Pain, 61 (1995) 251-260 251

© 1995 Elsevier Science B.V. All rights reserved 0304-3959/95/$09.50

PAIN 2704

The descriptor differential scale of pain intensity:

an evaluation of item and scale properties 1

Jason N. Doctor a,c,* Mark A. Slater a,b and J. H a m p t o n Atkinson a,b

a San Diego Veterans Affairs Medical Center and b University of California,
San Diego, CA 92161 (USA), and c University of California, San Diego and San Diego State University Joint Doctoral Program in Clinical
Psychology, San Diego, CA 92161 (USA)
(Received 11 April 1994,revisionreceived3 August 1994, accepted 17 August1994)

Summary Improved methods for pain measurement have both theoretical and clinical importance. This study
evaluated the Descriptor Differential Scale (DDS) of Pain Intensity, a recent methodology designed for assessing
pain reports in clinical samples. Experiment 1 evaluated the sensitivity of the measure to small changes in
electrocutaneous stimulation relative to a traditional visual analogue scale (VAS) of pain intensity. Additionally,
direct psychophysical scaling methods were employed to determine ratio-scale values for the DDS sensory items in
relation to the electrocutaneous stimuli. This ratio scale was cross-validated by comparison with previously
published ratio-scaled data from cross-modality matching pain intensity judgement studies. Experiment 2 evaluated
the performance of the measure in both experimental and clinical pain samples, as well as the similarity of
item-response patterns in each of these samples. Results indicate that the DDS of Pain Intensity is sensitive to small
changes in electrocutaneous stimulation, has consistent ratio-scale properties across two different psychophysical
methods, and demonstrates similar item-response patterns across divergent experimental and clinical samples. The
results support the validity of the sensory DDS as a measure of pain intensity.

Key words: Pain; Measurement; Psychophysics

Introduction d o m a i n s a m p l i n g and p s y c h o p h y s i c a l scaling, predomi-

Pain measurement has been a limiting factor in pain
research. Reliable and sensitive measures of clinical
pain are needed for adequate investigation of pain
mechanisms and determination of the effectiveness of
clinical interventions for pain reduction. Several differ-
ent instruments have been developed to measure clini-
cal pain, (see Turk and Melzack (1992) for an extensive
review of this literature). Two primary approaches,
* Corresponding author: Jason N. Doctor, Pain Management Pro-
gram (l16-B), San Diego VA Medical Center, 3350 La Joila
Village Dr., San Diego, CA 92161, USA. E-mail: jdoctor@
ucsd.edu.
i Thisworkwas supported in part by the United States Department
of Veterans Affairs.
nate current attempts at clinical pain measurement.
Domain sampling approaches (or category-scaled
measures) are common in the clinical pain literature.
The McGill Pain Questionnaire (MPQ) (Melzack 1975)
is probably the best known and most often used exam-
ple of this approach to pain measurement. These scales
have the advantage of utilizing a subject's endorsement
of verbal pain descriptors as a way of quantifying the
pain experience. Such a technique gives researchers
and clinicians valuable information about the qualita-
tive aspects of pain, while still providing quantitative
information. However, domain-sampled scales are sus-
ceptible to several bias effects associated with ordinal
scaled verbal descriptor measures (see Gracely et al.
1978). Because of bias effects, differences in individual
scores on category scales may not be solely attributable

