Correlates of online changes in movement representations in 240ms

Mathew, James [UCL] Bastin, M [UCL] Lefèvre, Philippe [UCL] Crevecoeur, Frédéric [UCL]

Humans and animals use adaptive representations of movement dynamics, called internal models, which allow them to adapt motor commands to novel disturbances in the environment. To date this process has largely been described through learning curves characterizing changes in performance across movement repeats, yet recent evidence suggests that adaptive changes in reach representation may occur during a single movement. Here we sought to determine the latency of these rapid changes in movement representation based on electromyogram recordings. Healthy adults (n=18) performed standard reaching movement trials with a robotic handle randomly interleaved with force field trials (lateral force proportional to hand forward velocity) in unpredictable direction. First, we observed that the maximum hand displacement towards the direction of perturbation in a force field trial displayed very small improvement across trials, while the target overshoot was associated with a slower exponential decay. This result supports the presence of adaptive feedback control within each trial. Next, we identified in the main muscles involved in the motor corrections (shoulder flexors and extensors) a clear modulation explaining the reduction in target overshoot. Based on a sliding comparison of the difference between the first and last trials, we measured that the onset of the response tuned to the force field occurred on average at 237±15ms following reach onset (mean±SD). To further assess whether these corrections were tuned to the force field, we regressed the commanded force with the measured force and found highly significant improvements across trials. Moreover, the correlations became better than when measured force of a single trial was compared with commanded force of random trials. This is similar to what we have observed in a control experiment, when the force field is in a single predictable direction, indicating that the adjustments of feedback were tuned to the specific perturbation of each individual movement. These results confirm an online change in reach representation and provide an accurate measurement of the latency of adaptive control in the nervous system. The underlying process is a candidate mechanism for linking feedback control with motor learning.