Trends in the incidence of ovarian cancer among 23,239 cases reported to the Swedish Cancer Registry from 1960 through 1984 were analyzed. Trends in age-standardized incidence rates were best described by a nonlinear model with a highly significant negative second-order term, which showed an initial increase followed by stabilization and, in later years, a decline. In multivariate models, both time period of diagnosis and birth cohort were found to have significant effects. These separate models provided consistent evidence for an increasing risk in women born during the last decades of the 19th century, and a slowly decreasing risk in recent years and in the youngest birth cohorts. The very high population exposure to oral contraceptives, which may protect against ovarian cancer, is one conceivable explanation for the decreasing trend.