Reliability of pain and stiffness assessments in clinical manual lumbar spine examination.

The incidence of low back disorders is now such that the cost of this problem in the United States alone is estimated at between $20 and $50 billion annually.[1] Although exact figures are not available, physical therapy is probably the treatment most widely used for low back disorders,[2] with manipulative therapies being popular treatments among physical therapists.

One technique commonly used by physical therapists to examine and treat problems of the lumbar spine is the posteroanterior (PA) central pressure test.[3] This manual technique requires the application of an anteriorly directed force over the spinous process of a prone patient. During the performance of this maneuver, information is collected on the therapist's perception of stiffness for the movement and the patient's perception of pain. The perceived stiffness is then compared with the therapist's experiential model of what would be considered normal for that particular location in the spine.[3] The information on stiffness has traditionally been collected because it has been suggested that there may be a relationship between pain, reduced voluntary movement, and abnormal spinal joint stiffness and that the restoration of normal spinal stiffness may bring about a return of pain-free voluntary movement.[3-5] The information on pain and stiffness is used to help in the selection of a region for treatment, in the selection of appropriate manual treatment techniques, and for monitoring patient recovery.[6]

If the PA central pressure test is to provide meaningful information that will assist patient management, then the measurements obtained should be reliable and valid. The reliability and accuracy of the PA central pressure test in the assessment of stiffness, however, is currently in dispute. Some authors have concluded that assessments of PA stiffness using the PA central pressure test are either unreliable[6,7] or inaccurate,[8,9] whereas others have concluded that the test can produce reliable[10,11] and accurate results.[12] Only one study[6] has investigated the reliability of assessments of pain using the PA central pressure test, and the researchers concluded that the test was reliable for this purpose.

Matyas and Bach[6] presented the results of a number of graduate student research projects they had supervised. The authors concluded that the assessment of pain could be achieved with good reliability (Pearson's r=.48-.83), whereas for stiffness assessment poor reliability was consistently observed (Pearson's r=.09-.38, Kappa=.08-.34. The reliability of assessments of spinal stiffness also varied across spinal levels (Pearson's r for L-4 was .64, whereas the mean for all levels was .30). Viner et al[9] evaluated raters' ability to rank the PA stiffness of three asymptomatic subjects by comparing raters' rankings with the true rankings obtained by measuring PA stiffness with an accurate mechanical device. Only 47% of the raters correctly ranked all three patients, and the authors concluded that the accuracy of manual assessment of PA stiffness was not high.

A major concern with these studies is that their results may lack external validity. In many of the studies, the raters included students or inexperienced raters, the targets to be rated were asymptomatic subjects, and the raters were required to use a particular method of PA central pressure testing rather than the method of their choice. In Hardy and Napier's study,[8] for example, the raters, some of whom were undergraduate students, were required to perform the test on a block of rubber attached to a materials testing machine. Because of the experimental settings used in these studies, there is considerable difficulty extrapolating the results to the clinical assessment of patients. An exception is the study of Binkley et al,[7] who evaluated the ability of experienced orthopedic physical therapists to rate the mobility of the lumbar spine of patients with low back pain using the PA central pressure technique. The authors found that the intertherapist reliability of judgments of PA mobility of a single marked (though unspecified) lumbar level was poor, with an intraclass correlation coefficient (ICC[2,1]) value of only .25. The authors, however, did not report the reliability of judgments of PA mobility at the other lumbar levels, or the reliability of pain findings.

Researchers who have concluded that the PA central pressure test is a reliable or accurate measure of stiffness have often used experimental paradigms that limit the generalization of their results to the clinical assessment of patients with low back pain. For example, Minucci[11] studied asymptomatic subjects, whereas Trott et al[12] required therapists to rate a simulated spine with stiffness properties that could be altered by the investigators. Investigators who have measured the PA stiffness of the spine[13] have reported PA stiffness values markedly different from those obtained by Trott et al, so it is currently not clear whether Trott and colleagues' results would apply to the assessment of human spines.

