The effect of frequency of kinetic feedback on learning an isometric force production task in nondisabled subjects.

Background

Feedback in Physical Therapy

Feedback regarding a patient's movement performance is an integral component of many physical therapy approaches to rehabilitation of patients with movement dysfunction.[1-4] This feedback is provided verbally, through tactile cues transmitted by the therapist's hands or via augmented visual feedback, such as center of pressure information provided to the patient during balance activities. During physical therapy intervention, feedback may be provided continuously as the patient attempts to perform a movement. Continuous feedback is especially prevalent for those therapeutic techniques espoused by the Bobaths[2] and for proprioceptive neuromuscular facilitation (PNF),[4] which is used for both neurological and orthopedic problems. This feedback is provided to enhance the patient's movement abilities; however, recent studies in nondisabled populations have suggested that frequent feedback may be detrimental to retention or learning of motor skills (for a review, see Winstein[5]).

Two issues regarding feedback and motor learning that are pertinent to physical therapy have not been adequately addressed in previous studies. First, the effect of feedback frequency has been studied extensively in knowledge of results (KR) paradigms, but little work has been reported in which knowledge of performance (KP) was provided. As defined by Schmidt and Young,[6] the feedback physical therapists typically provide to patients is KP feedback rather than KR feedback. Knowledge of performance feedback is information about the kinematic or kinetic components of the movement being attempted by the patient. This type of feedback differs from knowledge of results feedback, which is defined as information about the outcome of the movement. Although physical therapists provide KR feedback when patients are able to complete a functional task, they work with many patients who are unable to move adequately to complete a task unassisted. Physical therapists tend to focus on the kinetic or kinematic components of the attempted movement (eg, torque produced during manual resistive exercises, amount of terminal knee extension during assisted gait). As a result, physical therapists often provide KP feedback rather than KR feedback.

Second, most motor learning studies have evaluated the effects of augmented (extrinsic) feedback that is provided to the subject after the task has been completed. During therapy sessions, feedback is often provided concurrently by the physical therapist as the patient attempts to perfonn a movement as well as after the movement attempt. Augmented visual feedback displays of center of pressure, joint torque, or electromyographic activity again are typically provided concurrently to patients during treatment sessions. The effect of concurrent feedback on performance during acquisition or reacquisition of motor skills and on retention of motor skills has not been adequately studied in patient or nondisabled populations.

Knowledge of Results and Motor

Learning in Nondisabled Subjects

The effect of feedback on motor skill learning has been evaluated using primarily KR feedback. Early motor learning studies documented the positive effect of KR (augmented information) during skill acquisition.[7-15] Although the evidence of the positive effect of KR on motor skill acquisition seemed conclusive, more recent studies raised methodological concerns regarding when changes in motor skills were evaluated. Specifically, early motor learning studies failed to differentiate the effects of KR on performance versus learning. Performance is defined as changes in the capability to produce a given response during practice when augmented feedback about the performance is provided. In contrast, learning is defined as relatively permanent changes in the capability to produce a given response when augmented feedback may or may not be present.[16] The early motor learning studies established that KR enhanced performance, but these studies were not designed to evaluate more permanent changes in the ability to perform a motor skill and therefore could not be used to conclude that learning had been enhanced.[17] To infer that changes in motor skill learning have occurred, researchers should administer either a retention test or a transfer test after the acquisition phase of the experiment has been completed. The acquisition phase consists of practice trials in which feedback is provided after some or all of the trials. The second phase is a retention phase that consists of trials in which no feedback is provided or in which the frequency of feedback is manipulated in some way. A retention test will differentiate the effect KR may have on the more permanent changes in a motor skill from the transient changes in performance that may be observed during the acquisition phase of the experiment.[18]

The frequency with which KR feedback is presented has been shown to produce differential performance and learning effects.[19] The acquisition phase consists of practice trials in which feedback is provided after some or all of the trials at some relative frequency. Relative frequency is defined as the frequency with which feedback about the preceding trial is provided to the subject. The relative frequency can range from 0% (no feedback provided after any trials) to 100% (feedback provided after each trial). The effect of relative frequency KR on performance during acquisition trials and retention trials can be evaluated by manipulating the frequency with which KR feedback is provided. Several previous studies[10,20-22] have shown that performance during the acquisition trials is enhanced with high relative frequency feedback (eg, 100%) when compared with less frequent feedback (eg, 25% or 50%). On retention tests, however, those subjects who received 100% KR during acquisition trials exhibited an inferior ability to produce the required motor task when compared with those subjects who had received a lower frequency of feedback during the acquisition trials. These findings suggest that although high relative frequency KR enhances performance during the acquisition phase of these experiments, reduced frequency KR during acquisition trials may enhance learning of the task.

