Effect of feedback on learning a vertebral joint mobilization skill.

Spinal manual therapy involves the assessment of resistance to movement. (1) Resistance variations can be quantified as a force-displacement relationship. The slope of the force-displacement curve is the stiffness. (2) The concept of spinal stiffness assessment through palpation is an important part of the manual therapy process. (1,3) One element of the spinal stiffness evaluation is the locating of the point "R1," defined by Magarey as "he point where resistance is first felt by the operator." (4(p299))

Commonly, the "grades" of manual therapy are defined in relation to the variations of resistance that are perceived by the therapist. (3,4) Particular grades of treatment have been recommended according to the therapist's palpation findings. (4) Such recommendations are not useful unless physical therapists agree about the location of key levels of resistance such as R1. Matyas and Bach measured the forces applied by seven physical therapists during spinal mobilizations to R1. (5) They found that the applied force varied between therapists by a factor of more than nine. The therapists, therefore, were giving very different treatments even though they believed that they were all performing the same mobilization.

The method by which mobilization skills are taught may contribute to the large variation in perception of aspects of spinal resistance. Learning an appropriate level of force to apply to reach Ri can be seen as a process in which a motor skill is developed. The nature of the feedback given to the student is one factor affecting such a skill-acquisition process. (6-10) The traditional method of giving feedback to the student about the level of force is a delayed qualitative method. The instructor uses visual observation to evaluate the force level and report this to the student. Alternatively, the "patient," who is usually another student, gives the feedback.

Feedback to the student about the level of force that he or she applies is given for two main reasons. Firstly, when physical therapy students are being taught joint mobilization by an educational institution, it is important that the institution is satisfied that the students can perform mobilization safely. The use of levels of force that are not excessive may be one criterion for safe joint mobilization. Secondly, each instructor's experience will suggest that particular force levels give the best results in therapy. Each instructor attempts to guide the students to apply this level of force during mobilization.

The accuracy of performance of a simple motor skill is influenced by the type of feedback provided. Quantitative feedback results in greater accuracy than qualitative feedback or no feedback. (9-11) Feedback that is given concurrently with motor skill performance may result in greater accuracy than feedback that is delayed, depending on the nature of the task. (9,10,12,13) This finding suggests that concurrent quantitative feedback of the level of force applied during joint mobilization may be more effective that, the traditional methods.

This study was designed to investigate whether physical therapy students can use concurrent quantitative feedback to increase their consistency and to alter the applied forces to be closer to those applied by their tutors. Olson has reported the successful use of concurrent quantitative feedback to train one therapist to apply an appropriate mobilizing force to an analog of a dog's forepaw. (14) There have been no published investigations into the use of this type of feedback in learning manual therapy procedures using living human patients. We hypothesized that the use of 45 seconds of concurrent quantitative feedback of the level of force applied would improve the performance of this aspect of a grade II lumbar mobilization. This improvement would be measured by changes in consistency among the student physical therapists and changes in agreement with their instructors.

Method

Subjects

At Cumberland College of Health Sciences (Sydney, New South Wales, Australia), physical therapy students are randomly divided into two groups for teaching purposes at the beginning of the course. Second-year students are taught peripheral joint mobilization. All third-year students, therefore, have achieved a satisfactory level of competence in peripheral joint mobilization skills. In the third year, each group is taught spinal mobilization techniques while the other group is away from the campus in various hospitals and clinics. The first group to be taught the spinal mobilization techniques was invited to take part in this study. Thirty-one students volunteered and were placed in the experimental group. After the participation of the experimental group in this investigation was complete and the two groups had exchanged places, 22 students from the second group volunteered to participate in the study and were placed in the control group. Written informed consent was obtained from all subjects. This method of selection resulted in unequal group numbers, but ensured minimal interaction between the experimental and control group subjects. The 53 subjects represented 48% of the 110 third-year students who were learning spinal mobilization techniques.

Procedure

The technique used was a grade 11 posterior-to-anterior (P-A) mobilization (3) applied to the third lumbar vertebra (L3). Grade 11 mobilization was chosen because the peak force used should correspond to the point R1. (4) The L3 was selected because its orientation would give mobilization forces in a nearly vertical direction. The location of the spinous process of L3 was determined by palpation and was marked on the skin of the "patient" with long-lasting skin-marking dye. For consistency, one of the researchers (ML) acted as the patient throughout the entire experiment.

