Twenty-three cases of aortic arch aneurysm, over the past 10 years, were divided into 3 groups of similar age and sex for comparison of operative technique. The aortic arch was reconstructed under normothermic femoral bypass via left posterolateral thoracotomy in 9 patients (group II), under hypothermic circulatory arrest via left posterolateral thoracotomy in 7patients (group II) and under hypothermic circulatory arrest with ascending aorta cannulation and brief retrograde systemic perfusion immediately after finishing reconstruction of the aortic arch via median sternotomy in 7 patients (group III). Surgical mortality in groups I, II and III was 33.3% (3/9), 14.3%(1/7)and14.3%(1/7), respectively. Cerebrovascular complications occurred in 1 of 8 patients (12.5%), excluding an intra-operative death, in group II, and in 4 of 7 cases (57%) in group II, of whom 2 were restored to consciousness. There were no cerebrovascular complications in group III (p< 0.005 v. group II). The patient in group III who died of a lung complication 11 days postoperatively was restored to consciousness. Circulatory arrest time was 30.7 ± 17 and 33 ± 17 minutes in groups II and III, respectively. This study demonstrated that normothermic bypass (group II) resulted in an unsatisfactory surgical outcome, and femoral artery cannulation (group II) caused increased cerebrovascular complications compared with ascending aortic cannulation (group III), suggesting that the retrograde aortic flow, but not circulatory arrest per se, predisposed cerebrovascular morbidity. Vascular embolism by debris could be prevented by normograde perfusion via the ascending aorta and brief retrograde systemic perfusion immediately after finishing reconstruction of the aortic arch.