Estimates of the risk of stroke for men with non-rheumatic atrial fibrillation were obtained from two large cohort studies--the Whitehall Study of London Civil Servants and the British Regional Heart Study. The first cohort provided an estimated relative risk of stroke of 6.9 compared with controls. This increased risk confirms that of the other prospective estimate, 5.6, found in the Framingham study. In the second cohort only one of the men at risk had a stroke, and the risk estimate did not differ significantly from unity. The absolute rates of stroke in both cohorts were lower than those reported in the Framingham study, implying a lesser potential benefit from preventive measures. These lower rates should be considered in the planning of trials. The risk of stroke was significantly associated with raised systolic or diastolic blood pressure but not with age or coronary heart disease.