Another factor that contributes to the current confusion regarding the reliability of judgments made using the PA central pressure test is the use of reliability indexes such as Pearson's r and percentage of agreement. Unfortunately, both of these statistics can provide misleadingly high estimates of reliability. For example, Minucci's[11] exclusive use of percentage of agreement as a reliability index for categorical data is not recommended because the index ignores chance agreement, which can be high if few categories are used by the raters.[14] Because in Minucci's study subjects were asymptomatic and a three-point scale was used, chance agreement would have a high probability. Pearson's r is not recommended as a reliability index for continuous data because it is not sensitive to a systematic observer bias and because it reflects covariance rather than agreement. Because of these characteristics, these two indexes cannot provide convincing evidence of reliability.

Our study sought to clarify the uncertainty regarding the reliability of judgments made using the PA central pressure test. The primary purpose of the study was to determine the inter tester reliability of judgments made by experienced raters. A secondary purpose was to determine whether judgments of pain have different reliability from judgments of stiffness and whether reliability varies across the lumbar levels examined. The data were collected as part of the normal assessment and treatment of patients seen in physical therapy clinics. This procedure was followed to allow the rating sessions to be as natural as possible and to better reflect the reliability of judgments actually made in clinical practice.

Method

Project Overview

The design of the experiment required three pairs of physical therapists to each sequentially assess 30 patients with low back pain; thus, a total of 90 patients were examined. The physical therapists were required to rate the stiffness of each lumbar level, and the maximum pain reported during assessment of each level, using a standardized rating system. Three physical therapy clinics in the Sydney (New South Wales, Australia) metropolitan region participated in the study. All subjects signed informed consent statements.

Subjects

The subjects in this study were 90 patients who were attending physical therapy practices for treatment of nonspecific mechanical low back pain. The term "nonspecific mechanical low back pain" is used to describe the large group of patients with low back pain for whom it is not possible to specify the exact cause of their symptoms.[1] The term excludes patients with neurological signs; patients seen postsurgically; and patients for whom the cause of their symptoms can be established and a specific diagnosis such as spinal stenosis, metastasis, or visceral disease can be made. Subjects were classified as having nonspecific mechanical low back pain if their routine management had not resulted in a specific diagnosis for the cause of their symptoms. Subjects were included in the study if their normal clinical assessment would include the use of the PA central pressure test to assess the lumbar spine. Those whose condition might be exacerbated by two sequential assessments or who possessed any contraindication or precaution to PA central pressure testing[3] were excluded from the study.

In all Australian states, physical therapists are primary contact practitioners, and patients are allowed direct access to physical therapy services without the requirement of a medical referral. This arrangement ensured a wide variety of patients in the study. A description of the patients is given in Table 1.

Table 1. Characteristics of the
Patients (N=90)

Variable           X         SD       Range

Age (y)              45.37     14.16     21-78
Height (cm)         169.89      9.11    152-188
Weight (kg)          70.49     12.00     45-105
Time since
  onset of LBP(a)
  (d)                45.20    100.00      1-730
Previous history
  of LBP                       82%

Gender
  F                  56
  M                  34

Patients could be assessed at their first visit or at a subsequent visit to the physical therapist; however, to allow the patients sufficient time to read the project information sheet and not interfere with the practice schedule, the assessment generally took place on a subsequent visit. Subjects who were unable to speak English or use a numerical pain rating scale were not asked to participate.

Raters

The raters in the study were experienced manipulative physical therapists who were members of the Manipulative Physiotherapists Association of Australia. Each volunteer rater had completed an undergraduate program in physical therapy and had received a Graduate Diploma of Applied Science in Manipulative Physiotherapy. Each rater had also complied with the mandatory continuing education program that is a condition of continued membership of the Manipulative Physiotherapists Association of Australia and also had at least 5 years of clinical experience as a manipulative physical therapist. These raters regularly used the PA central pressure test to assess patients with low back pain. The experience of each of the raters is shown in Table 2. Rating sessions were included as part of the patients, normal assessment and treatment, so pairs of physical therapy raters who worked in the same practice were used.

