We examined the effect of parity on breast cancer risk on the basis of information from 107,146 married female nurses followed prospectively between 1976-1980. For everparous women, a significant protective effect against breast cancer was observed with increasing parity (chi 12 for trend = 14.2, p less than 0.001). Adjustment for age at first birth and other potential confounders by multiple logistic regression did not materially affect this trend. After controlling for these risk factors, the overall relative risk of breast cancer for women with four or more pregnancies lasting 6 months or longer was 0.68 (95% confidence limits 0.51-0.90) compared with women with only one pregnancy. When nulliparous and everparous women were compared, an interaction of age and everparous status in 1976 was observed. The logistic regression coefficient for the interaction was 0.3419 (95% confidence limits 0.0374-0.6464). Specifically, for women of parity 2 relative to nulliparous, the relative risk of breast cancer and the 95% confidence intervals were 1.85 (1.09, 3.14), 1.32 (0.92, 1.89), and 0.93 (0.62, 1.41) for 30-39, 40-49 and 50-55-year-old women, respectively. This points to the need of presenting age-specific, rather than age-adjusted, relative risks in future studies of breast cancer and parity.