Therapists' conceptualization and characterization of the clinical concept of spinal stiffness.

A number of research teams have acknowledge that the current protocol for the post-teroanterior (PA) pressure test does not enable physical therapists to reliably judge spinal stiffness[1-4] and have initiated studies in an effort to solve the problem. Research endeavors have both sought to develop instruments that may eventually replace manual assessment and attempted to identify the factors responsible for the disagreement between different raters' manual assessment results.

Instruments to measure PA stiffness have been developed by Lee and Svensson[5] and Latimer and colleagues.[6] Although these instruments are not yet suitable for routine clinical use, they have allowed the quantification of the effect of numerous variables on spinal PA stiffness (See Lee et al[7] for a review). Mechanical spinal models have been developed by Simmonds and colleagues[8] and have been used to study the PA loading strategies used by therapists. Maher and Adams[9-11] have studied the factors that affect the perception of stiffness. In addition, treatment plinths have been linked to force transducers and used to study the force-time characteristics of PA pressure testing.[12-14]

Overall, this work constitutes the background for attempts to enchance the reliability of measurement of PA pressure through standardization of the test protocol. For example, it seems logical that test reliability would be enhanced if factors that have been shown to affect the measured PA stiffness of the spine or to affect the perception of stiffness were controlled in a revised protocol. There is, however, one problem with this approach. All of the above work is based on unstated assumptions that (1) the clinical concept of PA stiffness is equivalent (ie, worldwide) agreement in therapists' understanding and communications of the stiffness construct. Neither of these assumptions may be true.

In 1995, Maher and Adams[9] hypothesized that the clinical concept of spinal stiffness may be multidimensional. To investigate this hypothesis, Maher, in an unpublished dissertation project,[15] conducted an audit of the writings of manipulative physical therapy authors and recorded the phrases and words that were used to described the results of the PA pressure test. This search located a large number of descriptors, most of which and no definition associated with their usage in texts and articles, so the actual meaning of the words was not clear. To facilitate the study of these descriptors, Maher grouped them according to their context of use, and this grouping is presented in the Appendix. Some words were used in two contexts and consequently appear in two groups.

Although some of the descriptors listed in the Appendix are probably synonyms (eg, "free-running" and "friction-free") and thus could be replaced with a single descriptors, other terms appear to be describing quite different properties (eg, "loose" and "boggy"). Rather disagreement is particularly likely with descriptors that are ambiguous (eg, "thick") and with descriptors that are similar in meaning but not exact synonyms (eg, "boggy," "spongy," and "squashy"). How each of the descriptors that have been used to described the feel of the spine (either through range of motion or at end of range) relates to the terms used by therapists to described their overall impression of the spine remains unclear. For example, it is not clear what individual descriptors of feel or pattern of feel must be present for a therapist to label the pine that they are examining as "disadvantaged."

Many of the words and phrases apparently have been borrowed either from other disciplines or from common usage. Maitland[16] and Jull and colleagues[17] have proposed that terms used by biomechanists to describe the response of soft tissues to mechanical loading are appropriate for physical therapists to use to describe what they feel during manual examination. Terms such as "toe," "compliance," "stiffness", and "friction", therefore, have entered the manipulative physical therapy literature. These authors also have advocate the use of words in common usage such as "thicker," "boggy" and "squashy" to describe aspects of what is felt during manual examination. However, it has yet to be established that therapists have the sensory acuity to feel, identify, and discriminate between such characteristics as boggy and squashy. Most importantly, almost none of these terms have widely accepted definitions, and the terms are used without accompanying definitions, resulting in ambiguity that may obfuscate understanding and theory in manual therapy.

An example of such ambiguity is Jull and Bullock's use of the world "motion" to describe the "combined measure of both displacement and resistance to that displacement."[18](p71) This use of the world is inconsistent with its use in biomechanics and its meaning in common usage as reflected in standard dictionaries (eg, The Australian Oxford Dictionary[19]). Jull and Bullock seem to be using the world "motion" as a synonym for stiffness. Similarly, Maitland[16] uses the words "resistance" and "stiffness" interchangeably, whereas these two terms have quite different meanings in biomechanics. Furthermore, it is not clear whether what Grieve[20] refers to as "elastic resistance" is the same as what he refers to as "elastic resilience."

