Is sleep per se a zeitgeber in humans?

It is not clear whether shifting of sleep per se, without a concomitant change in the light-dark cycle, can induce a phase shift of the human circadian pacemaker. Two 9-day protocols (crossover, counterbalanced order) were completed by 4 men and 6 women (20-34 years) after adherence to a 2330 to 0800 h sleep episode at home for 2 weeks. Following a modified baseline constant routine (CR) protocol on day 2, they remained under continuous near-darkness (< 0.2 lux, including sleep) for 6 days. Four isocaloric meals were equally distributed during scheduled wakefulness, and their timing was held constant. Subjects remained supine in bed from 2100 to 0800 h on all days; sleep was fixed from 2330 to 0800 h in the control condition and was gradually advanced 20 min per day during the sleep advance condition until a 2-h difference had been attained. On day 9, a 25 to 27 h CR protocol (similar to 0.1 lux) was carried out. Phase markers were the evening decline time of the core body temperature (CBT) rhythm and salivary melatonin onset Q pg/ml threshold). In the fixed sleep condition, the phase drift over 7 days ranged from +1.62 to -2.56 h (for both CBT and melatonin rhythms, which drifted in parallel). The drifts were consistently advanced in the sleep advance schedule by +0.66 +/- 0.23 (SEM) h for CBT (p = 0.02) and by 0.27 +/- 0.14 h for melatonin rhythms (p = 0.09). However, this advance was small to medium according to effect size. Sleep per se may feed back onto the circadian pacemaker, but it appears to be a weak zeitgeber in humans.