Effectiveness of two forms of feedback on training of a joint mobilization skill by using a joint translation simulator.

Joint mobilization, a form of passive movement with the application of force to the joint in a specific direction, is a common therapeutic technique used by orthopedic physical therapists to reestablish the arthrokinematics, osteokinematics, and function of a joint. (1-3) The same principles and techniques also are used to assess joint dysfunctions such as hypomobility and pain. (4-6)

To perform effective, consistent, and safe mobilization movements, the applied external forces should be within a range that would stretch, but not cause damage to, the capsuloligamentous structures. The determination of the characteristic tissue resistances is important because the specified magnitudes of forces or grades of movement perceived during the assessment will be used as the basis for prescribing treatment. (2) Few reports on the magnitude of applied external force during joint mobilization performance have been published. (7-10) In these studies, data were obtained by experienced therapists, and the magnitude of the external force was the main focus. Certain psychophysical characteristics of the forces applied by the therapists appeared to indicate high intertherapist variability, and thus poor intertherapist reliability, in the magnitudes of applied forces, even when mobilization was applied to the same subject or specimen. These variations were attributed to the differences in therapists' perceptions of the initial resistance (first stop) and final resistance (final stop) during the mobilization or assessment procedures. (7,8)

During passive movement, the initial resistance is the reactive force encountered at the moment when the therapist starts to feel resistance while the slack of target tissues is being taken up. (2) This initial resistance defines the end of the grade II mobilization movement and, theoretically, should be located at the toe-region of the load-displacement curve when the resistance begins to increase against displacement. Lee et al (11) reported that initial resistance predetermined by the instructors during spinal mobilization occurred in the linear elastic region of the stress-strain relationship. Using a shoulder specimen mounted on a materials testing machine, Hsu et al (8) collected load-displacement data from experienced therapists during glenohumeral joint mobilization and linked these data to the load-displacement curve. The results revealed that the initial resistance detected by therapists fell mostly on the inflection point of the toe region of the load-displacement curve and the final resistance fell on the upper portion of the toe region and extended into the lower portion of the linear elastic region.

The proper execution of the mobilization technique depends on the therapist's ability to perceive the resistance and his or her skill level. Learning how to recognize or perceive resistances in response to the applied force can be seen as a process by which a motor skill is developed. (11) If skill learning takes place, changes in performance parameters, kinematic or kinetic in nature, will be measurable. (12) The most common method of assessing whether improved performance due to skill acquisition persists is to administer a retention test immediately following training (immediate retention test) and after a period of time (delayed retention test).

Skill acquisition usually involves the use of different modes of feedback to enhance motor learning. (12-15) The traditional methods of learning joint mobilization include instructor demonstrations and students practicing on each other. Information, including the perceived resistance and movement by instructors, is conveyed to the student while the technique demonstrated on the student. Augmented feedback about the characteristic resistances is usually provided subjectively and verbally to students in a delayed and qualitative way.

These qualitative methods of teaching likely contribute to large variations in perception of joint movement or, more precisely, the relative movement of the adjacent bones. Previous motor learning studies have documented the positive effects of augmented feedback during skill acquisition (16,17) and force-targeting tasks. (11,14,18,19) The nature of feedback given to the subject is one of the empirical factors affecting a skill-acquisition process (12,14,20) and its outcome. (11,20-23) Previous reports (24,25) indicated that quantitative feedback may result in better performance than its qualitative counterpart or no feedback at all.

Augmented feedback can be provided at different times. (12) Concurrent feedback is augmented feedback that is given while the learning task is in progress. Terminal feedback is summary feedback that is given after every skill has been performed. (12) Based on the guidance hypothesis, (14,26) the guiding properties of augmented feedback are beneficial for motor learning when used to reduce error. "Heavy" guidance, however, might be detrimental to motor learning if the learner depends on the availability of feedback information. Thus, training with concurrent feedback may lead to better immediate practice performance, but might lead to a poorer performance during the retention test when the feedback is no longer present. (19,27) In comparison, terminal augmented feedback has been reported to be more effective for skill learning. (14,19,28) In these studies, superior performance was demonstrated by the leaner who trained with terminal feedback.

Vander Linden et al (19) investigated the effects of terminal kinetic feedback and concurrent kinetic feedback during the learning of a 5-second isometric force-production task in adults who were healthy. The results showed that terminal feedback group exhibited 39% and 26% less error in the immediate retention test and the delayed retention test, respectively, than did the concurrent feedback group. The authors concluded that augmented feedback about task performance provided after a trial is more desirable for helping subjects learn the task than feedback presented concurrently. To clarify the effects of augmented feedback on learning mobilization skills, this study used both concurrent and terminal augmented feedback while subjects trained to apply specific grades of mobilization.

