BACKGROUND: The use of warfarin anticoagulation for several weeks before cardioversion results in a 90% reduction in the incidence of cardioversion-related thromboembolism. The mechanism of this benefit, however, is unknown; it has been widely attributed to organization and adherence of atrial thrombi, a finding observed among pathological studies of patients with rheumatic valvular disease. METHODS AND RESULTS: Serial transesophageal echocardiography was performed in 14 patients with nonrheumatic atrial fibrillation after identification of atrial thrombi on initial transesophageal study. All patients received warfarin anticoagulation and were followed clinically for signs of thromboembolism. Eighteen atrial thrombi were identified on initial transesophageal study, including 14 thrombi confined to the left atrial appendage, 2 in the body of the left atrium, 1 in the right atrial appendage, and 1 in the body of the right atrium. Thrombus size varied from 5 to 20 mm, and 6 were considered mobile. After a median of 4 weeks of warfarin, 16 of 18 atrial thrombi (89%; 95% CI, 73% to 100%) had completely resolved on transesophageal echocardiographic study. In addition, no new thrombi were identified on follow-up study, and no patient had a clinical thromboembolic event between studies. CONCLUSIONS: These data strongly support the hypothesis that among patients with nonrheumatic atrial fibrillation, the mechanism of clinical benefit with 3 to 4 weeks of warfarin before cardioversion is related to thrombus resolution and prevention of new thrombus formation rather than thrombus organization.