The laboratory reliability and validity of sex hormone measurements were examined at multiple levels, including lower levels characteristic of children and postmenopausal women. Serum was drawn from four adult male and four adult female healthy volunteers. From each individual's serum pool, a medium- and a low-dilution pool were created. Biochemical analyses for total and non-sex hormone-binding globulin (SHBG)-bound estradiol, estrone, estrone sulfate, progesterone, and SHBG were performed on female samples. Male samples were analyzed for total and non-SHBG-bound testosterone, dihydrotestosterone, androstenedione, and dehydroepiandrosterone sulfate. Two aliquots from each pool were assayed twice in each of two labs. All assays except SHBG in one lab used RIA procedures. Reliability was assessed by variance components analyses and estimated coefficients of variation (CVs). Validity was assessed by comparing observed measurements versus expected values based on known dilution ratios. For the testosterone and dihydrotestosterone assays, CVs were usually less than 10%. For estradiol and progesterone, CVs were usually less than 15%. Assays with larger estimated CVs included androstenedione, dehydroepiandrosterone sulfate, estrone, and estrone sulfate. Absolute levels differed markedly between labs for most assays. Observed measurements generally agreed with values expected from the dilution ratios. A notable exception was the estrone assay at the lowest dilution level, where observed measurements were 2-4 times those expected. A similar but less pronounced overestimation bias for the low levels of estradiol was also suggested. This intra- and interlaboratory variability and apparent low dilution overestimation should be accounted for in studies relating hormones to cancer risk, especially those involving children and postmenopausal women.