Is the clinical concept of spinal stiffness multidimensional?

Diagnosis that specifies the anatomical source of the symptoms is not possible for the majority of patients with low back pain (LBP); instead, these patients are often labeled as having nonspecific low back pain (NSLBP).[1] Prior to applying manipulative techniques for patients with NSLBP, physical therapists typically use a range of manual tests to identify the level of the spine to be treated.[2] One test, the posteroanterior (PA) central pressure test,[2] requires the physical therapist to apply a graded force over the spinous process of the prone patient to judge the quality of the movement that results and to note any reports of changes in symptoms. Although a number of terms have been used to describe the quality of the perceived movement, the term "stiffness" is the most common descriptor found in the osteopathic,[3,4] chiropractic,[5] and physical therapy[2,6] literature.

Although judgments of PA stiffness continue to be emphasized in the physical therapy literature (for example, see the recent article by Jull et al[6]), a recent study of ours[7] has confirmed the results of earlier research8 in showing that the use of current protocols for assessing PA stiffness results in judgments of poor reliability. We reported very low intertherapist reliability for judgments of stiffness made by experienced manipulative physical therapists examining patients with NSLBP in their own clinics. At present, it is not clear what part or parts of the process of judging PA spinal stiffness are responsible for the relatively poor ability of manipulative physical therapists to rate this property in the clinic; however, a review of the psychophysical research literature suggests to us that it is unlikely to be due to poor stiffness discrimination.

Psychophysics is that branch of psychology concerned with the relationship between the physical properties of a stimulus and the sensations it evokes. In the case of stiffness perception, research is concerned with the relationship between the stiffness of some material (expressed in terms of how much force is required to produce a given deformation) and the sensations evoked when the material is deformed by a subject. One psychophysical method for characterizing a subject's ability to judge stiffness of stimuli is to determine the discrimination threshold for stiffness perception. The discrimination threshold describes how different in proportional terms two stimuli must be before one stimulus can be consistently discriminated as noticeably different from the other. The "just noticeable difference" can be expressed as a percentage of difference that allows 75% accurate discrimination, and is commonly referred to as the Weber fraction.

Psychophysical research over the last 50 years has shown that humans possess a good ability to discriminate between stiffness Stimuli,[9-12] with estimates of the discrimination threshold (Weber fraction) being around 12%. We recently investigated stiffness discrimination using a protocol that simulated clinical practice.[13] In our study, 72 subjects used the pisiform grip2 when judging a set of stimuli (metal springs set in the mechanical device shown in the Figure) in the range of 6 to 11 N/mm. The relative increment in stiffness needed so that two stimuli could be differentiated 75% of the time was an 11% increase, a value consistent with the previous research. With a discrimination threshold of 11% for manual assessment of stiffness and PA spinal stiffness values that range from 5.3 to 27 N/mm,[14,15] it would seem that physical therapists should possess sufficient acuity of stiffness discrimination to rate the PA stiffness of their patients.

We contend that poor stiffness discrimination is not the reason for the poor reliability of clinical judgments of PA stiffness. The psychophysical literature suggests two more likely options, both of which are consistent with good stiffness discrimination ability. The first of these hypotheses is that the scaling procedure used to judge PA stiffness is the problem. Manipulative physical therapists generally judge stiffness relative to an experiential model of normal or average stiffness. Jull and colleagues,[6] for example, have suggested that alterations in segmental tissue stiffness that are related to symptoms can be detected with manual tests and that this information is important for treatment planning. To record such information, Jull and colleagues have developed a scale in which the categories are anchored to the rater's expectation of what constitutes average or normal. For example, grade 1 on the scale has the following critetia:

No resistance is perceived through

average displacement; displacement is

greater than aterage range. The normal

elastic limiting resistance is not encountered

at end of displacement [italics

added].[6(pl65)] When physical therapists attempt to scale PA spinal stiffness, insofar as these scaling attempts are similar to scaling other stiffness stimuli, it might be expected that they will be influenced by the same kinds of factors that have been shown elsewhere to influence such ratings, that is, recent stimuli in the series, stimuli experienced in the past, and the rater's patterns of attention.[16-19] The effect of past experience on psychophysical judgments led Helson[20,21] to develop his theory of adaptation level, which states that such judgments are made relative to an individual's adaptation level, which is a weighted value determined that individual's past experience with stimuli on the relevant dimension. This theory would suggest, for example, that physical therapists who routinely examine patients with stiff spines will have different adaptation levels than therapists who only examine spines that are somewhat less stiff. Unless physical therapists have similar past experience, there are likely to be differences in what they would consider to be the midpoint or "normal" level of PA stiffness, and as a consequence their clinical judgments would be unlikely to agree.

