Despite the importance of the issue for AIDS-related research, there is a paucity of studies that have addressed the issue of the validity of self-reports of patterns of risk behavior over time. This article frames issues of measurement in the context of cognitive psychology and presents a research paradigm that investigators can use to explore issues of accuracy, bias, and stability in self-reports. The paradigm also permits the analysis of transitory influences on risk behavior, influences which have been minimized in current AIDS research. Preliminary data are presented that support the feasibility of the method and which are suggestive of important considerations in the development of valid measures in national surveys.