Measuring levels of steroid hormones in epidemiologic studies is difficult because pulsatile release can cause the levels of many hormones to vary markedly over short intervals, leading to a loss of precision in between-subject comparisons. Clinicians often control this variation by collecting several samples from each subject at defined intervals and pooling these samples for assay. The number of samples per subject that would adequately control such variation in an epidemiologic study has not been fully investigated. This study examines the effects of collecting 1, 2, or 3 samples per subject on the variances of 11 hormones and sex hormone binding globulin in men and 6 hormones in women. Three samples were collected at 30-minute intervals from each of 20 men and 59 women and were assayed separately. Variances that would be obtained in studies collecting one, two, or three samples per subject were then estimated. Collecting more than one sample substantially reduced the variances of several hormones in men but not in women.