In contrast to previous observations, which had indicated an increased prevalence of depressive symptoms among the elderly, recent epidemiological studies have shown a decreasing current and lifetime prevalence after age 20 or 30. The question whether depression is more or less common among older adults can best be resolved by turning to some of the core tasks of epidemiology: the identification of cases of depression, the distribution of cases in the population, and historical trends in the incidence of cases. In case identification there is agreement among clinicians on the signs and symptoms of depression, but controversy regarding depressive syndromes. The latter may be seen as forming a continuum (with little change in prevalence across the life cycle) or as including differing categories (with some, such as depressive reactions to physical illness, being more significant among the elderly). DSM-III and DSM-III-R do not capture all of the important subtypes of depression. Scepticism has been aroused over the recent finding of case distribution showing lower prevalence among the elderly because the lifetime prevalence has also been reported as lower. It is suggested here that a "cohort effect" may account for this finding. Recent data bearing upon historical trends suggest that the current 65- to 85-year-old group has been protected against major depression and suicide. While suicide rates do increase with advanced age in white males, current older men have lower suicide rates than previous cohorts at the same age (60 years). This may reflect better health and economic status of the current group or, perhaps, an increase of depression among younger adults now. Better case definition of depression and further epidemiological studies including consideration of cohort effects are indicated.