The circadian rhythm of deep body temperature consists of both an endogenous component and evoked components resulting from exogenous influences. Previous studies of the sleep-evoked effect have failed to control confounding influences, so that the effect of sleep per se has not been established. In the present study, eight good sleepers had their rectal temperatures recorded for 24 hours in each of two laboratory conditions employing a constant routine to control exogenous influences. Sleep was allowed at night in one condition. Following sleep onset, body temperature dropped more rapidly and remained lower than when wakefulness continued over the same time, resulting in a mean sleep-evoked decrease of 0.31 +/- 0.09 degree C. Fourier regression analysis showed a significant 24-hour (circadian) temperature rhythm, together with a 12-hour harmonic rhythm, in each condition. Circadian rhythm parameters were also altered by the sleep-evoked (or masking) effect, with the amplitude increased and the mean decreased when subjects slept at night in the constant routine. It was suggested that a constant routine methodology be used in studies of circadian rhythm differences and that Fourier regression be used in preference to simple cosine curve fitting to give a better approximation of the temperature rhythm.