Pain, 27 (1986) 117-126 117

Elsevier

PA1 00954

The Measurement of Clinical Pain Intensity:

a Comparison of Six Methods

Mark P. Jensen, Paul Karoly and Sanford Braver

Department of Psychology, Arizona State University, Tempe, AZ 85287 (U.S.A.)

(Received 24 October 1985, revised received and accepted 12 March 1986)

Summary

The measurement of subjective pain intensity continues to be important to both

researchers and clinicians. Although several scales are currently used to assess the
intensity construct, it remains unclear which of these provides the most precise,
replicable, and predictively valid measure. Five criteria for judging intensity scales
have been considered in previous research: (a) ease of administration of scoring; (b)
relative rates of incorrect responding; (c) sensitivity as defined by the number of
available response categories; (d) sensitivity as defined by statistical power; and (e)
the magnitude of the relationship between each scale and a linear combination of
pain intensity indices. In order to judge commonly used pain intensity measures, 75
chronic pain patients were asked to rate 4 kinds of pain (present, least, most, and
average) using 6 scales. The utility and validity of the scales was judged using the
criteria listed above. The results indicate that, for the present sample, the scales
yield similar results in terms of the number of subjects who respond correctly to
them and their predictive validity. However, when considering the remaining 3
criteria, the lOl-point numerical rating scale appears to be the most practical index.

Introduction

Current conceptualizations of clinical diagnosis argue for the importance of

simultaneously assessing the varied components and effects of pain [3,4,9,13,22,23].
Of the components that should be assessed, subjective intensity is probably the one
most often measured in both clinical work and in treatment outcome research.

To whom correspondence and requests for reprints should be addressed.

0304-3959/86/$03.50 0 1986 Elsevier Science Publishers B.V. (Biomedical Division)

118

However, there are several scales now in common use to assess this construct, and it
is unclear which, if any, of these is the most efficacious.
Six examples of various pain intensity measures are illustrated in Fig. 1. These
include: (a) the Visual Analogue Scale (VAS); (b) a IOl-point Numerical Rating
Scale (NRS-101); (c) an 11-point Box Scale (BS-11); (d) a &point Behavioral Rating
Scale (BRS-6); (e) a 4-point Verbal Rating Scale (VRS-4); and (f) a 5-point Verbal
Rating Scale (VRS-5).
The Visual Analogue Scale (VAS) consists of a 10 cm line anchored by 2
extremes of pain. In the VAS illustrated in Fig. 1, the extremes are no pain and
pain as bad it could be. Patients are asked to make a mark on the line which
represents their level of perceived pain intensity, and the scale is scored by
measuring the distance from the no pain end to the patients mark.
The NRS-101 consists of asking the pain patient to rate his or her perceived level
of pain intensity on a numerical scale from 0 to 100, with the 0 representing one
extreme (e.g., no pain), and the 100 representing the other extreme (e.g.. pain as
bad as it could be). The number stated by the patient as representing his or her
level of pain intensity is the basic datum for the NRS-101.
The BS-11 consists of 11 numbers (0 through 10) surrounded by boxes. The
patient is told that the 0 represents one extreme of pain and the 10 represents the
other, and is asked to place an X through the number representing his or her pain
level. The BS-11 illustrated in Fig. 1 is adapted from that used in a study by Downie
et al. [6].
The BRS-6 illustrated in Fig. 1 is an adaptation from a scale developed by
Budzynski et al. [2] to measure the intensity of head pain and used in another pain
measurement comparison study [15]. We refer to it as the Behavioral Rating Scale
because patients are asked to rate the intensity of their pain in terms of its
behavioral effects. Each set of descriptive words on the BRS is given a score from 0
(for the description indicating no pain) to 5 (for the description indicating incapaci-
tating pain), and the patients intensity score equals the score associated with the
chosen description.
Verbal Rating Scales (VRSs) consist of a list of adjectives which describe
different levels of pain. The least intense descriptor is usually given a score of 0, the
next a score of 1, and so on until each adjective has a number score associated with
it. The patients intensity score is that number associated with the word he or she
chooses as most descriptive of his or her pain level. The 4-point VRS illustrated in
Fig. 1 is commonly used in scale comparison studies [e.g., 6,11,20], and the 5-point
VRS ilfustrated in Fig. 1 is commonly used in treatment outcome studies, since it is
a part of the popular McGii1 Pain Questionnaire 1171.
There are 5 major criteria on which the various pain intensity scales have been
judged: (a) the ease of administration and scoring; (b) rates of correct responding;
(c) the relative sensitivity of the scales as defined by the number of response
categories they provide; (d) the relative sensitivity of the scales as defined by their
ability to detect treatment effects; and (e) the magnitude of the relationship between
each scale and a best possible combined measure of subjective pain intensity.
Regarding the criterion of ease of administration and scoring, from the descrip-
 119
The Visual Analogue Scale (VAS).

