In this study the torque output of intermittently stimulated paralyzed human knee extensor muscles during imposed isokinetic cyclical lower leg movements was investigated in four paraplegic subjects. During prolonged (10 min) experiments the influence of knee angular velocity and stimulation parameters on fatigue-induced torque decline was studied. Pulse width and amplitude were set to obtain maximal recruitment. The cycle time was maintained constant at 2 s, comparable to a walking cycle. The maximum torque and averaged torque per cycle were estimated to determine the muscle's performance during sustained intermittent stimulation. The overall loss in time of these parameters had a typical exponential decay reaching asymptotic values. Additionally, larger knee velocities resulted in a significantly faster and relatively larger decay of maximum and averaged torque. Also, the rate and relative decrement of torque output during concentric contractions increased with increasing number of pulses in a cycle. Identification trials, determining the (isometric) torque-angle and (isokinetic) torque-angular velocity relation, were performed. The relations appeared to change due to fatigue. The results might be valuable in the design of optimal control systems for functional electrical stimulation which pursue minimization of muscle fatigue. They may contribute to the derivation of a cost criterion, describing muscle fatigue as a function of both joint movement and stimulation parameters.