Ambiguity is also possible where words in common usage are used to describe test results. Although Jull et al[17] describe test results in terms of "thicker tissue stiffness," "reactive tissue stiffness," and "normal tissue stiffness," it is not clear exactly what each term is intended to refer to or how the terms are related. Similar problems apply to Maitland's description of variations of resistance such as "elastic" resistance, "physical" resistance, "spasm-free" resistance, and resistance "due to crepitus."[16] Although it is plain that Grieve[20] considers the presence of a boggy, squashy, or soggy feel to be an indicator of pathology, he does not make clear what are the referents for each type of feel and how this is different from the feel of a normal spine. Thus, although these authors[16,17,20] suggest that nature of each dimension is unclear from their writings because of the lack of definitions. These terms are used to describe the feel of the spine when a physical therapist applies PA pressure, an action that cannot be judged by two therapists simultaneously. There are other factors that are routinely assessed by physical therapists, such as spinal active range of motion, that can be judged simultaneously by two independent observes, so confusion need not arise.

This semantic confusion may be seen as an understandable attribute of an initial, descriptive scientific phase for manual therapy, but a more clearly defined, parsimonious set of descriptors is required if therapists are to be able to meaningfully communicate the results of PA stiffness testing. To clarify the number of dimensions or underlying characteristics of the concept of "PA stiffness," some form of classification that is generalizable is needed. If this classification is to reflect the views of practicing physical therapists, we believe that it needs to be done in a manner that is not restricted by investigators' preconceptions about the number and types of dimensions contained in the construct "PA stiffness." The potential for investigator bias can be avoided with cluster analysis and multidimensional scaling, techniques that allow the participating therapists to determine the dimensions of PA stiffness simply by making similarity judgments of PA pressure descriptors. This process allows the participating therapists to generate the types and number of dimensions underlying their concept of PA stiffness; thus, the dimensions are discovered by the experimenters, not imposed by them.[21]

This study was confined to the investigation of published terms used to describe the feel of the spine through the range of motion. The primary purpose of our study was to determine whether the clinical construct of PA stiffness is a multidimensional concept. A second purpose was to determine whether this construct is conceptualized and characterized in a similar manner internationally. The two hypotheses investigated in the study were:

1. The clinical concept of PA stiffness is multidimensional.

2. The clinical concept of PA stiffness is not the same for all physical therapists.

Method

Overview

The relationships among the PA pressure descriptors were examined using cluster analysis. Cluster analysis can be used to explore the relationships among a set of multidimensional stimuli, which in our study were PA pressure descriptors. Cluster analysis is used to form groups or clusters of similar descriptors[22,23] with the intent of describing the inherent structure of the data.[24] The clustering procedure starts with as many clusters as there are descriptors and then proceeds in a stepwise fashion, joining individual descriptors or lower-order clusters into larger clusters of similar descriptors. As with the factors identified in factor analysis, the clusters identified in cluster analysis can be evaluated and named by the investigation, but the names applied are those of the experimenter. The procedure can be used for reducing data, understanding complex relationships among variables, and developing classification systems.[25]

Cluster analysis is used to minimize experimenter bias because the participating therapists determine the attributes underlying the set of descriptors through their similarity judgments.[21] This would not be the case if the experimenters asked the therapists to rate the PA pressure descriptors in terms of normality or a mechanical characteristic such as stiffness. In these cases, the dimensions underlying the PA pressure descriptors are imposed by the experimenters rather than being generated by the therapists from the similarity of their judgment.[21] Some influence of the experimenters persists, even within similarity judgment, because the experimenters select the PA pressure descriptors to be judged. To minimize this effect, the descriptors used in our study were obtained from an extensive search of the published manual therapy literature, including MEDLINE and CINAHL searches for the years 1966 through 1996 to locate relevant journal articles.

Cluster analysis requires a matrix of dissimilarities to describe how different each pair of descriptors are, and this matrix has typically been obtained through the use of rating, sorting, and perceptual confusion tasks.[26] These methods are, however, somewhat inefficient, requiring each therapist to make many judgments. For example with 31 PA pressure descriptors, each therapist would have to make 465 pair-wise dissimilarity ratings. Pilot testing revealed that this method was impractical, with therapists needing 3 to 4 to compete the task.

