We compared the cortisol and melatonin circadian and ultradian rhythms in normal men using two approaches: 1) the men were exposed successively to two conditions, one normal and a second chosen to alter differently each of the hormones, i.e. complete bedrest for 34 h (supine, fasting, and under dim light), and 2) analyses of the rhythms using a combination of curve smoothing for the description of the 24-h rhythm, and peak detection and spectral analysis for the measurement of periodic phenomena. Blood was sampled every 30 min from 0700-0700 h. A diurnal rhythm was detected for both hormones, with different underlying frequencies. Plasma cortisol had an ultradian rhythm of 8 h. From 0000-0800 h (night) and 0830-1600 h (early day), the pulsatile activity and baseline values of cortisol were high, while from 1630-2400 h (late day), these variables were low. During complete bedrest, pulsatile activity and baseline values were even higher during the night period, and the nocturnal peak of cortisol, usually present between 0300-1000 h, was split in two, with an early peak at 0000-0400 h. There were two specific events during the day associated with synchronous, high amplitude pulses: awakening and eating at noon. No such pulses occurred at suppertime or when the men fasted. Melatonin secretion was organized around a 5.5-h period. In the rest condition, plasma melatonin values were higher during the night. The 24-h rhythms of cortisol and melatonin were temporally related. Plasma melatonin began to rise when plasma cortisol was at its lowest, it peaked when cortisol began to rise, and it began to decrease when cortisol reached its peak, with a 5-h phase delay between plasma cortisol and melatonin rise at night. In summary, melatonin and cortisol rhythms have different ultradian frequencies, suggesting an intrinsic difference in the mechanisms controlling their secretion. In addition, their responses to restricted physical activity in an environment with dim light were completely different; for plasma melatonin, the change was primarily quantitative, with an increase in total production especially at night, while for plasma cortisol, there was more of a qualitative change, with different patterns of pulsatile activity and possible splitting of the nocturnal peak. The differences in the ultradian organization of these two hormones imply that the correlation between their peaks must depend on a third factor, which is likely to be the 24-h organization of the day.