Sidaway, Ben PT, MA, PhD; Ahn, Soyeon PT; Boldeau, Pansy PT; Griffin, Scott PT; Noyes, Bethany PT; Pelletier, Kristin PT
Physical therapists are frequently called on to treat asymmetrical weight-bearing in a variety of neurological and orthopedic diagnoses. Asymmetrical stance is a common impairment in many patients with hemiparesis with such patients bearing as much as 61%–80% of their body weight through their nonparetic lower extremity.1 These patients are unable or reluctant to bear 50% of their body weight through their involved lower extremity due to sensory, cognitive, or motor impairments.2 This inability to bear weight evenly may affect gait and other activities of daily living.3–5
Traditionally, some physical therapists have employed techniques to normalize weight-bearing status that rely on manual facilitation or guidance to improve balance and optimize function.6,7 Such facilitatory techniques are one of the cornerstones of the neurodevelopmental treatment approach (NDT)8 originally developed by Berta Bobath.6 An important goal of this therapeutic facilitation is to move or position the patient so that more appropriate proprioceptive feedback might be available to them. It is suggested that providing the feel of normal movement patterning will greatly increase the speed at which a skill is acquired. More recently, the theoretical underpinnings of such manual guidance (GD) have been questioned, and authors have criticized the lack of long-term carryover engendered by this approach.8,9 For example, it has been argued that the handling techniques used might dissuade the patient from engaging in some of the problem-solving processing that is thought to be crucial to motor relearning.10 Unfortunately, despite the prevalence of GD in the treatment of neurologically involved patients, there have been relatively few controlled experiments comparing the therapeutic efficacy of such guidance with other treatment techniques.11–13
Within the motor learning literature, the effect of physical guidance on learning motor skills has been investigated in a number of empirical studies.14–16 Typically in these experiments, learners are correctly moved through the goal pattern, thus providing them with the correct proprioceptive sensations associated with that movement. Some have suggested that such guidance techniques provide the learner with a clear image of the goal movement, reduce errors, and may offer safety in potentially dangerous situations.16 However, others have suggested that this guidance may produce less than optimal motor learning. For example, in a study teaching subjects a bimanual coordination task, Tsutsui and Imanaka14 found that physical guidance resulted in less effective retention performance than when learners were given knowledge of results (KR). In an earlier study examining an angular aiming task, Winstein et al15 gave subjects a physical stop that presented a proprioceptive feel of the target location. When this type of guidance was received at high frequencies during practice, subjects demonstrated poorer retention performance than subjects provided with lower frequencies of guidance or KR in practice.15
Other means of providing informational guidance have also been investigated within the motor learning literature. In an influential review of the KR literature, Salmoni et al17 suggested that high relative frequencies of KR during acquisition promote accurate performance while such feedback is available, but performance deteriorates when this feedback is removed in tests of retention. The authors proposed a guidance hypothesis to explain these findings in which high relative frequencies of KR during practice guide the learner to the goal response but in so doing obviate the necessity for the learner to develop his or her own intrinsic error detection and correction processes. Retention performance therefore suffers because such cognitive processes are necessary to maintain performance when KR in no longer available.17 Sidaway et al18 later argued that in fact the frequency of KR presentation (the rate at which KR is presented) rather than the relative frequency of KR (the percentage of total trials that have accompanying KR) that influences retention performance. High KR presentation frequencies guide the learner during acquisition, resulting in poorer retention performance when compared to learners given a lower frequency of KR presentation during acquisition.18
Researchers have tested the predictions of this KR guidance hypothesis in healthy subjects by reducing the frequency of KR during acquisition. These investigations have typically revealed beneficial learning effects of lower acquisition frequencies of KR. For example, Winstein and Schmidt19 examined the ability of participants to produce a goal movement pattern in 800 milliseconds when provided with either 100% or 50% relative frequency of KR. When tested in retention without KR, the 50% KR group produced significantly smaller errors than the 100% KR group, suggesting that lower frequencies of KR might enhance learning because during the no-KR trials, the learner is engaged in cognitive processes related to error detection and correction.19
Winstein and Schmidt19 provided their subjects with terminal feedback; feedback after completion of the skill. In contrast, concurrent feedback is provided while the task is being performed.20 Research investigating concurrent feedback has shown it to be a powerful tool in guiding the learner to the correct response and in minimizing errors. Park et al21 investigated the effect of different frequencies of concurrent and terminal feedback on learning. In one experiment, they assessed the ability of groups to perform a waveform production task when provided with different frequencies and combinations of concurrent and terminal feedback. Their results revealed strong guiding effects for concurrent feedback in acquisition. However, in retention where performance was tested without any feedback, groups receiving concurrent feedback in acquisition demonstrated relatively poor performance. Similar to previous research, these results show that while concurrent feedback is effective in guiding participants across trials, it does not prove to be a helpful tool in the retention of motor skills.21–23
In a paradigmatically similar study of balance by Winstein et al,24 both concurrent and terminal feedback were again investigated during acquisition and retention phases. Similarly, the researchers found that while concurrent feedback provided more guidance in acquisition than terminal feedback, the learners given concurrent feedback in acquisition performed with greater error in retention when compared to those who practiced with terminal feedback.24 While there have been studies comparing the effects of physical guidance and KR in upper extremity tasks3,25 and while there have been investigations of the role of concurrent and terminal feedback in balance tasks,24 a specific examination of the effects of varying the frequency of physical guidance and KR in balance tasks has yet to be undertaken.
