A system has been developed for performing 31P-n.m.r. studies on cerebral tissues superfused in vitro, and gives results comparable with those reported from studies in vivo. Under optimal superfusion conditions [10 mM-glucose and O2/CO2 (19:1)] the tissue concentrations of phosphocreatine and ATP were calculated to be approx. 3.1 and 1.3 mumol/g respectively. When the glucose of the superfusing medium was lowered to 0.5 mM, slightly decreased sugar phosphate peaks were observed, but there was no detectable change in [ATP] or [phosphocreatine]. At 0.2 mM-glucose, significantly decreased concentrations of phosphocreatine and ATP were observed. Substitution of pyruvate plus malate for glucose did not decrease levels of phosphocreatine and ATP. When the superfusing medium was gassed with air/CO2 (19:1; 'mild hypoxia'), there was an appreciable fall in sugar phosphates and phosphocreatine with no detectable effect on ATP. In the presence of N2/CO2 (19:1; 'severe hypoxia', since O2 was not completely excluded), concentrations of phosphocreatine fell considerably, but with little effect on ATP. The results demonstrate the feasibility of studying cerebral energy metabolism in vitro using the non-invasive 31P-n.m.r. technique and are discussed in relation to the sensitivity of cerebral tissues to metabolic insults in vitro and in vivo.