[TABULAR DATA OMITTED]

Procedure

For each patient, the treating physical therapist served as the first rater; thus, the order of rating within each pair alternated. The first rater performed the usual assessment of the patient up to and including the PA central pressure test and also marked the skin over the lumbar spinous processes so that the second rater could rate these same vertebrae. During this time, the second rater was in a different room and could not see or hear the first rater assessing the patient. The first rater recorded the results of the PA central pressure assessment on a standard form. When the assessment was completed, the first author or an assistant checked that the rating form was filled in and then filed the form so that it was out of sight of the second rater. The second rater was then asked to rate the patient using the first rater's skin markings to identify the lumbar levels. After the patient had been rated by both physical therapists, the first rater returned and continued the treatment.

Therapists were asked to rate the patient using only information derived from the PA central pressure test and not to incorporate any other information about the patient. It was explained to each rater that it would be counterproductive to incorporate any information about the patient from a previous encounter as this information would not be available to the peer rater and so would only contribute to unreliability. Because the study was intended to simulate clinical procedure, raters were instructed to use the technique they would normally use in clinical practice to rate patients using the PA pressure test and were allowed to collect the information in any order they wished. There was no time limit for assessment. In general, participation in the study lengthened the treatment session by 5 minutes.

Therapists used numerical rating scales to record pain and stiffness findings (Appendix). The patient's report of pain intensity in response to the application of the PA central pressure sure test was requested using a 0- to 10-point scale. This type of scale was familiar to all the raters and most of the patients. Subjects were referred to the end points of the scale and asked to use the numbers on the pain scale as seemed appropriate to them to represent their pain. The maximum intensity of local lumbar pain produced by PA central pressure testing at each level was the value recorded. Local lumbar pain was defined as pain in an area bounded by the lower ribs and the crease of the buttocks. If no pain was reproduced in this area, the rater was asked to record a 0 for that level.

Stiffness judgments were recorded on a scale developed for this study. The scale ranged from -5 (markedly decreased stiffness) to 5 (markedly increased stiffness), with 0 representing normal stiffness. Raters were asked to judge the stiffness of each lumbar level based on what they would expect to be normal for that patient and spinal level. None of the raters had previously used this stiffness rating scale, but as experienced manipulative physical therapists, each rater was familiar with making stiffness judgments. Similar numerical scales have been used successfully in psychological research to study the perception of sensations such as heaviness, thickness,[15] and firmness[16] and by Binkley's team when studying the reliability of judgments of PA mobility.[7] The rating scale was explained prior to data collection, and raters were allowed 2 weeks to practice using the scale before data collection began. To reduce any possible recall bias on the part of the patients when they provided their reports of pain intensity, they were not told that the study was a reliability study but rather that the study would investigate a new method of recording examination findings.

Data Analysis

Data for the three rating pairs were pooled, and the pain and stiffness ratings for all 90 patients were compared at each of the five lumbar levels. This comparison, we believe, provides a better estimate of the population value for intertherapist reliability than the results of a single pair of raters and so was used to evaluate judgment reliability. Intertherapist reliability of the individual pairs was also evaluated by comparing the two ratings for the 30 patients rated by that pair.

Intraclass correlation coefficients (type 1,1) with 95% confidence intervals[17] were calculated to express interrater reliability. The ICC (1,1) index reaches a maximum of 1.00 when raters are in complete agreement, with 1-ICC representing the percentage of variance due to the disagreement between raters. With two raters, the theoretical lower limit for the ICC (1,1) is -1.[14] Negative values, however, are uncommon and indicate a raterxsubject interaction.[18,19] Agreement between raters was expressed by noting the number of occasions for which ratings were in complete agreement and expressing this value as the percentage of exact agreement (PEA).

