OBJECTIVE: To determine the ability of transesophageal echocardiography to accurately identify or exclude left atrial thrombi. DESIGN: Prospective cohort study. SETTING: University hospital. PATIENTS: 231 consecutive patients having transesophageal echocardiography before elective repair or replacement of the mitral valve or excision of a left atrial tumor. Fifty-six percent of patients had a history of atrial fibrillation, and 17% had a history of thromboembolism. MEASUREMENT: Identification of left atrial thrombi during transesophageal echocardiographic examination and comparison with direct near-simultaneous visualization during cardiac surgery. RESULTS: Transesophageal echocardiography identified 14 left atrial thrombi in 14 patients (6%). Thrombus size range from 3 to 80 mm. Surgery confirmed 12 of 14 thrombi (86%), including 9 thrombi confined to the left appendage. No additional thrombi were found on direct inspection of the atria (sensitivity, 100% [95% CI, 74% to 100%]; specificity, 99% [CI, 97% to 99.9%]; positive predictive value, 86% [12/14]; negative predictive value, 100% [217/217]; for a population that had a 5.2% prevalence of thrombi). All 12 surgically confirmed thrombi were identified by two independent observers. Neither thrombus seen by only a single observer on transesophageal echocardiography was confirmed during direct inspection of the atria at surgery. CONCLUSION: Transesophageal echocardiography is highly accurate for identifying left atrial thrombi and can be used clinically to exclude left atrial thrombi.