Earlier investigations have suggested that variables derived from quantified electroencephalographic (EEG) sleep analysis might predict good clinical response in an early phase of antidepressant treatment. In this report we evaluate the predictive value of all-night sleep EEG spectral analysis during the washout period before treatment. We compared the spectral EEG sleep profiles of major depressed inpatients divided into two groups according to an improvement > or = 50% on the Hamilton Rating Scale for Depression. Findings in this population demonstrate the presence of specific characteristics of the responder group compared with the nonresponder group. Delta band relative power was increased in the former group, while theta, alpha, and beta relative power were decreased. All the bands showed decrease in absolute power in the responder group. These results can be interpreted as enhanced sleep intensity in the responder group. All-night sleep EEG spectral variables are valid baseline markers of the functional differences between treatment responders and nonresponders and thus might permit prediction of clinical outcome.