Despite the importance of determining the initial and final resistances in the performance of joint mobilization, few studies have investigated the processes of learning the mobilization skill. (11) Previous studies (11,29) showed that physical therapist students were able to apply forces of specified magnitudes after appropriate training, and the improved performance persisted even during the retention test.

Due to the lack of a protocol applicable to the teaching or learning of joint mobilization, this study set up an instrumented system for learning purposes by providing different types of joint stiffness and different modes of quantitative augmented feedback. We also investigated whether quantitatively augmented feedback could enhance the accuracy of perceived resistance in learning mobilization. We hypothesized that both concurrent and terminal feedback would lead to immediate increases in the accuracy and consistency of performance. We also hypothesized that subjects who received concurrent feedback would exhibit less error during the acquisition phase and those who received terminal feedback would perform with greater accuracy during the retention tests.

Materials and Methods

Subjects

Thirty-six (15 male, 21 female) undergraduate physical therapist students with a mean age of 21.8 [+ or -] 1.4 ([+ or - ] SD) years participated in this study. These students were all sophomores and juniors and had limited exposure to peripheral joint mobilization in class. The demographic characteristics of the subjects are presented in Table 1. After completing the informed consent form, subjects were randomly assigned into 1 of 3 groups: a control group, a concurrent feedback group, and a terminal feedback group. The control group received training by using a joint translation simulator (JTS) without receiving any quantitative feedback; the concurrent feedback group received quantitatively augmented feedback from a monitor simultaneously during skill acquisition; and the terminal feedback group received feedback immediately after each trial. Written descriptions and diagrams, as described by Kaltenborn, (2(pp18-30) were given to the participants before testing in order to standardize practice technique, such as posture and hand placement, as well as to provide standardized definitions of mobilization grades. Subjects were given opportunities to ask questions about the contents of the written descriptions but were not informed about the hypotheses that were being tested.

Instrumentation

The JTS consisted of a force transducer (FAK), * an analog-to-digital converter (AD574A), ([dagger]) an amplifier, and a circuit board with 2 microcontrol units (89C52), ([double dagger]) and it incorporated an AC servomotor linear translator (model 6CC401G-3DEBE) ([section]) and its servodriver (model H15CC401). ([section]) The JTS was constructed to simulate the responses of the glenohumeral joint to a joint mobilization task. Figure 1 displays the design and schematic of the JTS.

[FIGURE 1 OMITTED]

A uniaxial force transducer, mounted on an L-shaped platform and connected to the linear translator, was calibrated by standardized weights (ranging from 1 to 20 kg) prior to the experiment. The equation derived from the relationship between loads and resulting output voltages (Fig. 2) was incorporated by the controller unit of the JTS for translating the magnitude of the applied force. A sheet of foam rubber (1-cm thick) covered the top surface of the force transducer to make the texture of the contacting surface similar to the soft tissues over the joint. A servodriver was used to direct the servomotor-driven linear translator to allow a specific amount of translation according to the amount of force applied by the subject. The JTS was rigidly clamped to a mobilization table.

[FIGURE 2 OMITTED]

To serve as model templates for the JTS, load-displacement relationships, which were established in the resting position and in the end-range position from 3 fresh cadaver glenohumeral specimens of varying laxities, were used. (8) These load-displacement curves were obtained at a loading rate of 16 N/s to approximate those applied by the therapists. (8) With the humerus oriented horizontally in the plane of scapula, the glenohumeral joints were tested in the resting position (40[degrees] of abduction) and at the end position of abduction. An anteroposteriorly directed force, simulating the movement of dorsal glide, was applied by the piston of a biaxial material testing system (MTS 858 Mini Bionix) ([parallel]) through a torque arm.

Multisectional curvilinear regression analyses were performed on the force applied (y) and displacement (x) for each model. For resting position models, linear relationships, with 95% confidence intervals, were established for the neutral zone and the linear elastic region. The toe region of the load-displacement curve was divided into 2 equal parts, and each was fitted by a second-order polynomial equation using the least squares method. For the end position models, linear regressions were used to establish the load-displacement relation for the whole curve. Figure 3 shows the data obtained from the material testing system and the fitted curves ([R.sup. 2] > .95) calculated for the resting position and end position, respectively. These equations were programmed into the JTS to determine the amount of translation allowed by the JTS in response to the amount of the force applied by the subject.

