We examined the distribution and long-term reliability of serum measurements of the two main human lignans, enterolactone and enterodiol, and the isoflavonoid phytoestrogens daidzein, genistein, equol, and O-Desmethylangolensin in the New York University Women's Health Study, a prospective cohort study of sex hormones and breast cancer. Serum samples collected at three yearly visits in 30 premenopausal and 30 postmenopausal women who had not been diagnosed with cancer or cardiovascular disease were included in the study. Assays were carried out by ion-exchange chromatography and capillary gas chromatography-mass spectrometry. Levels of isoflavonoid phytoestrogens were low, often at or below the sensitivity level of the assay. The reliability coefficients for these compounds were also low (< or =0.30). The median levels of enterodiol and enterolactone were 1.52 nmol/liter and 20.2 nmol/liter, respectively, and were comparable with the levels observed in omnivorous Finnish women living in the Helsinki area. A substantial number of women, though, had fairly high levels: for instance, 15% of the assays showed levels of enterolactone greater than the mean level observed in vegetarian Finnish women, i.e., 89.1 nmol/liter (H. Adlercreutz et al., Cancer Detec. Prev., 18: 259-271, 1994). The reliability coefficient of a single measurement of enterolactone was moderately high (0.55), suggesting that serum measurements of this compound could be a useful tool in prospective epidemiological studies with access to repeated blood or serum specimens. For instance, the reliability coefficient of the average of three measurements of enterolactone would be 0.79, a level considered acceptable in light of the other sources of error that are present in epidemiological studies (W. Willett, Stat. Med., 8: 1031-1040, 1989).