Our research over the past 5 years has addressed two interrelated aims: 1) to characterize the psychobiologic correlates of response to cognitive behavior therapy (CBT); and 2) to clarify the state dependence (i.e., reversibility) or state independence (i.e., irreversibility) of selected neurobiologic disturbances across the course of the illness. We have found that clinical severity, an abnormal EEG sleep profile, and increased hypothalamic-pituitary-adrenocortical (HPA) activity are associated with poorer response to CBT. Further, remitted patients showed only partial normalization of these presumably state-dependent abnormalities. These findings suggest that some unipolar depressions are characterized by a level of psychobiological disturbance that requires somatic treatment in order to correct or control dysregulated CNS processes. We now propose to extend this work in a controlled study of 150 depressed outpatients randomly assigned to CBT or double-blind treatment with either fluoxetine (FLX) or placebo (PBO). Patients will be assessed with a comprehensive psychobiological battery after 14-days of single-blind PBO "washout" (T1). Only persistently depressed patients manifesting a prespecified level of EEG sleep disturbance will enter the controlled trial. Psychobiological studies will be repeated after 3 months of acute-phase therapy (T2); treatment responders will receive 6 months of continuation therapy. We predict (Hypothesis 1) that there will be a gradient of treatment efficacy, such that FLX > CBT > PBO. Further, we predict (Hypothesis 1a) that response to CBT and PBO, but not FLX, will be inversely related to measures of HPA activity. We also predict (Hypothesis 2) that the FLX group will show significantly greater changes in selected psychobiologic abnormalities at T2 when compared to both CBT and PBO. In Hypothesis 3, we propose that these neurobiologic disturbances are linked to cognitive/phenomenological changes (i.e., impaired executive cortical function, decreased working memory, blunted mood reactivity, and diminished hedonic capacity) that functionally inhibit patients' ability to use CBT. This proposed research, the first to integrate serial, hypothesis-guided psychobiologic assessments within a PBO-controlled trial of pharmacologic and psychotherapeutic treatments, has substantial conceptual and public health significance.