Epidemiologic studies have shown that atrial fibrillation raises the risk of stroke approximately fivefold, and that, because atrial fibrillation is so common among the stoke-prone elderly, it accounts for a substantial fraction of all strokes. The set of five recently completed randomized trials have established that warfarin, at low dose, can prevent most of the additional stroke risk caused by atrial fibrillation. The trials have also demonstrated that warfarin therapy can be safe if careful monitoring is implemented. The trials have not, however, settled the anticoagulation decision for all patients. Warfarin remains a demanding and risky therapy, which many patients and physicians do not find attractive. This review analyzes the epidemiologic observations and randomized trials that provide our new understanding of atrial fibrillation and stroke. It also reviews possible risk factors for stroke in atrial fibrillation and alternative antithrombotic regimens such as aspirin. At the present time, low-dose warfarin should be considered for all patients with chronic atrial fibrillation. Younger patients with no other clinical heart disease may be at such low risk for stroke that warfarin might be avoided. The evidence is not substantial that aspirin is an effective alternative.