[FIGURE 3 OMITTED]

To determine the criterion grades of mobilization, the "initial resistance" (the maximal resistance felt during the execution of a grade II mobilization task) and the "final resistance" (the maximal resistance felt during a grade III mobilization) had to be defined theoretically. We used the relationships among the initial and final resistance of the resting and end positions based on the results of Hsu et al. (8) For resting position model, grade II force was defined by finding the inflection points. Because no obvious inflection point was shown on the end position curve, grade III force was determined based on that obtained from resting position curve:

eG2 = 1.62 x (rG2 - 0.73), [R.sup.2] = .81

where rG2 and eG2 are the grade II forces in the resting position model and the end position model.

Accordingly, final resistance resulted from resting position model and end position model were determined:

rG3 = rG2 + 2.36, [R.sup.2] = .67

eG3 = 1.28x (eG2 + 7.02),

[R.sup.2] = .94

where rG3 and eG3 are the grade III forces in the resting position and the end position models respectively. The resulting criterion values used in this study are listed in Table 2.

As shown in Figure 1B, 2 microcontrol units were employed to calculate the displacement corresponding to the force registered by the force transducer, based on the load-displacement equations of the joint model. This information was directed to the servodriver, which in turn drove the servomotor to translate the platform in real time at the exact amount of distance calculated. Quantitative visual feedback, including the criterion force and the force applied by the subject, was presented to the subject in the form of vertical bar graphs (Fig.1B) through a monitor either in real time for the concurrent feedback group or at the end of the trial for the terminal feedback group. The control group did not receive feedback. The data acquisition software and user interface were programmed in Visual Basic 6.0. (#) When manual force was applied on the translator, the mobilization platform was positioned in real time, based on the selected joint model and its embedded equations. The frequency of system processing was set at 10 kHz to bring about the simulated tissue response as quickly as possible.

Procedure

Before testing, the height of platform was adjusted using a remote control and the starting position was reset for each subject to allow for proper body mechanics. A monitor was placed in front of the subject to provide augmented feedback displays. The subject was asked to place the hypothenar eminence of his or her dominant hand on the platform and then apply specific grades of mobilization force downward through the hypothenar and pisiform region in a consistent manner. The subject's shoulder was directly over the contact point on the JTS with the elbow slightly flexed.

During each phase (pretest, acquisition phase test, and retention tests), the resting position joint model and end position joint model obtained from one specimen were used and presented to the subject in a random order. To familiarize themselves with the equipment, subjects were allowed to practice on the JTS without feedback for 2 minutes prior to the experiment. The investigator initiated the trial by giving a single "beep" to the subject and the data collection process began. All subjects were asked to assess the stiffness and mobility of the simulated joint (JTS) once before applying a specific grade of force and performing mobilization as consistently as possible.

Pretest. After assessing joint stiffness once, the subjects were asked to perform a grade II mobilization on the JTS for 3 repetitions as close to their understanding of the grading principles as possible. A grade III mobilization was then executed in the same manner. Magnitudes of force performance were collected and served as initial condition.

Acquisition phase. During skill acquisition, the feedback groups received graphic feedback information (Fig.1B) on their applied force and the criterion force (the designated initial resistance for grade II mobilization or final resistance for the grade III mobilization) for each trial through a monitor. For the subjects in the concurrent feedback group, both the criterion force level and the force actually applied by the subject were displayed on the monitor during training. After assessing joint mobility once, the subject then executed the mobilization and was asked to perform a grade II mobilization as close as possible to the criterion force level demonstrated on the monitor. Three 25-trial blocks were executed and the subject was allowed to rest for I minute for every 3 blocks. Grade III mobilization was executed in the same way subsequently.

The same procedure was followed by the terminal feedback group, except that the criterion force level and the subject's performance (ie, the magnitude of force applied by the subject) were given immediately after each training trial. For the control group, the subject underwent the same learning procedure without receiving augmented visual feedback from the monitor.

None of the subjects were provided with information regarding the number of repetitions and the retention tests, and they were not informed that there would be no visual feedback during retention tests. These arrangements were designed to eliminate errors due to anticipation of the end of the experiment and to avoid bias among the experimental groups. At the end of acquisition phase, data from three 25-trial blocks for the grade II and grade III movements in the resting position model and the end position model were collected for a total of 75 repetitions (10 minutes) for each grade of each joint model. The total duration for the skill acquisition phase was about 40 minutes for each subject.

Retention tests. The same procedures performed during the pretest were repeated for each grade of each model during retention tests. Subjects also were asked to assess joint mobility once before performing grade II or grade III movements 3 times. An immediate retention test was conducted 10 minutes after the acquisition phase without visual feedback. All subjects returned 5 days later for a delayed retention test without feedback.

