It was predicted from free running and ultradian cycle studies that sleep-onset insomniacs would have endogenous circadian rhythms that were phase delayed compared to good sleepers. Thirteen sleep-onset insomniacs and nine good sleepers were selected to differ only in their sleep-onset latencies as confirmed by polysomnography. their rectal temperatures were measured over a 26-h constant routine and analyzed with best-fit Fourier curves including 24-h fundamental and 12-h harmonic components. The temperature rhythm markers of the insomniacs' rhythms were approximately 2.5 h later than the respective phases of the good sleepers. The usual bedtimes of the insomniacs fell within the "wake maintenance zone" of their delayed temperature rhythm. The good sleepers had typical bedtimes several hours after their "wake maintenance zone" and closer to their body temperature minimum. It was suggested that manipulations to phase advance the insomniacs' rhythms would reduce their sleep-onset latencies. It was also predicted that early morning insomnia results from phase advanced circadian rhythms and that sleep maintenance insomnia results from an abnormal phase relationship between the 24-h temperature rhythm and 12-h sleep-alert rhythm.