Previous reports have suggested that duplex ultrasonography might supplant arteriography as a guide to operative decision making in selected patients with cerebrovascular disease. This study was undertaken to test that tenet in patients with focal carotid territory symptoms. Seventy-two patients having independently interpreted arch and selective carotid arteriography and duplex scanning underwent 78 carotid endarterectomies. Operative specimens were analyzed in all cases and used as the standard in evaluating the accuracy of the preoperative studies. All patients had disease found at the time of operation. The sensitivity of duplex scanning was 99% vs. 91% for arteriography (p = 0.06). In seven cases the scan accurately predicted disease in patients with normal arteriograms and in a single case the scan was read as normal in a patient with a smooth minimally stenotic plaque read as an irregular 30% stenosis on arteriography. The accuracy of duplex scanning was markedly superior to arteriography in detecting intimal surface abnormalities (92% vs. 64%, p less than 0.001) and ulceration (90% vs. 54%, p less than 0.001). There was no difference between duplex scan and arteriography (p = 1.0) in predicting a greater or less than 50% stenosis (accuracy, 94% for arteriogram; 92% for duplex scanning). Of the patients with preoperative potentially reversible symptoms, 97% were free of symptoms at a mean follow-up of 9 months after operation. Eighty-nine percent (17 of 19 patients) of patients with concomitant, ipsilateral, intracranial, or intrathoracic cerebrovascular disease were free of symptoms after carotid endarterectomy.