One explanation for the conflicting performance and learning effects found across different relative frequency KR conditions is known as the "guidance hypothesis."[17] This hypothesis suggests that during acquisition trials in which feedback is provided, subjects develop a reliance on KR feedback to complete the required task. A dependency on this augmented feedback then negatively affects the subjects' ability to complete the task when feedback is withdrawn during the retention phase, and according to Schmidt's definition of learning,[16] has a negative impact on learning. For example, in a 100% relative frequency condition, subjects are provided KR after each attempt or preceding each subsequent trial. Subjects rely on the augmented information to evaluate their performance rather than develop an internal reference of correctness. Consequently, the ability to complete the task during acquisition trials is enhanced, although more permanent changes in behavior fail to occur. In comparison, subjects in a 50% relative frequency condition are significantly better able to complete the required task during a no-KR retention test than subjects in the 100% relative frequency condition. Subjects in a reduced feedback condition apparently are less likely to develop a reliance on KR and instead develop an internal reference for correctness in the absence of frequent KR.[19] Thus, during the retention test in which an KR is withheld, subjects who receive less feedback during acquisition trials demonstrate better learning of the required motor task than those subjects in the higher frequency feedback conditions.

Not all studies designed to differentiate the effects of relative frequency of KR on performance from the effects on learning fully support the guidance hypothesis. Winstein and Schmidt[23] (experiment 1 of the three experiments in their study) manipulated relative frequency KR feedback while subjects attempted to learn a complex spatial-temporal movement pattern, which required three elbow extension-flexion reversals in 800 milliseconds. Young, nondisabled university students performed the task for 22 blocks of 9 acquisition trials each (198 total trials) and 3 blocks of 9 retention trials each (27 total trials). Relative frequency KR, as provided by the visual display of average absolute error for a trial in comparison with the target template, was varied across two groups for acquisition trials (33% and 100%) and across four relative frequency groups for retention trials (0%, 33%, 66%, and 100%). Analysis of the acquisition and retention data failed to differentiate performance or learning between the 33% and 100% relative frequency groups. Regardless of the amount of feedback provided, the amount of error the subjects exhibited as they attempted the required task during acquisition or retention trials was equivalent across groups.

Winstein and Schmidt[23] (experiment 2 of the three experiments in their study) further investigated the guidance hypothesis by varying the amount of KR early and later during acquisition trials. They argued that for two relative frequency groups (50% and 100%), providing more KR early in practice versus later in practice (faded KR) may distinguish the contribution of the guidance hypothesis. The critical finding was better performance of the required task on the second day of acquisition trials for the 50% condition, when comparing only those trials in which feedback was provided. Subjects in this faded KR condition exhibited better performance during the acquisition trials and completed the required task more accurately during a delayed no-KR retention test. The greater relative frequency KR during the acquisition trials did not enhance performance, even though the guidance hypothesis suggests that more feedback should guide the learner toward correct actions. Subjects in the faded 50% KR feedback group also exhibited less error during the no-KR retention test than did the 100% KR feedback group. Winstein and Schmidt concluded, however, that the guidance hypothesis was a viable explanation in that the faded 50% KR feedback group was able to use frequent early feedback to develop an appropriate response capability, yet did not become reliant on the feedback and as a result demonstrated superior learning of the motor task as measured by the retention test.