The purpose of the present experiment then was to examine the independent effects of both KR, in the form of terminal verbal feedback, and experimenter-provided GD in the learning of a weight-bearing skill. In this initial experiment, we sought to examine the underlying theoretical premise of a manual facilitatory approach that suggests that the presentation of appropriate proprioceptive feedback patterns will facilitate motor learning. To this end, this experiment required neurologically uninvolved subjects to learn to support 70% of their body weight on one leg when provided with either KR or GD during practice. The frequency of presentation of both GD and KR was also varied across subjects to examine predictions emanating from the Salmoni et al17 guidance hypothesis. If the guidance hypothesis holds true for KR provision in this study, we predict that subjects presented with less frequent KR in acquisition will exhibit less error than those given KR on every trial in acquisition in tests of retention. Similarly, if frequent GD inhibits motor learning, we hypothesize that subjects who are passively moved into the target weight-bearing position on every trial will not remember the position as well as those subjects who are given such GD less frequently.
Forty students volunteered to participate in the experiment. Subjects were 20 males and 20 females whose age ranged from 18 to 45 years (mean = 21.8, SD = 4.9 years). Subjects with self-reported lower extremity musculoskeletal lesions, balance impairments, or neurological dysfunctions were excluded from participation. Subjects were quasi-randomly assigned to one of four experimental groups with the restriction that each group be composed of five males and five females. Prior to participation, all subjects read and signed institutional review board–approved consent forms.
Four experimental groups were created by crossing two types of behavioral information for learning (verbal KR and GD) with two frequencies of information presentation (100%, 33%). The performance of the four groups was examined in pretest, acquisition, and retention phases of the experiment. The task required all subjects to learn to bear 70% of their weight on one foot, in other words, to distribute their weight at a 70:30 ratio between their feet.
At the start of the experiment subjects were weighed on a large-scale analogue bathroom scale. During the pretest, subjects stood with one randomly chosen foot on the bathroom scale with the other foot placed on a wooden platform of equal height to the scale. The subjects were required to complete 12 trials during which they attempted to place 70% of their weight on the scale for five seconds. Subjects were told to look straight ahead so they were unable see the scale. In addition, a cardboard screen placed at approximately knee height shielded the scale from view by the subjects. No feedback was given to the subjects during the pretest.
Following the pretest, all subjects completed the acquisition phase in which they performed 10 blocks of 12 trials. A five-second rest was given between each trial, and a 30-second rest was given between each trial block. Before practice began, all subjects were reminded that their goal was to learn the 70% weight-bearing skill and that they would be tested on this ability in retention tests 10 minutes, one day, and one week after practicing. The same instructions were given to all subjects except that subjects in the KR groups were told that they would receive feedback on how close to the goal of 70% they were, while subjects in the GD groups were told that they were to be moved into the position necessary to generate 70% weight-bearing on the foot placed on the scale.
