An Evaluation of the Autonomy Support Predictions and Recommendations in OPTIMAL Theory

Title: An Evaluation of the Autonomy Support Predictions and Recommendations in OPTIMAL Theory
Authors: McKay, Bradley James
Date: 2020-02-14
Abstract: Research in the field of motor learning focuses on how practice variables impact the retention of movement skill. The Optimizing Performance Through Intrinsic Motivation and Attention for Learning (OPTIMAL) theory is a modern and influential theory of motor learning, hypothesizing that three factors underlie optimal practice: autonomy support, enhanced expectancies, and an external focus of attention (Wulf & Lewthwaite, 2016). The purpose of this dissertation was to test the OPTIMAL theory predictions and recommendations related to the autonomy support factor as well as a key prediction about the relationship between practice performance and learning. Specifically, OPTIMAL theory predicts that providing autonomy support, usually by presenting learners with choices during practice, enhances a) expectancies for successful performance, b) motor performance during practice, and c) motor learning, as measured after at least a 24-hour delay. In addition, OPTIMAL theory predicts that, in general, conditions that enhance performance enhance learning. Emanating from these predictions are recommendations that instructors ask learners to choose when they receive feedback as well as when to practice specific skills in order to support their autonomy. These recommendations put OPTIMAL theory in conflict with previous theories of motor learning that predict certain feedback (guidance hypothesis) and practice schedules (schema theory) may be optimal for performance but suboptimal for learning. These conflicts were addressed by relying on the OPTIMAL theory prediction that instructionally-relevant choices, such as when to receive feedback, and instructionally-irrelevant choices, such as which colour golf ball to putt, each support autonomy and affect motor learning and performance similarly. In this way, it was possible to manipulate autonomy support, feedback schedule, and practice schedule independently, and test whether scheduling variables have an effect when learners have their autonomy supported. This dissertation tested the autonomy support predictions and recommendations in OPTIMAL theory by conducting two increasingly high-powered factorial experiments and a bias-corrected meta-analysis. In the first experiment, a total of 72 participants were assigned to one level of two independent variables: autonomy support (choice, yoked) and knowledge of results (KR) frequency (100%-KR, 50%-faded-KR). The experiment involved putting golf balls on an indoor putting surface with the goal of stopping the ball on a specific point on the green. Participants putted from behind a screen that occluded their vision of the outcome of their putts, forcing them to rely on augmented KR to learn to improve their putting accuracy. Baseline putting accuracy, expectancies, and autonomy measures were recorded and then participants completed ten blocks of five practice trials. Participants in the autonomy supportive conditions were asked to choose from three different colours of golf ball (green, yellow, red) for each putt during practice. Conversely, participants in the yoked conditions were matched to the golf ball colour schedule chosen by a counterpart in the autonomy support conditions. Participants also received KR pertaining to the final location of their putts. In the 100%-KR conditions, participants received KR following all 50 practice putts. Conversely, participants in the 50%-faded-KR conditions received KR after half of their practice putts, beginning with higher frequencies of KR that were gradually reduced across practice blocks. Following practice, participants completed autonomy and expectancy measures and then returned the following day to complete a retention test and a transfer test to a new, farther target than had been practised. The results displayed no statistically significant effects of autonomy support or feedback frequency on expectancies, learning, or performance. Due to the null results in Experiment 1, the power was substantially increased in Experiment 2. A total of N = 128 participants were assigned to one level of two independent variables: autonomy support (choice, yoked) and practice schedule (variable, constant). Participants practiced a dart throwing task with their non-dominant hand for 12 blocks of three darts. Participants in the autonomy support conditions were asked to choose from three colours of dart flight for each of their practice blocks, while their counterparts in yoked conditions followed the same dart-colour schedule. The variable practice conditions rotated among three different targets during practice, while the constant practice conditions threw all their darts to the central target. The day after practice, participants returned to perform a retention test to the central target and a transfer test to a previously unpracticed target. Following a pre-registered analysis plan, the results revealed that the autonomy support groups reported higher levels of perceived autonomy need satisfaction following practice. Further, the yoked groups performed significantly more accurately than the autonomy support groups, while there were no significant differences between groups at transfer. The constant practice conditions performed more accurately during practice, but less accurately during transfer than the variable practice conditions and the change in relative effectiveness from practice to transfer was significant. These results were inconsistent with OPTIMAL theory and call into question some of its central predictions. Critically, the basic prediction that autonomy support is beneficial to motor learning was not supported in either experiment. In order to evaluate the evidence that providing choice during practice is beneficial for motor learning, as well as the hypothesis that all choices are equally effective, a meta-analysis of the so-called self-controlled learning literature was conducted. Following a pre-registered plan, a search for all randomized experiments that met the following criteria was conducted: 1) A self-control group in which participants were asked to make at least one choice during practice, 2) a yoked-group that experienced the same practice conditions as the self-controlled group, 3) a delayed ~24-hour retention test or test with longer delay interval, 4) an objective measurement of motor performance, and 5) publication in a peer-reviewed journal or acceptance as part of a Master’s or PhD thesis. A total of 79 experiments were identified that met the inclusion criteria but 23 were excluded because there were insufficient data to calculate effect sizes after contacting the authors. The resultant sample of k = 56 experiments were submitted to a naïve random effects model which estimated the average effect of self-controlled practice as g = .43. Publication status significantly moderated the self-controlled practice effect, accounting for 42.6% of the heterogeneity in the sample. While published experiments displayed a strong benefit of self-controlled practice, unpublished experiments found almost zero effect. To correct for selection based on statistical significance, a weight-function model was fit to the data with .025 one-tailed p-value cutpoint. Adjusting for selection provided a substantially better fit to the data and the adjusted estimated average effect size was g = .13 and not statistically significant. A suite of sensitivity and exploratory analyses were conducted and the results converged with the weight-function model estimate. Consistent with the OPTIMAL theory prediction, choice-type did not significantly moderate the effect of self-controlled learning. This finding was difficult to interpret however, as the more important finding of the meta-analysis was that the effect of self-controlled learning could not be distinguished from zero based on the available evidence. Overall, the results of this dissertation contradict the OPTIMAL theory predictions and recommendations tested.
CollectionThèses, 2011 - // Theses, 2011 -