Lee and colleagues[19] used a reciprocal tapping task to test the effect of feedback frequency on the guidance hypothesis. In each of their three experiments, subjects were required to time an intertap interval of 500 milliseconds for six consecutive movements (seven taps and six intertap intervals for each trial). Three relative frequency conditions were manipulated during the 50 acquisition trials (10 blocks of 5 trials): (1) 100% - KR after each trial; (2) 50% - KR after every other trial (only 25 KR trials); and (3) 50% - KR after every other trial, for 100 total trials (50 KR trials). In each experiment, subjects performed 20 retention trials in 4 blocks of 5 trials. in the first experiment, a metronome provided guidance for the tapping movement during the acquisition trials, with a no-KR retention test provided after the acquisition trials. Using this type of guidance (this was guidance for the task rather than feedback about the subjects' ability to complete the task accurately), no difference was found among the three relative feedback groups for the acquisition trials or for the retention trials. In two other experiments in which bandwidth KR was used, subjects in the 50% KR feedback group exhibited the same amount of error as the 100% KR feedback group during both acquisition and retention trials. The findings from the study by Lee and colleagues suggest that the type of feedback and the preciseness of the feedback (narrow or wide bandwidth) may complicate the principles of the guidance hypothesis somewhat.

Manipulation of KR feedback has been the primary protocol for evaluating the effects of relative frequency feedback on differences in the performance and learning of a motor skill. Although studies using KP have been designed to evaluate differences between KP and KR feedback,[6,24] no previous studies have investigated the effect of relative frequency of KP feedback. Schmidt and Young[6] found that KP in addition to KR resulted in significantly better learning of a movement task, as measured by a retention test, than KR alone. Further, Newell and colleagues[24] found that KP resulted in less error during the acquisition phase of an isometric force task and that although there was a trend toward less error during a no-feedback retention test, no significant differences were found between the KR and KP feedback groups. Newell and colleagues suggested that the lack of significant differences on their retention test may have been the result of a "ceiling effect" of the task that caused the KR feedback group to exhibit less error than may have been observed with a different task in which the potential for greater error was present. Because KP feedback is frequently used by physical therapists, more information regarding the effects of KP manipulation is needed. For the results of motor learning studies to be of use to practicing physical therapists, the type of feedback used and the task to be learned should be similar or at least relevant to the feedback used and to those tasks that patients perform in physical therapy clinics.

Relevancy and Justification of

Using an Isometric Force

Production Task to Evaluate

Kinetic Feedback Frequency

The ability to perform a functional task such as reaching with an upper extremity or pushing against a surface to assist in rising from a sitting to a standing position requires active elbow extension. To move the elbow into extension during these tasks, one must generate appropriate forces about the elbow (as well as other joints) at the appropriate time during the movement. The motor task used in our experiment, therefore, was designed to require subjects to modulate elbow extension force, where accurate amplitude and timing of force generation was required to complete the task with minimal error. This bi-amplitude task required that subjects be able to control force amplitude to two different levels and to control the relaxation of that force as well. To successfully complete the task, elbow extensor muscles must be recruited and derecruited in a controlled manner (ie, muscle contraction onset and offset must be learned). As patients with hemiplegia have been shown to have delayed termination of agonist activity during movement,[25] Modulation of agonist activity of the elbow extensors is likely to be one motor skill that these patients would need to relearn during physical therapy. An isometric rather than an anisometric task was chosen to avoid the confounding effects that hyperactive stretch reflexes may have on movement in patients with hemiplegia secondary to cerebrovascular accident.

Although the guidance hypothesis provides a rather compelling explanation for differences in performance and learning across different relative frequencies of feedback, the hypothesis is not fully supported in several of the studies discussed previously. Lee and colleagues'[19] findings in the experiment in which guidance rather than feedback was used are counter to what the guidance hypothesis would predict. In addition, the effects of KP feedback frequency (eg, kinetic feedback) have not been systematically studied. The purpose of our study, therefore, was to investigate the effects of the frequency of kinetic feedback as well as concurrent kinetic feedback on the performance and learning of an isometric force production task in young, nondisabled adults. We hypothesized that (1) during acquisition trials, the concurrent kinetic feedback group would perform better than the 50% or 100% relative frequency feedback group; and (2) during retention trials, the 50% relative frequency kinetic feedback group would display more accuracy than the 100% or concurrent feedback group.

