Regular physical exercise has beneficial effects in many human disease states, including cardiovascular diseases, cancer, and depression. Exercise training of genetically modified mouse models may provide insight into the molecular mechanisms that underlie the beneficial effects of exercise. Presently, there is relatively little understanding of the normal physiology of mouse exercise. In this paper, we describe a novel computerized voluntary wheel-running system capable of recording and analyzing individual wheel rotations. Using this system, we demonstrate that C57BL/6 mice run considerable distances during the night in short bouts and at a preferred speed: the cruising speed. We find that the vast majority of running occurs around this cruising speed, which is close to the maximum speed at which the animal can run but is significantly higher than the average speeds recorded by simple digital odometers. We describe how these parameters vary with exercise training and demonstrate marked sex differences in the patterns of voluntary exercise. The results of this study have important implications for the design and interpretation of both voluntary and forced exercise experiments in mouse models. The novel parameters described provide more physiological quantitative measures of voluntary exercise activity and training and will extend the physiological utility of exercise training as a phenotyping tool in genetic mouse models.