In 60 patients with a clinical diagnosis of cerebral infarction, all brachiocephalic vessels were completely visualized angiographically in 2 planes. The degree of stenosis of neck and intracranial vessels was determined using actual measurement at all stenotic points. Only 25% or greater narrowing of the lumen was counted. The site of vascular involvement was deduced from clinical signs and symptoms by 2 of us working independently. Interexaminer reliability was high. Data were analyzed both according to the clinical classification and the classification based on angiography. When classified clinically, 29% of patients were diagnosed as having involvement of the internal carotid artery; 49%, of the middle cerebral artery; and 22%, of the basilar artery. Twenty-five percent were thought to have involvement of multiple vessels. Clinical-angiographic comparisons showed the highest agreement in cases attributed clinically to occlusion of the carotid system (49%). A diagnosis of infarction in the vertebrobasilar system gave the lowest percent of agreement (24%). Patients in whom the angiogram confirmed the clinical localization were not necessarily those with severe stenosis. Some confirmations occurred with mild stenosis but usually the stenotic vessel was more severely involved relative to other vessels. Clinical syndromes attributed to occlusion in the left hemisphere were confirmed angiographically somewhat more frequently than those in the right hemisphere but the numbers were insufficient to show that the difference was statistically significant. The percent of agreement between clinical and angiographic localization did not increase with increasing age of the patients. The frequency of angiographic confirmation of the clinical localization was shown to vary inversely with the time between onset of the stroke and angiography. In the analysis based on an angiographic classification of cases, about one-fourth of the patients had a normal angiogram, one-fourth had involvement of a single vessel, and 50% had several vessels involved. The angiogram showed involvement of a vessel predicted clinically in only about one-half of the cases. Angiographically significant stenosis, i.e., greater than 25% narrowing, was more often recognized in extracranial vessels (carotid and vertebrobasilar arteries) than in intracranial vessels (middle, anterior, and posterior cerebral arteries). Despite the high frequency of middle cerebral artery syndromes clinically, the middle cerebral artery was uncommonly involved alone on the angiogram. No consistent relationship was found between severity of stenosis and frequency of clinical symptoms in the field of perfusion of the stenotic vessel. It was concluded that the angiogram as performed in this study showed limited correlation with clinically deduced sites of vascular stenosis or occlusion. In a high proportion of cases, angiographically demonstrable stenosis or occlusion appeared to have no clinically significant importance. However, the time lapse between occurrence of the clinical stroke and performance of angiography may have played a major role in accounting for the poor correlation in that the earlier the angiography, the higher the incidence of demonstration of clinically significant pathology.