One set of analyses examines the quality of the longitudinal retrospective self-report from narcotics addicts, including validity of recent narcotics use, reliability of various measures, stability of relationships among these measures, and pattern reliability among latent constructs, Results contribute strongly to confidence in the validity of the relationships among these data derived from addicts' self-report. The second set of analyses focuses on validity of self-reported drug use among high-risk groups, including samples from sexually transmitted disease clinics, hospital emergency rooms, and jails. Results suggest that the accuracy of self-report of recent drug use varies by the sample sources, drug types, and subject characteristics. Targeting these high-risk groups may improve prevalence estimation. Empirical validation of self-report is always necessary to enhance the utility of self-report data and to provide means of controlling for potential biases. Tables, references