The in vivo torque-velocity relationships of the knee extensors (KE), knee flexors (KF), ankle plantarflexors (PF), and ankle dorsiflexors (DF) were determined in 12 untrained subjects using an isokinetic testing device (Cybex II). These data were then matched to the predicted maximum forces and shortening velocities derived from muscle architectural determinations made on three hemipelvectomies (36). The torque-velocity curves of all muscle groups resembled that predicted by Hill's (19, 20) equation except at the higher forces and lower velocities. The peak torques occurred at mean velocities ranging from 41-62 rad X s-1 for the KE, KF, and PF. Although the peak torque of the DF occurred at the isometric loading condition, it was also lower than that predicted by Hill's equation. The muscle fiber length and physiological cross-sectional area measurements indicate that the architecture of the human leg musculature has a major influence on the torque-velocity characteristics. These data corroborate previous findings (24) that some neural inhibitory mechanism exists in the control of the leg musculature, which limits the maximum forces that could be produced under optimal stimulating conditions.