The present state of knowledge in manual therapy does not allow the prescription of an optimal level of force for mobilizations. Instructors use their own experience and knowledge to guide students to apply appropriate levels of force for a given patient. The average peak grade II force applied by the instructors to the chosen patient was designated as the "ideal" level of force in this experiment. This level of force would not necessarily be optimal for therapy or appropriate for other patients; however, it would be the level we would expect the students to achieve for this specific patient if the guidance from the instructors was effective.

Two individuals provided the majority of classroom teaching of these techniques. The peak grade II force applied to the chosen patient by each of these instructors was measured during 10 cycles of mobilization. The average peak force of 33.3 N was used as the "ideal" value of force for the mobilization.

The force applied by the therapist to the patient was estimated using an indirect method of measurement similar to that used by Matyas and Bach. (5) The equipment used for measuring forces is shown diagrammatically in Figure 1 [omitted]. The patient is positioned prone on a plinth supported by a steel platform that is bolted to a Kistler force plate.* After recording an initial ground reaction force (GRF) reading from the force plate, the therapist then applies the mobilizing force (F) to the patient. if the mobilizing force is vertical, then from Newton's second law of motion 

F + W[.sub.1] + W[.sub.2] + W[.sub.3] - GRF = (W[.sub.1] + W[.sub.2] + W[.sub.3]) a/g (1)

where W[.sub.1] is the weight of the patient; W[.sub.2] is the weight of the plinth; W[.sub.3] is the weight of the platform; a is the acceleration of the center of mass of the patient, plinth, and platform; and g is the gravitational acceleration.

Therefore

F = GRF - W[.sub.1] - W[.sub.2] - W[.sub.3] + (W[.sub.1] + W[.sub.2] + W[.sub.3]) a/g (2)

If the last term of equation 2 is small (that is, if the acceleration, or a, is small), then the force applied by the therapist can be estimated closely by measuring the GRF via the force plate. Using a similar procedure, Matyas and Bach concluded that the error involved in estimating the peak value of the mobilizing force was in the order of 1% to 3%.5

When learning mobilization techniques in practical classes, subjects from both groups received the traditional delayed qualitative feedback of their performance. The experimental group additionally received concurrent quantitative feedback of their performance on one occasion. The students' deviation from this "ideal" level of force was shown to the experimental group in the form of concurrent feedback (via an oscilloscope+).

Each subject in the experimental group was first familiarized with the equipment and experimental procedure and was then asked to perform a grade II P-A mobilization of the (marked) L3. The subjects practiced this mobilization for 30 seconds and then performed the mobilization for an additional 45 seconds with the GRF being recorded on the chart recorder, which was not visible to the subjects (pretest). They then repeated the 45-second mobilization with concurrent feedback from an oscilloscope, which displayed a moving trace, the deflections of which were proportional to the force applied by the subject to the patient. The subjects' performance during the time that feedback was given was not recorded. A line drawn on the screen showed the "ideal" force that the subject should aim to reach; a shaded band around the line corresponded to deviations of up to 10% from this force. After a 3-minute rest interval, each subject was retested immediate posttest) using the pretest procedure. That is, 30 seconds was spent in practice, followed by 45 seconds of recorded performance. Approximately one week after the immediate posttest, each subject repeated the mobilization procedure follow-up posttest). As stated previously, 30 seconds of practice was followed by 45 seconds of recorded performance.

The subjects in the control group performed the pretest, immediate posttest, and follow-up posttest mobilization procedures. These were the same procedures adopted with the experimental group, except that practice with oscilloscope feedback was omitted. No subject or instructor was given any information about the performances shown on the chart recordings until all subjects in each group had completed the experiment.

Data Analysis

The chart recordings showing the variation of applied force with time were all analyzed by one person (AM). That person was unaware of the level of force corresponding to the "ideal" level. Thus, the results are unlikely to be systematically influenced by any possible bias of the person who performed the analysis.

The recordings of force formed a series of regular cycles in which force increased to a maximum and then decreased to near the starting level. The mean number of cycles in the 45 seconds of recorded data for each subject was 52.8 (s = 16.7). From the recordings of force, the last 11 cycles were considered. These cycles were chosen to ensure that steady-state conditions were obtained. The final cycle of this 11-cycle sequence was excluded from the data analysis because of possible transient effects occurring as the student was asked to stop mobilization. Thus, the analysis of the data involved consideration of three variables: group control or experimental), test (pretest, immediate posttest, or follow-up posttest), and cycle number (from 1 to 10).

In this experiment, the variable of most interest was the deviation of forces from the "ideal" value. Therefore, the data were converted to error scores by subtracting 33.3 N from each force reading. The absolute value of the error scores was used for all statistical analysis.

Consistency of the students' performance was measured by the variance of the group's data. We used F tests to compare the variance between each pair of tests for both experimental and control groups.

