Several studies have established the value of anticoagulation for primary prevention of thromboembolic events in patients with non-rheumatic atrial fibrillation (NRAF). However, in patients with a recent transient ischaemic attack (TIA) or minor ischaemic stroke the preventive benefit of anticoagulation or aspirin remains unclear. Physicians in 108 centres from 13 countries collaborated to study this question. 1007 NRAF patients with a recent TIA or minor ischaemic stroke were randomised to open anticoagulation or double-blind treatment with either 300 mg aspirin per day or placebo (group 1, 669). Patients with contraindications to anticoagulation were randomised to receive aspirin or placebo (group 2,338). The measure of outcome was death from vascular disease, any stroke, myocardial infarction, or systemic embolism. During mean follow-up of 2.3 years, the annual rate of outcome events was 8% in patients assigned to anticoagulants vs 17% in placebo-treated patients in group 1 (hazard ratio [HR] 0.53; 95% confidence interval [CI] 0.36-0.79). The risk of stroke alone was reduced from 12% to 4% per year (HR 0.34; 95% CI 0.20-0.57). Among all patients assigned to aspirin (groups 1 and 2), the annual incidence of outcome events was 15%, against 19% in those on placebo (HR 0.83; 95% CI 0.65-1.05). Anticoagulation was significantly more effective than aspirin (HR 0.60; 95% CI 0.41-0.87). The incidence of major bleeding events was low, both on anticoagulation (2.8% per year) and on aspirin (0.9% per year). No intracranial bleeds were identified in patients assigned to anticoagulation. We conclude that anticoagulation is effective in reducing the risk of recurrent vascular events in NRAF patients with a recent TIA or minor ischaemic stroke. In absolute terms: 90 vascular events (mainly strokes) are prevented if 1000 patients are treated with anticoagulation for one year. Aspirin is a safe, though less effective, alternative when anticoagulation is contraindicated; it prevents 40 vascular events each year for every 1000 treated patients.