OBJECTIVE To determine whether replacement estrogen delays or prevents loss of cognitive function in elderly women. METHODS A 15-year prospective and cross-sectional study. METHODS Rancho Bernardo, a geographically defined community in Southern California. METHODS Eight hundred women (80% of local surviving women from the original Rancho Bernardo cohort) aged 65 to 95 years. Estrogen use was evaluated at baseline between 1972 and 1974 and at follow-up between 1988 and 1991. METHODS Twelve tests of cognitive function from eight standard instruments administered at follow-up between 1988 and 1991. RESULTS Almost half of this older, educated cohort had used estrogen at some time after menopause, and one third were current users. The age-related decrement in cognitive function was similar for women who were current, past, or never users of estrogen. Age- and education-adjusted comparisons also failed to show any consistent association between performance on tests of cognitive function and baseline, past, current, or never estrogen use; estrogen dose; or duration of use. Among 132 statistical comparisons, only five statistically significant differences were observed, less than the number expected by chance alone. Furthermore, these significant differences occurred with different tests of cognitive function, and in only one instance was the better test score associated with current estrogen use. No biases were identified that could explain these negative results. CONCLUSIONS No compelling or internally consistent evidence for an effect of estrogen on cognitive function was found in these older women. These data do not support the hypothesis that estrogen use after the menopause preserves cognitive function in old age.