Data Analysis

Accuracy of performances was measured as the absolute deviation of the grading force, which was the final force applied by the subject, from the corresponding criterion grading force. The resulting deviation in the applied force was then normalized by the corresponding criterion force value. The formula for calculating the normalized error (NE) was:

[NE.sub.ij] = [absolute value of [(AF).sub.ij] - [(SF).sub.ij]]/,[(SF).sub.ij]

where AF = applied force, SF = standard force, i = grade II or grade III force, and j = the number of the joint model.

The resulting NEs were used for all statistics, and a low NE value indicated that subject was able to apply force that approximated the criterion force level. During the pretest and retention tests, the performance was represented by the averaged NE of the 3 attempts. To assess effects of group (control, concurrent feedback, and terminal feedback groups), grade of mobilization (grades II and III), and learning (pretest, immediate retention and delayed retention tests) on the performance in terms of the mean NE, two 3-way analyses of variance (ANOVAs) were conducted on the measurements of the resting position model and those of the end position model, respectively. Learning and grade were set as within-subject factors and group as a between-subject factor. A 1-way ANOVA was then conducted when significant between-group and within-group interaction occurred. Tukey honestly significant difference (HSD) test was used for all post hoc comparisons when a significant F value was revealed by the ANOVAs.

To illustrate the performance during skill acquisition, the mean NE was obtained from every 5 trims. Two 3-way ANOVAs (group x block x grade) for repeated measures on the last 2 factors were conducted on measures of the resting position model and those of the end position model, respectively. The Tukey HSD test was applied when significant differences were revealed by ANOVA. All data analyses were conducted using the SPSS 12.0 software program. ** The [alpha] value was set at .05. In cases when separate 1-way ANOVA procedures and multiple comparisons were conducted for post hoc contrasts, the Bonferroni adjustment was made ([alpha]=.017).

Results

Effects of Feedback on Skill Acquisition

The resulting changes in performance from the pretest to the retention tests were analyzed to assess effects of feedback on skill learning. Figure 4 displays the mean NEs from graded mobilization during the pretest and retention tests in the resting position and end position models for each group. For the resting position model, significant main effects of learning (F=31.569, P<.001), grade (F=7.612, P = .011), and group (F=15.507, P<.001) on NE were found. Significant main effects for learning and group also were shown (F=60.027, P<.001 and F=7.229, P=.004) for the end position model.

[FIGURE 4 OMITTED]

Interactions between learning and group were found in both the resting position model (F=3.501, P=.014) and the end position model (F=4.883, P=.002). During the pretest, no differences among groups in NE values were shown either for the resting position model or the end position model. During retention tests, the post hoc test indicated that the control group displayed greater NE values than did the 2 feedback groups for either the resting position model (immediate retention test: F=77.518, P<.001; delayed retention test: F=66.426, P<.001) or the end position model (immediate retention test: F=24.482, P<.001; delayed retention test: F=30.749, P<.001).

Across all the graded mobilizations and joint models, significant differences in NE were found between the pretest and the immediate retention test for the 2 feedback groups (P<.001), but the NE did not change significantly between the 2 retention tests. The augmented feedback evidently led to improved mobilization performance, and this improvement was maintained 5 days later. Specifically, there was no significant difference in NE between the concurrent feedback and terminal feedback groups. Similar effectiveness of 2 forms of feedback on skill performance was revealed even when a different mobilization grade was executed on a different joint model

Performance During Acquisition Phase

The mean NE resulting from every 5-trial block during acquisition phase revealed significant main effects for grade (resting position model: F=342.808, P<.001; end position model: F=1277.225, P<.001), block (resting position model: F=7.433, P<.001; end position model: F= 18.726, P<.001), and group (resting position model: F= 20.569, P<.001; end position, model: F=2.451, P<.001). The Tukey HSD post hoc test was carried out across the group factor, and the results revealed that the control group produced more errors than the concurrent feedback group either performing on resting position model (P<.001) or end position model (P<.001).

The concurrent feedback group tended to execute more accurately during early skill acquisition while performing mobilization in the resting position model (block 1-block 5, P<.001). However, no significant differences in NE were found between the concurrent feedback group and the terminal feedback group across training for both resting position model and the end position model. Concurrent or terminal feedback seem to have similar effects on performance during acquisition. Figure 5 illustrates the mean NE for every 5-trial block of each group while performing mobilization in the resting position model and the end position model, respectively.

[FIGURE 5 OMITTED]

Discussion

The present study used a JTS, which was based on the load-displacement relations derived from testing of glenohumeral joint specimens, to assess the effectiveness of quantitatively augmented feedback on learning specific grades of mobilization. The magnitudes of the forces corresponding to the initial and final resistance were used as the criterion grade II and grade III mobilization forces. (8) Accordingly, a low NE value indicated that the subject was able to apply a force level closer to the criterion grade II or grade III forces during mobilization.