Application of Motor Learning to

Physical Therapy Practice

The results of feedback studies may have implications for physical therapists who provide feedback as they are attempting to train or retrain a specific motor skill in a person who has movement dysfunction.[1,5,16,26] Studies in patients with movement dysfunction are needed, therefore, before physical therapists can be certain that the motor learning principles established from studies in nondisabled subjects can be applied to patient populations. Studies using nondisabled persons with intact central nervous systems, however, may allow therapists to identify and better understand mechanisms and processes related to motor learning and skill acquisition.

Although the research reported in this study used nondisabled persons as subjects, the study was designed to address issues that are relevant to physical therapy and to answer questions that arose from clinical practice. Findings from this and other studies using nondisabled persons as subjects cannot be directly applied to patients, but perhaps these findings will help therapists to more critically evaluate the feedback techniques they use when working with patients.

Method

Subjects

Twenty-four young, nondisabled, righthand-dominant subjects (18 male, 6 female) volunteered for testing. Subjects were recruited from a population of convenience that included undergraduate and graduate students in exercise science and physical therapy. To be included in the study, subjects had to be free of any history of right upper-extremity pathology. The subjects, who had a mean age of 22.5 years (SD=4.1, range=19-33), were randomly assigned to one of three relative frequency kinetic feedback conditions with the restriction that there be an equal proportion of male and female subjects (ie, 6 male, 2 female) in each group. Subjects were naive to the experimental task and the hypotheses tested. After the procedure was explained, each subject provided written, witnessed, informed consent to participate in the study.

Equipment

A straight-backed chair was positioned directly in front of a storage oscilloscope(*) (50 cm away) to allow the seated subjects to easily view the oscilloscope screen. A semirigid forearm splint (one of two available to accommodate differences in forearm size) was used to position and stabilize each subject's right forearm in a neutral position midway between supination and pronation. The elbow was positioned at 90 degrees of flexion, and the shoulder joint was in a neutral, resting position, midway between flexion and extension, adduction and abduction, and medial and lateral rotation. The forearm splint was attached to a 50-lb Geinisco(dagger) oad cell by link chain, which could be adjusted such that the subject's forearm was parallel to the ground; the elbow was flexed at 90 degrees; and the shoulder was in a neutral position between flexion and extension, medial and lateral rotation, and adduction and abduction. The load cell was attached securely to a bracket above the subject's forearm.

The load cell measured isometric elbow extension force, which was amplified by a direct-coupled amplifier,(double dagger) fed back for display on the storage oscilloscope, and sampled by a 12-bit analog-to-digital board(section) on a personal computer(perpendicular) at a rate of 500 samples per second. Data-Pac II software(#) was used to store and process the data. Calibration of the load cell, with loads from 0 to 50 lb, demonstrated that the response was linear (.999), with a maximum error of less than 2.02% of full scale. The resolution of the analog-to-digital board was 2.44 mV/bit.

Procedure

Subjects were positioned in the chair, and adjustments were made to ensure that the elbow was flexed to 90 degrees and the shoulder was in a neutral position so that any elbow extension force was generated in line with the load cell. Standardized verbal instructions were given to the subjects, and questions were answered prior to each test session. Subjects were told that the motor task they were to accomplish required them to use their right elbow extensors to generate force in such a way as to replicate as closely as possible a predetermined bi-amplitude force template (Fig. 1). The force template was drawn on clear acetate with a permanent marker and placed over the oscilloscope screen during the experiment. The largest force amplitude required during the task was 38 N. Subjects were not shown the force template during the instructional period and were allowed to see it for the first time immediately prior to the first trial.

Subjects were given no information regarding the number of acquisition or retention trials, nor were they informed that retention trials would be performed without feedback. This was done to reduce errors that might be introduced as subjects perceived the end of the experiment was near and to avoid bias among the feedback groups. To become familiar with how the elbow extension force changed the vertical position of the oscilloscope light beam, subjects practiced moving the light beam to different amplitudes on the oscilloscope screen during five trials that were 5 seconds in duration. During this practice, the force template was not displayed on the oscilloscope screen.