A more time-efficient method is the spatial arrangement method, which requires therapists to arrange items on a surface (eg, a computer screen or white board) so that their proximity is proportional to their similarity. Goldstone[26] has shown that the spatial arrangement method provides data that are highly correlated to data obtained with the more traditional pair-wise rating tasks but obtained much more quickly. In other study, the participating therapists typically completed the spatial arrangement task in 20 to 30 minutes.

Subjects

Data were collected from 29 physical therapists in Sydney, New South Wales, Australia, and from 23 physical therapists in Houston, Tex. All of the Australian therapists were members of the Manipulative Physiotherapists' Association of Australia. The American therapists were a heterogeneous group. All therapists in the US group were licensed physical therapists who had completed postgraduate continuing education courses. Formal credentialing, however, was limited. Two therapists in the US group were certified orthopedic specialists, 1 therapist was a member of the International Federation of Orthopaedic Manipulative Therapists, and 5 therapists taught physical therapy techniques to physical therapist students within the university. Both groups were familiar with the PA pressure concept and used manipulative treatment techniques to treat patients with back pain. Characteristics of both groups of therapists are given in Table 1. All therapists gave informed consent prior to participation in the study.

[TABULAR DATA 1 NOT REPRODUCIBLE IN ASCII]

Apparatus

Thirty-one PA pressure descriptors (see column 1 of the Appendix) were printed onto white magnetized rubber strips, a material commonly used to make refrigerator magnets for advertising purposes. Each magnet was 70 mm long, 14 mm wide, and 1 mm thick. The words were printed in black ink (Arial typeface, size 24 font) in the middle of each magnet. To standardize test stimuli, all the magnets were printed in Australia, and a set was sent to the author (MS) in Texas. The surface used for arrangement of the magnets was a teaching white board measuring 1,000 [mm.sup.2], including a 15-mm aluminum border.

Procedure

The participating therapists could choose to stand or sit during data collection. The white board was placed in front of them so that it was both vertical and at a convenient height. The word magnets were shuffled and placed in a container adjacent to an alphabetical list containing all 31 PA pressure descriptors. The following instruction were then read to the therapists:

The words on the list have been collected from manipulative therapy texts where they were used to describe what physical therapists feel through range as they perform the PA central pressure to a lumbar level. I would like you to arrange the word-magnets on the white board so that the distances between the words are proportional to their dissimilarity in this context. The more dissimilar two words are, the further apart they should be. Words that are similar in meaning should be placed close together. The more similar two words are, the closer they should be. If the words are exactly the same, you may put them on top of each other. When arranging the words, you should only consider their meaning as it applies to what you feel through range as you perform the PA central pressure to a lumbar level. There are no wrong or right answers, as I am interested in finding out how you as an individual physical therapist compare these words. Do you understand?

The therapists were then allowed to place the word-magnets on the board, with an investigator close by in case they required clarification of the procedure. When the therapists had completed the task, they were asked whether they used any other words to describe what they felt through the range of motion as they, performed PA central pressures to a lumbar level. They were also asked to identify words that they would not normality. The experiments would then write down the therapists' verbal response. To assist the therapists in answering question 2, they were provided with an alphabetical list of the 31 words.

Data Analysis

The positions of the magnets were recorded by computing Cartesian coordinates for each magnet. The x coordinate was recorded as the distance (in centimeters) from the left edge of the board to the midpoint of the front edge of the magnet. The y coordinate was the distance (in centimeters) from the bottom edge of the board to the midpoint of the bottom edge of the magnet. Distances were measured using a retractable metallic tape measure in Sydney and a metal T-square in Houston, bottom of which enable measurement to the nearest millimeter.

Cluster analysis was performed using SPSS for Windows, Release 6.0.(*) The data from each test site were analyzed separately in the first instance, with cluster analysis first performed on the data from each group as a whole. Infrequently used words were then excluded, and the analysis was repeated. Finally, to confirm that the group cluster reflected individual therapists' clusters, cluster analyses of individual therapists' data were performed and graphically compared. It is recognized that confirmation analysis would require an independent data set.

