Abstract Background. Breastfeeding is considered to be an important factor for maternal and children's health. However, the epidemiological findings related to the effect of breastfeeding on women's health, especially with respect to breast cancer development, are inconsistent. Determinants of infant feeding method may contribute to the inconsistency. Methods. A total of 24,769 women aged 40–64 in Miyagi Prefecture, Japan, responded to a self-administered questionnaire survey in 1990. Using the data obtained from 22,085 parous women, we calculated odds ratios (ORs) for the choice of “breastfeeding only” during reproductive period. Results. Late age at menarche (≥16 years, OR = 1.57) and high body mass index (BMI) at 20 years of age (≥24, OR = 1.31) were associated with the choice of breastfeeding only. Late age at birth of first child (≥28 years, OR = 0.29), history of breast cancer in mother (OR = 0.68), and high educational level (more than a high school education, OR = 0.53) reduced the possibility of choosing breastfeeding only. Conclusion. The results indicate that the choice of infant feeding method is associated with several breast cancer risk factors. Based on this finding, we should construct appropriate breast cancer risk models for parous women and investigate the changes in the effects of breastfeeding and other breast cancer risk factors among these risk models. Especially in a risk model controlling for breastfeeding, the effects of other breast cancer risk factors should be reevaluated. Through comparisons among different risk models, we may find the best-fitted risk model and identify the true effect of breastfeeding.