Movement speed is biased by prior experience

How does the motor system choose the speed for any given movement? Many current models assume a process that finds the optimal balance between the costs of moving fast and the rewards of achieving the goal. Here, we show that such models also need to take into account a prior representation of preferred movement speed, which can be changed by prolonged practice. In a time-constrained reaching task, human participants made 25-cm reaching movements within 300, 500, 700, or 900 ms. They were then trained for 3 days to execute the movement at either the slowest (900-ms) or fastest (300-ms) speed. When retested on the 4th day, movements executed under all four time constraints were biased toward the speed of the trained movement. In addition, trial-to-trial variation in speed of the trained movement was significantly reduced. These findings are indicative of a use-dependent mechanism that biases the selection of speed. Reduced speed variability was also associated with reduced errors in movement amplitude for the fast training group, which generalized nearly fully to a new movement direction. In contrast, changes in perpendicular error were specific to the trained direction. In sum, our results suggest the existence of a relatively stable but modifiable prior of preferred movement speed that influences the choice of movement speed under a range of task constraints.

individuals tend to move at a preferred speed: some talk slowly, whereas others walk fast (Slijper et al. 2009). Theoretical models suggest that the chosen speed is a compromise regarding the importance or reward value of the goal state (Xu-Wilson et al. 2009), the effort or energy needed to execute the movement (Mazzoni et al. 2007), and the cost of the wait before the goal is reached (Haith et al. 2012; Shadmehr 2010; Shadmehr et al. 2010; Tanaka et al. 2006). The motor system chooses a speed that optimizes a cost function (Todorov and Jordan 2002) representing a combination of these factors. For example, speed of walking with an arthritic hip could be a compromise between reaching the goal of catching the bus and minimizing pain.

Here, we ask whether the choice of movement speed, rather than being an optimal solution to an internal cost function, is also determined by the habit of moving at a specific movement speed that has formed during prior repetitive training. Previous studies have shown that a use-dependent learning mechanism strongly influences spatial characteristics (Diedrichsen et al. 2010; Verstynen and Sabes 2011). When Verstynen and Sabes (2011) trained people to move repeatedly in one direction, the variability of reaches in the trained direction decreased substantially. However, this came at the cost of biasing movements in less-frequently cued directions toward the trained direction. Therefore, movement speed may also be habitual in that a certain speed may be chosen because recent movements have tuned the system to this trained speed. If this habitual preference can be imposed by previous training, this would have important implications for ongoing training and treatment programs.

Our hypothesis was that training movements at a specific speed could similarly bias the speed of subsequent movements. Over 3 consecutive days of practice, a group of healthy young volunteers practiced a center-out arm-reaching task; half of them were trained to move fast, whereas the others were trained to move slowly. Before and after the training sessions, participants were required to make reaching movements within four different time windows. One of these was the same as the trained task; the three others were different. At the end of training, people in both training groups made more accurate movements at less variable speeds, but this came at the cost of biasing the speed in the untrained movements. Those who had been trained to move slowly tended to move slower than before training, whereas those who had been trained to move fast did the opposite. The changes in speed preference generalized to other movement directions, indicating that the underlying adjustment is relatively global.