The mechanical efficiencies of pure positive (eta) and pure negative (eta-) work were investigated on a special "sledge ergometer" with 25 and 36 subjects, respectively. The work intensities varied in positive work between 40% and 90% and in negative work from 30% to 120% of the maximum concentric exercise. In 54 exercises of positive work, eta was 17.1% +/- 2.2%, and its value correlated negatively with the work intensity (r = 0.367, P less than 0.01) and with the average knee angular velocity, omega+ (r = 0.359, P less than 0.01). In 103 eccentric exercises, eta- was on the average 80.2% +/- 31.8%, correlating positively with the work intensity (r = 0.396, P less than 0.01). Both inter- and intrasubject variations were large (32%-163%). The integrated electrical activity (IEMG) of the leg extensor muscles increased with an increase of work intensity both in the positive and in the negative work situations. Less efficient MU recruitment in higher positive work rates is suggested to be the reason for the decrease in eta, whereas better stiffness regulation via increased preactivation is speculated to cause high values of eta- in higher work intensities in eccentric exercise.