Abstract: There is substantial evidence that men and women differ in their perception and experience of pain. However, research on sex differences in pain has mainly relied on self-report ratings,
whereas little is known about sex differences in facial expression of pain. The aim of the present
study was to investigate: 1) whether men and women differ in their facial expressiveness of pain; and
2) whether sex modulates the relationship between self-report and facial pain responses when tonic
experimental pain is applied. Forty young and pain-free individuals (male n 20, female n 20) were
investigated for their subjective and facial responses to tonic heat stimulation at both painful and
nonpainful levels. Tonic heat stimulation was tailored to the individual pain threshold. Self-report
was assessed via visual analog scales. Facial expression was objectively examined using the Facial
Action Coding System. Correlation analyses for the relationship between self-report and facial
expression of pain were conducted. Men and women differed neither in self-report ratings nor in
facial responses during tonic heat stimulation. However, sex had a considerable impact on the
relationship between these variables. Whereas no significant correlations at all were found for men,
we obtained several significant correlations in woman. For that reason, future studies investigating
the relationship between self-report and nonverbal pain behaviors should consider sex as an important modulating factor.
Perspective: The findings of the present study suggest that facial responses to pain can be used as
estimates of the intensity of subjective pain in women better than in men.
2006 by the American Pain Society
Key words: Sex, facial expression, self-report, pain, FACS.
bal expressions of pain. Compared with self-report ratings, nonverbal pain reactions have been found to be less
subject to conscious distortions because of their more
reflexive and automatic nature,15 and therefore the assessment of nonverbal pain responses might add useful
information to the understanding of sex differences in
pain perception.
The facial expression of pain has been considered to be
the most prominent nonverbal pain behavior.8,32 Sex differences in facial responses to pain have hardly been
investigated and only in very few studies on facial expression of pain was sex included in analyses. Whereas
in 2 studies no significant sex differences in facial expression of pain were found,7,33 Guinsburg et al14 reported
increased facial pain responses in female neonates. However, systematic investigations of sex differences in facial
expression of pain in adult subjects have been lacking so
far.
With regard to facial expression of other emotional
states, men and women have appeared to differ substantially in their emotional expressiveness, with women being facially more expressive.5,19 Furthermore, sex has
915
916
been found to have a considerable impact on the relationship between self-report and facial expression of
various emotions. Lang et al23 reported that facial expression of emotions corresponded much better to selfreport in women than in men. Because definitions of
pain suggest that affective reactions are a substantial
component of pain (eg, Merskey and Bogduk30), it seems
very likely that sex moderates the relationship between
self-report and facial responses to pain. This question has
not been answered so far.
In summary, the aims of the present study were to
investigate: 1) possible sex differences in facial expressiveness to painful stimulation; and 2) whether sex has an
impact on the relationship between self-report and facial expression of pain.
We used tonic heat stimulation at both nonpainful and
painful levels as the method of pain induction. Tonic
experimental heat pain offers many advantages. First, it
resembles clinical pain more closely than experimental
phasic pain because of greater similarities in quality,
temporal pattern, and underlying physiologic mechanisms.6,18,26 Second, tonic pain stimulation allows for investigating possible changes in facial expression of pain
over time, of which little is known. Third, tonic heat pain
stimulation allows for easy variation of stimulus intensity. Accordingly, response variations over time and
across stimulus intensities can be separately analyzed,
and covariations can be related to the exact source of
variance. Correlation analyses between self-report and
facial expression of pain were computed in 3 ways: across
stimulus intensities (type 1), over time (type 2), and between aggregated pain responses (type 3). Forth, tonic
heat pain stimulation is not likely to produce sex differences in facial expression of pain, which are simply due to
sex differences in the subjective experience of pain, because heat pain is the pain induction method with the
lowest probability of generating sex differences in the
subjective experience.34 Accordingly, any sex differences
in facial expression of pain can thus be attributed to a
specific, sex-related, mechanism.
Apparatus
Heat stimuli were delivered by use of a Peltier-based,
computerized thermal stimulator (Medoc TSA-2001;
Medoc, Ramat Yishai, Israel) with a 3 3 cm2 contact
probe.
Procedures
Heat stimulation. Tonic heat stimuli were administered either at painful or at nonpainful intensities to the
ventral surface of the right or left thigh. This was done
according to the Tonic Heat Pain Model protocol.26 Small
heat pulses with an amplitude of 1.3C were administered at a constant frequency of 30 pulses per minute.
