To determine the relation between estrogen replacement therapy and fractures. Prospective cohort study. Four clinical centers in Baltimore County, Maryland; Minneapolis, Minnesota; Portland, Oregon; and the Monongahela Valley, Pennsylvania. 9704 ambulatory, nonblack women 65 years of age or older. Estrogen use, medical history, and anthropometric data were obtained by questionnaire, interview, and examination. Appendicular bone mass was measured by single-photon absorptiometry. Incident fractures were validated by radiographic report. After adjustment for potential confounders, current estrogen use was associated with a decrease in the risk for wrist fractures (relative risk [RR], 0.39; 95% CI, 0.24 to 0.64) and for all nonspinal fractures (RR, 0.66; CI, 0.54 to 0.80) when compared with no estrogen use. Results were similar for women using unopposed estrogen or estrogen plus progestin, for women younger or older than 75 years of age, and for current smokers or nonsmokers. The effect of estrogen remained after adjustment was made for appendicular bone mass. The relative risk for hip fracture tended to be lower among current users (RR, 0.60; CI, 0.36 to 1.02) than among never-users. Estrogen was most effective in preventing hip fracture among those older than 75 years. Current users who started estrogen within 5 years of menopause had a decreased risk for hip fractures (RR, 0.29; CI, 0.09 to 0.92), wrist fractures (RR, 0.29; CI, 0.13 to 0.68), and all nonspinal fractures (RR, 0.50; CI, 0.36 to 0.70) when compared with women who had never used estrogen. Previous use of estrogen for more than 10 years or use begun soon after menopause had no substantial effect on the risk for fractures. Current use of estrogen appears to decrease the risk for fracture in older women. These results suggest that for protection against fractures, estrogen should be initiated soon after menopause and continued indefinitely.