No pain Pain as bad as it could be

The IOI-point Numerical Rating Scale (NRS-101).

Please indicate on the line below the number between 0 and 100
that best describes your pain. A zero (0) would mean no pain.
and a one hundred (100) would mean pain as bad as it could be.
Please write only one number.

The II-point Box Scale (KS-II).

If a zero (0) means no pain. and a ten (IO) means pain as bad as it
could be. on this scale of 0 to IO. what is your level of pain? Put an
x through that number.

The 6-ooinr Behavioral Rating Scale (BRS-6).

( ) No pain.
( ) Pain present. but can easily be ignored.
( ! Pain present. cannot be ignored. but does not
interfere with everyday activities
Pain present. cannot be ignored. interferes with concentration.
I i Pain present.
tasks except
cannot be ignored. interferes with all
raking care of basic needs such as
toileting and eating.
( ) Pain present. cannot be ignored. resr or bedresr
required.

The 4-point Verbal Rating Scale (VRS-4).

( ) No pain.
( ) Some pain.
( ) Considerable pain.
( ) Pain which could nor be more severe.

The 5-poinr Verbal Rating Scale (VRS-5).

( ) Mild.
( ) Discomforting
( ) Distressing.
( ) Horrible.
( ) Lxcruciaring.

Fig. 1. Six pain intensity measures.

120

tion of the 6 scales given above, the NRS, VRSs, and BRS would appear to be
extremely simple to administer and score. These scales can be given either in written
or in verbal form. Both the BS and the VAS, on the other hand, must be given in
written form. Two additional drawbacks are associated with the VAS. First, its use
involves 2 steps, the estimate of pain by the patient and the clinicians measurement
of the patients line. The second step involved in scoring the VAS can add a source
of error to the scale score [8]. The second problem with the VAS is that its users
must be careful not to photocopy it, since this may change the length of the 10 cm
line, making the comparison between distances measured on the original and the
photocopied scale more difficult [8,10]. Thus, the most practical instruments appear
to be the NRS, BRS, or VRSs.
Only 5 experiments have examined the relative rates of correct responding across
the scales. The results of these studies indicate that: (a) response problems can occur
with all scales when subjects are asked to complete them in unsupervised settings
[cf.. 1, 111; (b) more response problems may occur if the measurement task involves
more effort on the part of the pain patient (e.g, the employment of cross-modality
matching of the descriptors) [cf., 11; and (c) no scale is consistently associated with
more response problems than any other scale [1,11,14,15,20], although older patients
may have more difficulty using the VAS as compared to younger patients [14].
A third criterion by which pain intensity measures have been judged is their
relative sensitivity, as defined by the number of response categories offered by each
scale. In discussing sensitivity in this way, the conflict has usually been between
VASs (which provide an infinite number of response categories between two
extremes of pain) and VRSs, which usually provide only 4 or 5 response categories.
Because VRSs provide fewer response categories, they are usually assumed to be less
sensitive than VASs [11,18,19,21]. However, even though scales with more response
categories have the potential to be more sensitive, such scales are not necesssari(lx
more sensitive or valid by virtue of the number of response categories alone. For
example, requiring patients to rate their pain on a scale of 0-1.000,000,000 is not
necessarily a psychometrically superior approach as compared to asking them to
rate their pain intensity on a scale of O-100, despite the fact that the former NRS
offers greater room for variability of response categories.
A fourth criterion that has been used to judge pain intensity measures is their
relative ability to detect treatment effects. Unfortunately, only 3 studies have
examined the sensitivity of pain scales defined in this way, and these studies have
produced inconsistent results [11,18,20]. Although the paucity of the experiments
which analyze scale sensitivity to treatment effects makes definitive conclusions
difficult to draw. it appears that the different scales may be more sensitive under
different conditions, and that no scale consistently demonstrates more statistical
power than another.
Even if the results were consistent, however, there is still a major problem
inherent in using statistical power as a criterion for evaluating construct validity of a
scale. While it is clear that a scale which never detects a treatment effect (i.e., is
absolutely insensitive), should be considered invalid as a measure of pain intensity.
in general, greater sensitivity, as defined by a scales ability to detect treatment effects,
 121

