Age of onset of depressive episodes may serve as a useful marker of pathogenetic heterogeneity in late-life depression. Medical illness may play an important role in the pathogenesis of depression in the elderly, but its relationship to age of onset has not been carefully examined. We prospectively studied 110 older inpatients with DSM-III-R major depression. Using multiple regression techniques, we found that medical illness was not independently associated with age of onset. Independent predictors of older age of onset were age, male sex, absence of substance abuse history, and absence of melancholia. Our discussion reconsiders the usefulness of age of onset as a primary research variable for elucidating heterogeneity of late-life depression.