The purposes of this investigation were to examine among a sample of hemiparetic stroke patients the nature and reliability of selected muscle performance measures and the correlation between the measures and gait speed. Fourteen subjects who met specific entry criteria participated. The Lido Active Rehabilitation System was used to obtain bilaterally all muscle performance measures of knee extension (velocity, isometric torque-peak and time specific, and time to peak torque and 90% peak torque). Gait speed was measured under both "comfortable" and "maximum safe" conditions. In addition to lower torques, the subjects demonstrated slower knee extension speeds and longer times to peak torque and 90% peak torque on the paretic than on the nonparetic side. High intrasession reliabilities were demonstrated for all muscle performance measures except the torque at 0.1 sec on the nonparetic side and the time to peak torque and 90% peak torque measures. Significant correlations were demonstrated between both gait speeds and paretic knee extension torque (except at 0.1 sec), nonparetic knee extension torque at 0.1 sec and 0.2 sec, and time to nonparetic peak torque. Muscle performance measures other than peak torque appeared to offer little advantage (alone) over paretic peak torque for predicting gait speed. Over 74% of the variance in gait speed, however, could be predicted by the peak isometric peak torque of the paretic side and the time to peak torque on the nonparetic side, together. Thus, there may be benefits to testing the muscle performance of both lower extremities when examining patients with stroke.