For all acquisition and retention trials, the investigator (TAG) initiated each trial by pressing a switch that turned on a red light, initiated the light beam sweep on the oscilloscope that represented the subject's force production, and began data acquisition. The red light was the cue to the subject that a trial had begun. The oscilloscope was set such that the light beam swept across the screen in 5 seconds. A new trial was begun every 30 seconds.

During the acquisition trials, the relative frequency of the kinetic feedback was manipulated and displayed on the oscilloscope screen. The three relative frequency conditions varied by when the force production information was presented. The conditions were (1) 100% relative frequency - visual kinematic feedback after each acquisition trial, (2) 50% relative frequency - visual kinematic feedback after every other trial, and (3) concurrent relative frequency - continuous visual kinematic feedback during response execution and 100% after each trial. For the 100% and 50% relative frequency conditions, a shield was placed over the oscilloscope to prevent the subjects from observing the force produced during performance. Subjects were then allowed to view their performance in comparison with the target template for 8 seconds after the trial was completed. The concurrent relative frequency group was able to view their performance in comparison with the template during the trial as well as for 8 seconds after the completion of each trial. The acquisition trials were completed in 10 blocks of 10 trials each. A 1-minute rest period was given between blocks, with a 5-minute rest between the fifth and sixth blocks. The 1-minute rest periods were designed to reduce the effects of mental fatigue and allowed the investigator to store the computer data on disk. The longer rest period was included to again minimize mental fatigue and enable the subjects to stand, stretch, and walk about for a few minutes.

Upon completing the acquisition trials, subjects rested for 5 minutes, and an immediate retention test was administered. The retention test involved switching all subjects to a common level of feedback, a no-kinematic feedback condition, and administering 30 additional trials in three blocks of 10 trials each. A second retention test of 30 trials was administered after 48 hours had elapsed. The delayed retention test also required that subjects complete the isometric force production task without any kinetic feedback. During both retention tests, a 1-minute rest period was provided after each block of 10 trials.

Data Analysis

An overall accuracy measurement known as root mean square error (RMSE) was calculated by determining the absolute difference between the trial force and the template force at each data point (2-millisecond interval) and averaging across the 2,500 data points for each trial. The RMSE value thus describes the average absolute error (in millivolts) for the 5-second trial. Mean RMSE values were calculated for each block of 10 trials. Separate mixed-design analyses of variance (ANOVAs) (feedback group x trial block) for repeated measures on the second factor were conducted for RMSE for each of the three tested phases: (1) acquisition, (2) immediate retention, and (3) delayed retention.

The findings reported in this study were significant with both the traditional F test and the conservative degrees of freedom (Greenhouse-Geisser) adjustments for repeated measures.[27] Alpha was set at .05 for all statistical tests, and Tukey's Honestly Significant Difference (HSD) procedure was used for all post boc mean comparisons when significant F values were revealed by the ANOVAs.

Results

Acquisition Trial Blocks

The 10 trial blocks of acquisition data were analyzed by a 3 x 10 (feedback group x trial block) ANOVA for repeated measures on the second factor. A significant main effect for feedback group and trial block and a significant feedback group x trial block interaction were found (Tab. 1.) Figure 2 shows the pattern of change for the kinetic feedback groups across the trial blocks. Tukey's HSD follow-up procedure indicated that the concurrent relative frequency feedback group displayed less RMSE than did the 50% or 100% KR group at each of the 10 trial blocks (P<.05). No significant differences were found between the 50% and 100% feedback groups at any of the 10 acquisition trial blocks.

[TABULAR DATA OMITTED]

Immediate and Delayed

Retention Trial Blocks

The immediate retention data were analyzed by a 3 x 3 (feedback groupxtrial block) ANOVA for repeated measures on the second factor. Significant main effects were found for feedback group and trial block (Tab. 2). Tukey's HSD follow-up procedure on the relative frequency kinetic feedback groups showed less RMSE for the 50% and 100% relative frequency conditions than for the concurrent feedback condition (P<.05) (see Fig. 2, immediate retention). In addition, the 50% relative frequency group displayed less error than did the 100% relative frequency group (P<.05). Post hoc analysis of the trial block main effect revealed more RMSE for the second block (X = 676, SD = 137) and the third block (X = 719, SD = 144) than for the first block X=598, SD=122). The RMSE values were equivalent for the second and third blocks.

