To compare the diagnostic value of transesophageal and transthoracic echocardiography in infective endocarditis, paired transesophageal and transthoracic echocardiograms were obtained prospectively for 66 episodes of suspected endocarditis in 62 patients. Echocardiographic results were compared with the presence or absence of endocarditis determined by pathologic or nonechocardiographic data from the subsequent clinical course. All echocardiograms were interpreted by an observer told only that the studies were from patients in whom the diagnosis of endocarditis was suspected. The diagnosis of endocarditis was eventually made in 16 of the 66 episodes of suspected endocarditis (14 by pathologic and 2 by clinical criteria). In 7 of 16 transthoracic and 15 of 16 transesophageal echocardiograms, endocarditis was diagnosed at a probability level of "almost certain," giving a sensitivity of 44% and 94%, respectively (p less than 0.01). For the remaining episodes, 49 of 50 transthoracic and all transesophageal studies yielded normal results, giving a specificity of 98% and 100%, respectively. This study suggests that transesophageal echocardiography is highly sensitive and specific for the diagnosis of infective endocarditis and significantly more sensitive than transthoracic echocardiography. Although echocardiography cannot rule out endocarditis, the high diagnostic sensitivity of transesophageal echocardiography results in a low probability of the disease when the study yields negative results in a patient with an intermediate likelihood of the disease.