It is well known that augmented feedback provided during or after trial completion can be manipulated to enhance motor learning. Following a movement, the learner is able to evaluate performance based on internally generated, intrinsic feedback provided through the perceptual and sensory systems. Augmented feedback, in contrast, is provided by an external source, such as a trainer or therapist, and provides error information that can be used in addition to the learner's own intrinsic error signals. Knowledge of results (KR) is one form of augmented feedback and is defined as "verbal (or verbalizable), terminal (i.e., post-movement) feedback about the outcome of the movement in terms of the environmental goal"
(Schmidt and Lee, 1999, p. 325). KR can be provided after every practice trial, after only a portion of trials, termed reduced relative frequency of KR, or when performance falls outside a pre-determined acceptable range, called bandwidth KR. Here we review the effect of reducing the relative frequency of KR as an example of how manipulating augmented feedback can impact motor skill learning.
Relative frequency of KR refers to the percentage of total practice trials for which KR is provided, such as 50% (half of the trials) or 25% (one-quarter of the trials). The prescribed relative frequency or scheduling can be accomplished in different ways. For example, a 50% relative frequency schedule may be achieved by providing KR on every other trial. It may also be achieved by use of a faded schedule whereby an average relative frequency of 50% is obtained by providing more frequent KR during early practice (say 100%) and progressively reducing the frequency over trials (e.g., to 25%) (Winstein and Schmidt, 1990; Schmidt and Lee, 2005).
While several studies found no benefit of a decreased KR frequency on learning a linear positioning task (Sparrow and Summers, 1992; Sparrow, 1995), multiple studies have found a beneficial effect on the learning of other motor tasks in healthy young adults (Ho and Shea, 1978; Wulf and Schmidt, 1989; Lee et al., 1990; Winstein and Schmidt, 1990; Vander Linden et al., 1993; Wulf et al., 1993, 1994; Winstein et al., 1994; Lai and Shea, 1998; Goodwin et al., 2001). This benefit seen when KR is provided only after a percentage of practice trials has also been demonstrated in older, healthy adults (Behrman et al., 1992; Swanson and Lee, 1992) except in one study where a reduced frequency KR schedule did not benefit either younger or older subjects (Wishart and Lee, 1997). Why might reducing the frequency of augmented feedback benefit learning? KR is beneficial in that it provides useful information that leads to improved performance on the next trial and, therefore, facilitates learning of the motor task. If feedback is provided after every trial or at high relative frequencies though, the learner may become dependent on the KR such that it can be detrimental to performance on trials without KR (i.e., retention test) and actually degrade learning. In essence, an optimal schedule of KR provides an opportunity for practice that attenuates dependence on the extrinsic KR, promotes the development of intrinsic error-detection capabilities, and allows active engagement in information processing activities required for future skillful action (Salmoni et al., 1984; Schmidt, 1991). Additionally, it is imperative for the learner to be able to perform a given task without augmented feedback as this is how tasks are performed in the "real world". This is an important point when discussing feedback in rehabilitation; patients must be able to perform tasks once they leave the sheltered rehabilitation setting. Conditions of practice that promote the development of intrinsic error-detection capabilities are important for self-maintenance and persistence of skilled performance in the future.
While a reduced KR frequency may benefit learning of relatively simple laboratory tasks, it is yet unclear if such a KR schedule also benefits learning more complex real-world skills that may require several days to learn and involve multiple degrees of freedom (Wulf and Shea, 2002; Guadagnoli and Lee, 2004). Research that used summary KR, where a summary of performance is provided after a specified number of trials, suggested that more frequent feedback may be needed when practicing a complex motor skill (Wulf and Shea, 2002). Additionally, Wulf et al. (1998) found that reducing KR frequency did not benefit the learning of a complex ski simulation task. In fact, the group that practiced the task while receiving continuous, 100% KR performed better on a retention test than did the 50% KR group. Therefore, it seems that there is something different about learning a complex task compared with a simple task. One hypothesis is that the cognitive processing demands may be inherently greater for the learning of a complex task (Wulf and Shea, 2002). If a reduced KR frequency schedule requires more processing by the learner compared to one with feedback after every trial, then we might imagine that the aggregate processing demands could be quite large for the learning of a complex task under conditions of reduced KR frequency. In this case, the processing demands may be too high resulting in diminished learning. There is a great deal yet to be learned about the interaction of task complexity and augmented feedback scheduling to enhance motor learning.
Several studies have demonstrated that individuals post-stroke have the ability to learn a new motor task (Platz et al., 1994; Hanlon, 1996; Pohl and Winstein, 1999; Winstein et al., 1999). Only one study to date has directly investigated the effect of reducing the relative frequency of KR on learning in individuals post-stroke. Winstein et al. (1999) used a lever task that required subjects to learn a series of elbow flexion-extension movements with specific amplitude and timing requirements using the less-involved UE. While their performance was not as accurate as age-matched controls, the subjects with stroke were able to demonstrate learning of the task as measured by a delayed retention test. Both control and stroke Participants that practiced the task under a faded feedback schedule (67% average KR frequency) condition demonstrated similar learning when compared to their peers who received KR after every trial (100% KR frequency). While the faded schedule of KR did not enhance learning, it was not detrimental either. Additionally, the subjects with stroke demonstrated very similar performance curves to control subjects when compared across feedback conditions. Therefore, the results of this study suggest that the literature on relative frequency of KR with healthy control subjects may be applicable to individuals with stroke (Winstein et al., 1999).
Further work is needed to better understand the benefits of augmented feedback for motor learning in individuals with central nervous system pathology. While the Winstein et al. (1999) study suggests that individuals with stroke benefit from feedback similarly to controls, this research needs to be replicated and expanded to include other tasks. Recent work that manipulated explicit instructions (Boyd and Winstein, 2004) found that when individuals post-stroke were provided with explicit information during practice of an implicit lever task, performance and learning was disrupted. This was not the case for age-matched control subjects who actually demonstrated better learning across days when explicit information was provided than when it was not. Although explicit prescriptive information about the movement strategy is distinctly different from post-response augmented feedback, this work in general suggests that there may be differences in the way in which motor skills are learned between individuals post-stroke and age-matched controls. The information provided, including augmented feedback or explicit task information, may be processed in different ways after brain damage. It is hoped that these ideas will invoke new approaches and hypotheses for rehabilitation science as researchers from multiple disciplines including psychologic and physiologic science collaborate to gain a better understanding of the neural correlates of motor learning (Miller and Keller, 2000).
Was this article helpful?