During practice, two groups received KR provided by an experimenter at either a 100% or 33% rate. The remaining two groups received manual GD at either a 100% or a 33% rate. Groups provided with KR were told to place 70% of their weight on the chosen foot on the bathroom scale for five seconds. An experimenter sitting in front of the subject read the scale during the last second of the trial. After the trial finished, subjects returned to a symmetrical weight-bearing position at which time an experimenter verbally provided them with their percentage of error in weight-bearing from the 70% goal and whether this percentage was greater or less than the 70% goal (eg, +8%, −7%). Prior to testing, a look-up table was calculated for each subject that converted pounds to percentage of error in order that error feedback could be quickly reported to the subject. In the KR100% group, this feedback was provided after every trial, whereas the KR33% group received feedback on every third trial. During no-KR trials, subjects attempted to produce the 70% weight-bearing goal without any feedback and without looking at the scale.
In the two manual GD groups, an experimenter sat behind the subject with her hands on either side of the subject’s pelvic girdle and positioned the subject asymmetrically by moving the pelvis laterally so that 70% of the subject’s weight was borne through the foot on the scale. A plastic marker on the edge of the scale was positioned at the pound equivalent of 70% weight-bearing for each subject. This allowed the experimenter to quickly and reliably move the subject to the position that would produce the correct amount of force on the scale. A second experimenter sitting in front of the scale was responsible for recording the scale reading when manual GD was not being provided. Similarly to the KR groups, the position on each trial was held for five seconds, and, again, the subjects were not allowed to read the scale. As in the two KR groups, this GD was provided at a frequency of either 100% or 33%. On trials without manual GD, subjects attempted to produce the asymmetrical weight-bearing position without any assistance. No feedback was provided to the GD groups at any time.
Upon completion of the acquisition phase, a 10-minute rest was given and then the first of three retention tests was administered. During all retention tests, subjects were required to perform two blocks of 12 trials without any KR or manual GD. The second retention test was one day following completion of acquisition, while the third retention test was performed one week after acquisition. Before each retention, test subjects were told that they had to reproduce the 70% weight-bearing skill that they had practiced.
To examine the differences in the ability of the four groups to learn the 70% weight-bearing skill, both accuracy and variability of performance were examined. For each subject on each block of 12 trials, the average percentage of error with regard to sign (±) from the 70% goal was calculated. If subjects produced a force greater than 70% of their body weight on a given trial, they received a positive percentage error score (eg, +8%) while forces less than 70% received a negatively signed percentage error score (eg, −7%). The averaged percentage of signed error for a block of trials is known as constant error (CE).26 When averaging CE across subjects within a group, it is customary to remove the sign from CE to prevent signs canceling each other out in calculation of the group mean.26 This measure, termed absolute constant error (ACE), was the primary measure of accuracy in producing the 70% target force in the present study. A second dependent variable, variable error (VE) was also calculated. VE is simply the standard deviation of a subject’s percentage of error scores on a given block of trials.26
Both ACE and VE were analyzed in the pretest, acquisition, and retention phases of the experiment. In the pretest, a one-way (group) analysis of variance (ANOVA) was calculated. For the acquisition phase, a two-way, 3 × 10 (group × block) ANOVA was calculated. The GD100% group could not be included in this analysis because there was no error in performance during acquisition and thus no variance within the data. The GD33% group also had no error on every third trial because on those trials, the experimenter positioned them at the 70% goal. Thus, only the trials in which the subject was responsible for producing the force without assistance were entered into the analysis. In turn, for comparison purposes, only the corresponding trials from the KR100% and KR33% groups were entered into the ANOVA.
In the analysis of the retention data, a four-way 2 × 2 × 3 × 2 (frequency × technique × retention test × block) ANOVA was used. Post hoc analysis of significant F ratios (P < 0.05) was conducted using Tukey’s honestly significant difference procedure.
The pretest ANOVA revealed no significant main effect of group (F values < 1) for both accuracy (ACE) and consistency (VE) of performance.
The group × block ANOVA performed on ACE during acquisition revealed no main effect for group (F2,27 = 2.77, P > 0.05) or block (F9,243 = 1.06, P > 0.05). There was, however, a significant interaction between block and group (F18,243 = 1.80, P < 0.05). Post hoc analysis indicated that error decreased in the KR100% group across blocks, while in the other two groups, performance remained relatively stable (Fig. 1).
The analysis of VE during acquisition revealed no main effect of group (F < 1). However, there was a significant main effect for block (F9,243 = 7.43, P < 0.001), with all groups showing an overall decline from the first (mean = 4.78%, SD = 3.16) to the last (mean = 2.93%, SD = 1.51) acquisition block (Fig. 2).
