The absence of spontaneous respiration is a crucial determinant in the diagnosis of brain death, but standardized criteria for apnea testing have not been established. Guidelines are proposed based on the results of 51 apnea tests and associated physiological measurements. In patients who fulfilled all other conventional criteria for brain death, three exhibited non-repetitive back arching and shoulder shrugging when CO2 pressures reached 41 to 51 mm Hg during apnea testing. These respiratory-like movements were ineffective for ventilation and were not reproducible on the following day at the same of higher pCO2. The nature of these movements, evoked potential testing, and autopsy results suggest that they were not triggered by normal medullary centers, and that these patients were, in fact, brain-dead. In four other patients with severe brain damage sparing only the medulla, normal spontaneous ventilation resumed at CO2 pressures of 30 to 39 mm Hg (mean 34 mm Hg). High arterial oxygen tension raised this apnea point slightly, but spontaneous breathing always began at CO2 pressures lower than 40 mm Hg. This level is therefore adequate to stimulate medullary respiration in patients with severe brain damage who are not brain-dead. In brain-dead patients, pCO2 rises slowly during apnea (2.58 +/- 0.85 mm Hg/min), in part because CO2 production is diminished (1.8 +/- 0.23 ml/min/kg). These data allow estimation of a desired length of an apnea test and standardized interpretation of results.