The ICC statistic is sensitive to a restriction in the range of ratings and in such a scenario can suggest poor reliability even when there is substantial agreement between raters. The use of both the ICC (1,1) and the PEA rather than the ICC index alone has been suggested to provide a more robust analysis of reliability that will alert the investigator to this potential problem.[20] The data analysis was performed using software developed at The University of Sydney for the calculation of ICCs and PEAs. The extent of any relationship between pain and abnormal spinal joint stiffness was evaluated by correlating the absolute (unsigned) value of the stiffness and pain judgments.

Results

The ICC (1,1) values with 95% confidence intervals for the pooled data of all raters are illustrated in the Figure. Values for judgments of pain for the group as a whole ranged from .67 to .73, with the confidence intervals predicting that the population value would fall within the range of .55 to .81. The values for stiffness for the group as a whole were less, ranging from .03 to .37 with the confidence intervals predicting that the population value would fall within the range of .18 to .53.

Inspection of the ICC (1,1) values for pain and stiffness at each level (Figure) reveals that the confidence intervals do not overlap and thus provides support for the hypothesis that assessment of pain during lumbar spine examination is more reliable than assessment of stiffness. Inspection of the values across levels reveals that for both pain and stiffness, the level tested does not significantly affect the reliability of the test.

The PEA values ranged from 21% to 29% for stiffness judgments and from 31% to 43% for pain judgments. Although these PEA values seem to be of similar magnitude and to contradict the ICC values, the distribution of the pain and stiffness judgments were very different, as shown in Table 3. The raters tended not to use all of the 11-point scale when rating stiffness, with most raters effectively collapsing the scale into a 6-point scale. This was not so with the pain scale, where the raters generally used all of the 11 points on the scale. With this in mind, it is apparent that 21% to 29% agreement on a 6-point scale actually represents poorer agreement than 31% to 43% agreement on an 11-point scale. The low PEA figures for stiffness judgments also show that the low ICC (1,1) values for stiffness judgments were not simply the result of a lack of variability in the patients to be rated.

Table. 3 Distribution of Stiffness and Pain Judgments Recorded
by Each Rater

             Rater 1                Rater 2
             X      SD     Range    X      SD     Range

Stiffness
  Pair 1     2.5    1.3    0-5      3.0    1.2     0-5
  Pair 2     2.3    1.4    0-5      2.3    1.3     0-5
  Pair 3     2.0    1.2    0-5      1.7    1.6    -2-5

Pain
  Pair 1     3.9    2.6    0-10     4.2    2.5     0-10
  Pair 2     2.7    2.8    0-10     2.8    2.9     0-10
  Pair 3     2.6    2.1    0-10     3.2    2.3     0-9

The reliability indexes for each rating pair are shown in Table 4. There was great variability in the performance of different rating pairs for judgments of stiffness, with ICC (1,1) values ranging from -.40 to .73. Ratings of pain were more consistent, although ICC values also showed a wide range, from. 27% to .85. The PEA values showed similar variability, ranging from 13% to 43% for stiffness judgments and from 27% to 57% for pain judgments.

[TABULAR DATA OMITTED]

The Pearson r values obtained from correlating the pain and absolute value of stiffness judgments are shown in Table 4. These values ranged from .27 to .40.

