Although the sensitivity and specificity of the Duke criteria for the diagnosis of infective endocarditis (IE) have been validated by investigators from Europe and the United States, several shortcomings of this schema remain. The Duke IE database contains records collected prospectively on >800 cases of definite and possible IE since 1984. Databases on echocardiograms and on patients with Staphylococcus aureus bacteremia at Duke University Medical Center are also maintained. Analyses of these databases, our experience with the Duke criteria in clinical practice, and analysis of the work of others have led us to propose the following modifications of the Duke schema. The category "possible IE" should be defined as having at least 1 major criterion and 1 minor criterion or 3 minor criteria. The minor criterion "echocardiogram consistent with IE but not meeting major criterion" should be eliminated, given the widespread use of transesophageal echocardiography (TEE). Bacteremia due to S. aureus should be considered a major criterion, regardless of whether the infection is nosocomially acquired or whether a removable source of infection is present. Positive Q-fever serology should be changed to a major criterion.