Accuracy of performance was measured by how close to the "ideal" value the students' forces were. An increase in accuracy would correspond to a decrease in the mean value of absolute error scores, The significance of error score differences was tested using an analysis of covariance (ANCOVA). The ANCOVA adjusted for possible between-group differences that existed at the time of the pretest by using the pretest error levels as the covariate. The pretest error values were established by averaging the absolute error values for the 10 cycles of data corresponding to this test. The ANCOVA was used to establish whether the error scores at the two posttests were significantly smaller for the experimental group compared with the control group.

That is, if the feedback was effective in reducing error, then the effect of group would be significant. If the improvement was maintained between the two posttests, then the effects of test and group-test interaction would be nonsignificant.

Results

The mean force readings at each test for the control and experimental groups are shown in Figure 2 [omitted]. This graph shows that after the feedback was given, the experimental group's mean force closely approximated the "ideal" value of 33.3 N, whereas there was little change in the control group's mean force level. Furthermore, at the follow-up posttest, the experimental group's mean force level remained close to the "ideal" value with a slightly increased standard deviation (from 11.9 to 13.3). Again, little change was noted in the control group's performance, although the standard deviation had decreased (from 28.7 to 18.8). The actual time interval between the immediate posttest and the follow-up posttest was, on average, 7.9 days (s = 1.5) for the control group and 6.3 days (s = 0.9) for the experimental group. The average values of the absolute error scores are shown in Table 1 [omitted] and in Figure 3 [omitted]. These values follow a similar pattern to the force readings.

The results of the F-test comparisons are shown in Table 2 [omitted]. There were significant differences in variance between the pretest and the immediate posttest for the experimental group (p <.01), but this variance did not change significantly between the immediate posttest and the follow-up posttest. The control group's variance did not change significantly between the pretest and the immediate posttest but did decrease significantly at the follow-up posttest (p < .05).

The pretest and immediate posttest mean values were compared using an ANCOVA with two within-subjects factors cycle and test) and one between-subjects factor (group). The ANCOVA showed that the effect of group was significant (p < 0002). This result indicates that the addition of quantitative concurrent feedback did have a significant effect compared with the control group procedure. In addition, the absence of a significant effect attributable to test or group-test interaction shows that this learning had been retained over the time interval between the immediate posttest and the follow-up posttest. There was no significant variation of force with cycle number within each test.

Discussion

The most important finding of this experiment was that the addition of 45 seconds of concurrent quantitative feedback to the traditional process of learning manual therapy can result in an immediate improvement in consistency and accuracy of performance, which is still retained for a substantial period of time later. This result is consistent with previously published findings on the acquisition of simple motor skills. (9-11)

The particular timing of the feedback session in the overall manual therapy learning process may be one reason for its success in this experiment. The students involved had previously learned the basic skills of joint mobilization as applied to the peripheral joints. At the time of this experiment, they were engaged in extending these skills to the joints of the vertebral column. Perhaps this background of basic skills allowed the students to focus on the level of force more than would have been possible if the whole complex motor act was being taught for the first time.

Both experimental and control groups received delayed qualitative feedback during classroom teaching of the manual therapy techniques. The source of this feedback was either the patient or the observing instructor. Other investigators have concluded that this type of feedback tends to decrease error, but to a lesser extent than more specific quantitative feedback. (9-11) That outcome was demonstrated in this experiment where the control group's mean error decreased and became more consistent with time, showing a 45% decrease in standard deviation of error scores over the duration of the experiment. The use of concurrent quantitative feedback provided a much greater change in consistency, with a 68% decrease in the standard deviation of the experimental group's error. One explanation for the relatively poor efficacy of the qualitative feedback could be that it is based on subjective judgment and is therefore likely to show significant variability. This variability may be reduced if the methods of this experiment were used to provide quantitative feedback in the classroom.

Currently, there is no evidence to suggest the existence of a unique physical event that occurs at levels of force commonly applied during spinal accessory mobilizations to R1. Future research or theoretical developments may allow a workable, objective definition of R1 that can be translated into a particular level of force. The feedback techniques used in this experiment could be applied to train therapists to apply that force. Until that time, these techniques can only be applied to other patients" and other spinal levels after first establishing the appropriate "ideal" force level. The average level of force used by experienced therapists is probably the most appropriate standard for such "ideal" forces at present.

