Atrial fibrillation is associated with a sixfold increased risk for stroke. More than a dozen published randomized trials of anticoagulants or antiplatelet agents for stroke prevention provide solid evidence on which to base antithrombotic prophylaxis. Adjusted-dose warfarin reduces risk for stroke by about 60% compared with placebo, aspirin reduces this risk (primarily for nondisabling stroke) by about 20% compared with placebo, and warfarin reduces it by about 40% compared with aspirin. Warfarin provides maximal protection against stroke at international normalized ratios of 2.0 to 3.0. Risk stratification of patients with atrial fibrillation identifies those who potentially benefit most or least from anticoagulation; this is important because a substantial percentage of patients with atrial fibrillation have relatively low rates of stroke if they are given aspirin. Many elderly patients with recurrent intermittent atrial fibrillation experience high rates of stroke and benefit from anticoagulation. The value of precordial or transesophageal echocardiography in addition to clinical risk stratifiers for stratifying stroke risk is controversial. Altered hemostasis favoring thrombosis may contribute to formation of atrial appendage thrombus, but these conditions remain ill defined. The past decade has brought unprecedented progress toward understanding thromboembolism in patients with atrial fibrillation and has changed the clinical perspective of a prevalent cardiac arrhythmia into an important opportunity for stroke prevention. Making the most of this promise calls for appreciation of the epidemiology of atrial fibrillation and the concept of risk specificity in the face of diverse therapeutic options.