Differing opinions about the value of ventilation-perfusion lung scanning have created controversy concerning the correct approach to the diagnosis of pulmonary embolism. In a prospective study of 305 consecutive patients with clinically suspected pulmonary embolism and abnormal perfusion lung scans, we evaluated the role of ventilation-perfusion lung scanning, pulmonary angiography, and objective testing for venous thrombosis in the diagnostic process. Segmental or greater perfusion defects with ventilation mismatch have a high probability (86 percent) of pulmonary embolism. Contrary to current clinical practice, however, the approach of ruling against pulmonary embolism by a "low probability" scan pattern is incorrect, even with an improved technique for ventilation imaging; the frequency of pulmonary embolism in these patients ranged from 25 to 40 percent. Objective testing for venous thrombosis provides a practical alternative to performing pulmonary angiography in the diagnostic work-up; by providing an endpoint for commencing anticoagulant therapy, a positive result obviates the need for further testing in 20 to 30 percent of patients.