For painful heat, the pulses were tailored to have a base
of 0.3C below the individual pain threshold and a peak
temperature of 1C above it. For nonpainful heat, the
procedure was the same with the exception that the
peak was 1C below and the base 2.3C below the individual pain threshold. There were 3 experimental blocks
for painful heat and 3 blocks for nonpainful heat. In each
block, the duration of tonic heat stimulation was 10 min.
The temperature increased from baseline with a heating
rate of 0.5C/s to the preset temperatures. The same rate
applied for return to baseline.
Because tonic thermal stimulation was tailored to
the individual pain threshold, assessment of heat pain
threshold preceded tonic stimulation. Heat pain
threshold was determined using the method of adjustment. Subjects were asked to adjust a temperature
starting from 38C, using heating and cooling buttons,
until they obtained a level which was barely painful. A
constant press of the buttons resulted into a heating
or cooling rate of 0.5C/s. Following a familiarization
trial, there were 6 trials and the average of the last 5
trials were used to determine the threshold estimate.
Heat pain thresholds were assessed separately for each
body side.
The order of stimulation was: for the right body side,
assessment of heat pain threshold (block 1), painful
heat (block 2), nonpainful heat (block 3), and nonpainful heat (block 4); and for the left body side, assessment of heat pain threshold (block 5), painful heat
(block 6), painful heat (block 7), and nonpainful heat
(block 8). We always started with the right body side.
Body sides could be assumed to be comparable in our
study, because no side differences were observed regarding pain thresholds [T(39) 1.72; P .15]. To
avoid local sensitization the site of heat stimulation on
the ventral surface of each thigh was changed after
ORIGINAL REPORT/Kunz et al
each stimulation. The 4 sites per thigh were defined by
dividing the skin surface into a distal and proximal
part, and then dividing each of these into a right and a
left section. The stimulation sites were changed in a
clockwise fashion always starting at the distal sites.
The intervals between blocks were approximately 5
min.
Self-report. The subjects were asked to rate the
intensity of the tonic heat stimuli in intervals of 1 min
during tonic heat stimulation (between seconds 55
and 60 of each 1-min epoch). This procedure resulted
into 10 ratings per block, thus yielding 30 ratings for
painful heat and 30 ratings for nonpainful heat in total. Self-report was assessed by a horizontal visual analog scale (VAS) of 100 mm. The scale was labeled with
a verbal anchor of faintly painful in the center so
that all nonpainful sensations were asked to be rated
below 50 and all painful ones above 50.28
Facial expression of pain. The face of the subject
was videotaped throughout the entire session. The camera was placed in front of the subject at a distance of
approximately 4 meters. Before each block with tonic
heat stimulation, subjects were reminded to look at a
fixation point (an abstract picture being positioned behind the camera) to ensure a frontal view of the face.
Subjects were also instructed not to talk during pain induction unless they were asked to provide the self-report
ratings.
The Facial Action Coding System (FACS)9 was used to
assess facial expression. The FACS is based on anatomic
analysis of facial muscle movements and distinguishes
44 different action units (AUs). However, we used an
abbreviated version of FACS, focusing on 4 AUs: brow
lowering (AU 4), tightening of the orbital muscles surrounding the eye (AUs 6/7), nose wrinkling/upper lip
raising (AUs 9/10), and eye closure (AU 43). These 4 AUs
have been found to represent a common facial response to pain. Prkachin33 reported that the bulk of
facial information about pain is transmitted by these
AUs when considering the results obtained by different pain induction methods. The intensity for each AU
was rated on a 5-point scale. The intensity and duration of these 4 AUs were rated by 2 coders who had
been trained by an official FACS coder. The rate of
correspondence between the 2 coders was assessed
and the training was continued until the rate of correspondence reached the level of trained coders (.80).
Time segments of 55 seconds (facial expressions were
not coded during self-report rating) were selected for
scoring. In total, 30 segments of painful heat (3 times
tonic heat pain stimulation of 10 minutes) and 30 segments of nonpainful heat stimulation (3 times tonic
heat stimulation of 10 minutes) were analyzed.
The Observer Video-Pro (Noldus Information Technology) was used to facilitate FACS coding. The Observer is a
professional software package for the analysis of observational data. It allows coding behavior by means of keystrokes that correspond to a user-defined coding
scheme.
