Accumulating evidence suggests that exercise may have both rapid and delayed effects on human melatonin secretion. Indeed, exercise may acutely (i.e., within minutes) alter melatonin levels and result in a shift of the onset of nocturnal melatonin 12 to 24 h later. The presence and nature of both acute and delayed effects appear to be dependent on the timing of exercise. The presence of a detectable acute effect also depends on the duration, intensity, and type of exercise. Late evening exercise during the rising phase of melatonin secretion may blunt melatonin levels. High-intensity exercise during the nighttime period, when melatonin levels already are elevated, consistently results in a further (nearly 50%) elevation of melatonin levels. No effect of low-intensity exercise performed at the same circadian phase could be detected. Irrespective of intensity, exercise near the offset of melatonin secretion or during the daytime has no consistent acute effect on melatonin secretion. Nighttime exercise, whether of moderate or high intensity, results in phase delays of the melatonin onset on the next evening. In support of the concept that a shift of the melatonin onset on the day after nighttime exercise represents a shift of intrinsic circadian timing is the observation that similar phase shifts (in both direction and magnitude) may be observed simultaneously for the onset of the circadian elevation of thyrotropin secretion. The observation of exercise-induced phase shifts of the onset of melatonin secretion is, therefore, interpreted as evidence that, in humans as in rodents, increased physical activity during the habitual rest period is capable of altering circadian clock function.