The association between oral contraceptive (OC) use and the risk of ovarian cancer was analysed in a case-control study, conducted between 1985 and 1989 on 505 epithelial ovarian cancer cases under 60 years of age, and 1375 controls in hospitals for a spectrum of acute conditions, not gynaecological, hormonal or neoplastic, apparently unrelated to OC use. 41 (8.1%) women with epithelial ovarian cancer and 192 (14.0%) controls reported OC use. The multivariate relative risk (RR) for ever use was 0.7 (95% confidence interval (CI) = 0.5-1.0). The risk decreased with duration of use: compared with never users the multivariate RRs were 0.9 and 0.5 respectively for less than 2 years and 2 years or more users (chi 2(1) trend = 6.17, P = 0.01). The risk of ovarian cancer decreased with recency and latency of use: the estimated RR were 0.5 and 0.9 in women reporting last OC use less than 10 or 10 years or more from the diagnosis of the disease, and 0.6 and 0.8 in those reporting first OC use less than 10 or 15 or more years before. The protective effect of OC was consistent in separate strata of selected covariates, including parity and other major known or suspected risk factors for ovarian cancer. There was some indication that the protection declines with advancing age, but the risk estimates were similar in premenopause and postmenopause.