Recently, we have demonstrated that in normal men the nocturnal testosterone rise antedated the first rapid eye movement (REM) sleep episode by about 90 min and was correlated with REM latency. To further elucidate whether the diurnal testosterone rhythm is a sleep-related phenomenon or controlled by the circadian clock, we determined serum testosterone levels in 10 men during the ultrashort 7/13 sleep-wake cycle paradigm. Using this schedule, subjects experienced partial sleep deprivation and fragmented sleep for a 24-h period. Serum testosterone levels were determined every 20 min between 1900-0700 h with simultaneous sleep recordings during the 7-min sleep attempts. The results were compared with those obtained in men during continuous sleep. Although mean levels and area under the curve of testosterone were similar in both groups, fragmented sleep resulted in a significant delay in testosterone rise (03:24 h +/- 1:13 vs. 22:35 h +/- 0:22). During fragmented sleep, nocturnal testosterone rise was observed only in subjects who showed REM episodes (4/10). Our findings indicate that the sleep-related rise in serum testosterone levels is linked with the appearance of first REM sleep. Fragmented sleep disrupted the testosterone rhythm with a considerable attenuation of the nocturnal rise only in subjects who did not show REM sleep.