The antidepressant response to partial sleep deprivation early in the night (PSD-E) was compared with the response to partial sleep deprivation late in the night (PSD-L) in 16 drug-free depressed inpatients using a balanced order crossover design. PSD-L had a significantly greater antidepressant effect that PSD-E. The response to PSD-L was sustained and enhanced by a second night of treatment. Patients had significantly shorter sleep durations and reduced REM sleep on PSD-L that did not occur in the PSD-E situation. There was a significant negative correlation between response to PSD and sleep duration, and in particular, REM sleep duration, in the late sleep deprivation situation. Thus, the amount and timing of sleep appear to be factors in the response to PSD, but additional studies are needed to evaluate the relative importance of these parameters.