Although it is not clear whether the training the subjects received in the present study was long enough in duration, our results demonstrated that subjects who received augmented feedback during training performed more accurately and consistently during acquisition and retention tests compared with the control group (Figs. 4 and 5). No differences, however, were found in the accuracy of performance during the retention tests between the 2 feedback groups. In the present study, the control group received training similar to the traditional method of learning joint mobilization skills. The control group displayed a greater variability in force application even though the joint models used during training and testing were the same as the other 2 groups with the exception of feedback. The control group had high intersubject variability throughout the experiment, and the results supported the idea that the individual subject is the primary source of variability in the perception of resistance.

Previous studies (8,10) have reported large intertherapist variation in the magnitude of the applied force. During spinal mobilization, Simmonds et al (10) reported a wide range (between 57.59 N and 178.27 N) for the mean peak forces across grades and stiffness levels. Hsu et al (8) also found that the mean forces applied during dorsal glide translational mobilization of the glenohumeral joint ranged from 18.36 N to 38.76 N (coefficients of variation: 40.97%-77.49%).

As expected, the feedback groups had superior performance during the acquisition phase compared with that of the control group. Guiding effects of concurrent and terminal feedback, provided by the JTS, were seen. The results indicated that the visual display of the force production and the criterion force provided a reference for correcting performance during skill acquisition. Similar results (11,28) have been reported previously, indicating that augmented feedback is associated with improved performance of simple motor skills.

During learning, augmented feedback plays 2 roles. The first role is to facilitate the achievement of the task by providing information that the learners can use to determine whether they are performing the skill correctly. (12) The second role is to motivate the learner to achieve a goal by comparing performance to the goal. The results of our study confirmed that quantitatively augmented feedback provided immediate guiding effects and enhanced the learning of joint mobilization techniques across trials during the acquisition phase.

It is generally believed that concurrent feedback provides more powerful, immediate guiding effects on skill learning during acquisition than terminal feedback does. (4,28,30) Our study demonstrated no significant differences in performance between these 2 modes of augmented feedback. Although the concurrent feedback group initially performed better on the resting position model (Fig. 5A), no difference was discovered in the performance between these 2 feedback groups during the retention tests (Fig. 5). Students who received either concurrent or terminal feedback showed improved performance during acquisition phase.

During retention tests, our results were similar to previous studies, (11,19,29) which reported significantly improved performances in the learning of force-production tasks after training with augmented feedback. An important variable in learning a task such as joint mobilization is the feedback provided to the learner (ie, the extrinsic information regarding task performance) in the form of a visual display of the force the learner achieved along with the criterion force or the difference between them. This display allows errors to be corrected immediately or in the next trial. Thus, the learner achieves better performance as practice continues. Furthermore, the practice given to learners has been shown to interact with many variables affecting motor skill learning, (14,25,30-34) such as types, timing, and frequency of feedback.

One important issue is whether concurrent or terminal presentations of augmented feedback are effective for facilitating acquisition of joint mobilization skills. In general, the terminal feedback seems to be universally effective in almost any skill-learning task, whereas the concurrent feedback is reported to be most effective on lower-level, task-intrinsic learning situations. (12,24,26-28) According to the guidance hypothesis, (14,27) reliance on feedback negatively affects the subjects' ability to form internal references to the task when feedback is withdrawn during the retention test. Concurrent feedback usually facilitates performance when it is present during skill acquisition phase but is detrimental to performance during the retention test. (19,24,26,27,35,36)

Our results, however, did not further support the guidance hypothesis because of the lack of differential influence on performances between the concurrent feedback group and terminal feedback group during the retention test. Previous studies (25,37) have demonstrated a similar failure to replicate earlier findings based on this hypothesis. Results of the present study appear to indicate that, as far as joint mobilization skill acquisition is concerned, both concurrent and terminal feedback are equally effective in facilitating motor performance.

In the interest of patient safety, it is critical to train students to gauge the appropriate force levels for the initial and final resistances. In the present study, beginning students tended to overestimate the grading forces during the pretest. In some instances, forces applied by these subjects were excessive--more than 3 times the magnitude of the criterion force--and may have been closer to injury producing levels. Clinically, determining the characteristic resistances is important because the specified magnitudes of forces or grade of movement perceived during assessment will be used as basis for prescribing treatment. (2)

During the pretest, the applied force varied greatly among subjects even though the simulated load-displacement relationship remained unchanged. For the feedback groups, performance of mobilization appeared to improve during the retention tests after training with quantitatively augmented feedback. Students appeared to be more accurate and consistent in executing grade II and grade III movements by using the JTS as a tool to create their own references about the task and the characteristic resistances provided by the JTS. Despite the apparent complexity of familiarization with varied resistances from different joint conditions, the trained tasks appear to have been learned successfully and maintained 5 days after training.

