One hundred and twenty-nine aortic aneurysm patients (true 68 and dissection 61) underwent aortic arch repair from January 1987 to December 1995. Postoperative brain damage was evaluated regarding both preoperative brain lesions and techniques of selective cerebral perfusion (SCP) in that one pump for SCP until April 1992, then two pumps were employed, one for right and the other for left hemisphere of the brain. Overall hospital mortalities were 21% in true and 13% in dissecting aneurysms. Ten patients were complicated with postoperative brain damages (coma 8 and hemiplegia 2), registering 7.8% of total patients. Both history of stroke and silent cerebral infarction (SCI) detected by CT and/or MRI were considered to be positive in the mean of having preoperative brain lesions. Although there was no significant difference between incidences of postoperative brain damage in true and dissecting aneurysms, registering 11% and 4% respectively, preoperative brain lesions in true aneurysm (68%) was significantly greater than dissecting aneurysm (32%). Furthermore, the incidence of postoperative brain damage was 22% in one-pump SCP which was significantly greater than 3% in two-pumps SCP in the patients with a true aneurysm. But there was no postoperative brain complication in patients with SCI even using either one two pumps for SCP. The present data suggest two-pumps SCP is better technique for cerebral protection compared to one-pump SCP for aortic arch repair.