Male subjects had their sleep schedules shifted 12 h to accommodate eight or more consecutive simulated night shifts. They were exposed to artificial light (approximately 5,000 lux, 3- to 6-h durations) each night, slept at home in very dark bedrooms during the day, and wore dark welders goggles whenever they went outside during daylight. Body temperature was continuously measured, and daily questionnaires provided estimates of sleep time and mood. The circadian temperature rhythm phase shifted, usually approximately 2 h/day, in 21 of 24 subjects. The temperature rhythms of the other three subjects appeared to remain entrained to the 24-h zeitgebers. In the subjects whose temperature rhythms shifted, the direction of shift (advance or delay) depended on the timing of light exposure relative to the baseline temperature phase, consistent with a phase-response curve. Sleep was disturbed, and fatigue was increased during the first few days after the shift of the sleep-wake schedule.