The same procedure was followed for each analysis. The group's x and y coordinates were used as variables to describe the 31 PA pressure descriptors and so served as the basis for cluster formation. The SPSS program used the x and y coordinates to determine the squared Euclidean distance between each descriptor pair (ie, the sum of the squared differences of the x and y coordinates for the descriptor pair). The squared Euclidean distances then served as the dissimilarity matrix data for subsequent cluster analysis. Ward's method, which uses the minimum variance within clusters as the basis for cluster membership,[24] was used as the method for combining clusters. This method for combining clusters was chosen over linkage methods because linkage models can produce clusters with large variance or result in ungrouped descriptors with very little in common,[24] two traits that were deemed undesirable when trying to determine which descriptors describe a common element of PA feel.

Results

The cluster analysis procedure can be understood by inspection of the results for one Australian therapist, as illustrated in Figures 1 and 2. Figure 1 is a scattergram illustrating how this therapist arranged the words on the board, and Figure 2 is a dendrogram showing the results of the clustering procedure. Dendrograms identify the descriptors or subclusters being combined in the clustering procedure and the distance between the descriptors or subclusters that are joined. In this case, the actual distances have been rescaled to numbers between 0 and 25. Each descriptor or subcluster is identified by a horizontal line, and the formation of a cluster is identified by a vertical line that joins a group of horizontal lines. The dendrogram shows that at the first level of clustering, six clusters were formed. The first cluster, at the top of the dendrogram, contained five descriptors: "active-recoil," "springy," "elastic," "spongy," and "reactive." The second cluster contained fine descriptors, the third cluster contained seven descriptors, the fourth cluster contained five descriptors, the fifth cluster contained one descriptor, and the sixth cluster contained eight descriptors. The sixth cluster of descriptors ("blocked," "hypomobile," "stiff," "limited," "tight," "firm," "restricted," and "thick"), at the bottom of the dendrogram, makes intuitive sense because these eight words were grouped together on the board. Similarly, each of the other five clusters that were identified were composed of descriptors that were grouped together on the board.

[FIGURES 1 & 2 ILLUSTRATION OMITTED]

The dendrogram for the data of the American therapists is shown in Figure 3, and the dendrogram of the data of the Australian therapists is shown in Figure 4. Both dendrograms suggested an eight-cluster solution at the first level of clustering. The proximities of these clusters easily reduced the number of clusters to three and then to two superclusters, with the data from each test site being remarkably similar. The three-supercluster solution was selected as optimal because of ease of interpretation. To assess the stability of the obtained solution, the data matrix was split in half and the cluster analysis was repeated on each half. Because the same clusters emerged, this process confirmed the stability of the solution. Another feature that can be understood from the dendrograms is that the distances between many of the descriptors is very small, evidence that many of the descriptors are synonyms.

[FIGURES 3 & 4 ILLUSTRATION OMITTED]

The results of the cluster analysis are shown in Tables 2 (US therapists) and 3 (Australian therapists), along with our interpretation of each cluster. At the first level of clustering, there appeared to be eight readily identifiable subclusters, which seemed to cluster on the basis of range of motion (high or low), resistance to motion, and resilience. With the data of the US therapists, we labeled six subclusters: "reduced mobility," "high resistance to movement," "increased mobility," "low resistance to movement," "low resilience," and "high resilience." We were not able to satisfactorily name the final two clusters. A similar result was found with the data of the Australian therapists, except there were two clusters for both high and low resilience. The subclusters merged to form form, three, and then two superclusters. We adopted the three-cluster solution because of ease of interpretation and its international usage. We labeled these superclusters "limited mobility," "increased mobility," and "viscoelasticity." The reduction from three to two superclusters merged the viscoelasticity and increased mobility clusters and resulted in clusters that describe limited or increased mobility.