SSDI 0304-3959(94)00180-4
252
to differences in underlying pain state. Additionally,
parametric statistical analyses are problematic with cat-
egory scales because these scales have ordinal proper-
ties that often violate underlying statistical assump-
tions.
Although laboratory based psychophysical scaling
methods do not provide extensive qualitative informa-
tion about the pain experience, they do have several
advantages over domain-sampled scales in the assess-
ment of pain. Most notably, they provide quantitative
data that are less susceptible to the bias effects associ-
ated with category scales. Moreover, clinical responses
on these scales can be compared directly with patient
responses to ratio-scaled physical stimuli (e.g., thermal
stimulation and electrocutaneous stimulation; Price and
Harkins 1992). The major drawback to using psy-
chophysical scaling approaches for clinical pain mea-
surement is that these approaches usually involve sin-
gle-item analogue scales that may be less stable esti-
mates of a clinical pain sensation than a scale com-
posed of multiple items. Additionally, these scales pre-
sume subjects' ability to directly scale their sensations.
The gap between the methodology of domain sam-
pling and psychophysical techniques has been a major
obstacle in the advancement of pain measurement.
Several laboratories have made extensive efforts to
bridge this gap (e.g., Gracely and Kwilosz 1988; Gracely
et al. 1978; Price and Harkins 1992; Tursky et al. 1982).
One promising development is the Descriptor Differ-
ential Scale (DDS) (Gracely and Kwilosz 1988), an
instrument that applies psychopbysical scaling princi-
ples to clinical pain assessment and measures both the
sensory and affective components of pain (Gracely et
al. 1979). With this scale, subjects are asked to estimate
the magnitude of their clinical pain relative to 12
graded descriptors of pain intensity (sensory dimen-
sion) and 12 graded descriptors of pain unpleasantness
(affective dimension). Patients' pain ratings relative to
each of the 12 descriptors are averaged within each
dimension of pain (intensity and unpleasantness) to
produce a total score for pain intensity and for un-
pleasantness. The DDS has the advantage of being
composed of multiple items like a Category scale, but
(as will be discussed later) may not be as susceptible to
the bias associated with category scales. Although the
measure has yet to be tested thoroughly it has been
thoughtfully designed to satisfy the criteria for pain
measurement proposed by both Gracely and Dubner
(1981) and Price and Harkins (1992). These criteria are
that an ideal pain assessment procedure should: (1)
separately assess the sensory intensity and affective
dimensions of pain; (2) provide immediate information
about the accuracy and reliability of the subjects' per-
formance on the scaling responses; (3) be relatively
free of biases inherent in different physical methods;
(4) be simple to use for pain patients and nonpain
patients in both clinical and research settings; (5) be
reliable and generalizable; (6) be sensitive to changes
in pain intensity; (7) have ratio-scale properties; and
(8) be useful for both experimental and clinical pain
and allow for reliable comparisons between both types
of pain.
The DDS meets 2 of these criteria (i.e., criteria 1
and 2) by design without empirical testing. As men-
tioned earlier, the measure was designed to assess
differentially the sensory intensive and affective com-
ponents of pain (criteria 1) by separating pain intensity
items from pain unpleasantness items, based on previ-
ous research findings (Gracely et al. 1979). Addition-
ally, because the measure is composed of multiple
items with descriptors anchored in the center of the
magnitude estimation line, it provides an index for the
consistency of the subject's responses, thus giving im-
mediate information about the accuracy and reliability
of scaling stimuli (criteria 2) (see Gracely and Kwilosz
1988).
Three of the 8 criteria have been demonstrated to
some extent in previous validation studies of the DDS
or DDS items. Initial validation of the DDS suggests
that it may satisfy or partially satisfy criteria 3, 4, and 5.
With regard to criteria 3, Gracely et al. (1978)
utilized a relatively bias-free ratio-scaling procedure,
known as cross-modality matching, for items that would
later constitute DDS. This procedure reduces bias ef-
fects such as stimulus frequency, range, distribution,
and category effects (Gracely et al. 1978). Other evi-
dence suggests that because the DDS is composed of
multiple items, it may reduce floor and ceiling bias
associated with single-item scales (Slater et al. 1991).
Criteria 4 ('simplicity of use') involves 2 issues: (1)
the time it takes to complete the pain measure and (2)
subject comprehension of the response operation. Be-
cause the DDS-I is composed of 12 items it takes
longer to complete than a single-item VAS. Moreover,
because the response operation is more complex than
that of a VAS, additional time is needed to administer
instructions. The DDS-I takes approximately 5 min to
administer in a clinical setting. Thus, the time neces-
sary for DDS administration is longer than VAS ad-
ministration, but is still reasonable for most clinical
populations. Previous research has addressed the issue
of subject comprehension of the response operation.
In the first DDS validation study, Gracely and
Kwilosz (1988) found that some subjects had difficulty
scaling their pain consistently, in a clinical sample.
Inconsistent scaling is a good indicator of poor com-
prehension of the response operation. In this study,
17.6% of the clinical sample were classified as 'incon-
sistent' responders on the sensory scale of the DDS
and 28.6% were classified as 'inconsistent' responders
on the unpleasantness scale. Additionally, Good et al.
(1991) found similar proportions of inconsistent re-

sponders in a sample of 18 chronic pain patients given
the DDS with instructions to, "Rate your sensation in
relation to each word with a check mark." However, in
a separate sample of 12 chronic pain patients given the
DDS with more detailed instructions and feedback on
2 sample items, 100% consistency in responding among
subjects for both intensity and unpleasantness scales
was observed (Good et al. 1991). This suggests that the
DDS may satisfy the 'simplicity of use' criteria in
chronic pain samples, provided efforts are made to
ensure patients understand the response operation.
While the DDS may be simple to use in chronic pain
samples when clear instructions and sample items are
given, the DDS may be more difficult for certain
clinical samples. Because the response operation for
the DDS is technically complex relative to other pain
measures, it may not be simple to use for patients with
cognitive difficulties (e.g., head injury patients) or pa-
tients of compromised health requiring more intensive
hospital care (e.g., post-operative pain samples).
Finally, with regard to criteria 5, reliability has been
demonstrated with an experimental pain paradigm
(Slater et al. 1991) and in a clinical sample of acute
dental pain patients (Gracely and Kwilosz 1988). The
generalizability of the DDS across a variety of pain
populations has yet to be demonstrated. To date DDS
psychometric characterisitics have only been evaluated
on experimental samples (Slater et al. 1991), acute
dental pain patients (Gracely and Kwilosz 1988), and
non-malignant chronic pain patients (Good et al. 1991).
The present paper attempts to take an initial look at
how well the DDS of Pain Intensity (DDS-I) satisfies
criteria 6, 7, and 8. Specifically, Experiment 1 evalu-
ated the sensitivity of the DDS to small changes in
electrocutaneous stimulation (criteria 6). Moreover,
Exp. 1 examined the ratio-scale properties of the DDS
verbal descriptors using a novel psychophysical
paradigm (criteria 7). Experiment 2 took an initial look
at how useful the DDS is in both an experimental and
a clinical pain sample and how well scaling assump-
tions of the DDS-I hold up across these two popula-
tions (criteria 8).
Experiment 1
Method
Subjects
Ten healthy subjects (5 men and 5 women; aged 20-54 years)
completed Exp. 1 and received $20 for their participation. Subjects
provided informed consent and were instructed that they could
decline to participate at anytime.
Apparatus and stimulus materials
Electrocutaneous stimulation was provided by a Grass Instru-
ments $44 stimulator that was linked to a Grass Instruments SIU5
253
stimulus isolation unit. Before the experiment each subject's leg was
cleaned and mildly abraded such that resistance would not exceed 10
kI2. Two Ag/AgCI electrodes with 2-way adhesive collars, one large
(32-ram diameter collar with 10-mm diameter across the conductance
surface), and one small (19-mm diameter collar with 2-mm diameter
across the conductance surface), each filled with TECA electrode
electrolyte gel, were used to deliver the stimulation. The electrodes
were separated by the width of their collars, such that there was
25.5-mm distance between the center of the 2 electrodes. The
electrodes were then applied to the abraded surface of the non-
dominant tibia just above the ankle. The direct current simulation
consisted of 100-msec constant voltage monophasic square-wave
pulses, delivered at 11 levels of stimulation (1-11 mA) by a research
technician trained in the use of the instrument. Following the meth-
ods of Slater et al. (1991), subject responses to each level of stimula-
tion were recorded separately on 10× 15 cm cards (see Fig. 1 for
example). Each card represented a single-item from the DDS-I or a
VAS. The VAS was 10 cm in length, at each end of the 10-cm line
was a perpendicular line extending 2 mm above and below it.
Directly below the far left end of the line were the words N o Pain
and directly below the far right of the line were the words Pain as
Intense as Possible. Visual analogue scales (VAS) that are between 10
and 15 cm in length and that use anchors that clearly indicate
extremes appear to have the greatest sensitivity and are the least
susceptible to bias or distortions in rating (Scott and Huskisson 1976;
Price and Harkins 1992).
Procedure
All instructions were provided by the research technician and
followed a strict scripted format. The scripted instructions covered
the following topics respectively: (1) overview of procedure, (2)
instructions on how to respond to DDS items followed by a test of
H o w INTENSE is your pain?
Less Intense The Same More Intense
than Intenslty as than
MILD MILD MILD
m
m
More -
Than -
MILD MILD MILD
Lmm
Than
r-
Fig. l. An example of the graphic representation of the DDS-I item
'mild' used in Exp. 1.
254

