A randomised, prospective study was performed to compare local (LA) and general anaesthesia (GA) in carotid surgery with special emphasis on complications and the need for intra-operative shunting. Fifty-six patients were randomised to LA and 55 to GA. Eight patients in the LA group required a GA for various reasons. During the same period 14 patients were not randomised. Seven perioperative neurological deficits occurred (5.6%), four in the LA group, two in the GA group, and one in the non-randomised group (NS). Selective shunting was used, in the Ga group according to stump pressure or in cases with a previous stroke and in the LA group according to the appearance of neurological symptoms. In the GA group 25 patients were shunted and in the LA group five patients (P less than 0.001) needed a shunt. If strict pressure criteria for shunting had been used in the LA patients, ten would have been shunted and three of the patients who developed symptoms during clamping would not have been shunted. During surgery the highest recorded systolic pressure was significantly higher in the LA group (210 mmHg versus 173 mmHg, P less than 0.001). LA for carotid endarterectomy is comparable with general anaesthesia regarding peroperative complications but produces significantly higher blood pressures than general anaesthesia. On the other hand it allows the possibility of neurologic monitoring of the patient and leads to significantly less use of an intra-operative shunt.