[TABULAR DATA 2 & 3 NOT REPRODUCIBLE IN ASCII]

The results from Australia and the United States were remarkably similar. Although some differences in word assignment appeared within the subclusters, these differences essentially disappeared in the superclusters. Only three words were assigned differently by the Australian and US therapists. The terms "boggy," "spongy," and "squashy" were assigned to the viscoelasticity cluster in Australia and to the increased mobility cluster in the United States. Another difference between the data sets was that a more distinct four-cluster stage could be identified in the Australian data. In the Australian data, two characteristics of viscoelasticity -- viscosity and elasticity -- were identified separately. One cluster describes viscosity ("boggy," "squashy," "spongy," and "rubbery."), and the other cluster describes elasticity ("active-recoil," "reactive," "elastic," and "springy,").

Although the group cluster analysis results were very similar between the two sites, there were some members of each group who produced results that were different from the group results. For example, the dendrogram of the Australia therapist shown in Figure 2 differs from) that of the Australian group in a number of ways. At the first level of clustering, six rather than eight subclusters were identified, with this therapist combining the resistance to movements and mobility descriptors, whereas these descriptors were separate in the group results. This therapist also grouped the descriptor "spongy" with "active-recoil," "springy," "elastic," and "reactive," whereas the group as a whole clustered "spongy" with "squashy" and "boggy."

Table 4 slows the new descriptors offered by the therapists. Although an extra 12 words were offered by, the Australian therapists, only 2 words ("spasm" and "normal.") were offered by more than two therapists. The American therapists offered 8 additional words, but only 1 word ("normal") was offered by more than two therapists.

Table 4. New Descriptors Offered by Participating Therapists(a)

US Physical Therapists    Australian Physical Therapists

Crunchy (1)               Cement-like (1)
Crackly (1)               Grotty (1)
Decrease (1)              Hard (1)
Increase (1)              Immobile (1)
Reactive (1)              Jumpy (1)
Guarding (1)              Mobile (1)
Normal (3)                Normal (2)
Spasm (1)                 Resistance (1)
                          Spasm (7)
                          Spasmy (1)
                          Stuck (1)
                          Woody (1)

(a) New descriptors offered by therapists. With number of therapists offering each descriptors in parentheses.

Table 5 shows the words that the therapists would not typically use. These infrequently used words were not assigned a separate dumping cluster. Instead, they were assigned to each of the three superclusters. To ensure that such, words were not confounding the results, the analysis was repeated with the infrequently used words excluded. Six descriptors ("free-running," "well-oiled," "active-recoil," "yielding," "friction-free," and "undue give") were excluded on the basis that they were not normally used by more than half of the Australian therapists; however, the results in terms of the number and types of clustering were essentially unaltered.

Table 5. Number of Therapists Who Did Not Normally Use a Descriptor

                      Frequency
                      Australian
                      Physical         US Physical
Descriptor            Therapists       Therapists

Free-running          24               16
Well-oiled            21               15
Active-recoil         18               13
Yielding              16                5
Friction-free         16                7
Undue give            15               11
Squashy               14               11
Nonspringy            14                8
Soft                  13                1
Noncompliant          13                7
Reactive              13                5
Rubbery               12                5
Resistance-free       10                3
Boggy                  9                6
Inelastic              9                4
Spongy                 8                4
Elastic                8                3
Thick                  8                5
Spasm-free             7                3
Loose                  6                1
Compliant              6                5
Smooth                 3                1
Springy                2                1
Hypermobile            2                0
Firm                   2                0
Hypomobile             2                0
Limited                1                0
stiff                  1                0
Tight                  1                0
Blocked                1                1
Restricted             0                0

Discussion

Clinical judgments of PA stiffness are fundamental to manual therapy examination and treatment. At present, the poor reliability of such clinical judgments renders them questionable at best and totally meaningless at worst. Given the fundamental nature of stiffness judgments, the lack of reliability is of paramount concern. The apparent complexity of the clinical stiffness construct and the manner in which it is conceptualized and characterized may account for part of the problem. To date, to research has addressed these fundamental questions.

The primary purpose of our study was to determine whether the clinical construct of PA stiffness is a multi-dimensional concept. A second purpose was to determine whether this construct is conceptualized and characterized in a similar manner internationally. Because just one supercluster did not emerge with either the American or the Australian cluster data, the data here support the contention that clinicians conceptualize PA stiffness in a multidimensional manner. Moreover, the similarity in results between two international sites suggests that this multidimensional manner of characterizing PA stiffness occurs similarity in both groups of therapists. Within each group, however, there were individuals who conceptualized the construct in a different manner than their peers.

