Prospective data from Busselton, Western Australia, collected during triennial surveys from 1966-81 with follow-up of subjects to 1983, showed that atrial fibrillation (AF) was frequent in elderly people and associated with increased mortality. Of 1770 people aged over 60 years, 40 were in atrial fibrillation when first seen and a further 47 developed it during follow-up. Atrial fibrillation was positively associated with angina, history of a myocardial infarction and left bundle branch block. Relative mortality in those with atrial fibrillation compared with those without it, was 1.92 for all causes, 1.82 for death from cardiovascular causes (excluding stroke) and 3.78 for deaths from stroke, after adjustment by proportional hazards regression for confounding effects of age, sex, history of a myocardial infarction, an abnormal electrocardiogram, angina, cholesterol level systolic blood pressure and Quetelet's Index (weight/height2). The excess relative mortality declined with increasing age for both women and men. This raised relative mortality remained constant with time from the first detection of AF for all causes and cardiovascular causes but appeared to increase with time from detection for stroke death. The risk of death from stroke was greatest in the younger women. The observed risk of death from stroke in patients with AF suggests that anticoagulant use should be considered in selected patients.