is not necessarily associated with greater co~st~ct ~aii~it~. This fact is illustrated best
in Ohnhaus and Adlers [lg] discussion of their results. Despite the greater statistical
power of the VRS employed in their study, they suggested that this greater
sensitivity was due to an artificial augmentation of the drug effect (p. 383), and so
concluded that the VAS had greater validity. Precisely because these studies do not
tell us the degree to which the different scales measure true subjective pain
intensity, the results of statistical power comparisons are not sufficient in themselves
for evaluating the validity of an intensity measure.
A fifth method which may be used to judge the merits of pain intensity measures
is to compare the relationships between each intensity scale and a derived composite
which represents the best possible assessment of the construct. Those measures
which show the closest association to the composite measure could be judged to be
the most valid. This method is particularly useful when assessing the validity of
self-report constructs (such as subjective pain intensity), since it can produce a
criterion measure which is more accurate than any single measure alone [cf., 121.
The potential accuracy of composite measures made from subjective ratings has
been demonstrated by Dawes [5], who used factor analysis to combine 5 subjective
ratings of height (using 2 VRSs and 3 VASs) into a single composite measure.
Because the composite measure he created was so accurate, correlating 0.98 with
actual height, Dawes concluded that rating scales could be used as a reasonable
substitute for rulers as measures of height.
Downie et al. 161 used factor analysis to judge the relative validity of 4 pain
intensity measures (a vertical VAS, a horizontal VAS, a vertical BS-11, and an
NRS-4) in a group of patients suffering from rheumatic diseases. Their rest&s
provided strong support for the conclusion that the 4 scales did indeed measure a
single underlying construct (as shown by the large eigenvalue of the first factor and
the small eigenvalues of the remaining factors), and that each of the scales is a valid
measure of the construct of subjective pain intensity in patients suffering from
rheumatic diseases.
In sum, 5 criteria have been used to judge pain intensity measures. While every
pain intensity measure should meet each of these criteria at some level, 3 criteria are
especially important. The criterion of ease of administration and scoring is im-
portant because scale users, such as clinicians and researchers, are often short of
time and resources. Similarly, the relative rate of correct responding is important to
consider, since patients must be able and willing to use a scale if it is to be useful.
Finally, a strong relationship between a scale and a best possible composite
measure is important, since this would provide evidence for the construct validity of
the scale. The criterion of statistical power should be considered less important
when comparing scales (providing each scale has demonstrated some minimal
degree of statistical power), since the measure that is most sensitive to treatment
effects is not necessarily the single most valid instrument. Similarly, while some
minimum number of response categories is clearly necessary to provide a valid
measure of pain intensity, there is likely an upper limit of response categories
beyond which the scales may become, at best, redundant, and at worst, confusing
for the pain patient.
122

The purpose of the present study is to examine the usefulness of 6 different pain
intensity measures in a group of chronic pain patients, with respect to the criteria
discussed above. In particular, the experiment focuses on two of the criteria deemed
to be the most important: (a) the relative rate of correct responding: and (b) the
strength of the relationship between each scale score and the first factor which
emerges when a number of pain intensity measures are factor analyzed.

Method

Subjects
The subjects for this study consisted of 75 chronic pain patients (i.e., had pain for
more than 6 months duration) who were consecutively admitted to the St. Josephs
pain unit, a lo-bed inpatient program located in a large metropolitan hospital
designed to teach patients to control and cope with their pain. Of these subjects, 31
were male (mean age = 49.0; S.D. = 12.6) and 44 were female (mean age = 50.3;
SD. = 15.2).

