This paper describes a series of questions designed to improve the accuracy of age-of-onset reports in the US National Comorbidity Survey (NCS) and empirically evaluates the impact of these questions on reports about age of onset of major depressive episodes. The logic underlying the series of question is traced to cognitive psychological research on autobiographical memory. Data are presented showing that the new question series yielded more substantively plausible age-of-onset reports than those obtained a decade earlier in the Epidemiologic Catchment Area (ECA) Study. The test-retest consistency of age-of-onset reports was also higher in the NCS than the ECA. Despite these improvements, considerable inconsistency in age-of-onset reports remains in the NCS test-retest data. The paper closes with a discussion of potentially promising future directions to improve retrospective age-of-onset reports in new psychiatric epidemiological surveys. Copyright © 1999 Whurr Publishers, Ltd.