comprehension using 2 hypothetical painful situations as stimuli Intensity Rating for DDS-I and VAS as a
('pinprick' and 'slamming hand in car door') and 2 non-DDS verbal 20- Function of Stimulus Level

pain descriptors as comparison stimuli (Just Noticeable and Excruci-

ating), (3) instructions on how to respond to a VAS, (4) presentation
of 3-sample electrocutaneous stimuli representing the lowest, middle, ¢1
*r

and highest levels of stimulation to be received during the experi-

10
ment, and (5) instructions on how the experiment will proceed.
Subjects were instructed to remove the top card from the stack to
their right, read it silently and indicate they are ready for stimulation
by saying 'ready'. The research technician then triggered a warning
signal that indicated there were 500 msec until stimulation. This
0
signal was followed by electrocutaneous stimulation for a duration of 0 1 2 3 4 5 6 7 8 9 10 11
1130 msec. Subjects then rated the stimulation and placed the card
face down. This procedure was repeated until all the stimulus
parameters had been evaluated relative to each DDS intensity item Stimulus Level (mA)
and the VAS. Instructions for the VAS were as follows: "" ... • denotes p < .01 for pairwise within subjects (paired sample) t-test.
indicate how intense the sensation was by marking a point on the Fig. 2. Intensity ratings for the DDS-I and VAS as a function of
line between the far left end which represents no pain, and the far stimulus level. Asterisks represent significant differences between
right end, which represents pain as intense as possible. Remember, two stimulus levels.
that in all the rating scales we are interested only in pain intensity."
In response to 11 levels of electrocutaneous stimulation (1-11
mA) spaced in 1-mA increments, subjects rendered 12 DDS intensity
item ratings and 1 VAS rating per stimulus level, for a total of 143 grouped by stimulus level, doubled and divided by 10
electrocutaneous stimulations (13 items x 11 levels of stimulation). to provide a comparable scaling range to the DDS (i.e.,
Stimulation was delivered in 3 blocks (of 48, 48, and 47 stimulations) 0-20), and evaluated for sensitivity using the same
separated by two 5-min rest periods. In order to reduce the likeli-
statistical procedure. Mean response values and stand-
hood of contrast and assimilation sequence effects (Lockhead 1992),
each subject received a different random sequence of the 143 item- ard deviations associated with the DDS-I and the VAS
stimulation pairs. at the 11 different stimulus levels are provided in Table
The research technician was blind to order of item presentation. 1. Fig. 2 illustrates that the DDS-I reliably discrimi-
Additionally, the research technician was positioned out of view of nated ( P < 0.01) 6 of the 1-mA intervals; by contrast,
the subjects during the experiment and was instructed not to talk to
the VAS responses were only significantly different at
the subjects other than during the scripted instruction period.
one such interval (between 3 and 4 mA). All 6 1-mA
intervals between 1 and 7 mA of electrocutaneous
Results
stimulation were differentiated statistically ( P < 0.01)
by the DDS-I. After 7 mA, when the psychophysical
How sensitive is the DDS-I to small (1 mA) changes in
function flattens, two of the four l-mA intervals were
painful electrocutaneous stimulation relatir, e to the VAS?
discriminated ( P < 0.05) and all 2-mA intervals were
This question was evaluated using pairwise within-
discriminated ( P < 0.05).
subjects (paired sample) t tests comparing subject re-
From these results we conclude that the DDS is very
sponses at 1-mA intervals. Subjects' responses on the
sensitive to small (as small as 1 mA) differences in
DDS intensity items were averaged, producing a DDS
electrocutaneous stimulation, whereas single VAS rat-
intensity total score for each subject at each stimulus
ings were not as sensitive to such small stimulus differ-
level. Total scores were then grouped by stimulus level
ences. These results are likely a function of the fact
(1-11 mA). VAS responses for each subject were also
that the DDS total intensity is an average of 12 obser-
TABLE I
vations whereas each VAS intensity score is a single
observation, producing smaller standard deviations for
M E A N ( ± SD) DDS-I A N D VAS PAIN R A T I N G S A T l l LEVELS
O F ELEC~FROCUTANEOUS S T I M U L A T I O N
the DDS-I than the VAS (see Table I). Had subjects
made 12 VAS judgments per stimulus level the sensitiv-
Stimulus level Sensory DDS Sensory VAS ity of the VAS might also increase. In a clinical pain
(mA) Mean ( ± SD) Mean ( ± SD) setting, however, the DDS intensity total score is com-
1 0.97 ( ± 0.54) 0.04 ( ± 0.09) posed of 12 observations (i.e., clinical pain relative to
2 4.58 ( ± 0.83) 1.36 ( -4-__1.42) each descriptor) and the VAS provides only one obser-
3 6.80 ( __4-0.39) 2.68 ( ± 2.16) vation (i.e., "How intense is your pain?"). Memory of
4 7.82 ( ± 0.45) 3.96 ( +__1.92)
previous VAS responding prevents valid immediate
5 9.30 ( ± 0.44) 6.40 ( ± 2.58)
6 10.75 ( ± 0.67) 8.28 ( ± 4.08) re-administration of the VAS to patients judging their
7 12.48 ( __+1.49) 7.96 ( ± 3.07) clinical pain during a brief assessment interval. In
8 13.07 ( ± 1.47) 11.68 ( ± 3.67) order to maintain the external validity of this experi-
9 13.65 ( ± 1.90) 12.32 ( ± 4.09) mental analogue study, the number of observations for
10 14.05 ( ± 1.58) 11.72 ( ± 4.92)
each of the 2 measures was designed to match that
11 14.37 ( ± 1.43) 13.32 ( ± 3.75)
seen in a clinical pain setting.

