Research on the Diagnostic Interview Schedule, the chief instrument in contemporary studies in psychiatric epidemiology, has supported its utility in enabling lay interviewers to reproduce psychiatric interviews within an acceptable margin of error. Nonetheless, we propose that the Diagnostic Interview Schedule commits itself to dubious assumptions regarding the accuracy of human memory, shared by other history-taking efforts, by relying on retrospective reports of lifetime DSM-III symptoms and episodic dating of symptom spells. For more than a century, the fallibility of human memory has been the topic of intensive experimental and naturalistic study, a history which is relevant to the construction of instruments like the Diagnostic Interview Schedule. The continuing use of retrospective lifetime symptom reports suggests that this literature has been largely ignored in the development and administration of the Diagnostic Interview Schedule. Prospectively organized research is needed to disclose the limits of human memory for recent psychiatric events and the mediating conditions under which memory for such events can be accurately retrieved and improved.