Computerized self-report sleepiness scales were administered before and after cognitive testing in a 72-hour sleep deprivation study. The cognitive test battery was administered every 2 hours and took approximately 1-1.25 hours. Two computerized measures of subjective sleepiness were used, one a visual analog scale, the other a Hebrew version of the Stanford sleepiness scale. The results indicated that both accumulated sleep loss, circadian and ultradian (2 cycles/day) factors were significant in determining subjective estimates of sleepiness. The extent of the differences between subjective sleepiness ratings before cognitive testing and after testing was dependent upon the phase of the circadian cycle. These differences were greatest at approximately 0200-0600 hours and least around 1000 hours. A second low point occurred at 1800-2000 hours. Analysis by complex demodulation of the individual subjects' sleepiness rating curves indicated that the amount of variance accounted for by the circadian component increased significantly after cognitive testing.