Of immediate interest is the question of patient safety. in the present study, the applied force varied greatly between subjects at the time of the pretest. The greatest force was more than four times the average force applied by the instructors. In the interests of patient safety, it may be appropriate to use quantitative feedback to train students to perform more consistently with instructors and with the remainder of the group. It is difficult to determine what force levels are capable of causing injury. One in vitro study of lumbar motion segments has found that anterior shear forces of up to about 150 N can be applied without evidence of permanent deformation. (15) A similar study found that shear forces of up to 539 N could be applied without signs of overt failure. (16) To date, safe levels have not been established for forces that are applied in a P-A direction to the intact spine. We suggest that further research should be carried out to establish such levels.

The velocity of movement is another variable that determines the response of the patient's spine. We did not attempt to control or measure this variable. Further research may also focus on this topic.

Summary

Physical therapy students received concurrent quantitative feedback to augment the delayed qualitative feedback normally given during manual therapy instruction. The effect of this feedback was studied using grade II P-A mobilization of the L3. This additional feedback resulted in immediate improvement in accuracy and consistency of performance of the technique. The improvement was found to still be present one week later.

Acknowledgments

We wish to acknowledge the valuable assistance of Ray Patton in collection of the data and the helpful advice that was given by Roger Adams and Thomas Matyas.

Footnotes

* Model 9261A, Kistler Instrument Corp, 2475 Grand Island, NY 14072.

+ Model 5B12N, Tektronix Inc, PO Box 1700, Beaverton, OR 97075.

References

1 Grieve GP: Mobilization of the Spine, ed 4. New York, NY, Churchill Livingstone Inc, 1984, pp 108-112, 123-134

2 White AA, Panjabi MM: Clinical Biomechanics of the Spine. Philadelphia, PA, J B Lippincott Co, 1978, p 498

3 Maitland GD: Vertebral Manipulation, ed 5. London, England, Butterworth and Co (Publishers) Ltd, 1986, pp 93-99, 155-160, 282, 355-364

4 Magarey ME: Selection of passive treatment techniques. In: Proceedings of the Fourth Biennial Conference of the Manipulative Therapists Association of Australia. Brisbane, Queensland, Australia, 1985, pp 298-320

5 Matyas TM, Bach T: Reliability in clinical arthrometrics. Australian journal of Physiotherapy 31:175-199, 1985

6 Annett J: Feedback and Human Behaviour. Harmondsworth, England, Penguin Publishing Co, 1969, pp 139-153

7 Gentile AM: A working model of skill acquisition with application to teaching. Quest 17:3-23, 1972

8 Marteniuk RG: Information Processing in Motor Skills, New York, NY, Holt, Rinehart & Winston Inc, 1976, pp 177-184

9 Salmoni AW, Schmidt RA, Walter CB: Knowledge of results and motor learning: A review and critical appraisal. Psychol Bull 95:355-386, 1984

10 Schmidt RA: Motor Skills. New York, NY, Harper & Row, Publishers Inc, 1975, pp 85-96

11 Trowbridge MH, Cason H: An experimental study of Thorndyke's theory of learning. J Gen Psychol 7:245-258, 1932

12 Summers JJ: Motor programs. in Holding D (ed): Human Skills, New York, NY, john Wiley & Sons Inc, 1981, pp 48-61

13 Schmidt RA: Motor control and learning: A behavioral emphasis. Champaign, IL, Human Kinetics Publishers Inc, 1982, pp 534-544

14 Olson VL: Evaluation of joint mobilization treatment: A method. Phys Ther 67:351-356, 1987

15 Panjabi MM, Krag MH, Chung TQ: Effects of disc injury on mechanical behavior of the human spine. Spine 9:707-713, 1984

16 Miller JAA, Schultz AB, Warwick DN, et al: Mechanical properties of lumbar spine motion segments under large loads. J Biomech 19:7984, 1986

M Lee, BE, PT, is lecturer, Division of Biomechanics, Department of Biological Sciences, Cumberland College of Health Sciences, PO Box 170, East St, Lidcombe, Sydney, New South Wales, Australia 2141. Address correspondence to Mr Lee.

A Moseley, BAppSc, PT, is Physical Therapist, Lidcombe Hospital, Joseph St, Lidcombe, Sydney, New South Wales, Australia 2141.

K Refshauge, Graduate Diploma in Manipulative Therapy, PT, is Lecturer, School of Physiotherapy, Cumberland College of Health Sciences.

This article was submitted June 1, 1988,- was with the authors for revision for 30 weeks; and was accepted June 27, 1989.

Commentary

This is an interesting article that could cause a little confusion in manipulation circles, for it speaks of Maitland's mobilizations without making that point sufficiently clear. it also does not carefully define the RI" point of resistance to motion that the students in this study were being trained to duplicate. Such lack of definition could cause confusion with other resistances not referred to and quite well known to therapists trained iii the United States.

