Alterations of sleep can be observed polysomnographically in approximately 90 percent of depressed patients. Most of the registered sleep abnormalities in depression also occur in other psychiatric disorders. Only some types of REM sleep alterations-- short REM latency, increase of REM density and shortening of mean latency of eye movements--were reported as more specific for affective disorders. In the present study polysomnograms of 21 medication free patients with major depressive disorder (assessed with a structured interview for DSM-III-R and Hamilton Scale) and 21 healthy controls were analysed. REM latency (LREM), REM density (RD), latencies of eye movements (LEM) and mean latency of eye movements (M-LEM) were calculated for both groups. Depressed patients (compared with healthy controls) showed increased RD (38.2% vs. 28.2%, p < 0.0001), shortened M-LEM (35.7 s vs. 48.3 s, p < 0.04) and shortening of LEM in the 1st (28.9 s vs. 48.9 s, p < 0.007) and 4th (27.0 s vs. 59.1 s, p < 0.043) REM sleep periods. LREM was not shortened significantly in depressives (78.5 min vs. 91.3 min, ns). In healthy subjects a negative correlation between M-LEM and RD was found (rho = -0.47, p < 0.03). Since in the current study depressed patients differed from healthy controls, especially concerning phasic activity during REM sleep, presented data support the essential role of REM density for the assessment of sleep in depression. As a quick and easy manner to compute measurement, M-LEM is suggested as additional parameter for the assessment of phasic activity during REM sleep.