Circadian rhythms and sleep-wake cycles were measured in volunteers staying singly in temporal isolation unit where they were exposed to artificial short and long light-dark (LD) cycles for 7 days. The long day consisted of 16-h light and 8-h dark (LD 16:8) and the short day consisted of 8-h light and 16-h dark (LD 8:16). During the light period, bright light of approximately 5,000 lux was given from the ceiling and during the dark period there was no illumination. Sleep was monitored by bed sensors, wrist actiwatch, and polysomnography (PSG) on the first and last nights of the schedule. Sleep length was significantly longer under LD 8:16 than under LD 16:8 and the sleep quality estimated by PSG was worse under LD 8:16 than under LD 16:8, which were comparable to natural seasonality in sleep. The circadian rhythm in plasma melatonin was measured in dim light (10 lux) before and after the LD exposures. The nocturnal melatonin secretion (NMS) was significantly longer after LD 8:16 than after LD 16:8 due to differential phase shifts of the rising and falling phases of NMS. After LD 8:16, the falling phase was much advanced than the rising phase, whereas after LD 16:8 the rising phase was much delayed than the falling phase, resulting in the NMS compression. These results indicate that the light sensitivity in terms of phase shifting is different in the two circadian phases, supporting a dual oscillator hypothesis with different phase-response curves for light in the human circadian system.NEW & NOTEWORTHY The present study demonstrated differential light responsiveness of the rising and falling phases of nocturnal melatonin secretion in human subjects exposed to artificial long (LD 16:8) and short days (LD 8:16) and suggested the involvement of different oscillators under these phases. The findings well mimicked the seasonality of the circadian rhythms in nature and consisted with the evening/morning dual oscillator hypothesis proposed originally for nocturnal rodents, providing a new concept for the human circadian system.