Relations between previous lactation experience and risks of cancer of the breast and other sites were investigated after follow-up of 50,274 parous women from 1961 through 1980. Among women with complete information on lactation, 5102 developed cancer and, of these, 1136 were diagnosed with breast cancer. Analyses of associations with mean duration of lactation per birth and duration for each of the three first births suggested a nonlinear relation to breast cancer. The highest risk was observed for those with intermediate duration of breast feeding, whereas lower risks were found among those with very short or very long duration. For all nongenital cancers combined, decreased risks were observed among those with the longest duration of breast feeding. However, among cancers of specific sites, a significant inverse association was found for pancreatic cancer only. The overall impression given by our data is that breast feeding is not strongly related to risks of breast cancer or any other common cancer.