It has been hypothesised that regional endothermy has evolved in the muscle of some tunas to enhance the locomotory performance of the fish by increasing muscle power output. Using the work loop technique, we have determined the relationship between cycle frequency and power output, over a range of temperatures, in isolated bundles of slow muscle fibres from the endothermic yellowfin tuna (Thunnus albacares) and its ectothermic relative the bonito (Sarda chiliensis). Power output in all preparations was highly temperature-dependent. A counter-current heat exchanger which could maintain a 10 degrees C temperature differential would typically double maximum muscle power output and the frequency at which maximum power is generated (fopt). The deep slow muscle of the tuna was able to operate at higher temperatures than slow muscle from the bonito, but was more sensitive to temperature change than more superficially located slow fibres from both tuna and bonito. This suggests that it has undergone some evolutionary specialisation for operation at higher, but relatively stable, temperatures. fopt of slow muscle was higher than the tailbeat frequency of undisturbed cruising tuna and, together with the high intrinsic power output of the slow muscle mass, suggests that cruising fish have a substantial slow muscle power reserve. This reserve should be sufficient to power significantly higher sustainable swimming speeds, presumably at lower energetic cost than if intrinsically less efficient fast fibres were recruited.