Speed-emphasized arm training is consistent with improved feedforward control of rapid elbow flexion in chronic stroke

Feedforward control is impaired after stroke, but whether practice improves it in the chronic stage remains unclear. In fast single-joint movements, feedforward control scales initial acceleration to movement amplitude. Feedback-error-learning theory predicts that fast practice should promote recalibration. We tested whether speed-emphasized reach training improved this scaling in the more-affected arm. Thirty-six individuals with mild-to-moderate chronic stroke completed 2,080 reaches over four sessions emphasizing Speed (n = 17) or Accuracy (n = 19). At baseline, 3 days, and 1 month after training, participants performed single-joint elbow movements of three amplitudes without online visual feedback. The slope relating initial peak acceleration to target amplitude quantified feedforward scaling. At baseline, scaling was reduced on the more-affected side for extension (Δβ = −17.1 ± 3.5 s⁻², P < 0.001) and flexion (Δβ = −21.9 ± 3.2 s⁻², P < 0.001). Speed training increased flexion scaling at 3 days (Δβ = 14.7 ± 3.6 s⁻², P = 0.00029), with 14 of 17 participants improving. The group-by-session interaction showed a greater increase after Speed than Accuracy training (Δβ = 14.13 ± 4.9 s⁻², P = 0.0053). Scaling declined by 1 month (P = 0.02); no training-related changes occurred for extension. Peak acceleration occurred within 100 ms. Time to peak velocity, an index of pulse width, showed no group-specific training effect. Thus, speed-emphasized training produced a short-term improvement in early flexion acceleration scaling that transferred to an untrained single-joint task and was consistent with improved feedforward control. Transfer primarily reflected acceleration pulse-height rather than pulse-width modulation.