Sleep disturbances and the failure to entrain circadian rhythms to the 24-hour day have been reported in the majority of totally blind subjects. The present case study of a totally blind man with a well-documented recurring sleep disturbance was designed to investigate the mutual relationship between sleep and the circadian timing system. The 63-year-old subject, a high school teacher with a regular work schedule, had suffered from cyclically recurring insomnia for the past 28 years. Analysis of a sleep log that he had kept for the past 15 years suggested that his circadian rhythms were not entrained to the 24-hour day. During a 3-month inpatient study, the period of the endogenous circadian pacemaker was assessed by analysis of ambulatory core body temperature, urinary excretion and a series of estimates of the phase of core body temperature cycles and plasma cortisol levels during constant routines. All circadian markers revealed periods in the range of 24.22-24.27 hours, with no evidence for a modulation of the observed periods by the sleep-wake cycle. During this 3-month inpatient study, a complete cycle of the subject's sleep disturbance and remission was polysomnographically documented while the subject lived on a regular 24-hour schedule. Because the subject's circadian rhythms were free-running and his sleep times were fixed, sleep occurred at virtually all circadian phases. Analysis of sleep latency, REM sleep latency, sleep duration, wake in sleep episode and REM sleep during sleep episode revealed a strong modulation by circadian phase. These findings in this blind man suggest that: 1) the period of his cyclically recurring sleep disturbance is directly related to the nonentrained period of an endogenous circadian pacemaker that drives circadian variation in core body temperature, urinary excretion, plasma cortisol and sleep propensity; 2) both his sleep structure and the severity of his daily sleep disruption are directly related to circadian phase and 3) his circadian pacemaker, which has an endogenous period that deviates only 0.2-0.3 hours from 24 hours, cannot be entrained by periodic daily exposure to nonphotic time cues, including a very regular 24-hour sleep-wake schedule.