A case-control study involving 405 cases of epithelial endometrial carcinoma (newly diagnosed between 1987 and 1990 in five US areas) and 297 population controls enabled evaluation of risk in relation to detailed smoking characteristics. Cigarette smokers were at a reduced risk of disease, with the effect primarily restricted to women whose diseases were detected postmenopausally (relative risk (RR) = 0.6, 95% confidence interval 0.4-0.9). Among postmenopausal women, current smokers showed the greatest reduction in risk (RR = 0.4, 95% confidence interval 0.2-0.7), with former smokers, including those who had recently stopped, being less affected (RR = 0.8). Other measures of smoking were highly correlated with currency of smoking, but there were no clear patterns of risk with either duration or intensity of smoking. Smoking appeared to reduce risk to the greatest extent in subjects who were multiparous, obese, or nonusers of exogenous hormones, but none of these relations was statistically significant. The results support the notion that smoking reduces the risk of endometrial cancer through extraovarian endogenous hormonal mechanisms, but further studies are needed to clarify why reduced risks are most pronounced among postmenopausal women and those currently exposed to cigarette smoke.