The primary purpose of this investigation was to study the eccentric and concentric torque-velocity characteristics of the quadriceps femoris in man using a recently developed combined isometric, concentric and eccentric controlled velocity dynamometer (the SPARK System). A secondary purpose was to compare the method error associated with maximal voluntary concentric and eccentric torque output over a range of testing velocities. 21 males (21-32 years) performed on two separate days maximal voluntary isometric, concentric and eccentric contractions of the quadriceps femoris at 4 isokinetic lever arm velocities of 0 degree.s-1 (isometric), 30 degrees.s-1, 120 degrees.s-1 and 270 degrees.s-1. Eccentric peak torque and angle-specific torques (measured every 10 degrees from 30 degrees to 70 degrees) did not significantly change from 0 degrees.s-1 to 270 degrees.s-1 (p greater than 0.005) with the exception of angle-specific 40 degrees torque, which significantly increased; p less than 0.05). The mean method error was significantly higher for the eccentric tests (10.6% +/- 1.6%) than for the concentric tests (8.1% +/- 1.7%) (p less than 0.05). The mean method error decreased slightly with increasing concentric velocity (p greater than 0.05), and increased slightly with increasing eccentric velocity (p greater than 0.05). A tension restricting neural mechanism, if active during maximal eccentric contractions, could possibly account for the large difference seen between the present eccentric torque-velocity results and the classic results obtained from isolated animal muscle.