In each of two population-based studies conducted on the Island of Guernsey between 1967-1976 and 1977-1984, respectively, single specimens of blood were taken from over 5000 normal women. From these two studies there were 1173 and 946 postmenopausal women in whom blood prolactin was determined and multivariate analysis was used to establish the association between blood prolactin concentration and possible determinants of risk of breast cancer. Since prolactin levels were log-normally distributed these analyses were done on log-transformed data. The age at menarche or menopause, age at first or last childbirth, length of reproductive life (i.e. time from menarche to menopause) or post-menopausal life (i.e. time from menopause to time of blood sampling), contraceptive use and history of breast cancer were not significantly associated with blood prolactin concentration. Of significance were age, parity, time of blood sampling and assay drift. Ponderosity (Quetelet's Index) was positively associated with prolactin concentration and this was significant using a one-tail criterion. Women with a mammographic pattern designated DY by Wolfe had significantly higher prolactin levels than those with N1 patterns. However, the main finding to emerge was that after standardizing for all the other variables increasing parity was related to a step-wise reduction in blood prolactin levels. Since this had occurred in women who had had their last child up to 35 years previously it implies this effect is permanent. It could therefore be that the protective effect on breast cancer risk of multiparity and early first pregnancy could be mediated by such a life-long reduction in blood prolactin levels.