The frequency × technique × retention test × block ANOVA on the ACE data in retention revealed significant main effects of technique (F1,36 = 7.80, P < 0.05), frequency (F1,36 = 5.57, P < 0.05), and retention test (F2,72 = 3.46, P < 0.05) (Fig. 1). These findings indicate that the verbal KR groups (mean = 3.59%, SD = 3.08) performed with less error than the GD groups (mean = 6.51%, SD = 5.3). Furthermore, the groups receiving information at a 100% rate (mean = 6.29%, SD = 5.39) performed with more error than groups receiving that information at a 33% rate (mean = 3.82%, SD = 3.16). The main effect of retention test indicated that with increased time from acquisition, all groups performed with more error: 10 minutes (mean = 4.29%, SD = 4.33), one day (mean = 5.13%, SD = 5.35), and one week (mean = 5.74%, SD = 3.82). Superseding these main effects was a significant technique × frequency × retention test interaction (F2,72 = 3.54, P < 0.05) (Fig. 1). Post hoc analysis of this interaction indicated that the accuracy of the GD100% group deteriorated immediately after practice finished and was significantly worse than all other groups in all retention tests. In contrast, the performance of the KR100% group was the most accurate in the 10-minute retention test but then also deteriorated rapidly across retention tests. The accuracy of the GD33% group also decreased with time from the end of acquisition but not at the same rate as that of the KR100% group. In stark contrast, the performance of the subjects in the KR33% group remained essentially unchanged across the three retention tests and was significantly more accurate than all other groups in the final retention test.
The four-way ANOVA on VE in retention revealed only a significant main effect of retention test (F2,72 = 3.61, P < 0.05). Performance increased in variability from the 10-minute retention test (mean = 2.79%, SD = 1.23) to the one-day and one-week retention tests (mean = 3.15%, SD = 1.53 and mean = 3.26%, SD = 1.4, respectively).
Previous research investigating the learning of basic motor skills has found that physical guidance facilitates acquisition performance but usually does not result in proficient retention performance.15,16 Similarly, high frequencies of KR have also been found to guide the learner during acquisition and yet seem to be detrimental to motor skill learning.17,19,21 The present study examined the effects of both of these types of information in the learning of a novel weight-bearing skill. Two frequencies of both types of information were examined in an attempt to see whether the potentially deleterious effects of high frequencies of such information might be avoided by making the information less readily available during acquisition.
Following the pretest, the guiding properties of both KR and GD were readily apparent in the present study as all groups rapidly decreased their error from pretest values with practice. Clearly, the greatest guidance to goal attainment was provided to the GD100% group as subjects in this group were physically guided to errorless performance during these practice trials. Subjects receiving 100% KR had the next best performance during acquisition and, as Figure 1 illustrates, subjects in this group gradually decreased their error with increasing practice. In comparison, performance during acquisition in the KR33% and GD33% groups was essentially stable across acquisition with the worst overall performance being exhibited by the KR33% group. Thus, examination of the acquisition phase in the current experiment provides clear evidence of the powerful guidance effect created by the provision of learning information on every trial.
Examination of group variability during acquisition revealed that all groups, except for the GD100% group, in which there was no opportunity for variability, became more consistent in their performance from practice sessions one to 10. There were, however, no significant differences between the groups, and, as such, there is no evidence in VE for the occurrence of maladaptive short-term corrections that Schmidt and Bjork27 have suggested might occur when high frequencies of KR are provided. These researchers suggested that high frequencies of KR might cause subjects to continually attempt to modify their response on every trial even though the perceptuomotor system is unable to detect and correct such minor errors being reported by the augmented KR. Thus, the stimulus to alter performance during each trial is hypothesized to increase the variability of the goal end state during acquisition and consequently lead to inferior retention performance. The present study found no support for this contention because the KR100% group did not perform with greater VE during acquisition than the other groups. It might be argued that the scale used was not sensitive enough to record the minute variations in weight-bearing caused by the presentation of KR on every trial. This might be the case, but it should be noted that the scale could be read to the nearest pound, and so such fluctuations must have been less than approximately 0.5% of an average subject’s body weight to go undetected. The scale was only read once after the subjects had stabilized their posture, and so the fluctuations in force leading up to the final posture could not be assessed. Such an assessment of the time course of weight transfer might have revealed differences in the control strategy of subjects as a function of the type and frequency of behavioral information being provided to the learner.