The clustering procedure allowed therapists to group words with similar meanings and to separate words with dissimilar meanings without imposing any limit on the number of clusters or on the distance between them. Eight clusters of words with similar meanings were identified at the first level of clustering. These clusters were formed on the basis of range of mobility, resistance to mobility, resilience, or some combination of these characteristics. These results make sense to us because they suggest that therapists consider a wide range of tissue and joint responses to PA pressures when making their clinical judgments. Herein lies the problem.

Assuming that therapists consider several characteristics of tissue response to PA pressure, what characteristic has the greatest influence on the clinical judgment? After all, the tissue and joint characteristics that contribute to stiffness are described in qualitative terms, and they occupy a continuum. The question remains whether therapists have the sensory acuity to discriminate among all the characteristics of tissue and joint responses (mobility, resilience, resistance) simultaneously and, more importantly, along a continuum of such characteristics.

Although the number and characteristics of the subclusters help to explain the poor reliability of PA pressure judgments, the use of superclusters may provide a reasonable basis for improving the reliability of stiffness judgments. Three superclusters were identified that described each end of the mobility continuum (increased or reduced) and the nature of the resistance felt in response to PA pressure testing (viscous or elastic). These superclusters probably represent fundamental characteristics of the clinical stiffness judgments used by therapists, and their identification allows for a more clearly defined basis for PA stiffness judgments. For example, PA stiffness judgments could be made on the basis of (1) mobility and (2) the nature of the resistance to pressure (ie, viscous versus elastic). Further research, however, is needed to test this notion before it is adopted clinically. It will also be necessary to perform a cluster analysis of the end-feel and overall impression descriptors (Appendix) so that all salient characteristics are considered when developing new rating protocols.

As is apparent from inspection of Figures 1 through 3, the results from Australia and the United States were vein similar. This finding suggests that, at least in two nations with well-established and accredited physical therapy education programs, the clinical concept of PA stiffness is generally understood and communicated in similar fashion. It is possible that this similarity, in results is based on the reading of similar literature. Colloquial expressions of stiffness (eg, "grotty") differed between the two countries but were individual expressions only. It must be acknowledged that the descriptors could have been clustered according to their dictionary, definition rather than the therapists' clinical perceptions of stiffness. Future studies should address the question of why therapist choose to cluster words in a particular manner.

Although the results were similar between the two countries, the Australian data revealed four distinct superclusters at an intermediate level. The Australian therapists characterized viscosity and elasticity separately. The Australian therapists, as a group, had longer formal education manual therapy and used the PA pressure technique more frequently. Whether viscosity and elasticity are reflected in tactile discriminability as well as conceptual discriminability needs to be tested. If such discrimination judgments can be made reliably, and the clinical relevance demonstrated, differentiation between viscosity and elasticity should be emphasized in the PA pressure protocol.

Few additional descriptors were offered by any therapists. This finding suggests that the original list was sufficiently extensive to allow the therapists to represent their concept of PA stiffness, and indeed that the moment is right to reduce, rather than further add to, the existing set of descriptors. The descriptor "normal" was offered by several therapists in Australia and the United States. This finding suggests that therapists have some concept of normal PA stiffness and that they use the descriptor "normal" as a reference base from which to formulate their clinical judgments. This is not unreasonable; however; variability in experiences makes "normal" a somewhat elusive clinical construct. The heterogenous nature of spinal problems and the fact that these problems are characterized by exacerbations and remissions add to the problem. Alternatively, if the descriptor "normal" is simply used to denote a response that is asymptomatic or pain-free, then it may be less problematic.

Because some words were not typically used by all therapists (eg, "free-running"), it is possible that some of the clusters identified could have represented and dumping ground for words that are not normally used. This is unlikely however, because although some descriptors within a cluster were used by only a few of the therapists, other descriptors within the same cluster were used extensively. Moreover, when these infrequently used words were excluded from the analysis, the results did not differ substantially, especially at the supercluster level.

