1. The mechanical power spent to accelerate the limbs relative to the trunk in level walking and running, W(int), has been measured at various ;constant' speeds (3-33 km/hr) with the cinematographic procedure used by Fenn (1930a) at high speeds of running.2. W(int) increases approximately as the square of the speed of walking and running. For a given speed W(int) is greater in walking than in running.3. In walking above 3 km/hr, W(int) is greater than the power spent to accelerate and lift the centre of mass of the body at each step, W(ext) (measured by Cavagna, Thys & Zamboni, 1976b). In running W(int) < W(ext) up to about 20 km/hr, whereas at higher speeds W(int) > W(ext).4. The total work done by the muscles was calculated as W(tot) = W(int) + W(ext). Except that at the highest speeds of walking, the total work done per unit distance W(tot)/km is greater in running than in walking.5. The efficiency of positive work was measured from the ratio W(tot)/Net energy expenditure: this is greater than 0.25 indicating that both in walking and in running the muscles utilize, during shortening, some energy stored during a previous phase of negative work (stretching).6. In walking the efficiency reaches a maximum (0.35-0.40) at intermediate speeds, as may be expected from the properties of the contractile component of muscle. In running the efficiency increases steadily with speed (from 0.45 to 0.70-0.80) suggesting that positive work derives mainly from the passive recoil of muscle elastic elements and to a lesser extent from the active shortening of the contractile machinery. These findings are consistent with the different mechanics of the two exercises.