Eight-five carotid endarterectomies were performed in 77 patients, under regional anaesthesia using 2 different techniques: cervical epidural anaesthesia (35 cases) and cervical plexus block (50 cases). The patients' mean age was 71 years; 80 per cent had arterial hypertension and 41 per cent coronary disease. Transoperative cerebral ischaemia was detected by a 5-minute carotid clamping test, the occurrence of a neurological event indicating that shunting was required. In 62 patients this test was combined with measurement of carotid back pressure. None of the patients needed general anaesthesia. Intraoperative neurological events occurred more frequently (P less than 0.01) when the carotid back pressure was 25 mmHg or less, and 12 temporary shunts were installed for that reason (14.1 per cent). Three neurological events occurred at the end of endarterectomy: no shunt was installed and complete recovery was observed immediately after declamping. No complications ascribable to the anesthetic techniques were recorded. Mortality was nil, and the only neurological morbidity was a brachio-facial deficit which left few sequelae. The frequency of intra- or postoperative arterial hypertension was similar in both groups. Intraoperative hypotension, frequent under epidural anaesthesia, was observed in only one patient who had brachial plexus block (P less than 0.01). The analgesia obtained was equally good with both anaesthetic techniques, but cervical plexus block anaesthesia is easier to perform, had less haemodynamic repercussions and therefore tends to be preferred to cervical epidural anaesthesia. The lack of mortality, low morbidity and absence of systemic complications in this series despite the high number of patients at risk are in favour of this type of anaesthesia, notably for such patients. Moreover, because vigilance is preserved attention can be paid to the quality rather than the rapidity of endarterectomy, which is the best way of preventing embolism.