An alternate explanation for the poor reliability of clinical manual PA stiffness judgments is that the concept of stiffness" physical therapists use when they assess the spine is not unidimensional, but rather multidimensional, and the kind and number of other dimensions interwoven may be idiosyncratic, and so differ from therapist to therapist. In support of this notion, there is evidence from within the physical therapy literature to suggest that what physical therapists perceive as stiffness (ie, sensory stiffness) may not be equivalent to mechanical stiffness. Maitland's text,[2] for example, uses fairly colorful and image-laden words to describe what is to be felt when a spine is assessed. Maitland's use of descriptors such as "leathery" or "spongy"[2(p74)]and his likening of the feel of normal passive joint movement to "wet soap sliding on wet glass"[2(p355)] hints that abnormal joint movement may involve a feeling of adhesiveness. The concept physical therapists refer to as "stiffness" may, therefore, have more than one dimension. Additionally, Maitland and other advocates of manual stiffness tests at times use mechanical terms imprecisely. Maitland uses the terms "resistance" and "stiffness" interchangeably[2(p351)] in a way that suggests that, for him, sensory stiffness is not equivalent to mechanical stiffness. Such confusion may increase the likelihood of individual concepts of stiffness being used during clinical assessment.

Our hypothesis is that the poor reliability of clinical judgments of PA stiffness is unlikely to be due to poor stiffness discrimination but rather is due to either a lack of consensus among raters as to what constitutes normal" stiffness or idiosyncratic multidimensional stiffness constructs. The first possibility can be examined by the use of a dismantling methodology for reliability indices proposed by Rey et al.[22] These authors noted that the intraclass Correlation coefficients ICC[2,1] and ICC[3,1]) from Shrout and Fleiss[23] and the Pearson Product-Moment Correlation Coefficient (r) form a linked set, in that the ICC (2,l) is sensitive to both additive and multiplicative rater biases, the ICC(3,1) is sensitive to multiplicative rater biases only, and Pearson's r is not sensitive to either type of rater bias. (The ICC[1,1], however, is not part of this "set" because it is based on the mean squares obtained from a one-way analysis of variance and so cannot be related to the other indices in such a manner. Differences between these reliability indices can thus be used to assess a data set for the presence of different kinds of rater bias. If the physical therapists differ in terms of their adaptation levels for stiffness, this should be apparent in a diference between the computed values of the ICC (2,l) and the ICC (3,l).

The second possibility can be examined by comparing the reliability of judgments of multidimensional stimuli made on a scale that was anchored to an experiential model of normal stiffness with the reliability of judgments of unidimensional stiffness stimuli using a scale anchored to reference stimuli. A difference in judgment reliability would provide evidence in support of the hypothesis that stiffness is multidimensional.

We investigated these alternate explanations by asking the raters who took part in our reliability study[7] (when they rated human spines) to rate a set of unidimensional stiffness stimuli (springs) that were transported to their clinics. Their performance when rating springs on a relative-to-reference scale (a scale anchored to three reference stiffness stimuli) was compared with their previous performance rating human backs on a relative-to-experience scale (a scale anchored to a rater's experience of "normal" stiffness).

Method

Overview of Reliability Study[7]

Three pairs of manipulative physical therapists with a minimum of 5 years of experience were asked to rate pain and PA stiffness in a total of 90 patients with NSLBP. Each pair of therapists assessed 30 patients within their own clinic, with each rater allowed to use a preferred technique to perform the PA central pressure test at the five lumbar levels. Each pair of therapists recorded their ratings of stiffness on a scale that ranged from -5 (markedly reduced 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. Reliability of stiffness judgments was evaluated by ICCs and percentage of exact agreement scores, both of which showed poor reliability of stiffness judgments.

Current Study

The raters involved in the study were the same experienced manipulative physical therapists who had taken part in our reliability study[7] and were all 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 physical therapy. 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 further had at least 5 years clinical experience as a manipulative physical therapist. The mandatory continuing education program requires members to accrue five points within 4 years, with points awarded for teaching, learning, and research activities related to manipulative physical therapy. For example, publishing a research article in a peer-reviewed journal would accrue two points. These raters reported that they regularly used die PA central pressure test to assess their patients with LBP. The experience of each of the raters at the time of participating in the original study (completed 1 year prior to this study) is given in Table 1.