Turning to the retention phase of the experiment, the detrimental effect of guidance, whether it be physical guidance or guidance with 100% KR, was strikingly apparent. Although the subjects in the GD100% group were experimenter-positioned in the 70% weight-bearing position on every trial, and thus experienced the appropriate proprioceptive feedback for that position, as soon as the GD was removed, their performance immediately and greatly deteriorated. During acquisition, the GD100% group was guided to errorless performance, but 10 minutes after the cessation of practice, the subjects in this group were performing with over twice the error of all other groups and demonstrated levels of error similar to those seen in the pretest. It appears that these subjects had almost no ability to perform the weight-bearing task without assistance despite being provided with 120 trials of practice with the sensory feedback commensurate with 70% weight-bearing. Clearly, providing subjects with the correct sensory feedback for the 70% weight-bearing skill did not result in the ability to reproduce this weight-bearing status once the GD was no longer available.
The results also highlight the danger of providing too much guidance through the provision of frequent KR. Similar to many previous investigations of the effect of KR frequency on learning,17,19,28 the present study found that although 100% KR engendered relatively accurate weight-bearing at the end of acquisition, retention performance was comparatively poor. Interestingly, as Figure 1 shows, the performance of the KR100% group was the most accurate in the 10-minute retention test, but one week later, their performance had deteriorated markedly. The analysis of performance variability in retention indicated that although both GD100% and KR had a powerful influence on performance accuracy, such information did not influence the consistency of response production. The relatively poor accuracy in retention of learners provided 100% KR during acquisition has been attributed to suppression of intrinsic error detection and correction processes by the saliency of KR during acquisition.17 The cognitive processes necessary to support performance during no-KR retention are therefore underdeveloped.17
Although the current study found clear differences in the learning of a weight-bearing skill as a function of both the type and frequency of information provided during practice, it is important to remember that all subjects were healthy without any neurological or orthopedic impairment. Given that the ability to maintain symmetrical weight-bearing in stance, in the absence of any significant muscle weaknesses, is a purview of the nervous system, any attempt to generalize the current findings to patients with neurological involvement should be done cautiously. It is interesting to note, however, that in recent years a number of motor learning principles originally based on research examining healthy subjects, including the use of KR and knowledge of performance (KP) as feedback, have also been found to hold in patient populations with various neurological involvements.3,4,25,29 For example, the use of KR and KP have been found to be beneficial in the rehabilitation of reaching movements after stroke,25 while balance training with visual feedback has been shown to be effective in improving stance and gait symmetry in children with hemiplegic cerebral palsy.4
There are also a number of studies that have specifically examined the role of feedback in improving the symmetry of weight distribution after stroke.5,12,30–34 Although a wide variety of methods have been used in these studies, they generally have found augmented feedback to be efficacious in reducing weight-bearing asymmetry. However, not all studies have found the addition of feedback to improve outcomes above that of conventional balance training. Walker et al,32 for example, reported that visual feedback provided via force platforms did not improve static and activity-based measures of balance over that generated by conventional physical therapy. Walker et al,32 however, chose to provide feedback on every trial and to provide such feedback concurrently with every trial. Typically, concurrent feedback21–23 and high frequencies of feedback presentation19,21,28 create a dependence on such feedback during practice resulting in ineffective learning of a skill. The fact that patients with neurological involvement have been shown to benefit from feedback provision in skill learning does hold the promise that the feedback manipulations used in the current study might also be effective in promoting skill learning in such patients.
Extrapolating the findings from the current study to patients with orthopedic lesions is perhaps less risky. In such patients, the central nervous system is relatively unaffected and so one might speculate that the techniques for motor learning used here would generate similar acquisition and retention effects in these patients. Similarly, although the subjects tested here were young adults, it is likely that the same pattern of results would be found in an elderly population given the same KR manipulations. This prediction is based on the fact that research has generally found that the effects of KR on motor learning are similar in young and elderly adults.35,36 Unfortunately, the ability of older adults to use GD has not been investigated and so it is not possible to speculate on the effect that manipulating frequency of guidance might have on such individuals.