These other resistances to motion include that resistance felt when giving distraction to unweighted joint surfaces during long-axis distraction. A second resistance is that felt once the end of range has been reached, referred to as first stop" by Kaltenhorn (F Kaltenhorn, personal communication). Then there is a third resistance to be overcome in order to arrive at the end of the physiological range, or "second stop." The nature of the resistance between first and second stop is termed end feel or barrier feel. (1) These are all points of resistance that are very different from the R1 that the article discusses.

My understanding of R1 is that if a patient were placed in a prone position and the knee was then passively oscillated as it is moved through flexion, a resistance RI) would be felt at about the midpoint of the range. This RI would not be felt if the knee were simply moved in a steady motion through the range. I believe, therefore, that RI is some kind of a neuromuscular resistance stimulated by oscillatory motions (P Edgelow, personal communication).

To my reading, the article has one principal flaw. it does not address the major issue of the students' ability to feel the R1. instead it appears that rather than learning to feel R1, the students have learned to duplicate a tracing on an oscillograph. What would have been more interesting would have been to have used a second patient to determine whether the instructors and students were any closer to agreement.

Stanley V Paris, PhD, PT

75 Comares Ave, 3C

St Augustine, FL 32084

References

1 Patla E: Reliability of interpretation of the Paris Classification of Normal End Feel for Elbow Flexion and Extension: A Pilot Study. Master's Dissertation. Atlanta, GA, Emory University 1988

Author Response

We would like to thank Stanley Paris for his comments. We appreciate the opportunity to clarify some issues and to discuss possible directions for future research in this area.

Paris has suggested that there may be some confusion between the resistance concepts of Maitland and those of Kaltenborn. The point "R1," as defined by Maitland (1) and Magarey (2) and used in our experiment, seems to be very closely related to Kaltenborn's "first stop," as defined by Paris. In a "physiological movement," one that the patient can perform voluntarily, there is usually a substantial amount of movement that is relatively free of resistance, followed by a region where resistance increases rapidly (for example, see Engin's article on biomechanics of the shoulder complex-3 for the pattern of resistance in the glenohumeral joint). We believe that the place where resistance starts to increase would he labeled by Maitland as "R1 " and by Kaltenborn as "first stop."

A problem arises when this concept is applied to "accessory movements," those that the patient cannot perform voluntarily. An example of such an accessory movement is a posterior-to-anterior (P-A) movement in the lumbar spine, which was the movement considered in our investigation. For accessory movements, the resistance begins very early in the range of movement and continues to increase in a linear way with further movement. (4) The true onset of resistance occurs at levels of force and displacement that are very small in terms of clinical practice. It has been observed that in accessory movements, therapists tend to locate RI part-way through the range of motion, in the linear region of the force-displacement curve. (4) The reason for locating RI in this region rather than at the true onset of resistance is not known.

Because the grades of treatment are defined in relation to the RI point, if that point seems to be arbitrarily located in accessory movements, it is important that there is some method for teaching students to perform appropriate grades of mobilization. We believe that our investigation has shown that objective, concurrent feedback of the level of force applied can be used to teach students to perform appropriate mobilizations.

It is possible that this method may not be needed for mobilization in "physiological" movements because the pattern of resistance is fundamentally different than for "accessory" movements. However, we know of no investigations that correlate physical therapists' location of R1 with true onset of resistance in physiological movements.

We believe that it is now imperative that further research be conducted to establish the optimum levels of force for use in all forms of mobilization and manipulation treatment. The variation of level of applied force with different types of patients would be an important part of this research. in the short term, surveys of the level of force applied by experienced therapists may provide some indication of an appropriate level, although we would expect wide variations to be seen among experienced therapists.

We look forward to future discussion of these issues and further attempts to develop a rational basis for treatment of joints using manual therapy.

Michael Lee 

Anne Moseley

Kathryn Refshauge

References

1 Maitland GD: Peripheral Manipulation, ed 2. Boston, MA, Butterworth Publishers, 1977, pp 344-353

2 Magarey ME: Selection of passive treatment techniques. In: Proceedings of the Fourth Biennial Conference of the Manipulative Therapists Association of Australia. Brisbane, Queensland, Australia, 1985, pp 298-320

3 Engin AE: On the biomechanics of the shoulder complex. J Biomech 13:575-590, 1980

4 Lee M, Moseley A: Dynamics of the Human Body: Lecture Notes for Physiotherapy Students. Sydney, New South Wales, Australia, Michael Lee, 1989, pp 123-138