Stiffness Stimuli

The unidimensional stiffness stimuli were provided by a mechanical device (Figure). The point of contact for the subject's hand was a raised metal pad on a 50-cm metal lever that was free to rotate about a bearing at one end and rested on a compression spring at the free end. The spring was mounted in the base and provided the resistance to downward movement of the lever.

The stiffness of the downward movement at the point of contact could be altered either by changing the spring or by changing the position of the spring in the mounting base plate and thereby altering the effective lever arm the subject used. The effective stiffness value of a given spring could be altered by moving the spring to a new hole in the base plate or by changing the base plate to a new one with a different arrangement of mounting holes. Prior to data collection, the device was calibrated using weights of known mass to provide the force and a dial indicator (model 2050F(*)) to measure the resultant displacement of the point of contact for each spring and base-plate combination used in the study. A dial indicator is a precision instrument used to measure small displacements. The model used in our study has a range of 0 to 20 mm in 0.01-mm increments.

The 30 stiffness stimuli generated for the study were chosen so that they would be typical of PA stiffness values that may be observed in patients with LBP. Lee and colleagues,[16,17] measured the PA stiffness of a total of 41 asymptomatic subjects and reported PA stiffness values ranging from 5.3 to 27 N/mm. Latimer and colleagues, in a master's research project24 and in an unpublished study, measured the PA stiffness of 51 patients with LBP and reported a similar range of PA stiffness values, so these values were used to select 30 stiffness stimuli from the range that could to be generated by the stifness device. A description of the PA stiffness values of the 51 patients with LBP and the 30 stiffness stimuli used in this study is provided in Table 2.

Table 2. Comparison of Stiffness Values Used in This Study With Typical Posteroanterior (PA) Stiffness Values of Patients With Nonspecific Low Back Pain NSLBP)(a)

           Typical PA
           Stiffness
           Values for       Stiffness Stimuli
Stiffness  Patients         Used in This
(N/mm)     With NSLBP       Study (N=30)
X          14.73            15.19
Median     14.6             14.6
SD          3.79             4.41
Minimum     8.87             7.83
Maximum    26.65            27.04
Range      17.78            19.21
(a) Lumbar PA stiffness values were obtained
from two studies (a master's research project[24]
and unpublished research) that measured the
PA stiffness of a total of 51 patients with
NSLBP. All values are expressed in newtons
per millimeter.

Scaling Method

Stiffness judgments were recorded using the same scale as that used in our reliability study.[7] The scale ranged from -5 (markedly decreased stiffness) to 5 (markedly increased stiffness), with 0 representing normal stiffness. When rating the stiffness of human spines, the raters were asked to judge the stiffness of each lumbar level relative to what they would expect to be normal for that patient and spinal level (ie, the scale was anchored relative to their experiential model of normal PA stiffness). In contrast, when the raters rated spring stiffness, they used a relative-three reference scale with three reference spring stiffness values made available to them prior to the data collection. The reference values for -5, 0, and 5 were the minimum, mean, and maximum stiffness values of the 30 stiffness stimuli.

Prior to data collection, the raters were allowed to practice pushing on the stiffness device to familiarize themselves with the reference stimuli and were instructed that during the rating session they would be required to judge the stiffness of the 30 stimuli relative to these reference stimuli. Raters took between 5 and 10 minutes to familiarize themselves with the three reference stiffness stimuli, which were separately made available to them in the order they requested.

Data Collection

The subjects were asked to press on the stiffness device and make a judgment of the relative stiffness of a given stimulus. Subjects pushed on the lever using the pisiform grip version of the PA central pressure test, a test advocated by Maitland[2] to assess the PA stiffness of the lumbar spine. With this grip, pressure is applied through the heel of the hand to a point of contact, in this case the contact pad of the stiffness device, but in clinical practice pressure would be applied to points on the spine (eg, over the spinous process of a lumbar vertebra).[2] Subjects were allowed to push on the stiffness device with whatever version of the pisiform grip they preferred and for as long as they wished. Raters usually pushed on the device two or three times and then made a judgment of the magnitude of the stimulus presented. Data-collection sessions took about 30 minutes, with ratings recorded by one of the experimenters. During practice and rating sessions, the subjects wore a blindfold. The order of presentation of the 30 different stimuli was the same for all raters. The random sequence, with the values expressed in newtons per millimeter, was as follows: 13.03, 17.13, 13.69, 19.87, 13.69, 15.06, 12.39, 18.93, 14.37, 12.39, 14.60, 11.76, 14.60, 21.80, 15.77, 10.56, 16.26, 15.77, 17-24, 9.42, 15.42, 23.83, 10.82, 7.83, 20.82, 18.02, 13.02, 11.15, 27.04, and 9.42.

