Creatinine-adjusted levels of estrone, estradiol and estriol were determined in overnight urine specimens from 220 post-menopausal women, aged 54 to 66 years, from Boston (121 women) and Athens (99 women). The associations of individual and total estrogens with various characteristics of the women (age at menopause, years since menopause, parity, age at menarche, schooling, height, weight, tobacco smoking and alcohol drinking) were studied by multiple regression procedures, to explore possible relationships and evaluate their compatibility with the descriptive epidemiology of female breast cancer. Weight was consistently and significantly related to higher levels of total estrogens and of each individually; the pattern of these associations was similar in American and Greek women, suggesting that the known qualitative differences in nutrition between the two groups of women are not important modifiers of the associations between weight and estrogens. Estrogen levels in these post-menopausal women were about 40% of the corresponding follicular levels among teen-age women. They did not decrease with age. There were suggestive but non-significant negative associations between estrogen levels on the one hand, and parity and age at menarche on the other. There was no consistent association of estrogen levels with alcohol consumption. The data provide only modest support for the hypothesis that estrogens are important in breast cancer etiology.