Section I of the essay briefly discusses attempts to estimate drug prevalence in earlier generations and introduces important epidemiological concepts. Section II discusses the two major self-report surveys, the NHSDA and the HSSS; presents current findings; and considers criticisms of both surveys. Sections III and IV discuss the DAWN and DUF systems. Section V examines recent attempts to integrate these findings by using synthetic estimation techniques. Synthetic estimation attempts to correct for biases in self-report surveys and offers plausible ways to compensate for errors in data sources and to integrate data drawn from various populations and obtained by distinct methodologies. The essay concludes that the collection of drug prevalence data can be improved by advancing the comparability of surveys between different populations, by making data sets publicly available for secondary analysis, and by supporting such activity both logistically and financially. Prevalence estimators should continue to develop rigorous estimation procedures, such as synthetic estimation, so as to integrate data from disparate sources. Current data, although imperfect, represent a net asset for the study of drug prevalence and related issues and can usefully inform policymakers, even if that information must be treated cautiously. 5 tables, 17 figures, and 70 references