Correlates of outcome were studied in a sample of 142 unmedicated patients with major depressive disorder treated with cognitive behavior therapy (CBT) in outpatient (n = 110) or inpatient (n = 32) protocols. Outpatients received up to 20 sessions of therapy over 16 weeks; inpatients received up to 20 sessions over 4 weeks. Across all three protocols, nonresponse was associated with unemployment, higher levels of pretreatment severity, and an abnormal electroencephalographic (EEG) sleep profile. Chronicity was associated with poor outcomes in male outpatients, whereas high scores on a measure of dysfunctional attitudes were associated with a trend for poor outcome only in female patients. Among inpatients, male gender, diagnostic comorbidity, and elevated urinary free cortisol levels also were associated with poor outcome. Although a majority of the unmedicated patients experienced full or partial remissions, our findings suggest that assessments of both clinical (e.g., severity, chronicity, and comorbidity) and psychobiological (EEG sleep and adrenocortical function) factors may identify those depressed patients who are less likely to benefit from CBT alone. Such patients may be preferentially responsive to treatment with pharmacotherapy, either alone or, in more complicated cases, in combination with psychotherapy.