Variability in motor learning: relocating, channeling and reducing noise

Abstract

Variability in motor performance decreases with practice but is never entirely eliminated, due in part to inherent motor noise. The present study develops a method that quantifies how performers can shape their performance to minimize the effects of motor noise on the result of the movement. Adopting a statistical approach on sets of data, the method quantifies three components of variability (Tolerance, Noise, and Covariation) as costs with respect to optimal performance. T-Cost quantifies how much the result could be improved if the location of the data were optimal, N-Cost compares actual results to results with optimal dispersion at the same location, and C-Cost represents how much improvement stands to be gained if the data covaried optimally. The TNC-Cost analysis is applied to examine the learning of a throwing task that participants practiced for 6 or 15 days. Using a virtual set-up, 15 participants threw a pendular projectile in a simulated concentric force field to hit a target. Two variables, angle and velocity at release, fully determined the projectile's trajectory and thereby the accuracy of the throw. The task is redundant and the successful solutions define a nonlinear manifold. Analysis of experimental results indicated that all three components were present and that all three decreased across practice. Changes in T-Cost were considerable at the beginning of practice; C-Cost and N-Cost diminished more slowly, with N-Cost remaining the highest. These results showed that performance variability can be reduced by three routes: by tuning tolerance, covariation and noise in execution. We speculate that by exploiting T-Cost and C-Cost, participants minimize the effects of inevitable intrinsic noise. Keywords motor learning; variability; noise; sensitivity; skill acquisition It is widely acknowledged that improvement in skilled performance with practice includes a decrease in variability. However, even in expert performance, variability is never entirely eliminated. One explanation for the ubiquitous presence of variability is that the nervous system is inherently noisy. The sensorimotor system is a complex dynamical system with processes Correspondence Address: Dagmar Sternad, Departments of Biology, Electrical & Computer Engineering, and Physics, 134 Mugar Life Science Building, Northeastern University, 360 Huntington Avenue, Boston MA 02115, Phone: 617-373-5093, email: dagmar@neu.edu. 2 A manifold is a mathematical term from differential geometry generally referring to a collection or set of objects (James & James, 1992, 5 th ed.). The set of solutions in the skittles task is a smooth differentiable manifold whether or not the execution variables have different units. The interested reader can find definitions at any level of detail in relevant mathematical texts. For example, James and James (1992, 5 th ed): "A topological manifold of dimension n is a topological space such that each point has a neighborhood which is homeomorphic to the interior of a sphere in Euclidean space of dimension n...

Investigating optimal technique in a noisy environment: Application to the upstart on uneven bars

“…In this sense the variability is described as having a functional role. However, from the results of the present study it can be argued that the gymnast adopts a technique that lies within a solution space that is able to cope with the resultant noise from the motor system (Cohen and Sternad, 2009;Wolpert 2007). In other words the solution space surrounding the chosen technique is relatively insensitive to the inherent kinematic variability.…”

“…Human movement variability has been described as having a functional role (Hamill, van Emmerick, Heiderscheit and Li, 1999;van Emmerick, Hamill, and McDermott, 2005;Bartlett et al, 2007), but it has also been described as having a component of "noise" (Wolpert, 2007;Cohen and Sternad, 2009) and as being due to redundancy in the sensorimotor system (Newell and Corcos, 1993). As variability will always be present in human movement (Bartlett et al, 2007) it will have an important influence on the success of a given task and therefore the technique adopted by athletes.…”

“…Cohen and Sternad (2009) used a decomposition of movement variability into components of noise, tolerance (how far the movements were from the successful solution space) and covariation (variation that was compensated for by redundancy) in relation to the task outcome. As it was not possible to break the variability down into its components and since the present simulation model was not capable of making feedback corrections, the assumption was made that only the noise component of variability was being considered.…”

“…The average standard deviations in the temporal and spatial measures were 10 ms and 2.3º respectively for a senior elite male gymnast. The measures of kinematic variability (Hiley and Yeadon, 2012b) would have included components of noise and any feedback corrections made by the gymnast (Cohen and Sternad, 2009;Harris and Wolpert, 1998). When the gymnast's giant circle technique was optimised, realistic solutions were obtained only when the variability present in the actual performances was included.…”

Hand selection in a preferential reaching task: The effects of object location, orientation, and task intention in preadolescent children

“…The difference was particularly large for young children, and more apparent between ages 5–6 and 7–9. The decrease in variability and increase in smoothness was interpreted as evidence of automatization (Cohen & Sternad, 2009) with increasing age (Rudisch et al. (2018).…”