This study established the reliability of a novel upper-limb trajectory-tracking task for assessment of perceptual motor control in hemiparetic adults. Eleven persons with chronic poststroke hemiparesis (mean 58.6 months) and eleven nondisabled control subjects performed an elbow flexion-extension task against a low-resistance isotonic load at three speeds: 25 degrees/s, 45 degrees/s, and 65 degrees/s. Both arms (paretic and nonparetic or dominant and nondominant) were tested during two identical sessions separated by 1 week. Relative reliability (intraclass correlation coefficient [ICC]) ranged from 0.5 to 0.8 and absolute reliability (standard error of measurement [SEM%]) ranged between 19% to 36% across both subject groups. No systematic errors between test sessions were revealed. Smallest real differences (SRDs) were determined to be +/- 2 degrees to 3 degrees in nondisabled, +/- 2 degrees to 5 degrees in nonparetic and +/- 9 degrees in paretic arms. Responsiveness ratios derived with the use of the SRDs ranged between 1.91 to 2.45, indicating that this instrument is sensitive to clinically important change and suitable for demonstrating effects on upper-limb motor performance following clinical intervention.