We compared sleep variables in 14 drug-free endogenous depressives and in 14 age- and insomnia-matched, nondepressed controls before and after brief rapid eye movement (REM) sleep deprivation by awakenings. Before REM sleep deprivation, compared with controls, depressives had lower REM latency, higher REM frequency, and--a new finding--an abnormal temporal distribution of REM sleep. Depression improvement by REM sleep deprivation correlated with the ameliorative effect of brief REM sleep deprivation on on indicator of the abnormal temporal distribution of REM sleep. Several findings suggest that the depressive abnormalities represent a "damaged," weakened sleep cycle "oscillator" and its correlate, a circadian rhythm disturbance, and that REM sleep deprivation improved depression to the extent that it stimulated the oscillator and corrected one manifestation of the circadian rhythm disturbance.