Discussion

Judgments made by experienced manipulative physical therapists when using the PA central pressure test in their own clinics to examine patients with low back pain were found to be affected by the type of judgment made, that is, whether the judgment was related to pain or stiffness, but not by the spinal level tested. Our results are consistent with those of a number of studies that similarly demonstrated that manual therapy tests that attempt to assess factors such as range of motion (ROM), alignment, or joint compliance yield unreliable results,[21-25] whereas manual therapy tests that seek to provoke symptoms yield more reliable results.[22,23,26]

Pain judgments were found to be more reliable than stiffness judgments in this study, but the reliability of the pain judgments was still less than that possible with many other clinical tests used by physical therapists. For example, ICC values above .80 have been achieved for measurements of spinal ROM,[27-29] spinal curves,[30,31] and forward-bending ROM.[32] This finding means that if physical therapists seek to document recovery in a patient with low back pain, they may be better served by these more reliable measurements than by measurements obtained with the PA central pressure test. The case would be even more compelling if there is a high cost, either for the patient or for the therapist, associated with an incorrect decision (eg, incorrectly classifying someone as fit for work). In contrast, if a physical therapist is using the results of the PA pressure test to select the spinal level to treat, then the assessment of PA pressure may be a useful measurement, particularly as there are no real alternate clinical measurements that a physical therapist can use to guide treatment. The physical therapist should recognize, however, that pain judgments are more reliable than stiffness judgments and so may be a better guide to the level to treat.

Inspection of the ICC values for individual rating pairs (Tab. 4) revealed that the third rater pair had much less reliable ratings than did the other pairs. No factor could be identified that would satisfactorily account for this result. These two physical therapists were similar to the other pairs of raters with respect to clinical experience, time in practice together, and clientele. Early on in the data collection, it became apparent from inspection of the raw data sheets that this pair's ratings were not in agreement, and the rating scale was therefore reexplained to them. Both therapists, however, confirmed that they understood the nature of the scale, so data collection continued. Removing this pair's ratings from the pooled analysis does not change the major findings.

On average, the pain judgments of pairs 1 and 2 demonstrated good reliability, whereas their stiffness judgments were poor.

The poor reliability for stiffness assessment observed in our study is consistent with the majority of studies that have evaluated PA central pressure.[6-9] Although the comparison of ICC values across studies is problematic, the mean ICC value of .20 for stiffness judgments found in our study is similar to the value of .25 reported by Binkley et al.[7] The results of our study and of the study by Binkley et al[7] do not suggest that experienced manipulative physical therapists are able to perform any better in this regard than the generally less experienced raters used in other studies. For example, Matyas and Bach,[6], in their studies using less experienced raters, found Kappa values for stiffness assessment ranging from .08 to .34. Reliability testing of raters in their normal clinical environment and using the style of their choice to perform the PA pressure test also did not augment reliability.

At present, it is not clear what part or parts of the process of judging PA stiffness are responsible for the relatively poor ability of manipulative physical therapists to rate PA stiffness. The task of rating stiffness is complex, with the physical therapist required to judge the amount of the PA stiffness and then to compare it with some internal schema of normal stiffness for that patient and level. This process takes place within an environment with other competing stimuli present that may distract the physical therapist from the task of rating PA stiffness. Finally, the physical therapist has to communicate his or her judgments using the numerical scale. Following data analysis, debriefing sessions took place with the raters to get their opinions on the source of the unreliability.

All of the raters suggested that judging whether the perceived PA stiffness is normal for a particular spinal level was the most difficult step in the judgment process, as there are no published normative values for PA stiffness and therapists must therefore rely on their experience. Inspection of the distribution of ratings used by each rater (Tab. 3) suggests that at least for pairs 1 and 3, this problem did contribute to the poor reliability of stiffness judgments observed in this study. For these two pairs, one therapist's mean rating was higher than the partner's rating, suggesting the presence of a systematic bias. This finding could have occurred because there was no clear agreement between the raters in each pair on what constitutes normal stiffness at each lumbar level. In such a situation, the reliability of judgments may be enhanced if normative values for PA stiffness could be established.

Another potential source of unreliability is the perception of stiffness itself. The studies that have investigated the perception of stiffness[12,33-35] have produced widely different estimates of the ability of subjects to discriminate stiffness. One early study[35] demonstrated that stiffness stimuli could be discriminated when one stimulus is 9% greater than the other, whereas later studies have suggested 17%,[34] 23%,[33] and 100%[12] as the threshold of discrimination. We are currently attempting to clarify this situation by establishing the detection threshold and discriminability function for PA stiffness using a range of stiffness stimuli typical of those that would be encountered in the lumbar spine.

