Summary The purpose of the study was to induce in two different ways, a phase-angle difference between the circadian pacemaker and the imposed sleep–wake cycle in humans, we intended to: (i) shift the circadian pacemaker by exposure to bright light and keep the timing of the sleep–wake cycle fixed; and (ii) keep the timing of the circadian pacemaker fixed by a constant light–dark cycle and displace sleep. We monitored dim light melatonin onset (DLMO), core body temperature and sleep. DLMO was delayed significantly after 3 days of a 3-h delayed sleep-phase when compared with 3 days of sleep at a normal or 3-h advanced sleep-phase. The shifts in DLMO were not accompanied by shifts in body temperature, changes in waking-up time or by a change in the duration of the first rapid eye movement (REM) sleep episode. Three days of light exposure in the morning or evening resulted in shifts in DLMO of similar magnitude, but this was accompanied by shifts in the rhythm of body temperature, changes in waking-up time and in the duration of the first REM sleep episode. We conclude that the changes observed after light exposure reflect shifts in the circadian pacemaker. In contrast, we propose that the changes observed in DLMO after sleep displacement are not mediated by the circadian pacemaker. These results raise some doubts about the reliability of DLMO as a marker of circadian phase in cases of sleep disturbances. Finally, we initiate a search for changes in sleep that might be responsible for the unexpected effects on DLMO.