The purpose of this study was to establish test-retest reliability of measurement procedures for quantifying isokinetic concentric peak torque (PT) at the knee using normalization methods post-stroke. A second aim was to estimate the change required to show clinically significant improvements in knee muscles strength. The isokinetic normalized PT (NPT) values for the knee extensors and flexors were measured in each participant at two different angular velocities during two sessions 1 day apart. Thirty participants with mild to moderate hemiparesis after stroke who were able to walk were tested. The normalized PT measures for the knee muscles of the affected lower extremity were highly reliable (intraclass correlation coefficients ranged from 0.85 to 0.98; p<0.05). Size of relative changes (the percent smallest real difference, SRD%) for extensors NPT (ranged from 22.35% to 25.68%) were lower than flexors NPT (ranged from 74.01% to 76.31%), indicating that the affected isokinetic knee flexors had more random variation than the knee extensors. This study supports the use of isokinetic dynamometers for the assessment of knee muscle strength in participants with chronic mild to moderate post-stroke hemiparesis and to measure clinical improvements. Established measurement error and smallest real differences in normalized PT will aid interpretation of real changes in muscle strength in this clinical population.