If human death is defined by brain death, its diagnosis needs medicolegal criteria based on clinical examination and EEG. However, this evaluation could be difficult because technical or physiological limitations might impair the interpretation, especially after barbiturates and/or hypothermia. Since brain death is characterized by an intracranial circulatory arrest, methods assessing this phenomenon are warranted. Among these methods, conventional or isotopic cerebral angiography appears the most promising, but it cannot be easily performed everywhere. Because superficial blood flow in arteries is now accurately measured by the pulsed Döppler technique, this prompted us to test the specificity and sensitivity of common carotid blood flow data for brain death diagnosis. Two series of age-matched patients (36 yrs in mean) were studied. Series 1 (n = 28) was used to define the discriminant parameters between 14 severe coma patients and 14 brain-dead patients diagnosed by the classical criteria. Then these parameters were prospectively tested in a blind manner on a second series of 28 patients suffering from severe coma. The parameters allowing us to classify patients as brain dead or not with a 100% specificity and sensitivity were: end diastolic blood flow (QED in ml/min), end diastolic blood flow velocity (VED), and cerebral metabolic index (CMI = QED x AV D 02). The most powerful discriminant parameter was QED, allowing a strictly non-invasive diagnosis of brain death.