In our opinion, grading of mobilization (resistance encountered by the therapist) should have a consistent relationship to the load-displacement curve of the joint in the specific position and direction in which mobilization is applied. For example, a posterior-anterior mobilization movement is called grade HI when the final resistance, as defined by Kaltenborn, (2) is reached. Thus, this grade III posterior-anterior mobilization force must be in a specific position in the load-displacement curve of the joint.

We, however, do not assume or advocate that consistency in applying graded mobilizations is essential to the success of the intervention. As in vivo, a graded mobilization (especially grade III or higher) procedure may cause changes in the load-displacement behavior of the joint and thus changes in the amount of force applied by the therapist. In the JTS, the simulated load-displacement relationship will not change during successive bouts of graded mobilization when applied to the same model. As a result, a consistent magnitude of force was expected in the present study.

In this study, the standard magnitudes of the initial and final resistances were represented as grade II and grade III forces. Practically, the initial resistance defines the end of a grade II mobilization, and the final resistance refers to the maximal resistance felt during a grade III mobilization. The criterion grade II and grade III forces used in our study were less than 50 N (Tab. 2), and these force levels were smaller than those reported by previous investigations. (7-10) Basically, differences in applied force between our study and previous findings may result from different experimental designs. In our study, grade II and grade III mobilization forces were determined based on the results of material properties testing of joint specimens. (8) Compared with a live subject, the results of mechanical testing on cadaver specimens lacks active tension from muscles crossing the joint. Most force exerted by the therapists would be directly toward the joint capsule, and, therefore, smaller force values were required.

The usefulness of the JTS for learning joint mobilization is evident in the present study. Although we tried to mimic the clinical situation to render the simulated joint as close as possible to the real joint, some constraints associated with the current design of the JTS were inevitable. Sound and vibration from the servomotor during testing may have influenced the learning process. Most subjects stated that they tended to use this information as additional reference to guide their performance during practice. To reduce these influences, a well-insulated earphone and replacing the servomotor with hydraulic or pneumatic devices should be considered in the future.

It is difficult to mimic the viscoelastic properties of soft tissue. Although both linear and nonlinear characteristics of biologic tissue were simulated by the JTS, the load-displacement relationship, which was recorded from a cadaver specimen and used as the joint model, has only one degree of freedom present at one rate of loading. In the present study, the anterior-posterior translation load-displacement curves used were obtained from only 3 cadaver specimens. In actual clinical practice, normal variability among individual patients and factors such as pain, inflammation, muscle co-contraction, and joint pathology may result in varied characteristics of load-displacement patterns. Future studies should allow for simulation of different joint pathologies at different loading rates.

More models based on specimens of varied joint stiffness are necessary additions to the built-in model repository of the JTS to extend the diversity and complexity of the tasks available. Although a 1-cm piece of foam rubber was glued on the top plate of the force sensor to simulate the soft tissue covering of the joint, it still feels different from the real condition. Further modifications to the contact exterior are necessary to match the in vivo conditions. All biological tissues exhibit viscoelastic properties and will show their characteristic patterns of hysteresis with repeated loading. These characteristics also could be modeled and incorporated into the JTS to simulate conditions in which repeated loads are programmed into JTS. In addition, the resistance felt by the subjects also depends on the rate of force application. Although all subjects were asked to perform joint mobilization at a constant speed, the influences of speed on resistance perception could not be excluded.

Joint mobilization is a complex skill to teach and learn because it involves the application of force in a specific magnitude and direction and the perception of resistance and displacement when tissues respond. In clinical practice, accuracy in the reproduction of mobilization should be enhanced to minimize intertherapist variability. Up to now, there has been no protocol applicable to the teaching or learning of joint mobilization. The JTS, simulating joints with different stiffness, provides a quantitative and reproducible way for learning specific grades of joint mobilization by giving quantitatively augmented information.

Conclusion

The results of this study indicated that practice conditions with both concurrent and terminal feedback were more effective for learning joint mobilization tasks, as measured during skill acquisition and retention tests, than the condition without augmented feedback. However, our results did not provide further support for the guidance hypothesis that concurrent feedback may be beneficial for immediate performance but detrimental for long-term learning. Similar effectiveness on learning was shown in the concurrent feedback and terminal feedback groups. Further study is required to refine the JTS to simulate joint responses as close to the real situation as possible and to verify the effectiveness of alternative methods of training by using different types of feedback.