A further limitation in the present work lies in the fact that only one session of learning, comprising 120 practice trials, was provided. Typically in motor skill learning, the skill is practiced over multiple sessions. The design of the present experiment does not address how the learning techniques examined would affect learning if they were given over multiple sessions. It should be noted, however, that in other investigations of basic motor learning principles, retention findings after multiple acquisition sessions typically mirror those found after a single day of practice.37–39 It is clear, however, that further investigation of long-term practice with the techniques studied here needs to be carried out before such a conclusion can be more than speculation.
CONCLUSION AND CLINICAL IMPLICATIONS
The findings from the present study clearly illustrate the potential deleterious effects of providing too much guidance during the practice of motor skills. Subjects provided with either KR or GD on every practice trial demonstrated significantly poorer learning of the weight-bearing skill than those groups provided with this assistance less frequently. This is an example of the paradoxical motor learning effect described by Schmidt and Lee26 in which there is a reversal in the relative performance of groups between acquisition and retention phases. It is important to note, then, that from an applied perspective, superior performance in practice does not always translate to proficient learning. In the present study, quite the contrary was observed. It appears that if practitioners are to effectively facilitate motor skill learning, they need to ensure that the learner is engaged in the necessary processing during acquisition to support performance when he or she is removed from the practice environment. Providing frequent GD or KR does not appear to generate the necessary conditions for such processing to occur.
It is likely that the frequent provision of GD degraded learning of the weight-bearing skill due to the effect of this technique on cognitive processing much in the same way that frequently provided KR has been suggested to do. When frequent KR is provided to a learner in acquisition, it is proposed that the availability of such feedback suppresses the learner’s development of error detection and correction processes.26 The ability to detect performance errors and then to modify motor system output accordingly is vital to the maintenance of accurate performance once informational support has been removed. Supporting this notion is the finding that encouraging learners to estimate their own performance errors prior to receiving feedback during practice improves motor skill learning.40–42 For a therapist in the clinic, then, endeavoring to facilitate a patient’s motor skill learning, the therapist must ensure that the patient does not become dependent on the behavioral information, be it KR or GD, being provided. Frequently guiding a patient to the goal response may rapidly improve performance during treatment but may not develop the requisite processing ability necessary to maintain that performance after treatment.
Finally, it is clear that the notion that providing a proprioceptive template of response production is not sufficient, in neurologically intact individuals at least, to enable proficient skill learning to occur. It appears that a more effective technique to engender learning is to allow individuals to self-generate a response and then to provide occasional feedback on the successfulness of that response.
1.Sackley CM, Baguly BI. Visual feedback after stroke with balance performance monitor: two single case studies. Clin Rehabil.
2.Sackley CM. Falls, sway, and symmetry of weight-bearing after stroke. Int Disabil Stud.
3.Bourbonnais D, Bilodeau S, Lepage Y, et al. Effect of force-feedback treatments in patients with chronic motor deficits after a stroke. Am J Phys Med Rehabil.
4.Ledebt A, Becher J, Kapper J, et al. Balance training with visual feedback in children with hemiplegic cerebral palsy: effect on stance and gait. Motor Control.
5.Winstein CJ, Gardner ER, McNeal DR, et al. Standing balance training: effect on balance and locomotion in hemiparetic adults. Arch Phys Med Rehabil.
6.Bobath B. Adult Hemiplegia: Evaluation and Treatment.
3rd ed. Oxford: Heinemann Medical Books, 1990.
7.Davies PM. Steps to Follow: The Comprehensive Treatment of Patients with Hemiplegia.
2nd ed. New York, NY: Springer-Verlag, 2000.
8.Ostrosky KM. Facilitation vs motor control: motor learning models in rehabilitation. Clin Management
9.Horak FB. Assumptions underlying motor control for neurologic rehabilitation. In: Lister MJ, ed. Contemporary Management of Motor Control Problems: Proceedings of the II Step Conference.
Alexandria, VA: Foundation for Physical Therapy, 1991:11–27.
10.Majsak MJ. Application of motor learning principles to the stroke population. Top Stroke Rehabil.
11.Langhammer B, Stanghelle JK. Bobath or motor relearning programme? A comparison of two different approaches of physiotherapy in stroke rehabilitation: a randomized controlled study. Clin Rehabil.
12.Mudie MH, Winzeler-Mercay U, Radwan S, et al. Training symmetry of weight distribution after stroke: a randomized controlled pilot study comparing task-related reach, Bobath and feedback training approaches. Clin Rehabil.
