Somatosensory evoked potentials (SEPs) were monitored in the course of 368 carotid endarterectomies (CEAs) carried out in 312 patients. In an initial group of 26 patients the shunt was used routinely while in a second group, involving 342 CEAs, it was applied selectively on the basis of modifications which the SEP underwent during clamping. The criterion for shunting was the progressive reduction, up to 50%, of the N20-P25 amplitude. New postoperative neurological deficits appeared in 6 patients, all of whom displayed a transitory SEP flattening. The SEPs of 2 of these returned to normal by the time they awoke and both showed a clinical deficit homolateral to the operated side. In only 2 cases did the deficit fail to regress completely and their postoperative CT scans revealed ischaemic lesions. A positive relationship emerged between SEP changes and back pressure values; nonetheless, as many as 75% of the patients with low residual back pressure values (less than 25 mmHg) tolerated the clamping. SEP monitoring appears to provide a reliable basis for selectively applying a shunt when there is a high risk of haemodynamic ischaemia during clamping.