Simulation of sudden death in a potential heart graft donor was accomplished by producing acute anoxic arrest in canine hearts that were then maintained at cadaver temperature for periods of from 15 to 60 minutes prior to reperfusion with oxygenated blood. Functional recovery of reperfused hearts measured by contractile force was more rapid for hearts that had undergone shorter periods of anoxic arrest. The rates of recovery were nearly identical within the group for 15 to 25 minutes and for 30 to 45 minutes of anoxic arrest, but not for the 60-minute anoxic arrest group in which two of four hearts made no recovery at all. Ultrastructural reactions to anoxic arrest were evident at 15 minutes and found to progress in a predictable manner as the duration of anoxia increased. Changes occurring during anoxic periods of up to 30 minutes were for the most part reversed within 1 hour of perfusion during which time characteristic nuclear, cytoplasmic, and mitochondrial "recovery patterns" were encountered. With a "pump-dog" preparation, these reperfused hearts were well kept for up to 7 hours. Anoxic arrest of 45 minutes or more produced ultrastructural changes that were less consistently reversed by reperfusion of 1 hour. The sensitivity of mitochondrial structure to anoxic arrest and reperfusion was demonstrated, and suggested that irreversible cell damage (i.e., cell death) in anoxia is due to mitochondrial failure. The techniques of perfusion and ultrastructural sampling employed in these experiments should prove useful in studying the effects of hypothermia, drug application and other measures in the preservation and restoration of isolated hearts.