To determine whether bypass surgery would benefit patients with symptomatic atherosclerotic disease of the internal carotid artery, we studied 1377 patients with recent hemisphere strokes, retinal infarction, or transient ischemic attacks who had atherosclerotic narrowing or occlusion of the ipsilateral internal carotid or middle cerebral artery. Of these, 714 were randomly assigned to the best medical care, and 663 to the same regimen with the addition of bypass surgery joining the superficial temporal artery and the middle cerebral artery. The patients were followed for an average of 55.8 months. Thirty-day surgical mortality and major stroke morbidity rates were 0.6 and 2.5 per cent, respectively. The postoperative bypass patency rate was 96 per cent. Nonfatal and fatal stroke occurred both more frequently and earlier in the patients operated on. Secondary survival analyses comparing the two groups for major strokes and all deaths, for all strokes and all deaths, and for ipsilateral ischemic strokes demonstrated a similar lack of benefit from surgery. Separate analyses in patients with different angiographic lesions did not identify a subgroup with any benefit from surgery. Two important subgroups of patients fared substantially worse in the surgical group: those with severe middle-cerebral-artery stenosis (n = 109, Mantel-Haenszel chi-square = 4.74), and those with persistence of ischemic symptoms after an internal-carotid-artery occlusion had been demonstrated (n = 287, chi-square = 4.04). This study thus failed to confirm the hypothesis that extracranial-intracranial anastomosis is effective in preventing cerebral ischemia in patients with atherosclerotic arterial disease in the carotid and middle cerebral arteries.