Procedure
Upon admission to the pain program, all subjects were asked to rate 4 aspects of
their pain: (a) current pain; (b) the most pain experienced during the past week; (cl
the least pain experienced during the past week; and (d) the average amount of pain
experienced during the past week, using each of the 6 pain intensity measures
illustrated in Fig. 1. The scales were administered in a random order using a Latin
Square design.
The responses to the intensity measures were examined to determine the degree
to which the patients responded incorrectly to them. Correctness of each response
was noted for each scale, with correct responding defined as provision of a response
within the stated scale guidelines. An incorrect response was noted for a particular
scale when the response to that scale: (a) was non-existent (i.e., left blank); (b) fell
between 2 appropriate responses (i.e., placing an X between 2 words of a VRS,
using fractions of numbers on the NRS-101); (c) extended beyond the numbers
allowed in the NRS-101; (d) included 2 answers on the same scale; or (e) indicated a
range of responses rather than providing a single numerical estimate.

Results

The rates of incorrect responding to each scale are presented in Table I. As can
be seen, some incorrect responding occurred with each scale, with the rate ranging
from 2.7% for the VRS-4 to 8.0% for the BRS-6. A chi-square test for correlated
dichotomous data 124, pp. 303-3051 indicated that the difference in the rates was
not significant (x2 (5, N = 75) = 3.33, P > 0.05).
The correlation between subject age and the degree of incorrect responding was
computed for each pain scale. The results of the correlations between age and
 123

TABLE I
NUMBER AND PERCENT OF SUBJECTS WHO GAVE INCORRECT RESPONSES TO THE 6
PAIN INTENSITY MEASURES

Scale Number of subjects Percent of subjects

who gave who gave
incorrect responses incorrect responses
VAS 4 5.3
NRS-101 4 5.3
BS-11 5 6.7
BRS-6 6 8.0
VRS-4 2 2.7
VRS-5 3 4.0

response rates indicated that only incorrect responding to the VAS was related to
age (r = 0.31, n = 72, P < 0.01, two-tailed test), with increased age associated with a
greater frequency of incorrect responding. All the other correlations were non-sig-
nificant (in order, the correlation coefficients between age and incorrect responding
with the NRS-101, BS-11, BRS-6, VRS4, and VRSJ were - 0.01,0.06,0.02, - 0.10,
and 0.15, ail ns = 72 and P's> 0.05, two-tailed tests).
The strength of the relationship between each individual scale and the shared
variance of all of the scales was determined by inter~rrelating responses to the 6
measures (separately for each of the 4 kinds of pain), and perfor~ng a series of
principal axis factor analyses on these correlations. The principal axis extraction
procedure was chosen because, according to Mulaik [16, p. 1111, this method creates
the most representative score found in a group of scores and provides estimates of
the correlation between each individual score and this most representative esti-
mate. Subjects were excluded from these analyses if they responded incorrectly to
any one of the 6 measures. For current pain, 63 subjects responded correctly to all 6
measures, and the intercorrelation coefficients among the scales ranged from 0.65 to
0.88, with a median r of 0.74. For most pain, the coefficients ranged from 0.42 to
0.81, with a median of 0.56 (n = 64). The intercorrelation coefficients of the 6
measures of least pain ranged from 0.56 to 0.89, with a median of 0.70 (n = 67).
Finally, the coefficients for average pain ranged from 0.47 to 0.86 (median r = 0.61,
n = 64). All coefficients were signific~t (P < 0.001, two-tailed tests), suggesting a
large amount of shared variance among the intensity measures.
The results of the series of factor analyses on the 6 pain intensity measures for
each aspect of pain are presented in Table II. In each of these analyses, a single
factor emerged, as is evident by the large first eigenvalues of each matrix (4.72, 3.83,
4.51, and 4.13 for present pain, most pain, least pain, and average pain respectively)
and the smaller second eigenvalues (0.51, 0.75,0.47, and 0.64 for present pain, most
pain, least pain, and average pain respectively). These factor solutions were not
rotated because two factors must emerge for a rotation to be possible. While some
degree of consistency in the resuhs of the factor analyses can be seen (e.g., the
BRS-6 consistently had one of the two lowest loadings, and the BS-11 and NRS-101
124

TABLE II

RESULTS OF THE PRINCIPAL AXIS ANALYSES ON PAIN INTENSITY MEASURES ESTIMAT-

ING CURRENT PAIN (u = 63), MOST PAIN (n = 64) LEAST PAIN (n = 67). AND AVERAGE
PAIN (n = 64)

Factor(s) Eigen- Loading of each scale on the first factor

value(s)
VAS NRS-101 BS-11 BRS-6 VRS-4 VRS-5
Current pain
First 4.72 0.89 0.91 0.90 0.80 0.81 0.85
Second-sixth 0.51-0.10