If it can be established that physical therapists are able to discriminate stiffness stimuli in such a simple setting, then it could be justifiable to attempt to design a standardized protocol for the clinical assessment of PA stiffness. Such a protocol would seek to optimize the conditions for the assessment of PA stiffness. At present, it is not clear what factors affect the perception of PA stiffness. Psychophysical research, however, has shown that surrounding stimuli, experience of stimuli in the past, patterns of attention, and the observer's expectations can all affect judgments of stimulus intensity.[36] It is possible that a similar range of factors can be identified for the perception of PA stiffness. One strategy that may be useful is to institute a form of training program that provides raters with feedback on their performance to improve their ability to make judgments of stiffness. Such programs have been shown to reduce the error in specified force production using the PA central pressure test.[37] Similar programs providing feedback on the stiffness of the target may provide improvements in the accuracy of physical therapists' ratings of stiffness.

It is ironic that manipulative physical therapists have placed such great importance on the PA central pressure test when the test yields such unreliable information about stiffness. According to measurement theory, an unreliable test will not convey meaningful information and so win not help management decisions, yet manual therapists consistently use this test and other similar tests in clinical practice and presumably find them of value in their clinical assessment. This suggests that therapists may be attending to factors other than stiffness, and it is this other information that could make the tests of value to physical therapists. Perhaps the practitioners are attending to the more reliable pain reports of their patients when deciding where to direct treatment, and it is this aspect of the test that makes it clinically useful.

The Pearson r values for the correlation between the pain judgments and the absolute value of the stiffness judgments were modest, suggesting that at least in the group of patients used in this study, the relationship between pain and abnormal PA stiffness is either weak or inconsistent. This finding is of interest because many manual therapy texts have suggested that there is a relationship between pain and abnormal PA stiffness and have advocated treatments aimed at restoring normal PA stiffness in order to bring about a resolution of patient symptoms.[3-5] Although it may be argued that the results represent an attenuation of correlation due to unreliable measurements,[38] manual assessment is the same assessment method used by the manipulative therapists who have developed and supported this theory. This issue deserves further investigation because it is this assumption of a relationship between a patient's symptoms and abnormal PA stiffness that forms the philosophical basis of many manual therapy approaches and leads many manipulative therapists to assess PA stiffness.

Inspection of the distribution of stiffness ratings in Table 3 reveals that most raters did not use the reduced stiffness end of the stiffness rating scale. Only one rater used this end of the scale, and then only for six patients. This distribution of results in our study would seem to contrast with the opinions of Derosa and Porterfield,[39] who state that instability or reduced stiffness is more of a problem in the average clinic than increased stiffness. Our results do not provide support for Derosa and Porterfield's opinion. This situation may reflect a difference between the types of patients with low back pain seen in Australia and in the United States, or it may reflect some expectation on the part of the raters.

Clinical implications

The results of our study suggest that it may be unwise to base a patient management plan predominantly on judgments of PA stiffness. The poor reliability of our stiffness judgments means that there is the possibility that this information provides a false impression of meaningfulness that hinders rather than helps treatment selection and patient management. In our investigation, there were a number of occasions when a patient was considered to have increased stiffness by one rater and reduced stiffness by the other rater. As most manual therapy texts suggest a different treatment approach for a patient with low back pain who has increased PA stiffness at a symptomatic level compared with a patient with reduced stiffness, these contradictory findings would have led the two physical therapists to institute very different treatment programs for the same patient.

The view that manual therapy tests that seek to provoke symptoms tend to yield the most reliable results has been supported by several authors.[13,40-42] These authors have suggested that the more reliable pain provocation tests should form the basis for clinical decisions rather than the tests assessing joint compliance, alignment, or muscle tension. Our findings support this suggestion.