Data Analysis

Interrater reliability was evaluated for each rating pair from their stiffness ratings for the 30 stiffness stimuli (metal springs), and the value obtained was compared with the reliability of judgments from our reliability study (human spines). Intraclass correlation coefficients (1,1), (2,1), and (3,1)[23] and Pearson's r were calculated to express interrater reliability. These forms of the ICC index reach a maximum of 1.0 when raters are in complete agreement, with 1-ICC representing the percentage of variance due to the disagreement among raters.[25] With these forms of the ICC index and two raters, the ICC formns (1,1) and (3,1) have a lower limit of -1 and ICC form (2,1) has a lower limit of minus infinity.[26] Pearson's r is an index of association and varies between - 1 and 1, with 0 representing no association. The difference in the mean ratings (DIMR) of each rater pair was also calculated. The extent of the relationship between raters' estimates of the magnitude of the stimuli and the measured stiffhess of the stimuli was assessed using Pearson's r.

Results

The reliability indices for rating stiffness of springs and human spines are shown in Table 3. The average interrater reliability for rating stiffness of springs was found to be markedly greater than that for rating stiffness of human spines. The improvement in reliability was most marked for rater pair 3, who achieved reliability indices around .70 when rating spring stiffness, whereas previously they had negative ICC values when rating the human spine stiffness. For rater pairs 1 and 3, the four reliability indices were of similar magnitude, suggesting that no additive or multiplicative bias was present. For rater pair 2, however, the situation was quite different when they rated spring stiffness. Although the Pearson's r and ICC (3,1) indices were of similar magnitude, the ICC (2,1) value was much lower, suggesting that an additive bias contributed substantially to the poor reliability for this rater pair. This interpretation is confirmed by inspection of the DIMR for this rater pair, which was 3.47 units on an 11-point scale, whereas for the other rater pairs the DIMR values were substantially smaller.

The correlations between the judged stimulus magnitude and the measured stiffness magnitude are shown in Table 4. The average correlation was .80, with individual correlations ranging from .71 to .91.

Table 4. Correlation of Judged Stiffness Magnitude (on 11-Point Scale) With Measured Value of Stiffness Stimulus (in Newtons per Millimeter) for Each Rater

Pearson's                   Pearson's
Pair             Rater      r
1                1          .79
                 2          .82
2                3          .91
                 4          .76
3                5          .80
                 6          .71
Average                     .80

Discussion

When the Rey et al[22] reliability index breakdown method was applied to the human spine data from our reliability study,[7] there was no evidence that the low ICC values could be attributed to any form of bias. This was not the case, however, for all rater pairs when rating spring stiffness. Rater pair 2 showed a decline (of .33) from ICC (3,1) to ICC (2,1), indicating that the source of the poor reliability of their ratings was indeed additive bias (sometimes known as the "hard marker, soft marker" effect, where two examiners might preserve the same order in their gradings of a set of papers, but their grades differ by a constant amount on each one). Taking this form of bias into account, and assuming that such bias can be corrected by giving this rater pair more practice with agreed reference stimuli and the use of the scale, then the potential reliability of rating spring stiffness for rater pair 2 is represented by their ICC (3,1) value, the value of the ICC statistic that is insensitive to an additive bias. This would mean that the average potential reliability of therapist pairs' ratings of spring stiffness is of the order of .70, a value considerably better than we found in our reliability study.[7] Such a simple, training program, involving practice with agreed reference stimuli to correct rater bias, however, would not work to enhance the reliability of clinical judgments of lumbar PA stiffness, because there was no similar bias causing the poor reliability. The cause of the poor reliability of clinical judgments of lumbar PA stiffness is apparently a more complex issue than the relatively straightforward "hard marker, soft marker" effect.

The lack of evidence for rater bias in our reliability study (when the raters judged the stiffness of human spines) and the considerable improvement in reliability when the same raters judged unidimensional stiffness stimuli (springs) suggest that there is not complete agreement among raters on the concept of lumbar PA stiffness. We believe that the construct of "PA stiffness" could have more than one dimension and that idiosyncratic interpretation of the construct led to the poor reliability evident in our earlier study.[7] The development of a reliable measurement protocol for judgments of lumbar PA stiffness, we contend, will therefore not be possible till the dimensions of PA stiffness are clearly defined. Identifying the other dimensions involved, and establishing whether stiffness can be isolated from them, requires further research.