Can direct psychophysical scaling techniques be used to
develop valid ratio-scale values for DDS intensity items?
Criteria 7, put forth by Price and Harkins (1992),
indicates that a pain measure should have ratio-scale
properties. Ratio scales are interval scales with a known
zero value. A ratio-scaled measure of pain allows an
individual in pain to make meaningful and inter-
pretable statements about the magnitude of the painful
sensation they are experiencing (Gracely et al. 1978).
Ratio scales for verbal pain descriptors allow for accu-
rate quantification of verbal pain report and improve
upon the ability to detect true treatment effects in both
pain research and clinical care.
The ratio-scaling of verbal pain intensity descriptors
involves determining what ratio-scaled physical magni-
tudes best represent the intensity of verbal descriptors.
This can involve subjects judging the intensity of a
descriptor using physical magnitudes. Or, subjects can
judge the intensity of a physical stimulus relative to
verbal descriptors. The DDS is composed of items that
have been ratio-scaled in a previous study (Gracely et
al. 1978). This seminal study utilized a cross-modality
matching technique (Tursky et al. 1982) to scale sen-
sory and affective verbal pain descriptors. In cross-
modality matching, subjects estimate the magnitude of
the dimension (e.g., pain intensity) that a descriptor
conveys using a ratio-scaled response continuum (e.g.,
numerical estimation, line drawing or brightness of
white light). Ratio-scale values for each descriptor are
defined as the subject's response on the associated
ratio-scale modality. The procedure is validated by
demonstrating consistency across two response con-
tinua. Thus, in cross-modality matching of verbal pain
descriptors the brightness of a light serves as a ratio-
scale response continuum to verbal pain descriptor
stimuli.
Cross-modality matching has been a desirable tech-
nique because it does not require the direct adjustment
of noxious stimulation to match descriptor magnitude.
A direct verbal stimulus scaling approach utilizing nox-
ious stimulation as a response adjustment continuum
has been problematic in pain research for numerous
reasons, the most salient being the confounding effects
of subjects' apprehension, fear, and expectation associ-
ated with delivering high levels of noxious stimulation
to match verbal descriptors of great magnitude (e.g.,
'extremely intense'). If the technical problems can be
resolved, directly associating descriptors with con-
trolled noxious stimuli is a preferable approach.
The methodology employed in Exp. 1 provides the
basis for a direct technique for ratio-scaling verbal pain
descriptors that avoids the common problems associ-
ated with noxious stimulus adjustment. This technique
is referred to here as 'bimodality stimulus comparison'.
Subjects are presented with a broad spectrum of elec-
trocutaneous stimulation. With each noxious stimulus,
255
Stimulus Judgments Relative to 'Mild'
20
y = 0.81484 + 15.726*LOG(x) R^2 = 0 . 9 8 7
I
11°
/
1
,
2
. ,
a
. ,
4
• ,
s
. ,
6
. ,
r
. ,
e
• ,
a
• ,
10
• ,
H
.
12
Stlmuhm~ (reAl
Fig. 3. Mean pain ratings for 10 subjects in Exp. 1 relative to the
word 'mild'. P a i n r a t i n g s less than 10 indicate that the sensation was
less than mild. Responses greater than 10 indicate that the sensation
was greater than mild and responses equal to 10 indicate that the
sensation was equivalent to mild. Ratio values were obtained by
setting y = 10 and solving for x in the function.
subjects are presented with verbal descriptors on the
DDS anchored at the center of a magnitude estimation
line (see Fig. 1; Gracely and Kwilosz 1988). Subjects
are then asked to estimate the magnitude of the nox-
ious stimulus relative to the verbal pain descriptor.
Noxious stimulus order and descriptor are completely
crossed, then item-stimulation pairs are randomized
for each subject to avoid context effects (Lockhead
1992). Subject responses for each descriptor are aver-
aged within each stimulus level and plotted against
stimulus level. Fig. 3 illustrates the method for calculat-
ing the ratio-scale value for the sensory descriptor
'mild'.
The y axis represents the subjects' response to elec-
trocutaneous stimulation relative to the descriptor
'mild'. Responses were drawn on a scale from 0 to 20
(see Fig. 1), where a response of 9 or less indicates that
the stimulus was rated as less intense than 'mild', a
response of 11-20 indicates that the stimulus was rated
as more intense than 'mild' and a response of 10
indicates that the stimulus intensity was rated as equiv-
alent to 'mild'. The x axis represents the ratio-scaled
stimulus level in milliamperes. The response curve in
Fig. 1 can be mathematically modeled using a log base
10 equation to explain 98.7% of the variance in re-
sponding. With this equation we can confidently pre-
dict at what ratio-scaled stimulus level subjects would
endorse the word 'mild' (i.e., y = 10). Table II shows
the ratio-scaled values of DDS-I descriptors obtained
using the bimodality stimulus comparison procedure.
Fig. 4 illustrates where each word falls on a mil-
liampere ratio scale. As with cross-modality matching,
a log transformation of ratio values provides linearly
spaced and statistically useful ratio values. Fig. 5 illus-
trates descriptors with log base 10-transformed values
that are approximately equidistant. These 5 descriptors
 256

