The inconclusive findings of past analytic epidemiologic studies on diet and breast cancer may have resulted from the inability of these studies to assess early dietary exposures. The role of macronutrient intake during early life can be indirectly studied, however, by examining past and present body size. The authors identified by computer linkage a population-based historical cohort of 38,084 women born between 1918 and 1943, on whom information about weight and height had been recorded in Hawaii in both 1942-1943 and 1972. Linkage of this cohort to the Hawaii Tumor Registry resulted in the identification of 607 incident cases of breast cancer for 1972-1983. An average of 4.4 cancer-free controls were matched to each case on year and month of birth and race of the parents. A matched case-control analysis, conducted in each five-year birth cohort, revealed a negative association of adolescent body mass to premenopausal breast cancer. This negative association was statistically significant in girls aged 10-14 years in 1942 (p for trend, 0.004), was present in all ethnic groups, and was strongest among overweight young women who remained overweight in adulthood. Early-age weight, height, and body surface area were not associated with either pre- or postmenopausal breast cancer. Adult weight and gain in body mass since 1942 were positively associated with postmenopausal breast cancer risk. Adjustment for age at first birth, parity and socioeconomic indicators for 1942 and 1972 did not modify the results. This study provides evidence for a protective role of adolescent obesity against premenopausal breast cancer, and for an enhancing role of a positive energy balance during adult life on postmenopausal breast cancer.