Despite numerous studies investigating the pathophysiology of tourniquet ischemia, definitive data at the cellular level have been lacking and no consensus regarding safe tourniquet-application times in extremity surgery has emerged. In light of the particular vulnerability of skeletal muscle to ischemic injury, we determined the degree of muscular metabolic derangement and cell damage produced by seven different protocols of tourniquet application and release, each providing three hours of total tourniquet time. We performed thirty-six experiments on canine hind limbs, comparing the following time-patterns of tourniquet application: I--three sequential one-hour periods, II--two sequential one and one-half-hour periods, III--two hours followed by one hour, and IV--a single continuous three-hour application. Five and fifteen-minute reperfusion intervals between ischemic periods were compared for the first three time-patterns, creating a total of seven different tourniquet protocols. Muscular metabolic derangement and cell injury were evaluated by monitoring changes in the cellular bioenergetic state (high-energy phosphate profile), cell pH, post-ischemic leakage of creatine phosphokinase, and ultrastructural cell degeneration. At the intracellular level, the metabolic recovery of muscle during reperfusion was much faster than previous studies focusing on extracellular parameters have indicated. In all instances complete intracellular bioenergetic recovery occurred within five minutes after tourniquet release. The use of one or more five-minute reperfusion intervals significantly reduced the degree of ischemic cell injury, as indicated by a decrease in creatine phosphokinase leakage and myofibrillar destruction. No additional benefit was derived by extending the reperfusion periods to fifteen minutes. The longest period of continuous ischemia in each tourniquet-application protocol bore the closest relationship with the amount of cell damage produced. Within the spectrum of observed pathological changes, time-patterns I and II produced comparatively little muscle damage.