T A B L E II measurement put forth by Helmholtz (1887) and more

DDS-I D E S C R I P T O R S A N D P R E D I C T E D m A E Q U I V A L E N T S recently by Ellis (1966) and Narens and Mausfeld
11992).
Descriptor ]) Equivalent mA level R : : Y = a + b log x One method of evaluating invariance under change
Faint 2.83 0.973 of scale is to compare standard scores for objects
Very Weak 2.98 {).987 measured by different scales. If the objects are invari-
Very mild 3.15 [).981 ant under change of scale, then standard scores should
Weak 3.38 0.982
be equivalent (or close to equivalent given random
Mild 3.84 0.987
Moderate 4.41 I).978 error in measurement). If the verbal pain intensity
Barely Strong 4.99 0.979 descriptors used in the DDS-I are invariant under
Slightly Intense 6.27 0.968 change of scale, this would suggest that they occasion
Strong 7.84 [).937 quantitatively meaningful responses (i.e., they have ra-
Intense 9.06 ().944
tio-scale properties). In order to evaluate the ratio
Very Intense 13.64 0.918
Extremely Intense 15.61 /).908
properties of the DDS-! descriptors, standard scores
from ratio values obtained in Exp. 1 as well as those
presented in Gracely et al. (1978) were calculated
may be a useful subset of sensory items in research and separately for both approaches.
clinical situations where efficient scaling is necessary. Fig. 6 illustrates the remarkable similarity of ratio
values obtained using the different approaches. Eight
Consistency between ratio t,alues obtained through of the 12 descriptors obtain almost identical ratio val-
cross-modality matching and bimodality stimulus com- ues under each of the 2 procedures, and only ratio
parison values for 2 descriptors ('intense' and 'very intense')
Given that both of these psychophysical approaches vary somewhat from each other (< 0.33 SD). The simi-
(i.e., cross-modality matching and bimodality stimulus lar findings across two psychophysical scaling tech-
comparison) use markedly different methods to deter- niques provide further evidence for the validity of ratio
mine ratio values for verbal pain descriptors, consis- scales for DDS sensory descriptors. Differences on
tency between ratio values (or invariance of values ratio values may be due to random error associated
under changes to physical scale) would provide strong with either or both of the techniques.
support for the validity of ratio scales for the DDS
intensity descriptors. Invariance under change of scale Summary and conclusions
is an important concept in psychophysics and has been
used to determine whether specific measurement tech- Results from Exp. 1 indicate that the DDS-I is
niques have ratio-scale properties (Stevens 1946, 1951: sensitive to small changes in electrocutaneous stimula-
Narens and Mausfeld 1992). In fact, the technique of tion, outperforming the VAS, when a clinical analogue
cross-modality matching has theoretical foundations on is maintained. This demonstration of the sensory DDS'
this premise (Luce 1990). Stevens (1946, 1951) indi-
cates that relationships that are invariant under change
of scale are meaningful quantitative relationships and DDS Descriptors that Exhibit Equidistant
those that are not invariant are quantitatively meaning- Ratio Scale Properties
less. Support for Stevens (1946, 1951) assertion can be
obtained from logical arguments in the epistemology of
1.0.

Item Endorsement as a Function of Stimulus Level 09- • strong

18. _o
17-
0.8- • stightiy intense
16, extremelyintense
15. A
14. veryintense <
13. 0.7- • moderate
12.
11,
i 10. intense 0.6- • mild
9.
strong
| ~
-- 7- slightly intense
moderate 0.5-
5- barelystrong • very weak
4. mid
3-

1-
O.
2.
N- ~
~ry weak
~llllt
mild 0.4

E U

Fig. 4. Plotted points represent mean predicted values for the stimu- Fig. 5. Five verbal descriptors that exhibit equidistant ratio-scale
lus level (mA) at which subjects would endorse a verbal descriptor properties on a log milliampere scale when ratio values are obtained
displayed to the right. through bimodality stimulus comparison.
 257

Standard S c o r e s for Ratio Values O b t a i n e d TABLE III

T h r o u g h Blmodallty Stimulus C o m p a r i s o n and
C r o a s - M o d a l l t y Matching DDS-I ITEM QUARTETS

Low intensity Medium intensity High intensity

• BimodalltyStimulusCompal~x,on Faint Mild Strong
0 Crc~ts-ModalityMatching Very Weak Barely Strong Intense
Very Mild Moderate Very Intense
O
Weak Slightly Intense Extremely Intense
0

