We studied 243 patients in whom 248 pulmonary angiograms were performed because of suspected pulmonary embolism. Ventilation and perfusion lung scanning in 140 of them revealed 38 to be in low and high probability groups. Of 19 patients with subsegmental and nonsegmental perfusion defects that were matched with ventilation defects, none had pulmonary embolism. Conversely, angiography was positive in 17 of 19 patients with multiple segmental or lobar perfusion defects in areas of normal ventilation. Doppler flow examinations of the veins of the legs showed normal flow in 61 of 79 (77 percent) patients with pulmonary emboli and, therefore, were insensitive indicators of embolism. There was no mortality from angiography, and serious complications occurred in 2 percent of the patients. Anticoagulation in 83 patients was associated with bleeding in 25, two of whom died. The data indicate that ventilation-perfusion lung scanning can be used to separate many of the patients suspected of having pulmonary embolism who need anticoagulant treatment from those who do not. However, there is a considerable number of patients with nonspecific abnormalities on lung scan. For this group of patients with nonspecific abnormalities, the risk of complications from empiric treatment with anticoagulant drugs is probably greater than the risk of complications from pulmonary angiography. Further, our data show that patients with negative angiography have a very low risk of subsequent pulmonary embolism. In this group of patients, therefore, pulmonary embolism should be demonstrated by angiography before long-term anticoagulant therapy is prescribed.