Cerebral blood flow was measured both under conditions of normocapnia and hypercapnia in 22 diabetic patients and 20 normal control subjects, using either the intravenous 133Xe method or the closely comparable 133Xe inhalation method. While 19 out of 20 control subjects responded appropriately to hypercapnia with an increase in flow, eight of the diabetic patients failed to respond normally, this difference being significant (p = 0.03). Those manifesting an abnormal response included young, insulin-dependent patients with a short duration of diabetes and no clinical evidence of complications.