o 8
0
, o into 2 groups, 'high pain' and 'low pain' based on their MPQ total
8 score. This was accomplished by establishing a cutoff MPQ score half
-1 (P e 8 •
Faint '
•
,.**d
i
.;k, ' .,k : ,• I
.
v, ira
,
way between the two modes of the obtained bimodal MPQ total
score frequency distribution. Using a between mode cutoff, as op-
v. wqmk week b. itrong lilt. int. Ir,L ext.int.
posed to a median split, enhances the validity of the groupings and
De~aptor the statistical distributions of the groups.
The 12 intensity items on the DDS were divided into 3 item
Fig. 6. Standard scores for ratio values obtained through bimodality quartets representing low-, medium-, and high-intensity descriptors
stimulus comparison and cross-modality matching. Standard scores following Slater et al. (1991) (see Table III). Subjects' ratings of pain
were calculated separately for values obtained within each technique. relative to low-, medium-, and high-intensity descriptors were evalu-
ated for both the 'low pain' and 'high pain' groups. The results from
sensitivity to small changes in pain intensity provides this experiment were compared with experimental pain ratings ob-
tained in Exp. 1.
initial information suggesting that the DDS-I satisfies
criteria 6 of the above stated criteria for an ideal
measure of pain (Gracely and Dubner 1981; Price and Results
Harkins 1992). Additionally, the DDS intensity items
exhibit relatively consistent ratio-scale properties across Are the DDS-I scaling assumptions met for both experi-
two markedly different psychophysical scaling tech- mental and clinical pain (criteria 8)?
niques, thus lending support for the satisfaction of One testable assumption of the DDS is that pain
criteria 7, namely, that a pain measure should have ratings will be higher relative to low-intensity descrip-
ratio-scale properties (Price and Harkins 1992). tors than relative to high-intensity descriptors. In other
In order to evaluate applicability for both experi- words, a patient's pain rating should be higher relative
mental and clinical pain and for reliable comparisons to the word 'faint' than it should be to the word
between both types of pain (i.e., criteria 8), Exp. 2 'extremely intense'. If the DDS items represent a well-
evaluated DDS intensity item response patterns across balanced set of descriptors for measuring both experi-
different levels of pain sensation in both an experimen- mental and clinical pain, then pain ratings relative to
tal and clinical sample. low-intensity descriptors, should be higher than pain
ratings relative to medium-intensity descriptors, and
pain ratings relative to medium-intensity descriptors
should be higher than pain ratings relative to high-in-
Experiment 2
tensity descriptors across experimental and clinical
Method samples. To evaluate this question, results from this

Subjects Clinical Pain Ratings Relative to Low, Medium

Sixty-five male CLBP patients with an average age of 46.43 years and High Intensity Pain Descriptors on the ODS
(range: 23-64 years), average education 14.41 years (range: 8-20
years), and an average annual income of $30-40,000 (range: $10- 20

67,000) comprised the sample in Exp. 2. Mean pain duration was 18-
14.9 years (range: 8 months to 44 years). Standardized orthopedic 16-
diagnosis indicated the sample consisted of degenerative disorders ¢m 14-
(degenerative disc disease, degenerative joint disease, spinal stenosis)
of the spine (80%), herniated nucleus pulposus (5%), and other
orthopedic diagnosis (15%). No subjects involved in the experiment ~ . ~ e LOw~ m s
were using narcotic analgesics, antianxiety agents, or muscle relax- 6" Med ~rns

ants. This was confirmed by urine toxicology screens. 4" A H~

2 I 1

LOW HIGH
Procedure PAIN (UPS)
Fig. 7. Mean intensity ratings separated by low, medium, and high
Subjects were asked to rate their clinical pain using both the sensory DDS item quartet for chronic low back pain patients report-
DDS for Pain Intensity and the MPQ. Subjects were then divided ing high and low levels of pain on the McGill Pain Questionnaire.
258
experiment utilizing clinical pain data were compared
with results from the experimental pain data of Exp. 1.
An evaluation of the properties of the low-, medium-,
and high-intensity descriptors in a clinical sample indi-
cated that these descriptors produced significantly dif-
ferent responses. Additionally, as would be expected
the DDS discriminated patients with high levels of pain
from patients with low levels of pain, as determined by
the MPQ. A 3 × 2 split-plot analysis of variance
(ANOVA) with pain level ('high pain' or 'low pain') as
the between-subjects factor and descriptor intensity
level item quartets (high, medium, or low) as the
within-subject factor indicated a main effect for pain
level (F (1, 63) = 14.67, P < 0.0001) and a main effect
for descriptor intensity level (F (2, 62)=63.14, P <
0.0001) and no significant interaction effect (F (2, 126),
P > 0.01).
This suggests that the DDS intensity total score as
well as each item quartet differentiated the 'low pain"
group from the 'high pain group'. Additionally, pair-
wise comparisons of descriptor intensity level indicated
that low, medium, and high descriptor quartet scores
were significantly different from each other. Fig. 7
shows data from Exp. 2. Here we see that item quartet
intensity ratings are higher for low-intensity descrip-
tors, lower for high-intensity descriptors and moderate
for moderate-intensity descriptors across patients re-
porting both high and low levels of pain. These results
provide evidence in a clinical sample that the DDS of
Pain Intensity is a well-balanced measure that meets
the specific item response pattern assumption. To test
whether this finding is consistent in an experimental
pain sample, the DDS intensity item responses for
subjects in Exp. 1 were grouped into item quartets (see
Table IV) and plotted against level of electrocutaneous
stimulation. Fig. 8 shows data from Exp. 1, where DDS
intensity items have been separated into item quartets
of graded intensity. As can be seen from Fig. 8, pain
ratings relative to low-, medium-, and high-intensity
descriptors are high, medium, and low, respectively.
This is consistent with findings from a previous study
Experlrnental Pain Ratings Relative to Low,
Medium and High Intensity Pain Descriptors
20"
-- 10" "
& High items
Stlmu|ue Level (mA)
Fig. 8. Mean intensity ratings relative to low, medium, and high
sensory DDS item quartets as a function of stimulus level (mA) for
subjects in Exp. 1.
(Slater et al. 1991) that found similar response patterns
to DDS items at 3 levels of electrocutaneous stimula-
tion. Figs. 7 and 8 illustrate the consistent findings
across experimental and clinical samples.
General discussion
This study provides evidence for the validity and
applicability of the DDS of Pain Intensity. Exp. 1
demonstrated that the DDS-I was sensitive to small
differences in electrocutaneous stimulation. This find-
ing suggests that the sensory DDS may be very useful
in applied research. Given the measure's sensitivity at
low and moderate levels of stimulation, the sensory
DDS could be used to detect small changes in pain
intensity associated with pharmacological, surgical,
physical, and behavioral intervention effects. The find-
ings from Exp. 1 also suggest that the DDS intensity
items have ratio-scale properties. This has now been
demonstrated using a direct psychophysical scaling
technique, bimodality stimulus comparison, as well as
commonly used cross-modality matching methods
(Gracely et al. 1978). In addition, Gracely and Kwilosz
(1988) have shown consistency in responding to DDS
items for 1 subject, when these items are plotted against
ratio values obtained from cross-modality matching
experiments.
In Exp. 2, results indicated that the DDS-I differen-
tiated patients reporting high levels of clinical pain
from patients reporting low levels of pain on the MPQ.
Additionally, the findings from Exp. 2 suggest that the
DDS intensity items have item response properties
consistent with assumptions about the measure.
Namely, low-intensity items (e.g., 'mild') result in higher
magnitude estimations of pain intensity relative to the
item than higher intensity items (e.g., 'intense') across
both experimental and clinical samples. The results of
Exp. 2 suggest that response patterns are evenly bal-
anced across the 3 item quartets for people experienc-
ing both high and low levels of clinical pain. This is
important support for the internal validity of the DDS
intensity total score where all 12 items are averaged
using equal weightings.
Log base 10-transformed ratio values obtained
through bimodality stimulus comparison suggest that
particular descriptors (i.e., 'very weak', 'mild', 'mod-
erate', 'slightly intense', and 'strong') are approxi-
mately equidistant. These descriptors could be used as
an abbreviated measure in situations where timely re-
sponding is necessary. Additionally, these descriptors
have the advantage of providing interpretable magni-
tudes: someone reporting 'moderate' pain before an
intervention and 'very weak' after an intervention has
indicated twice the reduction in pain intensity than if

