Thirty-four endogenous and 18 reactive, depressed patients (hospitalized and nonschizophrenic) were treated in a double-blind, crossover study of the hypothesis that rapid eye movement (REM) sleep reduction (by awakenings) relieves depression. In the endogenous group-but not in the reactive group-subjects deprived of REM sleep for three weeks improved significantly more than control subjects awakened from non-REM sleep. Therapeutic efficacy of REM sleep reduction appeared similar to reported efficacy of imipramine hydrochloride treatment of depression. Eight of nine endogenous patients, unimproved by REM sleep deprivation, did not improve with imipramine. Results suggested (1) that substantial REM sleep reduction has antidepressant activity, and (2) since imipramine and other drug antidepressants reduce REM sleep much more so than nonantidepressant drugs, that an antidepressant "mechanism" of drugs resides in their capacity to substantially reduce REM sleep.