The implications for addiction research of recent knowledge about human memory are described. It is important that research using self-reported data understands the limits of such data. The nature of human memory and the selective, constructive processes of remembering provide one set of limits. Abandoning retrospective data entirely is not feasible in addiction research, for it would require the abandonment of current and prospective self-reported data as well, as they are also subject to memory biases. Because of memory distortions, self-reports, even by rigorous questionnaire, are biased narratives rather than incomplete but otherwise accurate evocations of past events. These limits necessitate caution and humility in the interpretation of findings, and cannot be eliminated by any particular set of research methods. There will never be a philosophers' stone which will convert self-reported data into absolutely accurate figures of quantity, frequency and timing. Nor is it straightforward to infer social and psychological causality from the organization and timing of events as remembered.