Our hypothesis that the construct of spinal stiffness has more than one dimension is supported by a recent article by Jull et al,[6] who suggested that abnormal PA spinal stiffness has two distinct features. These authors suggested that spinal stiffness contains elements of "reactive" tissue stiffness as a result of muscle spasm and "thicker through range" stiffness, which results from pathological articular or connective tissues. Although it is unclear whether it is possible to infer the source of stiffness, the comments do confirm that for these authors, at least, stiffness has more than one dimension. Jull and colleagues[7] suggestion to use more than one term to describe the perceived mechanical properties of spinal tissues does not seem unreasonable when it is considered that the discipline of biomechanics requires a variety of terms to characterize the measured mechanical properties of spinal tissues.[27]

Manipulative therapy texts continue to offer a wealth of terms to describe what is to be felt when the mechanical properties of the spine are assessed with manual tests (eg, "end-feel," stiffness," "range of motion"),[2] although reliability studies have consistently demonstrated that the current protocols for judging these terms usually yield measurements with poor reliability.[28] Most of these terms lack operational definitions, so the relationship between terms or the exact meaning of each term is far from clear. if individual therapists' experience has led them to incorporate certain characteristics such as "end-feel,"[2] "spasm,"[29] and "springiness"[3] into their "sfiffness" judgments, then these characteristics will be confounded with stiffness. It is our view that measurements obtained with manual examination tests that seek to evaluate these characteristics are likely to have poor reliability until the meaning of each of these terms and their interrelationships become clear through a program of research.

Research could be similar to that conducted within the food industry. Food workers were in a similar position as that of physical therapists when it was discovered that the texture judgments of expert bakers[30] and cheese makers[3l] were not reliable. Collaborative research between physicists (who developed instruments for measuring food texture) and psychologists (who developed methods for measuring the responses of raters grading the food) has allowed the development of protocols that allow raters to make judgments of food texture that correlate highly with instrumental measures of food texture.[32,33] The protocols require the raters to attend training sessions where the different attributes of the food to be judged are identified, clearly defined, and referenced.[34] If research shows that this approach will work, it could be adopted by manipulative physical therapists wishing to develop reliable methods for judging PA stiffness.

The high correlations found between the measured intensity of the stiffness stimuli and the raters' estimates of the mechanical stiffness magnitude demonstrate that physical therapists are able to judge unidimensional stiffness satisfactorily, in the range of stiffnesses likely to be encountered in the lumbar spine. The correlations obtained in this study are of similar magnitude to those of previous studies[12,17] and provide further support for the suggestion that the major source of poor reliability of manual clinical lumbar spine stiffness judgments is not poor unidimensional stiffness perception.

Conclusion

The results of this study show that physical therapists can judge the magnitude of stiffness stimuli (provided by metal springs). Their judgments are reliable and correlate with the measured stiffness magnitude. That this performance is not evident in the clinic when physical therapists judge spinal "PA stiffness" suggests that mechanical stiffness is not equivalent to the concept of PA stiffness that physical therapists use when they assess the spine. We suggest that PA stiffness may well be a multidimensional construct and that idiosyncratic interpretation of the construct leads to poor interrater reliability in the clinic.

(*)Mitutayo Corp, 31-19 Shiba 5-chome, Minato-ku, Tokyo 108, Japan.

