Effects of robotically modulating kinematic variability on motor skill learning and motivation

Abstract

It is unclear how the variability of kinematic errors experienced during motor training affects skill retention and motivation. We used force fields produced by a haptic robot to modulate the kinematic errors of 30 healthy adults during a period of practice in a virtual simulation of golf putting. On day 1, participants became relatively skilled at putting to a near and far target by first practicing without force fields. On day 2, they warmed up at the task without force fields, then practiced with force fields that either reduced or augmented their kinematic errors and were finally assessed without the force fields active. On day 3, they returned for a long-term assessment, again without force fields. A control group practiced without force fields. We quantified motor skill as the variability in impact velocity at which participants putted the ball. We quantified motivation using a self-reported, standardized scale. Only individuals who were initially less skilled benefited from training; for these people, practicing with reduced kinematic variability improved skill more than practicing in the control condition. This reduced kinematic variability also improved self-reports of competence and satisfaction. Practice with increased kinematic variability worsened these self-reports as well as enjoyment. These negative motivational effects persisted on day 3 in a way that was uncorrelated with actual skill. In summary, robotically reducing kinematic errors in a golf putting training session improved putting skill more for less skilled putters. Robotically increasing kinematic errors had no performance effect, but decreased motivation in a persistent way.

Keywords: motivation; motor learning; motor skill; movement variability; robotic training.

Fig. 1.

A: experimental setup for a participant playing the virtual putting game. Participants controlled the head of a virtual putter by using a 3 degrees-of-freedom haptic robot. The computer screen provided participants with visual feedback about their performance, which included a scoring system and a streak counter. B: two sample trajectories, one for the short target and one for the long target, in the velocity vs. position (state-space) domain. The subcomponents of the swing are labeled as follows: backswing, downswing, and follow through. The arrows around the long target represent the error reduction (ER) force field, while the arrows around the short target represent the error augmentation (EA) force field. Note that the force fields were only active during the downswing. [Photograph reprinted with permission.]

Fig. 2.

Experimental protocol. The experiment was carried out on 3 separate days for each participant. On day 1, all participants trained without force fields for 100 putts to become familiar with the task and have their skill level assessed. On day 2, which occurred at least 1 wk after day 1, participants putted 40 trials without haptic input to warm-up and measure pretraining skill, 90 trials with force fields (depending on their group), and 40 trials without force fields to measure short-term skill retention. Finally, on day 3, all participants putted 100 trials without force fields to measure their long-term skill retention. Throughout all days of practice, participants were periodically forced to take breaks of at least 1 min during which they responded to four questions (4Q) concerning motivation during practice; they also answered a set of 13 motivation-related questions (13Q) at the end of each day.

Fig. 3.

Effect of the force fields on kinematic variability. A: near target. B: far target. The ER force field significantly reduced (near target: −56.22 ± 5.80% SEM; far target: −41.68 ± 9.37% SEM) the kinematic variability of participants (paired-sample t-test, near target P < 0.001; far target P = 0.002). The EA force significantly increased (near target: 107.50 ± 24.29% SEM; far target: 101.27 ± 26.01% SEM) the kinematic variability of participants (paired-sample t-test, near target P = 0.002; far target P = 0.004). Stars denote significant differences. CTRL, control.

Fig. 4.

Linear mixed model to analyze motor skill retention. A linear mixed model with three training conditions, two target locations, two retention assessments, and the participants' pretraining skill level was used to measure the retention of motor skill (impact velocity variability in putting) after training with the ER and EA force fields. In this figure, positive changes reflect improvements in motor skill. The pretraining skill level was a significant regressor [β = 412.23, t(84) = −4.12, P < 0.001]; therefore, participants who were initially less skilled at the task benefited more from the training. An interaction between the pretraining skill level and the ER group [β = 246.45, t(84) = 1.97, P = 0.05] revealed that, relative to the CTRL group, initially less-skilled participants benefited more from training with reduced kinematic variability.

Fig. 5.

Offline learning across training groups. We defined offline learning as the change in performance from the end of day 2 to the start of day 3. A: no significant differences in offline learning were found for the short target. B: there was a significant difference in offline learning between training groups for the long target (ANOVA P = 0.03, follow-up Tukey test indicated pairwise difference between ER and EA, P < 0.05). Star denotes significant difference. Error bars represent the standard error.

Fig. 6.

Error-based learning and recalibration errors. A and B: signed errors incurred during the training phase of the experiment. A: near target. B: far target. For error-based learning, we expected errors to decrease for consecutive putts to the same target location. This was evident in the EA group at the short target (ANOVA, P < 0.001; A) and in the CTRL group at the long target (ANOVA, P = 0.037; B). In A and B, the stars indicate significant differences in the error size for a training condition. C and D: signed recalibration error (i.e., the error incurred after there is a change in target location) for the short (C) and long (D) target locations. We found a clear trend for the recalibration error to be in the direction of the previous putt. Participants had a tendency to overshoot the short target and undershoot the long target following a change in the target location. In C and D, the asterisks show significant differences between experimental phases for a given training condition; the stars show significant differences between training conditions for a given experimental phase. Error bars represent the standard error.

Fig. 7.

Responses to the subset of 4Q from the Intrinsic Motivation Inventory (IMI) during training on day 2. The top bar, labeled with ON, indicates when the force field was active (i.e., during breaks 2, 3, and 4). Stars denote significant differences by Kruskal-Wallis ANOVA. Participants in the ER group reported high levels of competence (A) and satisfaction (C), while the EA group reported the opposite. B: the ER group also reported higher effort. D: no differences were found in the reported levels of attention. Error bars represent the standard error.

Fig. 8.

A–F: responses to the subset of 13Q from the IMI given at the end of each day. Only those questions with significant or near significant differences between training groups are shown. There was a significant trend for participants in the EA group to report more negative feelings toward the training at the end of day 2 (A, B, C, and E). Some of these feelings persisted at the end of day 3 (A and B), even though the force field was not active during day 3. In addition, the EA group showed a tendency to report a lower level of effort put into the task on days 2 and 3. Error bars represent the standard error. Stars denote significant difference.