The recently proposed Duke criteria were confirmed to be more sensitive than the former Beth Israel criteria for the diagnosis of infective endocarditis (IE). To assess the specificity of the Duke criteria, we reviewed the records of 100 patients admitted to two internal medicine wards because of acute fever or fever of unknown origin (FUO). IE was considered a possible diagnosis for all patients who had had at least two blood cultures performed and one transthoracic echocardiogram obtained. The diagnosis of IE was rejected in all cases in accordance with the Duke criteria (i.e., a firm alternate diagnosis [n = 23], resolution of symptoms with no antibiotics [n = 39], or both reasons [n = 38]). To calculate the specificity of the Duke criteria, all 100 cases were then reclassified according to these criteria as if the diagnosis of IE had not been rejected. Only one patient, who probably did not have IE, was reclassified as having clinically definite IE. The specificity of the Duke criteria could therefore be calculated to 0.99 (95% confidence interval, 0.97-1). Although the design of the study may have resulted in a slight overestimation of the specificity rate, we conclude that the Duke criteria are highly specific for ruling out IE in patients with acute fever or FUO who are at low risk for IE.