Mosr pain
First 3.83 0.70 0.82 0.84 0.64 0.77 0.75
Second-sixth 0.75-0.18

Least pain
First 4.51 0.85 0.91 0.92 0.74 0.78 0.81
Second-sixth 0.47-0.0s

Auerage pain
First 4.13 0.79 0.84 0.95 0.69 0.68 0.81
Second-sixth 0.64-0.11

Average loading 0.81 0.87 0.90 0.72 0.76 0.81

consistently had the highest loadings), overall, each of the scales correlated substan-
tially (0.64 or above) with the first unrotated factors across the diverse measures of
pain. As can be seen in Table II, the average loadings of the intensity measures to
the first factors were all high, and showed little variation, ranging from 0.72 for the
BRS-6 to 0.90 for the BS-11.

Discussion

The results of the present study indicate that the 6 pain intensity measures are
more similar than they are different in terms of the rates of incorrect responding
and in terms of construct validity.
While some range in the incorrect response rates between the scales was noted,
the obtained differences were not significant, indicating that the subjects in this
sample of chronic pain patients generally did not show more incorrect responses to
one scale than to any other. These results are consistent with those researchers who
found no differential rates of correct responding for scales administered under
somewhat controlled conditions and on one occasion [15,20]. However, they differ
from those of several other studies [1,11,14]. Perhaps differential rates would have
emerged in the present analysis if the patients in our study had been asked to record
their pain intensity on the scales across many occasions, as in the Joyce et al. [ll]
and Ahles et al. [lf experiments. Perhaps differences in correct responding would
 125

have emerged if the patients had been given more supe~ision as they completed the
intensity measures. In this study, the measures were not examined until all the data
were collected, so that patients who did not accurately complete the measures may
have done so either because they lacked motivation, or because they could not
understand the instructions. Had all patients been asked to complete each scale that
had been incorrectly responded to, incorrect responding might have been limited to
just those occasions when patients could not comprehend the measurement task.
Encouragement and supervision may have thus produced a decrease in incorrect
responding for all except perhaps the VAS, since problems with this scale appear to
be associated with difficulty in understanding. This difficulty occurs especially
among older pain patients, as shown by the significant correlation between age and
incorrect responding to the VAS in the present study, as well as in the results
reported by Kremer et al. 1141.
The series of factor analyses indicated that each of the scales may be considered a
useful measure of subjective pain intensity. The scales demonstrated a large degree
of association, and each scale correlated substantially with the first factors which
emerged from the factor analytic procedures. The large correlations found between
the measures employed in this study is likely due to the fact that the 6 pain intensity
measures are very similar to one another (consisting of either magnitude estimation
or category scaling measures). Less association between scales may have occurred,
and possibly more than one factor may have emerged from the factor analyses, if
other scales such as the McGill Pain Questionnaire [17] or Gracelys verbal
descriptor scales [cf., 73 had been employed.
If, as the present results suggest, the 6 scales have similar rates of correct
responding and a similar degree of utility as measures of subjective pain intensity in
chronic pain patients, which measure should researchers and clinicians employ when
assessing the intensity construct? The first, and most obvious answer to this
question is that the results support the use of any measure in an inpatient chronic
pain population. However, if the other 3 adequacy criteria are taken into account,
the superior measure appears to be the NRS-101.
The NRS-101 has several practical advantages over the other measures. First, it is
extremely simple to administer and score, and can be administered either in written
or verbal form. Thus, it has an advantage over the VAS (which can only be assessed
in written form, takes two steps to score, and may be rendered invalid if it is
photocopied without care), and the BS-11 (which also can only be assessed in
written form). In addition, the NRS-101 has 101 response categories, and thus is
more likely to be accepted by clinicians and researchers who are concerned with the
limited response options of the BS-11, VRS4, VRS-5, and BRS-6. The average
factor loading of the NRS-101 (0.87) underscores its utility as a measure of pain
intensity. The NRS-101 does not appear to be associated with incorrect responding
more than any other scale. Finally, unlike the VAS, difficulty with the scale does not
appear to be associated with age. In sum, to the degree that a standard measure of
pain intensity is needed to facilitate comparisons of treatment outcome, and to
index chronic patients pain intensity levels at different times in their lives, it
appears that the NRS-101 would be a wise choice.
126