Conclusion

The results of this investigation demonstrate that judgments of PA stiffness made by experienced manipulative physical therapists using the PA central pressure test within a clinical setting are not acceptably reliable. The test, however, provides more reliable results when pain provocation is the goal.

Acknowledgments

We thank the manipulative physical therapists and their patients who participated in the study and Jane Latimer and Michael Lee for their comments on earlier drafts of the manuscript.

[TABULAR DATA OMITTED]

REFERENCES

[1] Nachemson A. Newest knowledge of low-back pain: a critical look. Clin Orthop. 1992; 279:8-20. [2] Koes BW, Bouter LM, Beckerman H, et al. Physiotherapy exercises and back pain: a blinded review. Br Med J. 1991;302:1572-1576. [3] Maitland GD. Vertebral Manipulation. 5th ed. London, England: Butterworth & Co (Publishers) Ltd; 1986. [4] Mennell J. Back Pain: Diagnosis and Treatment Using Manipulative Techniques. Boston, Mass: Little, Brown and Co Inc; 1960. [5] Stoddard A. Manual of Osteopathic Practice. 2nd ed. London, England: Hutchinson and Co; 1983. [6] Matyas T, Bach T. The reliability of selected techniques in clinical arthrometrics. Australian Journal of Physiotherapy. 1985;31:175-199. [7] Binkley J, Stratford P, Gill C. Intertherapist reliability of lumbar accessory motion mobility testing. In: Proceedings of the Fifth International Conference of the International Federation of Orthopaedic Manipulative Therapists; 1992: Vail, Colorado. 1992:150-151. [8] Hardy GL, Napier JK. Inter- and intratherapist reliability of passive accessory movement technique. New Zealand Journal of Physiotherapy. 1991;19:22-24. [9] Viner A, Mackey K, Porter D, Lee M. Perception of stiffness in spinal assessment. In: Proceedings of the Seventh Biennial Conference of the Manipulative Physiotherapists Association of Australia; 1991; Blue Mountains, New South Wales, Australia. 1991:238-243. [10] Jull G, Bullock M. A motion profile of the lumbar spine in an ageing population assessed by manual examination. Physiotherapy Practice. 1987;3:70-81. [11] Minucci A. Palpation of the thoracic spine (T1-T8). In: Proceedings of the Fifth Biennial Conference of the Manipulative Therapists Association of Australia; 1987; Melbourne, Victoria, Australia. 1987:367-376. [12] Trott PH, Evans DH, Frick R. Accuracy of manual palpation skills: the ability of manipulative physiotherapists to detect changes in force and displacement on a palpation simulator. In: Proceedings of the Sixth Biennial Conference of the Manipulative Therapists Association of Australia, 1989; Adelaide, South Australia, Australia. 1989:207-214. [13] Lee M, Latimer J, Maher C. Manipulation: investigation of a proposed mechanism. Clinical Biomechanics. 1993;8:302-306. [14] Bartko JJ, Carpenter WT. On the methods and theory of reliability. J Nerv Ment Dis. 1976; 163:307-317. [15] McNicol D, Pennington CW. Sensory and decision processes in anchor effects and after-effects. J Exp Psychol. 1973;100:232-238. [16] Roberston GL, Koopmanschap EA, Scrivens CA. Comparison of instrumental and sensory panel methods for measuring kiwifruit firmness. Journal of Texture Studies. 1984;15:275-283. [17] Shrout P, Fleiss J. Intraclass correlations: uses in assessing rater reliability. Psychol Bull. 1979;86:420-428. [18] Tinsley HEA, Weiss DJ. Interrater reliability and agreement of subjective judgements. Journal of Counselling Psychology. 