Summary Characteristics probably associated with the fetal hormonal milieu have recently been shown to increase (birth size indicators, prematurity, neonatal jaundice) or decrease (pregnancy toxaemia) breast cancer risk in the female offspring. However, it is unknown whether differences in pregnancy hormone levels may contribute to the marked geographical variation in breast cancer incidence. We have compared, in a highly standardized manner, pregnancy hormone levels in a population with high incidence and one with low incidence of breast cancer. Three hundred and four pregnant Caucasian women in Boston and 334 pregnant Chinese women in Shanghai were enrolled from March 1994 to October 1995. Levels of oestradiol, oestriol, prolactin, progesterone, human growth hormone, albumin and sex hormone-binding globulin were measured in maternal blood at weeks 16 and 27 of gestation and compared between the two study sites using non-parametric Wilcoxon's rank-sum test. Demographical, anthropometrical and pregnancy characteristics were ascertained through interview, and relevant variables concerning delivery and the newborn were abstracted from medical records and paediatric charts. During the first visit, median serum levels of all studied hormones were statistically significant, and in most instances substantially, higher among Chinese women, who have a low incidence of breast cancer, compared with American women, who have a high incidence of breast cancer. An analogous pattern was evident during the second visit, although the relative differences tended to be smaller. Further research is needed to identify lifestyle or other exogenous determinants of pregnancy hormone levels, as well as possible mechanisms by which they may influence carcinogenic processes in the breast and possibly other organs.