Serum testosterone concentrations from 4.462 military veterans, ages 32-44, were examined for age and seasonal variation. Testosterone concentrations were assayed from a single serum sample from each subject. All samples were drawn before breakfast, at about 8:00 a.m., from subjects recruited over a 16-month study period. Mean levels declined with age (P less than .001), from 864 ng/dl at age 32 to 602 ng/dl at age 44. Mean levels also varied with month of testing (P less than .01), with a seasonal peak in December (the seasonal peak was in November for men in their early 30s). The age effect was greater than the seasonal effect. Both effects may bear upon behaviour and should be treated as possible sources of error in studies of testosterone.