We analyzed 118 consecutive cases of pathologically proven infective endocarditis (100 cases of native valve endocarditis [NVE] and 18 cases of prosthetic valve endocarditis [PVE]) with use of the Beth Israel criteria, the Duke criteria, and our suggested modifications of the Duke criteria; we found improved diagnostic sensitivity with our modifications. These modifications included the following additional minor criteria: the presence of newly diagnosed clubbing, splenomegaly, splinter hemorrhages, and petechiae; a high erythrocyte sedimentation rate; a high C-reactive protein level; and the presence of central nonfeeding lines, peripheral lines, and microscopic hematuria. Analysis of the pathologically proven cases of NVE showed that 64% were probable by the Beth Israel criteria, 83% were definite by the Duke criteria, and 94% were definite by our modified Duke criteria. For the pathologically proven cases of PVE, 50% were probable by the Beth Israel criteria, 50% were definite by the Duke criteria, and 89% were definite by our modified Duke criteria. All cases of NVE and PVE rejected by the Duke criteria remained rejected by our modifications. Therefore, our modifications improved diagnostic sensitivity while retaining specificity.