BACKGROUND Limited evidence suggests that polysomnographic alterations may be more prominent early in a depressive episode. Whether the effects of episode duration extend beyond middle age and appear in late-life depression as well has important implications for treatment decisions and for understanding depressive illness across the life span. Furthermore, the impact of episode duration on sleep has not been examined in the context of other factors related to clinical history and psychosocial status. METHODS Eighty-three persons aged 60 years or older with recurrent depression were studied: 34 had been depressed for 2 to 16 weeks and 49 for longer periods. An age- and gender-matched group of 48 persons with no history of major depression served as controls. Initial univariate analyses examined duration effects on electroencephalographic (EEG) sleep measurements. Multivariate analyses considered the combined effects of episode duration, clinical variables, and psychosocial variables on EEG sleep profile. RESULTS Episode duration was strongly associated with sleep continuity, architecture, and rapid eye movement: subjects who were earlier in their depressive episodes had their sleep impaired more than those later in their episodes, who, in turn, were more impaired than controls. Moreover, clinical characteristics of subjects' depressive illness, demographic variables, and psychosocial stressors and supports had unique effects on the EEG sleep profile. CONCLUSION Episode duration appears to be a potent factor to consider when evaluating sleep during depression. The additional contribution of clinical and psychosocial characteristics to the prediction of the EEG sleep profile demonstrates the importance of incorporating these variables into models of the psychobiologic characteristics of depression. The results are relevant to the timing and focus of therapeutic interventions.