1975;22:358-376. [19] Armstrong GD. The intraclass correlation as a measure of inter-rater reliability of subjective judgements. Nurs Res. 1981;30:314-320. [20] Rey JM, Plapp JM, Stewart G, et al. Reliability of the psychosocial axes of DSM-III in an adolescent population. Br J Psychol. 1987;150: 228-234. [21] Chesworth B, Koval J, Vandervort A. A pilot study to compare the subjective and objective evaluations of passive ankle joint stiffness. Physiotherapy Canada. 1991;43:13-18. [22] Potter N, Rothstein JM. Intertester reliability for selected clinical tests of the sacroiliac joint. Phys Ther. 1985;65:1671-1675. [23] Keating JC, Bergman TF, Jacobs GD. The objectivity of a multidimensional index of lumbar segmental abnormality. J Manipulative Physiol Ther. 1990;13:463-470. [24] Mann M, Glasheen-Whray M, Nyberg R. Therapist agreement for palpation and observation of iliac crest heights. Phys Ther. 1984;64: 334-338. [25] Carmichael J. Inter- and intra-examiner reliability of palpation for sacroiliac joint dysfunction. J Manipulative Physiol Ther. 1987;10: 164-171. [26] Laslett M, Williams M. The reliability of tests for sacroiliac joint pathology. In: Proceedings of the Fifth International Conference of the International Federation of Orthopaedic Manipulative Therapists; 1992; Vail, Colorado. 1992:174-175. [27] Youdas J. Carey J, Garrett T. Reliability of measurements of cervical spine range of motion: comparison of three methods. Phys Ther. 1991;71:98-104. [28] Rheault W, Albright B, Byers C. et al. Intertester reliability of the cervical range of motion device. J Orthrop Sports Phys Ther. 1992;15: 147-150. [29] Beattie P, Rothstein JM, Lamb RL. Reliability of the attraction method for measuring lumbar spine backward bending. Phys Ther. 1987;67: 364-369. [30] Lovell F, Rothstein JM, Personius W. Reliability of clinical measurements of lumbar lordosis taken with a flexible rule. Phys Ther. 1989;69:96-105. [31] Hart D, Rose S. Reliability of a non-invasive method for measuring the lumbar curve. J Orthop Sports Phys Ther. 1986;8:180-184. [32] Gauvin M, Riddle DL, Rothstein JM. Reliability of clinical measurements of forward bending using the modified fingertip-to-floor method. Phys Ther. 1990;70:443-447. [33] Jones LA, Hunter IW. A perceptual analysis of stiffness. Exp Brain Res. 1990;79:150-156. [34] Roland PE, Ladergaard-Peterson H. A quantitative analysis of sensations of tension and of kinaesthesia in man: evidence for a peripherally originating muscular sense and for a sense of effort. Brain. 1977;100:671-692. [35] Scott-Blair GW, Coppen FMV. The subjective judgement of the elastic and plastic properties of soft bodies: the "differential thresholds" for viscosities and compression moduli. Proc R Soc Lond. 1939;128B:109-125. [36] Coren S, Porac C, Ward L. Sensation and Perception. New York, NY: Academic Press Inc; 1966. [37] Keating J, Matyas T, Bach T. The effect of training on physical therapists, ability to apply specified forces of palpation. Phys Ther. 1993; 73:38-46. [38] Cohen J. Statistical Power Analysis for the Behavioural Sciences. 2nd ed. Hillsdale, NJ: Lawrence Erlbaum Associates Inc; 1988. [39] Derosa CP, Porterfield JA. A physical therapy model for the treatment of low back pain. Phys Ther. 1992;72:261-272. [40] Walker J. The sacroiliac joint: a critical review. Phys Ther. 1992;72:903-916. [41] Riddle DL. Measurement of accessory motion: critical issues and related concepts. Phys Ther. 1992;72:865-874. [42] Maher C, Latimer J. Pain or resistance: the manual therapist's dilemma. Australian Journal of Physiotherapy. 1992;38:257-260.