In a randomly selected population of 9067 individuals, 32-64 years of age in 1967-1970, 25 (0.28%) had chronic atrial fibrillation (CAF). Eight had lone atrial fibrillation. In 1984 the cases were compared with an age- and sex-matched control group of 50 and found to have more cerebrovascular accidents (6 versus 2; P less than 0.05), congestive heart failure (9 versus 1; P less than 0.001), and valvular rheumatic heart disease (3 versus 0) or history consistent with rheumatic fever (6 versus 0; P less than 0.01). The mortality in the CAF group was 60% higher due to an excess in cardiovascular (relative risk 6.1; P less than 0.05) and cerebrovascular (relative risk 12.2; P less than 0.05) causes. The prevalence or incidence of ischaemic or hypertensive heart disease or the presence of coronary risk factors did not significantly differ in the two groups. By M-mode echocardiography the left atrial size, left ventricular enddiastolic dimension and left ventricular mass were increased in the CAF patients, while the systolic left ventricular shortening was significantly less. Thus, the prevalence of CAF is low in a randomly selected population 32-64 years of age and CAF is not strongly associated with ischaemic heart disease or hypertension. The CAF patients have an increased risk of dying prematurely particularly from cerebrovascular causes, even in the absence of valve disease.