A group of 63 men and 86 women 77 +/- 8 years of age institutionalized at the Berceni Clinical Hospital, Bucharest, Romania, on a diurnally active living schedule with rest at night, eating 3 comparable but not identical meals a day, were investigated over a 2-year span in 278 profiles extending over a 24-hour span each and distributed over all our seasons. On each day of sampling six samples of blood were collected at 4-hour intervals beginning at 0800 of the first day of the study to 0400 of the next. Plasma testosterone, dehydroepiandrosteronesulfate (DHEA-S), 17-hydroxyprogesterone (17-OH Prog), FSH and LH concentrations were determined by radioimmunoassay. Statistically significant circadian rhythms were detected and quantified by population mean cosinor for the men and women of the entire group and for the age groups of 51 to 70 (I), 71 to 80 (II) and above 80 years of age (III) separately. Circannual rhythms of the circadian means were sought and evaluated by one way analysis of variance. Statistically highly significant circadian rhythms were found in plasma testosterone, 17-OH Prog and DHEA-S, concentrations in men and women of all three age groups with a phase advance of over 2 hours in DHEA-S with advancing age. Circannual cycles were found for plasma testosterone concentration and for DHEA-S in men and women, with a likely phase difference between men and women in testosterone. In LH a circadian rhythm was found in the women and in FSH in men only and a circannual variation in LH only in men.