Limited information is available on the reliability of diagnostic assessments in community populations. This study analyzed the 18-month test-retest stability of lifetime major depression determined from the Schedule for Affective Disorders and Schizophrenia-Lifetime Version using the Research Diagnostic Criteria. Overall, the reliability among the 391 female subjects was poor. Clinical status during the 18-month interval influenced reliability, while demographic, psychosocial, and interviewer characteristics were unrelated. The women who reliably reported lifetime episodes of depression were consistent about details such as medication use, but were inconsistent about other features, eg, number of episodes, length of longest episode, and age at first episode. The results suggest the need for caution in analyzing data on the lifetime prevalence of depression in community samples.