Every other month, during the course of 14 months, the circadian rhythm in plasma testosterone (among others variables) of 5 healthy males (26 to 32 years) was documented. Subjects' synchronization: light-on at 7 ± 1 h; light-off at 23 ± 1.5 h during the whole year. Blood sampling 4-hourly, at fixed times during 28 h, for radioimmunological testosterone determinations. The cosinor method was used for statistical analyses. An annual rhythm in the testosterone 24-h mean is detected (P < 0.025) with a peak in October (from 16 July to 26 December, with 95% confidence limits). Annual mean (± 1 se) = 759 (± 26) ng/100 ml. Peaktrough difference = 197 ng/100 ml (from 106 to 290 with 95% confidence limits). The timing of the testosterone circadian peak, in the 24-h-scales also varies during the course of the year. The peak is found around 8 h in May and around 14 h in November, this phase-shift being statistically significant (P < 0.05).