A 5-year experience with 562 carotid endarterectomies, using electroencephalogram (EEG) monitoring and selective shunting, was reviewed. EEG changes occurred in 102 patients (18%). The frequency of EEG changes, as related to cerebral vascular symptoms, was as follows: transient ischemic attacks, seven per cent (19/259); completed strokes, 37% (36/98); vertebral basilar insufficiency, 24% (32/135); asymptomatic, 21% (15/71). Patients with contralateral carotid occlusion exhibited EEG changes in 37% (28/76) of operations. Fifteen patients suffered perioperative strokes (2.6%). Nine of the 15 were associated with a technical problem of either thrombosis of the internal carotid artery (five) or emboli (four). Technical problems were more common when shunts were used (five per cent) than when they were not (0.9%). Patients who suffered strokes prior to surgery were more at risk to develop a perioperative stroke (three per cent) than those not suffering prior strokes (0.3%). The EEG did not change in three patients who had lacunar infarcts prior to surgery and who awoke with a worsened deficit. Our series does not clearly establish the advantages of EEG monitoring, which is expensive (+375/patient) and may not detect ischemia in all areas of the brain. However, the use of shunts may introduce a risk of stroke due to technical error that is equal or greater than the risk of stroke due to hemodynamic ischemia. Since the need for protection is unpredictable by angiographic or clinical criteria, the benefit of EEG monitoring may be in reducing the incidence of shunting in those patients whose tracing remains normal after clamping. The decision to shunt, however, when there is electrical dysfunction after carotid clamping should be based not only on the EEG but also on the clinical signs and computed tomography (CT) scan. Our data does not show a net benefit in selective shunting unless the patient has sustained a stroke prior to surgery.