[TABULAR DATA OMITTED]

The three blocks of delayed retention trials (48 hours after acquisition) were analyzed using the same mixed-design analysis as for immediate retention trials. The 3 x 3 ANOVA indicated a significant main effect for feedback group (Tab. 3). Post hoc analysis revealed that RMSE was less for the 50% and 100% kinetic feedback groups than for the concurrent kinetic feedback group (P<.05) (Fig. 2). As for the immediate retention trial blocks, the 50% relative frequency group exhibited less RMSE than did the 100% relative frequency group (P<.05).

[TABULAR DATA OMITTED]

Discussion

The superior performance by the concurrent kinetic feedback group across the acquisition trial blocks was not unexpected and supports our first hypothesis for kinetic feedback, which has not been previously reported. When subjects were allowed to simultaneously view the target template and the accuracy of their generated trace, they were able to minimize overall error. Of the feedback conditions tested, it appears that the visual observation of the force produced provided the best reference for correctness during acquisition. This interpretation is consistent with the KR guidance hypothesis proposed by Salmoni and colleagues.[17] Specifically, higher-frequency KR or KP, such as the concurrent feedback condition in our study (ie, visual observation during execution and 100% relative frequency after execution), minimizes the evaluative processes of the subjects. Moreover, a dependency on the kinetic feedback information may develop.

The equivalent RMSEs for the 50% and 100% kinetic feedback groups for acquisition trial blocks are similar to the results of Winstein and Schmidt's[23] study (experiments 1 and 3). Winstein and Schmidt also reported no difference between 50% and 100% feedback groups during the acquisition phase of their study. Similar accuracy scores for the 50% and 100% feedback conditions for acquisition trial blocks may be a reflection of task difficulty. As in Winstein and Schmidt's study, the required motor responses in our study were complex and may have been sufficiently difficult to complete in that more frequent feedback was not helpful in decreasing the error during the acquisition phase of the study.

For both the immediate retention and delayed retention tests, an RMSE advantage was identified for the 50% and 100% kinetic feedback conditions when compared with the concurrent feedback condition. The RMSE of the concurrent feedback group exceeded the RMSE of both the 50% and 100% feedback groups by a substantial amount. The degradation of the concurrent feedback group for immediate retention was especially dramatic when compared with the low RMSE that was exhibited during the acquisition trial blocks. Although previous studies have compared relative frequency of KR, this is the first study that investigated concurrent feedback as well. The results of our study indicate that concurrent feedback may not be useful for training an isometric motor task when the goal is for the subject to learn to perform that task in the absence of feedback. The subjects in the concurrent feedback group apparently relied so heavily on the extrinsic feedback that they did not form any kind of internal reference for completing the task. This finding would appear to have implications for those physical therapy training protocols that provide concurrent center of pressure feedback to retrain standing balance. The postural muscles that control the center of mass over the base of support during this activity will shorten or lengthen very little as they contract and essentially are performing nearly isometric contractions. Although the muscles used in standing balance are different from those used in our study, one might expect concurrent feedback to have similar effects on subjects' ability to perform the task in the absence of feedback.

The better accuracy for the 50% relative frequency condition over the 100% relative frequency condition is similar to findings reported in previous relative frequency KR studies.[19,23,28,29] During retention tests, the less frequent feedback (eg, 50%) condition consistently excels. This finding supports the guidance hypothesis in that less feedback during acquisition when feedback is present results in enhanced learning, as measured by a retention test. Although this result has been previously established for KR feedback, no previous studies have reported the effect of relative frequency of KP feedback on learning a motor task.

The results of this study provide further support for the guidance hypothesis, as evidenced by significant differences in learning between the 100% and concurrent feedback groups on the retention tests. The only difference between these two groups involved observation of the template and the subjects' force trace during execution. Greater RMSE for the concurrent feedback group during retention as compared with the 100% feedback group indicates that subjects in the concurrent feedback group failed to learn the task as well as the 100% feedback group. While watching the light beam sweep across the screen and control the vertical displacement of the light beam with the force produced, the subjects appeared to be so guided by the feedback that they developed no motor program that would allow them to accomplish the task without feedback. Consequently, when the subjects were required to perform the task in the absence of feedback, they were unable to generate the force of the appropriate magnitude, and timing and large errors in accuracy resulted.

