OBJECTIVE To evaluate the accuracy of the revised PIOPED (Prospective Investigation of Pulmonary Embolism Diagnosis) criteria for categorization of ventilation-perfusion lung scans and to compare the diagnostic accuracy of the revised criteria with that of the original PIOPED criteria and subjective probability estimates. METHODS The ventilation-perfusion scans of 104 consecutive patients with suspected pulmonary embolism were reviewed. All patients had also undergone pulmonary angiography. The scans were categorized according to the original and revised PIOPED criteria, and a "gestalt" percent probability estimate was made. In addition, the official clinical interpretation (made with the original PIOPED criteria) was recorded. RESULTS The gestalt percent probability estimate was the most accurate for assessing the likelihood of pulmonary embolism (area under the receiver operating characteristic [ROC] curve = 0.836). The revised PIOPED criteria (area under the ROC curve = 0.753) were more accurate than the original PIOPED criteria. CONCLUSIONS The revised PIOPED criteria are more accurate than the original PIOPED criteria. Experienced readers of lung scans can achieve higher accuracy after applying formal criteria by using their experience and subjective judgment.