they had reported 'mild' pain after an intervention (see
Fig. 5).
The findings from Exps. 1 and 2 provide initial
evidence that the DDS of Pain Intensity satisfies crite-
ria 6, 7, and 8 for an ideal pain measure put forth by
Gracely and Dubner (1981) and Price and Harkins
(1992). However, several limitations to these studies
should be considered. To begin with, because of elec-
trical current delivery limitations of our constant cur-
rent unit, subjects received constant voltage rather
than the preferable constant-current stimulation that
control for small changes in skin impedance. However,
the effect of this random source of variation is reduced
by the skin abrasion procedure used prior to electrode
application. Secondly, the data from Exp. 1 demon-
strate that for a group of 10 subjects, the DDS of Pain
Intensity was more sensitive to differences in electrocu-
taneous stimulation than was the VAS. The findings do
not provide direct evidence that for individual subjects
the sensory DDS is more sensitive to differences in a
painful stimulus than the sensory VAS. Future studies
using a signal detection research paradigm (Swets et al.
1961) could better answer this question.
Additionally, while electrocutaneous stimulation has
been widely used as a phasic pain model, it is limited
by the fact that it does not exclusively stimulate affer-
ent fibers carrying nociceptive information. Electrocu-
taneous stimulation of low-threshold mechanoreceptive
afferents may suppress the excitatory effects of high-
threshold primary nociceptive afferents on the second-
order neurons in the dorsal horn. This inhibition of the
transmission of nociceptive information could vary
across different milliampere levels of stimulation, re-
suiting in small attenuations in painful sensation at
certain ranges of electrocutaneous stimulation. The
presence of such an effect, however, would still result
in an ordinal stimulus series and would have little
effect on result comparing VAS and DDS sensitivity
and the comparison between experimental and clinical
pain. In light of this limitation, future studies should
examine the sensitivity of the DDS of Pain Intensity
utilizing thermal stimulation which is a phasic pain
model that is less susceptible to this limitation.
Although Exp. 2 provides initial information regard-
ing criteria 8, additional studies need to be conducted
to examine the usefulness of the DDS of Pain Intensity
in both experimental and clinical pain and whether the
DDS of Pain Intensity allows for reliable comparisons
between both types of pain. Use of a triangulation
validation procedure (see Gracely 1989; Heft et al.
1980) would provide useful data with regard to these
questions. Such a procedure would involve a within
subjects comparison of clinical and experimental pain,
which would provide stronger evidence for criteria 8
than is offered by Exp. 2. Moreover, the present study
only examined the DDS sensory scale and does not
259
provide empirical evidence regarding the validity of the
DDS affective scale. Future studies need to evaluate
the DDS affective scale, as this is an equally important
dimension of the pain experience.
The DDS-I is not as simple to use as some other
pain measures (e.g., the MPQ and VAS), but appears
to be reasonably simple for most patient populations.
The DDS-I trades simplicity for both increased sensi-
tivity and an index for response consistency. While
many clinical populations would find the DDS simple
to use, some would not. It is questionable, however,
whether patients who cannot consistently scale their
pain on the DDS can provide accurate pain estimates
with more simple direct scaling methods. When clinical
settings demand a time limited pain measure with a
simple response operation, an abbreviated 5-item mea-
sure utilizing the descriptors in Fig. 5, would provide
meaningful ratio-scaled information about pain inten-
sity. Such an abbreviated measure would simply ask
patients to identify the word that best describes the
intensity of their pain.
In summary, the sensory DDS has thus far shown
promise in satisfying the 8 criteria of pain measure-
ment proposed by Gracely and Dubner (1981) and
Price and Harkins (1992). Gracely and Kwilosz (1988)
specifically designed the measure to satisfy criteria 1 by
creating a measure that differentiated pain intensity
from pain unpleasantness. Additionally, the measure
provides immediate information regarding the consis-
tency of subject responding, indicating the likelihood of
reliable or accurate responding (criteria 2) (Gracely
and Kwilosz 1988). Previous DDS validation studies
(Slater et al. 1991) and pain descriptor validation stud-
ies (Gracely et al. 1978) suggest that the measure and
its items are relatively free from bias (criteria 3), that
the measure is simple to use with additional instruction
(criteria 4) studied clinical samples, and that the mea-
sure is reliable and generalizable (criteria 5). The pres-
ent paper provides support for the measure as being
sensitive to changes in pain intensity (criteria 6), having
items with ratio-scale properties (criteria 7), and being