References

[1] Nachemson A. Newest knowledge of low-back pain: a critical look. Clin Orithop. 1992; 279:8-20. [2] Maitland GD. Vertebral Manipulation. 5th ed. London, England: Butterworth & Co (Publishers) Ltd; 1986. [3] Stoddard A. Manual of Osteopatbic Practice. 2nd ed. London, England: Century Hutchinson Publishing Ltd; 1983. [4] Mennell JM. Back Pain: Diagnosis and Treatment Using Manipulative Techniques. Boston, Mass: Little, Brown and Company Inc; 1960, [5] Nathan M, Keller T. Measurement and analysis of the in vivo posteroanterior impulse response of the human thoracolumbar spine. J Manipulative Pbysiol Ther. 1994;17:431-441. [6] Jull G, Treleaven J, Versace G. Manual examination: Is pain provocation a major cue for spinal dysfunction? Australian Journal of Physiotherapy. 1994;40:159-165. [7] Maher C, Adams R. Reliability of pain and stiffness assessments in clinical manual lumbar spine examination. Phys Ther. 1994;74:801-811. [8] Matyas T, Bach T. The reliability of selected techniques in clinical arthrometrics. Australian Journal of Physiotherapy. 1985;31:175-199. [9] Scott-Blair G, Coppen F. Differential threshold for compression modulus. Nature. 1939; 144:286. [10] Harper R, Psychological and psychophysical aspects of studies of craftmanship in dairying. Br J Psycholmonograph. 1952;28:1-63. [11] 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. [12] Mioche L, Auroy P, Lepetit J, Compagnon D. Oral perception of hardness in viscoelastic products. Journal of Texture Studies, 1991;22: 333-347. [13] Maher C, Adams R. A psychophysical evaluation of manual stiffness discrimination. Australian Journal Physiotherapy. In press. [14] Lee M, Latimer J, Maher C. Manipulation: investigation of a Proposed mechanism. Clin Biomecb. 1993;8:JO2--3O6. " [15] Lee M, Esler M, Mildren J, Herbert R. Effect of extensor muscle activation on the response to lumbar postero-anterior forces. Clin Biobmech. 1993;8:115-119. [16] Scott-blair G. The subjective assessment of firmness. Laboratoty Practice. 1954;3:3-9. [17] Sheppard D. Subjective assessments as scientific measurements. Laboratory Practice. 1953;2:488-493. [18] Sheppard D, Scott-Blair G. Subjective assessments of firmness. Q J Exp Psychol. 1952; 4:101-108. [19] Sheppard D. The importance of psychophysical errors in subjective judgements of firmness of Cheshire cheese. Journal of Dairy Research. 1952;19:348-355. [20] Helson H. Adaptation-level as a basis for a quantitative theory of frames of reference. Psychol Rev. 1948;55:297-313. [21] Helson H. Adaptation-level Theory: An Experimental and Systematic Approach to Behavior. New York, NY: Harper and Row; 1964, [22] Rey J, Plapp J, Richards I, Bashir M. Reliability in the psychosocial axes of DSM-III in an adolescent population. Br J Psychiatry. 1987;150:228-234. [23] Shrout P, Fleiss J. Intraclass correlations: uses in assessing rater reliability. Psychol Bull. 1979;86:420-428. [24] Latimer J. An Investigation of the Relationship Between Low Back Pain and Lumbar Stiffness. Lidcombe, New South Wales, Australia: School of Physiotherapy, The University of Sydney; 1995. Master's research project. [25] Bartko J. On various intraclass correlation reliability coefficients. Psychol Bull. 1976;83: 762-765. [26] Lahey MA, Downey RG, Saal FE. Intraclass correlations: There's more than meets the eye. Psychol Bull. 1983;93:586-595. [27] Keller T, Holm S, Hansson T, Spengler D. The dependence of intervertebral disc mechanical properties on physiologic conditions. Spine. 1990;15:751-761. [28] Maher C, Latimer J. Pain or resistance? the manual therapist's dilemma. Australian Journal of Pbysiotherapy. 1992;38:257-260. [29] Hickling J, Maitland G. Abnormalities in passive movement: diagrammatic representation. Physiotherapy. 1970;56:105-114. [30] Katz D. The judgements of test bakers: a psychological study. Occupational Psychology. 1938; 1 2:139 148, [31] Scott-Blair G, Coppen F. The subjective judgement of the elastic and plastic properties of soft bodies. Br J Psychol. 1940;31:61-79. [32] Finney E, Abbott J. Sensory and objective measurements of peach firmness. Journal of Texture Studies. 1972;3:372-378. [33] Perez C, Juliano B, Bourne M, Anzaldua-Morales A. Hardness of cooked milled rice by instrumental and sensory methods. Journal of Texture Studies. 1993;24:81-94. [34] Chambers E, Smith E. Effects of testing experience on performance of trained sensory panelists. Journal of Sensory Studies. 1993;8: 155-166.

C Maher, PT, is Lecturer, School of Physiotherapy, Faculty of Health Sciences, The University of Sydney, East St, POB 170, Lidcombe, New South Wales, Australia 2141 ([e-mail]). Address all correspondence to Mr Maher.

R Adams, Phd, is Senior Lecturer, School of Physiotherapy, Faculty of Health Sciences, The University Sydney.

This study was approved by The University of Sydney Human Ethics Committee.

This article was submitted August 15, 1994, and was accepted May 8, 1995.