We established reliability of upper-limb muscle performance in adults with post-stroke hemiparesis. Ten adults with post-stroke hemiparesis (51.5 +/- 34.5 months) performed isokinetic concentric shoulder flexion, elbow flexion, and extension at 3 criterion speeds (30 degrees/s, 75 degrees/s, and 120 degrees/s) on 3 separate occasions (Days 1, 7, and 49). As several participants were unable to reach criterion speeds, actual speed and power were also analyzed. Relative reliability (intraclass correlation coefficient-ICC) was excellent for torque and power (0.82 to 0.98) but less consistent for speed (0.63 to 0.99). Absolute reliability (standard error of measurement-%SEM) ranged between 0% to 34%. No systematic errors were observed across sessions. Smallest real differences (SRD) ranged between 4-11Nm for torque and 3-24W for power. Shoulder flexion, elbow flexion, and elbow extension torque, speed, and power can be measured reliably following stroke. Given that many of these individuals demonstrate inability to generate torque at preselected speeds, power may be a more valid indicator of muscle performance.