The sleep of thirty elderly patients with recurrent unipolar depression was examined at baseline (before acute treatment of the index episode) and again in a state of symptomatic remission with nortriptyline (mean steady-state level: 82.1 ng/ml). Continuation therapy with nortriptyline was associated with improvement of polysomnographic sleep maintenance (mainly in the third and fourth sleep cycles, to a level similar to that of controls), prolongation of rapid-eye-movement (REM) sleep latency (exceeding that of controls), and potentiation of slow-wave activity during the first non-REM (NREM) sleep period. Clinical improvement, as measured by the Hamilton Depression Rating Scale, was significantly associated with shift of delta activity toward sleep onset (p less than 0.002), prolongation of REM sleep latency (p less than 0.0001), and improvement in sleep maintenance (p less than 0.0002). Multiple regression analysis showed that the single best correlate of clinical change was prolongation of REM sleep latency (i.e., prolongation of first NREM period). Perceived sleep quality improved significantly during early continuation therapy with nortriptyline, but not to the level reported by a group of 30 age- and sex-matched healthy controls. The findings are consistent with the concept that anti-depressant drug efficacy may depend upon strengthening of the homeostatic regulation of sleep and upon changes in the REM-sleep regulation.