Ju-Ying Chang and Dr Hsu provided concept/ idea/research design, writing, data analysis, and project management. Ju-Ying Chang provided data collection and subjects. Dr Guan-Liang Chang, Dr Hsu, and Dr Chung provided fund procurement. Chang Chien and Dr Hsu provided facilities/equipment. Dr Guan-Liang Chang, Chang Chien, and Dr Chung provided consultation (including review of manuscript before submission).

Part of the manuscript was presented at Australasian Winter Conference on Brain Research; Queenstown, New Zealand; August 26-30, 2006.

This article was received June 5, 2006, and was accepted December 29, 2006.

DOI: 10.2522/ptj.20060154

References

(1) Heartling D, Kessler RM. Management of Common Musculoskeletal Disorders: Physical Therapy Principles and Methods. 3rd ed. Philadelphia, Pa: Lippincott; 1996.

(2) Kaltenborn F. Manual Mobilization of the Extremity Joints: Basic Examination and Treatment Techniques. 4th ed. Oslo, Norway: Olaf Norlis; 1989.

(3) Maitland GD. Peripheral Manipulation. 3rd ed. London, United Kingdom: Butter-worth-Heinemann; 1991.

(4) Ben-Sorek S, Davis CM. Joint mobilization education and clinical use in the United States. Phys Ther. 1988;68:1000-1004.

(5) Conroy DE, Hayes KW. The effect of joint mobilization as a component of comprehensive treatment for primary shoulder impingement syndrome. J Orthop Sports Phys Ther. 1998;28:3-14.

(6) Olson VL. Evaluation of joint mobilization treatment: a method. Phys Ther. 1987;67: 351-356.

(7) Harms MC, Bader DL. Variability of forces applied by experienced therapists during spinal mobilization. Clin Biomech (Bristol, Avon). 1997;12:393-399.

(8) Hsu AT, Ho L, Chang JH, et al. Characterization of tissue resistance during a dorsally directed translational mobilization of the glenohumeral joint. Arch Phys Med Rehabil. 2002;83:360-366.

(9) McQuade KJ, Shelley I, Cvitkovic J. Patterns of stiffness during clinical examination of the glenohumeral joint. Clin Biomech (Bristol, Avon). 1999; 14:620-627.

(10) Simmonds MJ, Kumar S, Lechelt E. Use of a spinal model to quantify the forces and motion that occur during therapists' tests of spinal motion. Phys Ther. 1995;75:212-222.

(11) Lee M, Moseley A, Refshauge K. Effect of feedback on learning a vertebral joint mobilization skill. Phys Ther. 1990;70:97-102; discussion 103-104.

(12) Magill RA. Motor Learning and Control: Concepts and Applications. 7th ed. Boston, Mass: McGraw-Hill; 2004.

(13) Zelaznik HN, ed. Advances in Motor Learning and Control. Champaign, Ill: Human Kinetics; 1996.

(14) Salmoni AW, Schmidt RA, Walter CB. Knowledge of results and motor learning: a review and critical reappraisal. Psychol Bull. 1984;95:355-386.

(15) Kilduski NC, Rice MS. Qualitative and quantitative knowledge of results: effects on motor learning. Am J Occup Ther. 2003;57:329-336.

(16) 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.

(17) Newell KM, Wade MG. Stabilometer trial length as a function of performance. Res Q. 1974;45:16-20.

(18) Blackwell JR, Newell KM. The informational role of knowledge of results in motor learning. Acta Psychol (Amst). 1996; 92:119-129.

(19) 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. 1993;73:79-87.

(20) Annett J. Acquisition of skill. Br Med Bull. 1971;27:266-271.

(21) Hebert EP, Landin D. Effects of a learning model and augmented feedback on tennis skill acquisition. Res Q Exerc Sport. 1994; 65:250-257.

(22) McConnell A. Effect of knowledge of results on attitude formed toward a motor learning task. Res Q. 1976;47:394-399.

(23) Sewall LP, Reeve TG, Day RA. Effect of concurrent visual feedback on acquisition of a weightlifting skill. Percept Mot Skills. 1988;67:715-718.

(24) Park JH, Shea CH, Wright DL. Reduced-frequency concurrent and terminal feedback: a test of the guidance hypothesis. J Mot Behav. 2000;32:287-296.

(25) van Dijk H, Hermens HJ. Effects of age and timing of augmented feedback on learning muscle relaxation while performing a gross motor task. Am J Phys Med Rehabil. 2006;85:148-155; quiz 156-148.

(26) Winstein CJ, Pohl PS, Lewthwaite R. Effects of physical guidance and knowledge of results on motor learning: support for the guidance hypothesis. Res Q Exerc Sport. 1994;65:316-323.