Conclusions

This study, provides evidence that the clinical concept of spinal stiffness is multidimensional, with therapists including characteristics such as mobility and the nature of resistance to movement in the construct. This result suggests that spinal stiffness is unlikely, to be satisfactorily rated on a single scale. The attributes identified in this study provide a basis for the development of protocols to rate spinal stiffness.

References

[1] Matyas T, Bach T. The reliability of selected techniques in clinical arthrometrics. Australian Journal of Physiotherapy 1985;31:175-199.

[2] Maher CG, Adams R. Reliability of pain and stiffness assessment in clinical manual lumber spine examination. Phys Ther. 1994;74: 801-811.

[3] Binkley JM, Stratford PW, Gill C. Interrater reliability of lumbar accessory motion mobility testing. Phys. Ther. 1995;75:786-192.

[4] Phillips D, Twomey L. A comparison of manual diagnosis with a diagnosis established by a uni-level lumbar spinal block procedure. Manual Therapy. 1996;2:82-87.

[5] Lee M, Svensson N. Measurement of stiffness during simulated spinal physiotherapy. Clinical Physical and Physiological Measurement. 1990;11: 201-207.

[6] Latimer J, Goodsell MM, Lee M, et al. Evaluation of a new device for measuring posteroanterior forces in a patient Population, part 1: reliability, testing. Phys Ther. 1996;76:158-165.

[7] Lee M, Steven G, Crosbie J, Higgs J. Towards a theory of lumbar mobilisation: the relationship between applied force and movements of the spine. Manual Therapy 1996;2:67-75.

[8] Simmonds MJ, Kunmar S, Lechelt E. Use of a spinal model to qualify the forces and motion that occur during therapists' tests of spinal motion. Phys Ther 1995;75:212-222.

[9] Maher CG, Adams R. Is the clinical concept of spinal stiffness multidimensional? Phys Ther. 1995;75:854-860.

[10] Maher CG, Adams R. Stiffness judgment are affected by visual occlusion. J Manipulative Physiol Ther. 1996;19:250-256.

[11] Maher CG, Adams R. A comparison of pisiform and thumb grips in stiffness assessment. Phys Ther. 1996;76:41-48.

[12] Viner A Lee M. Direction of manual force applied during assessment of stiffness in the lumbosacral spine. J Manipulative Physiol Ther. 1995;18:441-447

[13] Harms M, Cusick G, Bader D. Measurement of spinal mobilisation forces. Physiotherapy.95;81:599-604.

[14] Petty N, Messenger N. Can the force platform be used to measure the force applied during a PA mobilisation of the lumbar spine? Journal of Manual and Manipulative 1996:4:70-76.

[15] Maher CG. Perception of Stiffness in Manipulative Physiotherapy. Lidcombe, New South Wales, Australia: The University of Sydney 1996. Doctoral dissertation.

[16] Maitland GD. Vertebral Manipulation. 5th ed. London, England: Butterworth & Co (Publishers) Ltd; 1986.

[17] 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.

[18.] 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.

[19] The Australian Oxford Dictionary. West End, Queensland, Australia: Herron Publications; 1992.

[20] Grieve G. Mobilisation of the Spine: Notes on Examination, Assessment, and Clinical Method. 4th ed. Edinburgh, Scotland: Churchill Livingstone; 1984.

[21] Clark WC, Fletcher JD, Janal MN, Carroll JD. Hierarchical clustering pain and emotion descriptors toward a revision of the McGill Pain Questionnaire: In: Bromm B, Desmedt J, eds Pain and the Brain: From Nociception to Cognition. New York,NY: Raven Press; 1995:319-329.

[22] Norussis MJ. SPSS for Windows Professional Statistics Release 6.0 Manual Chicago, III: SPSS Inc; 1993.

[23] Snodgrass JG, Levy-Berger G, Haydon M. Human Experimental Psychology. New York, NY: Oxford University Press; 1985.

[24] Michel Y, Allred C. Studying multidimensional constructs through cluster analysis. West J Nurs Res. 1996;18:102-107.

[25] Raymond R. Applications of multidimensional scaling to research in the health professions. Evaluation and the Health Professions. 1989;12: 379-408.

[26] Goldstone R. An efficient method for obtaining similarity data. Behaviour Research Methods, Instruments, and Computers. 1994;26: 381-386.