OBJECTIVE: Differentiation of stable, trait-like characteristics from more episodic or state-dependent disturbances will be helpful in gaining a better understanding of the pathophysiology of depression. However, research in this area has been complicated by artifactual and clinical problems associated with pharmacologic treatment. In this investigation the authors used EEG sleep studies to assess medication- free depressed male patients before and after cognitive behavioral therapy. METHOD: Forty-five male patients with the diagnosis of major depression according to the DSM-III-R criteria and the Research Diagnostic Criteria underwent EEG sleep studies before and after 16 weeks of cognitive behavioral therapy, during which they were free of medication. In addition to the documentation of changes within these patients, the findings were compared with those for 47 healthy subjects, including 15 who were restudied 12-24 months after their baseline assessments. RESULTS: The EEG sleep profiles of the depressed patients showed a significant reduction in REM sleep density after treatment, suggesting "normalization" of an abnormal state-dependent process. By contrast, slow wave sleep and tonic REM measures, including reduced REM latency, were unchanged after treatment. CONCLUSIONS: These findings suggest that early in remission there is disaggregation of irreversible, trait-like correlates of depression (e.g., diminished slow wave sleep and reduced REM latency) from more reversible disturbances (e.g., increased REM density).