We retrieved breast mammograms for 370 women 40 to 74 years old with no history of breast cancer, for whom birth weight, birth length, placental weight and other birth characteristics were indicated in their standard birth records at the Uppsala University Hospital. Blind evaluation of the mammograms allowed these to be classified according to Wolfe's pattern. Logistic regression analysis was applied using as independent variables the recorded birth characteristics and as outcome variable, high risk (P2 and DY) versus low risk (N1 and P1) mammographic parenchymal pattern. After controlling for all the recorded variables, the odds ratio for a high-risk pattern (P2 or DY) increased consistently and significantly (P for trend 0.02) with the weight of the placenta, i.e. the main estrogen-producing organ during pregnancy. There were weak and non-significant positive associations with increasing birth weight (P for trend 0.53) and birth length (P for trend 0.52). These results are compatible with hypotheses suggesting that pregnancy estrogens or other perinatal characteristics may play a risk-modulating role influencing breast cancer in the offspring.