The recent demonstration of the efficacy of carotid endarterectomy in certain patients emphasizes the advantages of having a noninvasive, accurate means of evaluating the carotid arteries. Advances in magnetic resonance (MR) angiography now allow accurate depiction of the carotid arteries that may be adequate for surgical planning in many cases. This report examines the accuracy of MR angiography compared with that of conventional angiography in symptomatic patients undergoing carotid endarterectomy and compares them with surgical findings. Twenty-one carotid arteries in 20 patients were treated surgically for severe stenosis or occlusion. Preoperatively, all patients had both MR and conventional angiograms, which were interpreted on a five-grade scale by two independent neuroradiologists who were unaware of the patient's clinical history. The two studies were highly correlated, particularly in the case of severe stenosis and occlusion. There were no false-negative MR studies that missed surgically significant lesions. In two cases, MR angiography overestimated the stenosis by one grade. On MR angiography, surgically significant stenosis appears as focal areas of signal intensity loss at the level of stenosis with reappearance of the signal distally. If the distal signal intensity does not reappear, the artery is likely to be occluded. In symptomatic patients, MR angiograms that demonstrate a flow-void gap with distal reappearance at a site consistent with the symptoms may be adequate as the sole preoperative study. Three patients who underwent carotid endarterectomy on this basis are presented. The factors that contribute to artifactual and overestimated stenosis are reviewed.