In order to determine whether a decrease in the amplitude of heart rate circadian rhythm (HRCR) could represent a chronobiological marker of depression and to answer the question of the possible role of this alteration in the reduction of rapid eye movement sleep latency (REM latency) observed in depression, 22 major depressed patients (DSM III R, Montgomery Asberg Depression Rating Scale) and 11 healthy volunteers entered this study, which included the simultaneous recordings of sleep and heart rate (HR) during two consecutive 24-h periods. After a smoothing procedure, the HR data, obtained from a portable device providing mean HR/min, were computed with the single cosinor method for the classical circadian parameters. We also determined the cosinor fitting index (CFI = percentage to fit). When studied as a single group and compared to the control group, the depressed patients, who had difficulties in sleep initiation and whose REM latencies were within normal limits, were characterized by a dampening of the amplitude HRCR and of the day-night HR difference. This was linked to a lesser increase of HR during the day and a reduced decrease at night. No significant phase modification of HRCR was observed in this group of depressives. In a second step, the depressives were divided into a depressed arrhythmic group (DAG, CFI < 50%, 11 patients) and a depressed rhythmic group (DRG, CFI > 50%, 11 patients) who did not differ on MADRS scores. Even though the amplitude of HRCR was reduced in the DAG only, with HR higher at night and lower during the daytime than for controls, both DAG and DRG significantly differed from controls on those items related to sleep induction and maintenance. In the DAG only was there a tendency toward a reduction of total sleep time and REM sleep time while REM latency was normal. This finding suggests that sleep problems are not responsible for the differences in HRCR parameters between the two depressed groups. These differences in HR and sleep parameters between two groups of similarly depressed patients also suggest that a short REM latency does not systematically imply a disruption of all circadian clocks.