5,160 non-hysterectomised women aged 44-100 years completed a health survey questionnaire as part of a longitudinal study of a southern California retirement community begun in June 1981. As of 1 January 1987, 50 incident cancers of the endometrium had occurred among these women, who had contributed 23,786 years of follow-up. Women who had used oestrogen replacement therapy had a relative risk of endometrial cancer of 10 compared to women who had never used oestrogens (P less than 0.0001). Risk increased with increasing duration of use (chi 2 test for trend = 50.60, P less than 0.0001); women who had used oestrogens for 15 or more years had a relative risk of 20 (95% C.I. = 7.2, 54) compared to non-users. While current and recent users (i.e. those who had used oestrogen within one year of the initial survey) had the greatest risk (RR = 25, 95% C.I. = 9.2, 69), women who had last used oestrogens 15 or more years ago still had a significantly increased risk (RR = 5.8, 95% C.I. = 2.0, 17). No other variable studied had a major effect on risk, except smoking. Women who smoked at the time of menopause had a significantly reduced risk of disease (RR = 0.38, P = 0.005), which was essentially unchanged after adjustment for oestrogen use.