To evaluate the reliability of 'low-probability' ventilation-perfusion (V-P) scintigrams in excluding pulmonary embolism (PE), we reviewed the clinical records of 99 consecutive patients (74 inpatients and 25 outpatients) whose V-P studies had been interpreted as indicative of a low probability of PE. None of the 99 patients were referred for pulmonary angiography. Venography was performed in four patients and was positive in three. No patients were treated specifically for pulmonary embolism, but five received full treatment with anticoagulants for other reasons (three for venous thrombosis, one for a history of venous thrombosis, and one for a ventricular aneurysm). Seven of the hospitalized patients died during the index admission and seven additional hospitalized patients (including one treated with anticoagulants) died 1-5 months after discharge from the hospital. None were thought clinically to have died as a result of PE, and autopsy disclosed no PE in two. Follow-up information was obtained for 69 surviving patients (49 inpatients and 20 outpatients) not treated with anticoagulants. None of these patients were thought clinically to have had PE during follow-up of at least 2 weeks duration (greater than 2 months in 93% and greater than 6 months in 75%). Our results suggest that major short-term morbidity or death attributable to PE are quite infrequent in patients with low-probability V-P scintigrams.