The reliability of selective shunting based on computerized electroencephalographic (EEG) monitoring has not been addressed. In this study, 103 carotid endarterectomies were performed with selective shunting based on a two-channel computerized EEG monitor that processed the on-line, raw electroencephalogram (EEG) to produce a compressed spectral array (CSA). Ischemic EEG events were identified by amplitude attenuation of the raw EEG and/or loss of high-frequency activity on the CSA. Fourteen patients (13.6%) received a bypass shunt, and postoperative neurological examinations showed 97 patients (94.2%) to be intact. A correlation between total (cumulative) ischemic EEG time and the postoperative neurological exam was demonstrated (P less than 0.0001). Six postoperative deficits (5.8%) occurred, five in patients whose computerized EEGs demonstrated an ischemic EEG event late during carotid clamping, when it was no longer possible to place a shunt. The sixth deficit was found in a patient whose EEG did not demonstrate any patient whose EEG did not demonstrate any signs of cerebral ischemia. Five of these six new deficits resolved within 12 hours, and only one persisted for 72 hours, when the patient died of a pulmonary embolism (cerebral infarction and mortality rate of 1%). These results appear to demonstrate that two-channel monitoring of both the CSA and the unprocessed (raw) EEG simultaneously can be used as a reliable indicator of whether a bypass shunt is required during carotid cross-clamping in all patients, regardless of their preoperative neurological history or angiographic findings.