13.van Vliet PM, Lincoln NB, Foxall A. Comparison of Bobath based and movement science based treatment for stroke: a randomised controlled trial. J Neurol Neurosurg Psychiatry.
14.Tsutsui S, Imanaka K. Effect of manual guidance on acquiring a new bimanual coordination pattern. Res Q Exerc Sport.
15.Winstein C, Pohl P, Lewthwaite R. Effects of physical guidance and knowledge of results on motor learning: support for the guidance hypothesis. Res Q Exerc Sport.
16.Wulf G, Shea C, Whitacre C. Physical-guidance benefits in learning a complex motor skill. J Mot Behav.
17.Salmoni AW, Schmidt RA, Walter CB. Knowledge of results and motor learning: a review and reappraisal. Psychol Bull.
18.Sidaway B, Moore B, Schoenfelder-Zohdi B. Summary and frequency of KR presentation effects on retention of a motor skill. Res Q Exerc Sport.
19.Winstein CJ, Schmidt RA. Reduced frequency of knowledge of results enhances our motor skill learning. J Exp Psychol Learn Mem Cogn.
20.Magill RA. Augmented feedback. In: Motor Learning: Concepts and Applications.
7th ed. New York, NY: McGraw-Hill, 2004:268–304.
21.Park J, Shea C, Wright D. Reduced-frequency concurrent and terminal feedback: a test of the guidance hypothesis. J Mot Behav.
22.Annett J. Learning a pressure under conditions of immediate and delayed knowledge of results. Q J Exp Psychol.
23.Vander Linden DW, Cauraugh JH, Greene TA. The effect of frequency of kinetic feedback on learning an isometric force production task in nondisabled subjects. Phys Ther.
24.Winstein C, Pohl P, Cardinale C, et al. Learning a partial-weight-bearing skill: effectiveness of two forms of feedback. Phys Ther.
25.Cirstea CM, Ptito A, Levin MF. Feedback and cognition in arm motor skill reacquisition after stroke. Stroke.
26.Schmidt RA, Lee TD. Motor Control and Learning: A Behavioral Emphasis.
Champaign, IL: Human Kinetics, 2005.
27.Schmidt RA, Bjork RA. New conceptualizations of practice: common principles in three paradigms suggest new concepts for training. Psychol Sci.
28.Sparrow W. Acquisition and retention effects of reduced relative frequency of knowledge of results. Aust J Psychol.
29.Hanlon RE. Motor learning following unilateral stroke. Arch Phys Med Rehabil.
30.Cheng PT, Wang CM, Chung CY, et al. Effects of visual feedback rhythmic weight-shift training on hemiplegic stroke patients. Clin Rehabil.
31.Engardt M, Ribbe T, Olsson E. Vertical ground reaction force feedback to enhance stroke patients’ symmetrical body-weight distribution while rising/sitting down. Scand J Rehabil Med.
32.Walker C, Brouwer BJ, Culham EG. Use of visual feedback in retraining balance following acute stroke. Phys Ther.
33.Wannstedt GT, Herman RM. Use of augmented sensory feedback to achieve symmetrical standing. Phys Ther.
34.Yoo EY, Bo-in C. The effect of visual feedback plus mental practice on symmetrical weight-bearing training in people with hemiparesis. Clin Rehabil.
35.Swanson LR, Lee TD. Effects of aging and schedules of knowledge of results on motor learning. J Gerontol.
36.Carnahan H, Vandervoort AA, Swanson LR. The influence of aging and target motion on the control of prehension. Exp Aging Res.
37.Shea CH, Wulf G. Enhancing motor learning thorough external-focus instruction and feedback. Hum Mov Sci.
38.Swinnen SP, Schmidt RA, Nicholson DE, et al. Information feedback for skill acquisition: instantaneous knowledge of results degrades learning. J Exp Psychol Learn Mem Cogn.
39.Newell KM. Motor learning without knowledge of results through the development of a response recognition mechanism. J Mot Behav.
40.Hogan JC, Yanowitz BA. The role of verbal estimates of movement error in ballistic skill acquisition. J Mot Behav.
41.Liu J, Wrisberg CA. The effect of knowledge of results delay and the subjective estimation of movement form on the acquisition and retention of a motor skill. Res Q Exerc Sport.
42.Guadagnoli MA, Kohl RM. Knowledge of results for motor learning: relationship between error estimation and knowledge of results frequency. J Mot Behav.