Total sleep deprivation (TSD) leads to an immediate amelioration of depressed mood in approximately 70 % of patients with the melancholic subtype of depression. The clinical utility of this procedure is limited, as the improvement usually subsides after the next night of sleep. In the present study, 40 depressed inpatients, being free of psychoactive medication for at least 7 days and who had responded to a TSD were then distributed (according to a matched-pair design) to a sleep phase advance (SPA = time in bed scheduled from 1700-2400 hrs) or a sleep phase delay (SPD = time in bed from 0200-0700 hrs) with a succeeding shift back (for one hour in the SPA group per day) respectively shift forward (for 30 minutes in the SPD group per day), until the initial sleep phase (2300-0600 hrs) was reached after seven days again. Based on previous observations it was hypothesized that a phase advance of the sleep period should prevent responders to TSD from relapsing. Whereas 75% of the TSD responders were stabilized by the phase advanced condition and did not relapse over a period of seven days, only 40% of the patients in the phase delayed condition did not relapse. Polysomnography during the course of the study gave no evidence that the unusual sleep schedules caused prolonged sleep deprivation. Abnormalities of REM sleep persisted both in the clinical responders and non-responders after the sleep wake manipulation. It is concluded that the clinical effectiveness of TSD can be significantly improved by combining TSD with a following phase advance of the sleep period.