917
Statistical Analysis
Sex Differences in Pain Threshold,
Self-Report, and Facial Expression
The effect of sex on pain thresholds was investigated
by computing analyses of variance with repeated measurements with 1 between-subject factor (sex) and 1
within-subject factor (stimulation side [right, left]). To
evaluate sex differences in VAS ratings and facial expression during tonic heat stimulation, analyses of
variance with repeated measurements were used with
1 between-subject factor (sex) and 3 within-subject
factors (condition [nonpainful heat, painful heat],
stimulation block [3 blocks for each intensity], and
minute [1-10]).
918
Mean response
Mean response
Mean response
Mean response
Mean response
Self-report
FACS
Mean response
Mean response
Mean response
Mean response
Mean response
Mean response
Self-report
FACS
Heat trials
Self-report
FACS
Pain trials
Self-report
FACS
Type 3: Aggregated
Mean response
Self-report
Pain trials
FACS
Mean response
Figure 1. Schematic representation of the 3 types of correlation analyses conducted to investigate the relationship between
self-report and facial expression of pain. See text for more detailed information on the 3 types of correlation analyses.
Table 1.
FACTORS
SEX
VAS ratings
AU4brow lowering
Intensity
Duration
Product score
AUs 6/7eye tightening
Intensity
Duration
Product score
AUs 9/10nose wrinkling
Intensity
Duration
Product score
AU 43eye closing
Intensity
Duration
Product score
CONDITION
STIMULATION BLOCK
F(DF)
F(DF)
.1 (1,38)
.180
39.6 (1,38)
<.001
9.8 (2,76)
<.001
7.8 (9,342)
<.001
.1 (1,38)
.6 (1,38)
.9 (1,38)
.433
.433
.348
5.8 (1,38)
2.9 (1,38)
4.3 (1,38)
.021
.095
.049
4.7 (2,76)
1.6 (2,76)
1.8 (2,76)
.012
.203
.162
1.8 (9,342)
1.3 (9,342)
1.1 (9,342)
.101
.215
.345
.1 (1,38)
.1 (1,38)
.1 (1,38)
.783
.719
.698
12.1 (1,38)
5.3 (1,38)
5.7 (1,38)
.001
.026
.021
5.0 (2,76)
1.7 (2,76)
2.0 (2,76)
.009
.179
.138
1.2 (9,342)
1.4 (9,342)
1.4 (9,342)
.248
.157
.166
.3 (1,38)
2.0 (1,38)
1.8 (1,38)
.590
.158
.183
2.1 (1,38)
1.8 (1,38)
1.5 (1,38)
.147
.188
.237
.9 (2,76)
.2 (2,76)
.6 (2,76)
.386
.810
.506
.3 (9,342)
.8 (9,342)
.8 (9,342)
.965
.573
.604
.4 (1,38)
.2 (1,38)
.1 (1,38)
.548
.894
.813
2.4 (1,38)
1.8 (1,38)
1.3 (1,38)
.124
.181
.248
.8 (2,76)
.3 (2,76)
.8 (2,76)
.445
.679
.447
.4 (9,342)
.8 (9,342)
.7 (9,342)
.879
.598
.697
F(DF)
MINUTE
F(DF)
ORIGINAL REPORT/Kunz et al
919
Table 2. Correlations Between Self-Report of Pain and AU 4 (Brow Lowering) for Men, Women,
and All Subjects
MEN
WOMEN
.05
.01
.03
.594
.916
.717
.19
.32
.33
.04
.03
.03
.352
.540
.400
.03
.02
.02
.05
.17
.17
ALL
P
.034
<.001
<.001
.12
.15
.14
.076
.022
.033
.12
.11
.14
.004
.010
<.001
.08
.04
.07
.009
.137
.019
.542
.688
.685
.04
.10
.10
.295
.013
.013
.03
.07
.06
.240
.024
.053
.835
.468
.483
.08
.59
0.51
.726
.006
.023
.01
.10
.01
.968
.532
.935
uct interaction terms between sex and facial responses. For each analysis, hierarchic regression was
conducted, entering the predictor variables sex and
facial responses at the first step and then the product
terms of the two predictor variables at the second
step. A significant change in R2 between steps was
taken as an indicator of a significant interaction. To
reduce the number of analyses, these multiple regression analyses were only conducted when the differ-
Correlations Between Self-Report of Pain and AUs 6/7 (Eye Tightening) for Men,
Women, and All Subjects
Table 3.