References

1 Ahles, T.A., Ruckdeschei, J.C. and Blanchard, E.B., Cancer-related pain. Ii. Assessment with visual
analogue scales. J. psychosom. Res.. 28 (1984) 121-124.
2 Budzynski, T.H.. Stoyva, J.M., Adler. C.M. and Mullaney, D.J., EMG biofeedback and tension
headache: a controlled outcome study. Psychosom. Med., 35 (1973) 484-496.
3 Catchlove, R.F.H. and Ramsay, R.A., A protocol for the medico-psychosocial evaluation of patients
with chronic pain, Canad. anaesth. Sot. J.. 30 (1983) 531-535.
4 Chapman, C.R., Casey. K.L.. Dubner, R., Foley, K.M.. Gracely, R.H. and Reading, A.E., Pain
measurement: an overview. Pain, 22 (1985) l-31.
5 Dawes: R.M., Suppose we measured height with rating scales instead of rulers. Appl. Psychol. Meas.,
1 (1977) 267-273.
6 Downie, W.W., Leatham. P.A.. Rhind. V.M.. Wright. V.. Branco. J.A. and Anderson. J.A., Studies
with pain rating scales. Ann. rheum. Dis.. 37 (1978) 378-381.
7 Gracely, R.H.. McGrath. P. and Dubner, R.. Ratio scales of sensory and affective verbal pain
descriptors. Pain. 5 (1978) 5-18.
8 Grossi. E.. Borghi, C.. Cerchiari. E.L., Puppa, T.D. and Francucci, B., Analogue chromatic continu-
ous scale (ACCS): a new method for pain assessment, Clin. exp. Rheum.. 1 (1983) 337-340.
9 Heaton, R.K.. Getto, C.J., Lehman, R.A.W., Fordyce, W.E., Brauer. E. and Groban, S.E.. A
standardized evaluation of psychosocial factors in chronic pain, Pain, 12 (1982) 165- 174.
10 Huskisson. E.C.. Measurement of pain. J. Rheumatol., 9 (1982) 768-769.
I1 Joyce, C.R.B.. Zutshi, D.W., Hrubes, V. and Mason, R.M.. Comparison of fixed interval and visual
analogue scales for rating chronic pain, Europ. J. clin. Pharmacol., 8 (1975) 415-420.
12 Judd, C.M. and Kenny, D.A.. Estimating the Effects of Social Interventions, Cambridge University
Press. Cambridge, 1981.
Ii Karoly. P., The assessment of pain: concepts and procedures. In: P. Karoly (Ed.), Measurement
Strategies in Health Psychology. Witey. New York, 1985. pp. 461-515.
14 Kremer, E., Atkinson, J.H. and Ignelzi. RJ., Measurement of pain: patient preference does not
confound pain measurement, Pain, 10 (1981) 241-248.
15 Linton, S.J. and Gotestam, K.G., A clinical comparison of two pain scales: correlation, remembering
chronic pain. and a measure of compliance, Pain, 17 (1983) 57-65.
16 Mulaik, S.A., The Foundations of Factor Analysis. McGraw-Hill. New York, 1972.
17 Melzack. R.. The McGill Pain Questionnaire: major properties and scoring methods. Pain, I (1975)
277-299.
18 Ohnhaus, E.E. and Adler. R.. Methodological problems in the measurement of pain: a comparison
between the verbal rating scale and the visual analogue scale, Pain. 1 (1975) 3799384.
19 Scott, J. and Huskisson, E.C., Graphic representation of pain, Pain, 2 (1976) 1755184.
20 Seymour, R.A., The use of pain scales in assessing the efficacy of analgesics in post-operative dentat
pain, Europ. J. clin. Pharmacol.. 23 (1982) 441-444.
21 Sriwatanakul, K.. Kelvie. W.. Lasagna, L., Calimlim, J.F.. Weis. 0.F. and Mehta. G.. Studies with
different types cif visual analog scales for measurement of pain. Clin. Pharmacol. Ther.. 34 (1983)
234-239.
22 Turk, D.C. and Kerns, R.D.. Conceptual issues in the assessment of clinical pain. Int. J. Psychiat.
Med., 13 (1983) 57-68.
23 Turk, D.C., Meichenbaum, D. and Genest, M.. Pain and Behavioral Medicine: a Cognitive-Behavioral
Perspective, Guilford Press, New York, 1983.
24 Winer, B.J.. Statistical Principles in Experimental Design, McGraw-Hill. New York, 1971.