Relationship between prediction-based motor control during loading task and motor learning during lever-pressing task

Motor performance depends on feedback and feedforward motor control systems, and can be improved through learning processes. According to the “feedback error learning” model, the feedback of error signals improves “internal models” and refines the feedforward motor control. Thus, feedforward motor control plays a key role in improving motor performance. Feedforward motor control has been evaluated by measuring predictive components of movement in several tasks, including ball-catching, grasping, and weight-loading tasks. In the loading task, hand movement just before the start of loading （anticipatory response） is observed only when the timing of loading is predictable. Thus, this anticipatory response is assumed to reflect prediction-based feedforward motor control. On the other hand, a multi-lever-pressing task has been used to evaluate motor performance and its improvement by analyzing accelerometer signals. Correlation coefficients of accelerometer signals have been reported to increase with the number of trials, indicating that this measure can be used as an index of motor learning. In the present study, we examined the relationship between feedforward motor control and motor learning in 18 healthy volunteers using anticipatory responses in a loading task and correlation coefficients of accelerometer signals in a three-lever-pressing task. For the loading task, we used the Space Interface Device for Artificial Reality （SPIDAR）. The subject was asked to hold the ball-shaped grip of SPIDAR. When the subject pressed the start button, a force of 4.9 N was applied to the grip. The subject was instructed to maintain the initial position during loading. The loading task was repeated 10 times, and the amplitude of upward deflection （anticipatory response） just before the start of loading was measured. In the three-lever-pressing task, the subject was instructed to press three levers as rapidly as possible using the left hand （hand）, the left hand loaded with a weight （weight）, and a stick attached to the left hand （stick）. The three-lever-pressing task was repeated 11 times in sequence under each condition （hand, weight, stick）. The hand movement was monitored using an accelerometer attached to the dorsal surface of the left hand. We found that correlation coefficients of accelerometer signals were lower in the stick condition than in the other two conditions, indicating that the stick variation of the task requires more learning. We also found that the amplitude of anticipatory response was correlated with the correlation coefficients of accelerometer signals only in case of the stick variation. These results provide evidence for a relationship between prediction-based feedforward motor control and motor learning.