both useful and comparable reliably in experimental
and clinical samples (criteria 8).
Because of the promising laboratory and clinical
findings, the sensory DDS is likely to be a sensitive and
useful measure for the focal evaluation of clinical pain
intensity. Clinical pain interventions call for sensitive
measures of pain in order to observe true treatment
effects. Most psychological as well as pharmacological
analgesic procedures are thought to produce only mod-
est reductions in pain. Marchand et al. (1993) note that
even the well studied opiate analgesic morphine
changes the perception of pain only 1-2°C in experi-
mental thermal pain studies (cf., Price et al. 1985,
1986). The present investigation suggests that the sen-
sory DDS is sensitive to small changes in painful stimu-
260
lation and thus is a desirable measure in clinical trials
and clinical practice. Equally important is the confi-
dence that non-significant differences on the DDS-I in
clinical trials can likely be meaningfully interpreted,
given the high sensitivity to small differences in noxious
stimulation. The DDS should also be useful in concert
with other measures of pain-related disability, pain
behavior and psychological distress for the comprehen-
sive evaluation of clinical outcome in clinical pain
populations.
Acknowledgements
This research was supported in part by the United
States Department of Veteran Affairs and Sigma Xi
Grants-in-Aid of Research. We would like to thank
Nancy J. Powell and Judith C. Ortega for their assis-
tance in data collection and management.
References
Ellis, B., Basic concepts in measurement. Cambridge University
Press, Cambridge, UK, 1966.
Good, A.B., Slater, M.A. and Doctor, J., Validation of the Descrip
tot Differential Scale for Pain Measurement. Presented at the
10th Annual Meeting of the American Pain Society, New Or-
leans, LA, November 1991.
Gracely, R.H., Pain psychophysics. In: C.R. Chapman and J.D.
Loeser (Eds.), Advances in Pain Research and Therapy, Vol. 12,
Raven Press, New York, 1989, pp. 211-229.
Gracely, R.H. and Dubner, R., Pain assessment in humans a reply
to Hall, Pain, 1l (1981) 109-120.
Gracely, R.H. and Kwilosz, D.M., The Descriptor Differential Scale:
applying psychophysical principles to clinical pain assessment,
Pain, 35 (1988) 279-288.
Gracely, R.H., McGrath, P.A. and Dubner, R., Ratio scales of
sensory and affective verbal pain descriptors, Pain, 5 (1978) 5-18.
Gracely, R.H., Dubner, R. and McGrath, P.A., Narcotic analgesia:
fentanyl reduces the intensity but not the unpleasantness of
painful toothpulp sensations, Science, 203 (1979) 1261-1263.
tteft, M.W., Gracely, R.H., Dubner, R. and McGrath, P.A., A
validation model for verbal descriptor scaling of human clinical
pain, Pain, 19 (1980) 363-373.
Helmholtz, H.V., Z/ihlen und Messen erkennthis-theoretish betra-
chet. (Counting and measuring from an epistemological perspec-
tive). Philosophische Aufs~itze Eduard Zeller gewidmet. Leipzig,
Germany: Engelmann, 1887. (Reprinted from Gesammelte Ab-
handlungen, 3 (1895) 356-391).
Lockhead, G.R., Psychophysical scaling: judgments of attributes or
objects?, Behav. Brain Sci., 15 (1992) 543-601.
Luce, R.D., 'On the possible psychophysical laws' revisted: Remarks
on cross-modality matching., Psychol. Rev., 97 (1990) 66-77.
Marchand, S., Charest, J., Li, J,, Chenard, J., Lavignolle, B., and
Laurencelle, L., Is TENS purely a placebo effect? A controlled
study on chronic low back pain, Pain, 54 (1993) 99-106.
Melzack, R. The McGill Pain Questionnaire, Pain, 1 (1975) 272-299.
Narens, L. and Mausfeld, R., On the relationship of the psychologi-
cal and the physical of psychophysics, Psychol. Rev., 99 (1992)
467-479.
Price, D.D. and Harkins, S.W., Psychophysical approaches to pain
measurement and assessment. In: D.C. Turk and R. Melzack
(Eds.), Handbook of Pain Assessment, Guilford Press, New York,
1992, pp. 111 134.
Price D.D., Von der Gruen, A., Miller, J., Raffi, A. and Price, C.A.,
Psychophysical analysis of morphine analgesia, Pain, 22 (1985)
261 270.
Price, DD., Harkins, S.W., Rafii, A. and Price, C., A simultaneous
comparison of fentanyl's analgesic effects on experimental and
clinical pain, Pain, 24 (1986) 197-203.
Scott, J. and Huskisson, E.C., Graphic representation of pain. Pain, 2
(1976) 175-184.
Slater, M.A., Bacon, S.F. and Good, A.B., Experimental Validation
of a Clinical Pain Measurement Instrument. Presented at the
10th Annual Meeting of the American Pain Society, New Or-
leans, LA, November 1991.
Stevens, S.S., On the theory of scales of measurement, Science, 103
(1946) 677-680.
Stevens, S.S., Mathematics, measurement and psychophysics. In: S.S.
Stevens (Ed.), Handbook of Experimental Psychology, Wiley,
New York, 1951, pp. 1-49.
Swets, J.A., Tanner, W.P. and Birdsall, T.G., Decision processes in
perception, Psychol. Rev., 68 (1961) 301-340, 1961.
Turk, D.C. and Melzack, R., Handbook of Pain Assessment, Guil-
ford Press, New York, 1992.
Tursky, B., Jamner, L.D. and Friedman, R., The Pain Perception
Profile: a psychophysical approach to the assessment of pain
report, Behav. Ther., 13 (1982) 376-394.