Future sleep research in affective illness will probably continue the current evolution beyond cross-sectional to longitudinal studies, and beyond a largely descriptive emphasis to the testing of specific hypotheses and predictions derived from models of the pathophysiology of depression. These models are and will be variously neurochemical, chronobiological, genetic, and developmental in nature. Adequate testing of these models and predictions from them will require the use of pharmacologic and naturalistic probes and the use of sophisticated CNS imaging techniques. These probes will help further characterize the physiology of depression under conditions of disequilibrium or perturbation, such as following sleep deprivation, REM deprivation, phase advancement of the major sleep period, or the administration of antidepressant drugs with specific monoaminergic activity. Concurrently, if one is to understand further whether the sleep abnormalities of depression are part of a larger circadian rhythm disturbance, investigations will necessarily include 24-h measures of sleep-wake activity, psychomotor activity, and probably core body temperature rhythm under constant routine conditions. A complementary point of view would suggest that more intensive investigative efforts be focused on the first 100 min of sleep at night, since it is the first NREM-REM cycle that seems to show the greatest and most specific deviation in depressed patients from normal controls. Efforts to characterize further this part of the 24-h cycle, with respect to age- and gender-related variance as well as responses to physiologic, hormonal, pharmacologic, and naturalistic probes, are strongly warranted.