A seasonal modulation of the circadian time structure of circulating T and natural killer (NK) lymphocyte subtypes was documented in five healthy men aged 24-36 yr. Venous blood was obtained every 4 h for 24 h from each subject in January, March, June, August, and November 1984. Three subjects were also studied in April and/or August and/or November 1983 for the T subsets only. Mononuclear cells were isolated on Ficoll-Paque gradient and aliquots were incubated with OKT3, OKT4, OKT8, or HNK-1 monoclonal antibodies for characterizing all, T, T helper, T suppressor-cytotoxic, and NK lymphocytes, respectively, under an epifluorescence microscope. An effect of both sampling time and study month was statistically validated (P less than 0.01) with both two-way analysis of variance and cosinor for the peripheral counts in total, pan-T, T helper, and NK lymphocytes (cells per cubic millimeter). Seasonal changes affected both the circadian patterns and the 24-h mean values. Thus the double amplitude (total extent of variation) of the circadian rhythm in circulating total, T and T helper lymphocytes varied between 0 in March (P greater than 0.30; no rhythm) and up to 46-68% of the 24-h-mean (M) in November, with acrophases (times of maximum, 0) localized in the first half of the night (P less than 0.001). Maximal values were found at 8:30 h for both T suppressor-cytotoxic and NK lymphocytes; a smaller second peak was also found at 20:30 h, and a 12-h rhythm was validated by cosinor (P less than 0.0001), with no patient change in waveform along the year scale. A circannual rhythm was statistically validated by cosinor for total (0 in November), pan-T (0 in March), T suppressor-cytotoxic (0 in December), and NK lymphocytes (0 in October). A rhythm with a period equal to 6 mo was found for circulating T helper cells with 0 occurring both in April and October. Seasonal variations in the incidence of several immunologically related diseases may correspond to an endogenous circannual time structure.