Time-related effects of a pregnancy on breast cancer risk were examined in a population-based prospective study of 802,457 parous Norwegian women aged 20-56 years. The mean follow-up time was 16.4 years. A total of 4787 women were diagnosed with breast cancer. We observed a short-term increase in risk of breast cancer after a full-term pregnancy, with a maximum 3-4 years after delivery, followed by a long-lasting decrease in risk. The maximum risk was about twice the risk for women whose last delivery was 20 or more years previously (incidence rate ratio = 1.99, 95% confidence interval = 1.70-2.33). Compared with nulliparous women, those with one or two children were at higher risk in the first decade after the last pregnancy, whereas those with three or more children were at lower risk in most categories of time since the last birth. The positive association between breast cancer risk and age at last birth was markedly reduced after adjustment for time since last birth. We conclude that there is a non-linear relationship between breast cancer incidence and time since last birth. Part of the relation with age at last birth may be attributed to the association with time since last birth.