Clinical Implications:

Therapeutic Interventions

The findings from our study regarding concurrent feedback and learning of the motor task as measured by immediate and delayed retention tests were not unexpected, but we believe these findings have particular relevance to some of the treatment protocols and techniques physical therapists use. During acquisition trials, because subjects in the concurrent kinetic feedback group relied heavily on the force feedback, it can be argued that they did not develop an internal reference for correctness. Therefore, when no kinetic feedback was provided, large error values were found during both immediate and delayed retention phases. This finding implies that verbal, manual, or visual feedback provided during a patient's movement attempts will cause the patient to rely on the physical therapist for feedback, which may adversely affect the patient's ability to accomplish the correct movements when feedback is withdrawn. That is, the concurrent kinetic feedback dramatically facilitates performance while the feedback information is available. When kinetic feedback is removed, however, the evidence from our study is clear that the original task was not learned. For patients to function in home and community environments, they must be able to perform tasks in the absence of extrinsic feedback. Whether our findings would be present in subjects with motor deficits (ie, patients) can only be determined by future research.

Concurrent kinetic feedback in this study, which allowed force production to be highly accurate and consistent, resulted in small error of movement during the acquisition phase (Fig. 2). Even though subjects in the concurrent feedback group also received kinetic feedback for 9 seconds after movement execution and compared their performance with the template after the task was completed, the error may have been so small as to not be helpful in learning to perform the task in the no-feedback condition during the immediate and delayed retention phases. From this finding, we would suggest that those interventions that rely primarily on manual handling and whose goals it is to allow the patient only to move in a "normal manner" may allow so little error of movement during therapy that the patient may be unable to develop an error detection mechanism. In line with the guidance hypothesis, the results of our study using nondisabled individuals indicate that too much reliance on an external source for movement evaluation deters permanent changes in behavior. These findings suggest that physical therapists may wish to reexamine how kinetic feedback is used in rehabilitation settings.

Conclusion

Previous studies of motor learning have not systematically evaluated the effects of relative frequency KP feedback on performance and learning of motor skills. Our findings suggest that the guidance hypothesis is supported when kinetic feedback is used during an isometric force task. For young, nondisabled persons, less feedback during the acquisition phase resulted in greater task accuracy when no feedback was present for both immediate and delayed retention tasks. In addition, feedback that was presented after a trial had been completed was more effective in helping subjects learn the task than feedback that was presented concurrently as the trial was being performed. This study evaluated a task that is relevant to the retraining of elbow extensor control in patients and used feedback that is similar to the type of feedback used with patients. The results of this study provide the groundwork for studies of a similar nature in patients with movement dysfunction.

(*) Model 5111A, Tektronix Inc, Beaverton, OR 97005.

(dagger) Model AWB-50, Genisco Technology Corp, Compton, CA 90221.

(double dagger) Model FPM-544, Therapeutics Unlimited, 2835 Friendship St, Iowa City, IA 52245.

(section) Das 16, Metrabyte Corp, 440 Myles Standish Blvd, Taunton, MA 02780.

(perpendicular) Zenith Inc, Lansing, MI 48901.

(#) Run Technologies, Laguna Niguel, CA 92677.