(27) Schmidt RA, Wulf G. Continuous concurrent feedback degrades skill learning: implications for training and simulation. Hum Factors. 1997;39:509-525.

(28) Winstein CJ, Pohl PS, Cardinale C, et al. Learning a partial-weight-bearing skill: effectiveness of two forms of feedback. Phys Ther. 1996;76:985-993.

(29) Keating J, Matyas TA, Bach TM. The effect of training on physical therapists' ability to apply specified forces of palpation. Phys Ther. 1993;73:45-53.

(30) Patrick J, Mutlusoy F. The relationship between types of feedback, gain of a display and feedback precision in acquisition of a simple motor task. Q J Exp Psychol A. 1982;34(Pt 1):171-182.

(31) Winstein CJ, Pohl PS. Effects of unilateral brain damage on the control of goal-directed hand movements. Exp Brain Res. 1995;105:163-174.

(32) Winstein CJ. Knowledge of results and motor learning: implications for physical therapy. Phys Ther. 1991;71:140-149.

(33) Guadagnoli MA, Kohl RM. Knowledge of results for motor learning: relationship between error estimation and knowledge of results frequency. J Mot Behav. 2001;33: 217-224.

(34) Wishart LR, Lee TD, Cunningham SJ, Murdoch JE. Age-related differences and the role of augmented visual feedback in learning a bimanual coordination pattern. Acta Psychol (Amst). 2002;110:247-263.

(35) Wierinck E, Puttemans V, Swinnen S, van Steenberghe D. Effect of augmented visual feedback from a virtual reality simulation system on manual dexterity training. Eur J Dent Educ. 2005;9:10-16.

(36) Pringle RK. Guidance hypothesis with verbal feedback in learning a palpation skill. J Manipulative Physiol Ther. 2004;27:36-42.

(37) Wishart LR, Lee TD. Effects of aging and reduced relative frequency of knowledge of results on learning a motor skill. Percept Mot Skills. 1997;84(3 Pt 1):1107-1122.

* Transtronic Scale Co Inc, 23145 7F, No. 129, Lane 234, Baociao Rd, Sindian City, Taipei Hsien 231, Taiwan.

([dagger]) Analog Devices Inc, 3 Technology Way, Norwood, MA 02062.

([double dagger]) ATMEL, 8F-3 No 266 Sec 1, Wen Hwa 2 Rd, Lin Kou Hsiang, Taipei, Hsien 244, Taiwan.

([section]) Sinano Electric Co Ltd, 23-11 Sengoku 1-Chome, Bonky-Ku, Tokyo, 112, Japan.

([parallel]) MTS Systems Corp, 14000 Technology Dr, Eden Prairie, MN 55344.

(#) Microsoft, Civica Office Building, 205 108th Ave NE, Suite 400, Bellevue, WA 98004.

** SPSS Inc, 233 S Wacker Dr, Chicago, IL 60606.

JY Chang, PT, MS, is a PhD student, Institute of Biomedical Engineering, National Cheng Kung University, Tainan, Taiwan.

GL Chang is Professor, Institute of Biomedical Engineering, National Cheng Kung University.

CJ Chang Chien is a PhD student, Department Electrical Engineering, National Cheng Kung University.

KC Chung, PhD, is Associate Professor, Institute of Biomedical Engineering, National Cheng Kung University.

AT Hsu, PT, PhD, is Professor, Department of Physical Therapy, College of Medicine, National Cheng Kung University, No.1, Ta-Hsueh Rd, Tainan 701, Taiwan. Address all correspondence to Dr Hsu at: [e-mail].

[Chang JY, Chang GL, Chang Chien CJ, et al. Effectiveness of two forms of feedback on training of a joint mobilization skill by using a joint translation simulator. Phys Ther. 2007;87:418-430.]

Table 1.
Demographic Characteristics of Subjects in the Control, Concurrent
Feedback, and Terminal Feedback Groups (a)

              Group

              Control       Concurrent     Terminal
                            Feedback       Feedback

Variable      Mean    SD    Mean    SD     Mean    SD

Age (y)        21.4   0.9    21.9    1.3    21.2   1.6
Weight (kg)    52.6   6.4    60.8   12.0    58.1   8.4
Height (m)    166.6   6.6   168.3    8.7   168.6   7.6

(a) n = 12 per group.

Table 2.
Theoretical Grade II and Grade III Forces Determined From the
Load-Displacement Relationships Recorded From the Material Testing
System

                      Resting Position

                      Specimen 1   Specimen 2   Specimen 3

Grade II force (N)    20           22           18
Grade III force (N)   26           32           27

                      End Position

                      Specimen 1   Specimen 2   Specimen 3


Grade II force (N)    25           29           26
Grade III force (N)   39           44           40