The relations between age at first and last full-term pregnancy and breast cancer risk were investigated after 20 years of follow-up of 63,090 Norwegian women, among whom 1,565 breast cancer cases occurred. An association seen in preliminary analyses between early age at first birth and low risk of breast cancer was removed after adjustment for parity and age at last birth. Age at last birth showed initially no association with breast cancer. After adjustment for parity, however, a significant positive association emerged. The authors' observations suggest that the relation between age when a pregnancy occurs and breast cancer risk may be more complex than previously believed. Despite the overall association between increasing parity and lower risk, the women with many late pregnancies and those with few, widely spaced pregnancies had higher risk than nulliparous women, indicating that both the age when pregnancy occurs and the length of intervals between successive births may modify the protective effect. The findings are consistent with a dual effect of a pregnancy, causing a transient increase followed by a subsequent strong and long-lasting decrease in risk of breast cancer.