Forty men were tested with a computerized dynamometer for concentric and eccentric torques during arm flexion and extension at 0.52, 1.57, and 2.09 rad.s-1. Based on the summed concentric and eccentric torque scores, subjects were placed into a high strength (HS) or low strength (LS) group. The eccentric and concentric segments of the torque-velocity curves (TVCs) were generated using peak torque and constant-angle torque (CAT) at 1.57 and 2.36 rad. Angle of peak torque was also recorded. Compared to LS, HS had significantly greater estimated lean body mass (+10.2 kg) and approximately 25% greater average torque output. Reliability of the peak torque scores on 2 days in 20 subjects was r greater than or equal to 0.85. The difference between observed torques and the mathematically computed criterion torque scores averaged 1% for three validation loads that ranged from 11.4 to 90.4 kg. Statistical analysis revealed that torque output in LS plateaued at low concentric velocities and was also flattened with increasing eccentric velocities. Conversely, torque output for HS increased with decreasing concentric velocities and increased with increasing eccentric velocities. The method of plotting the TVCs for peak or CAT did not influence the pattern of TVC. Eccentric flexion peak torque occurred at a significantly shorter muscle length (1.88 rad) than concentric torque (2.12 rad). This difference was also present for extension; it was 1.88 rad for eccentric and 2.03 rad for concentric torque.(ABSTRACT TRUNCATED AT 250 WORDS)