Biomechanics (and energetics) of human locomotion are generally studied at constant, linear, speed whereas less is known about running mechanics when velocity changes (because of accelerations, decelerations or changes of direction). The aim of this study was to calculate mechanical work and power and to estimate mechanical efficiency in shuttle runs (as an example of non-steady locomotion) executed at different speeds and over different distances. A motion capture system was utilised to record the movements of the body segments while 20 athletes performed shuttle runs (with a 180° change of direction) at three paces (slow, moderate and maximal) and over four distances (5, 10, 15 and 20 m). Based on these data the internal, external and total work of shuttle running were calculated as well as mechanical power; mechanical efficiency was then estimated based on values of energy cost reported in the literature. Total mechanical work was larger the faster the velocity and the shorter the distance covered (range: 2.3-3.7 J m kg) whereas mechanical efficiency showed an opposite trend (range: 0.20-0.50). At maximal speed, over all distances, braking/negative power (about 21 W kg) was twice the positive power. Present results highlight that running humans can exert a larger negative than positive power, in agreement with the fundamental proprieties of skeletal muscles in vivo. A greater relative importance of the constant speed phase, associated to a better exploitation of the elastic energy saving mechanism, is likely responsible of the higher efficiency at the longer shuttle distances.