The "long nights" protocol was designed to evaluate sleep processes and circadian rhythm parameters in young humans. A total of 19 children (10 boys, ages 11.2 to 14.1 years [mean = 12.7 +/- 1.0], and 9 girls, ages 12.2 to 14.4 years [mean = 13.1 +/- 0.7]) took part in the study. Sleep/wake initially was assessed at home using actigraphy and diary for 1 week on each child's self-selected schedule followed by an 8-night fixed light-dark (LD) condition, while sleeping from 22:00 to 08:00 h and wearing an eye mask to exclude as much light as possible. Phase measurements included 4-night mean actigraphically estimated sleep onset and offset as well as 1-night dim light salivary melatonin onset (DLSMO) phase at the end of each condition. Subjects then lived in the laboratory for 6 consecutive cycles: Day 1 LD = 14:10 h, lights out 22:00 to 08:00 h; Days 2-4 LD = 6:18 h, lights out 18:00 to 12:00 h; Days 5-6 = constant routine in continuous dim light (about 20 lux); Night 6 = 14 h recovery sleep. Phase markers (sleep onset, sleep offset, DLSMO) were significantly less dispersed after the fixed LD as compared to the self-selected condition, indicating efficacy of the LD protocol. Phase markers were correlated at the self-selected assessment (sleep onset vs. sleep offset r = .72; DLSMO vs. sleep onset r = .82; DLSMO vs. sleep offset r = .76) but not on the fixed schedule, probably due to restricted range. The constant routine provided additional phase markers, melatonin offset and midphase. Offset phase of melatonin secretion was significantly correlated with age (r = .62) and Tanner stage (r = .62). In conclusion, these preliminary data indicate a relationship between adolescent development and circadian phase. Thus, the long nights protocol is a feasible way in which to assess circadian parameters in young humans as well as to examine intrinsic sleep processes.