References

[1] Winstein CJ. Motor learning considerations in stroke. In: Duncan PW, Badke MB, eds. Stroke Rehabilitation: The Recovery of motor Control. Chicago, Ill: Year Book Medical Publishers Inc; 1987:109-134. [2] Bobath B. Adult Hemiplegia: Evaluation and Treatment. London, England: William Heinemann Medical Books Ltd; 1970. [3] Carr JH, Shepherd RB. A Motor Relearning Programme for Stroke. Rockville, Md: Aspen Publishers Inc; 1987. [4] Voss DE, lonta MK, Myers BJ. Proprioceptive Neuromuscular Facilitation, 3rd ed. Philadelphia, Pa: Harper & Row, Publishers Inc; 1985. [5] Winstein CJ. Knowledge of results and motor learning: implications for physical therapy. Phys Ther. 1991;71:140-149. [6] Schmidt RA, Young DE. Methodology for motor learning: a paradigm for kinematic feedback. Journal of Motor Behavior 1991;23:13-24. [7] Adams JA A closed-loop theory of motor learning Journal of Motor Behavior 1971;3: 111-149. [8] Annett J, Kay HC, Knowledge of results and skilled performance. Occupational Psychology. 1957;31:69-79 [9] Bilodeau IM. Information feedback, In: Bilodeau EA, ed. Principles of Skill Acquisition. New York, NY: Academic Press Inc; 1969:255-285. [10] Bilodeau EA, Bilodeau IM. Variable frequency knowledge of results and the learning of a simple skill. J Exp Psychol. 1958;55:379-383, [11] Bilodeau EA, Bilodeau IM, Schumsky DA. Some effects of introducing and withdrawing knowledge of results early and late in practice. J Exp Psychol. 1959;58:142-144. [12] Briggs GE, Fitts PM, Bahrick HP. Learning and performance in a complex tracking task as a function of visual noise. J Exp Psychol. 1957; 53:379-387. [13] Lavery JJ. Retention of simple motor skills as a function of type of knowledge of results. Can J Psychol. 1962; 16:300-311. [14] Newell KM. Knowledge of results and motor learning. Journal of Motor Learning. 1974; 6:235-244. [15] Schmidt RA. A schema theory of discrete motor skill learning. Psychol Rev. 1975;82:225-260. [16] Schmidt RA. Motor learning principles for physical therapy. In: Lister MJ, ed. II Step Contemporary Management of Motor Control Problems. Alexandria, Va: Foundation for Physical Therapy Inc; 1991:49-63. [17] Salmoni AW, Schmidt RA, Walter CB. Knowledge of results and motor learning: a review and reappraisal. Psychol Bull, 1984;95: 355-386. [18] Schmidt RA. Motor Control and Learning. A Behavioral Emphasis. 2nd ed. Champaign, Ill: Human Kinetics Publishers Inc; 1988;423-455. [19] Lee TD, White MA, Carnahan H. On the role of knowledge of results in motor learning: exploring the guidance hypothesis. Journal of Motor Behavior. 1990;22:191-208. [20] Annett J. Learning a pressure under conditions of immediate and delayed knowledge of results. Q J Exp Psychol 1959; 11:3-15. [21] Ho L, Shea JB. Effects of relative frequency of knowledge of results on retention of a motor skill. Percept Mot Skills. 1978;46:859-866. [22] Wulf G, Schmidt RA. The learning of generalized motor programs: reducing the relative frequency of knowledge of results enhances memory. J Exp Psychol [Learn Mem Cogn]. 1989; 15:748-757. [23] Winstein CJ, Schmidt RA. Reduced frequency of knowledge of results enhances motor skill learning. J Exp Psychol [Learn Mem Cogn]- 1990;16:677-691. [24] Newell KM, Sparrow WA, Quinn JT. Kinetic information feedback for learning isometric tasks. Journal of Human Movement Studies. 1985;11:113-123. [25] Sahrmann SA, Norton BJ. The relationship of voluntary movement to spasticity in the upper motor neuron syndrome. Ann Neurol 1977;2:460-465. [26] Winstein CJ. Designing clinical practice for motor learning: clinical implications. In: Lister MJ, ed. II Step Contemporary Management of Motor Control Problems. Alexandria, Va: Foundation for Physical Therapy Inc; 1991:65-76. [27] Greenhouse SW, Geisser S. On the methods in the analysis of profile data, Psychometrika. 1959;21:451-467. [28] Gable CD, Shea CH, Wright DL. Summary knowledge of results, Res Q Exerc Sport. 1991; 62:285-292. [29] Lee TD, Carnahan H. Bandwodth knowledge of results and motor learning more than just a relative frequency effect. QJ Exp Psychol [A]. 1990;42:777-789. 29 Lee TD, Carnahan H. Bandwidth knowledge of results and motor learning: more than just a relative frequency effect. Q J Exp Psycbol [Al. 1990;42:777-789.