–retest reliability for the questionnaire data produced reliability coefficients of 0.88, 0.91, and 0.88, for alcohol, cocaine, and marijuana, respectively. Variation in inter-test interval had virtually no effect upon reliability coefficients. Interview data were compared to toxicologic analyses of blood and urine samples collected on admission. Overall, this comparison showed self-reports to be valid, with a 97% agreement between verbal report and laboratory data for alcohol, 93% for cocaine, and 84% for marijuana. The comparison of interview data with questionnaire responses also showed self-reports to be valid: 90% agreement for alcohol, 93% for cocaine, and 81% for marijuana. Level of cognitive function did not influence the validity of self-reports for any of the three substances. Recent consumption of alcohol also had no statistically significant effect on the validity of self-reported marijuana use, regardless of the operational form of validity tested. However, BAC-negative subjects produced a significantly greater validity coefficient for self-reported cocaine use (kappa = 0.87) than did BAC-positive patients (kappa = 0.43), when interview data were compared with toxicologic measures. A similar finding was not uncovered when interview and questionnaire data were compared. An interaction between admission alcohol status and cognitive function was uncovered for cocaine self-reports when interview data was compared with toxicologic measures. The rate of agreement for alcohol-negative subjects is quite high for both cognitively impaired and unimpaired subjects (M = 93% and M = 94%, respectively) as well as for alcohol-positive, cognitively unimpaired subjects (M = 94%), but not for alcohol–positive, cognitively impaired subjects (M = 67%). Results are discussed in terms of threats to the validity of self-report and strategies for the optimization of response accuracy.