Motor Learning through Induced Variability at the Task Goal and Execution Redundancy Levels

ABSTRACT

The authors examined the influence of introducing variability at two different levels in the learning of a striking task. Variability at the task goal level was introduced by changing target location, whereas variability at the execution redundancy level was introduced by using an intermediate target that constrained participants to use different paths from trial to trial to strike the same target. After practice, participants were transferred to 2 test conditions: (a) a fixed-target test, where the position of the target was unchanged; and (b) a variable-target test, where the position of the target was varied from trial to trial. The results from the manipulation at the task goal level were consistent with predictions from the specificity of practice hypothesis. In both the fixed- and variable-target tests, the best performance was achieved by the group that had practiced in the condition matching the test condition. At the execution-redundancy level, practicing multiple solutions to achieve the task goal did not improve performance in either the fixed- or variable-target tests. These results show that introducing variability at the task goal and execution redundancy levels has different effects on learning and generalization and that practice schedules that constrain the participant to use redundant solutions may not facilitate learning.

Keywords:

• motor learning
• path variability
• redundancy
• task goal
• variability of practice

ACKNOWLEDGMENTS

The authors thank Tim Benner for assistance with the software programming. They also thank Dagmar Sternad, Mark Latash, Joseph Cusumano, Luc Tremblay, and an anonymous reviewer for their helpful comments on an earlier draft of this manuscript.

Notes

1. It is often assumed that contextual interference effects are observed only when practicing variations from different tasks in a random practice schedule (i.e., when the variations are interspersed during learning). Although this has been the dominant paradigm in the literature, interference during learning can be introduced by a variety of factors that include manipulating contextual variety (e.g., practice schedules) or task similarity (for a detailed discussion, see Magill & Hall, 1990).

2. On the suggestion of a reviewer, we analyzed if there was learning during the pretest by splitting the block into 5 subblocks of 10 trials. The percentage of target hits showed a significant main effect of subblock (F(3.1, 87.4) = 30.9, p < .001) with Subblocks 1 and 2 (i.e., Trials 1–20) showing significantly lower number of hits (4% and 17%, respectively) compared with Subblocks 3–5 (34%, 37%, and 42%, respectively). There were also significantly more hits in Subblock 2 compared with Subblock 1. The effects of Group and Subblock × Group were not significant.

3. The computation of constant error (CE) and variable error as measures of bias and variability (Schutz & Roy, 1973) becomes rather arbitrary for two-dimensional tasks and their interpretation becomes especially problematic when the targets change position from trial to trial because these errors are computed relative to the target. For instance, the variable error is no longer a direct indicator of motor variability because it is influenced both by the variability in the motor response as well as the variability in the target location. Therefore, we used the absolute error (measured from the edge of the target) as an index of task performance, and the spatial variability of the paths as an index of motor variability. On the suggestion of a reviewer, we computed CE (measured along the y-axis of the target) and found that apart from the initial block of practice (CE was significantly higher in Block 1), it did not play a significant role in any of the effects found. This suggests that the differences found were driven primarily by changes in variability and not by systematic bias.