MEN
WOMEN
ALL
.20
.13
.15
.032
.161
.097
.39
.34
.33
<.001
<.001
<.001
.29
.24
.25
<.001
<.001
<.001
.08
.06
.06
.048
0.133
.181
.07
.12
.13
.084
.002
.001
.08
.09
.09
.009
.002
.002
.10
.07
.08
.011
0.088
.054
.31
.33
.31
<.001
<.001
<.001
.20
.20
.20
<.001
<.001
<.001
.13
.05
.07
.577
.823
.769
.57
.46
.42
.009
.043
.064
.33
.24
.25
.039
.124
.119
920
Table 4. Correlations Between Self-Report of Pain and AUs 9/10 (Nose Wrinkling) for Men,
Women, and All Subjects
MEN
WOMEN
ALL
.07
.05
.04
.437
.578
.707
.21
.17
.18
.025
.069
.053
.14
.11
.12
.028
.089
.072
.03
.01
.03
.482
.776
.531
.03
.08
.07
.547
.056
.071
.03
.05
.05
.359
.063
.064
.08
.07
.10
.053
.075
.014
.12
.09
.13
.004
.017
.001
.09
.07
.09
.002
.025
.003
.01
.04
.13
.974
.881
.586
.14
.12
.16
.563
.630
.490
.07
.06
.09
.656
.709
.567
ence in correlation coefficients between males and females (Tables 2-5) was larger than .40.
With regard to analyses of variance and multiple regression analyses, findings were considered to be statistically
significant at .05. However, because of the large number of simple correlations being computed, an alpha level
of .001 was required to be judged significant in this case.
Results
Effect of Sex on Self-Report
Heat pain thresholds in men (right side: 44.5 1.5; left
side: 44.1 1.8) and women (right side: 44.2 1.3; left
side: 44.0 1.5) did not differ significantly from one
another [F(1,38) 0.3; P .590]. Furthermore, VAS rat-
Table 5. Correlations Between Self-Report of Pain and AU 43 (Eye Closing) for Men, Women,
and All Subjects
MEN
WOMEN
ALL
.12
.07
.07
.203
.444
.440
.17
.18
.21
.072
.056
.024
.14
.10
.10
.035
.123
.115
.08
.02
.03
.046
.593
.549
.11
.14
.14
.007
.001
.001
.10
.09
.09
.001
.003
.002
.07
.02
.02
.083
.611
.680
.08
.05
.08
.067
.268
.059
.07
.03
.04
.015
.282
.228
.04
.02
.01
.859
.948
.953
.06
.05
.15
.816
.825
.532
.04
.02
.04
.794
.887
.830
ORIGINAL REPORT/Kunz et al
70
VAS ratings
60
921
Male
Female
Non-Painful Heat
= 16,90
= 16,42
= 17,30
= 17,90
= 18,94
= 13,84
faintly painful
50
40
30
70
Painful Heat
VAS ratings
60
faintly painful
50
40
= 19,92
= 18,62
= 20,80
= 17,08
= 20,41
= 18,94
30
1 - 10
1 - 10
1 - 10
minutes
Figure 2. Mean values of visual analog scale (VAS) ratings of perceived intensity during tonic heat stimulation at nonpainful and
painful intensities.
ings during tonic heat stimulation did not differ significantly between sexes (Table 1 and Fig 2). Neither were
any of the interactions involving sex significant (interaction effects are not displayed in Table 1).
We did find a significant effect for condition (Table
1), with painful heat stimulation leading to higher pain
ratings than nonpainful heat (Fig 2). Stimulation block
also had a significant effect on self-report ratings (Table
1), with ratings tending to increase from the first stimulation of 10 min to the second and then to decrease from
the second to the third both at nonpainful and at painful
levels. However, this tendency to increase and decrease
was more pronounced for nonpainful heat stimulation
as indicated by a significant interaction between condition and stimulation trail [F(2,76) 6.4; P .003].
There was also a significant effect of the factor minute
(Table 1) and a significant interaction effect for
minute and condition [F(9,342) 16.7; P .003],
with self-report ratings decreasing for nonpainful heat
and increasing for painful heat over the 10 min of stimulation (Fig 2). None of the other interactions were significant.
922
Male
Female
Non-Painful Heat
AU 4 (product score)
14
12
= 3,93
= 3,60
10
= 3,84
= 3,23
= 8,70
= 1,86
= 11,31
= 4,26
= 9,99
= 2,35
8
6
4
2
0
18
Painful Heat
16
AU 4 (product score)
14
= 12,52
= 12,83
12
10
8
6
4
2
0
0 - 10
0 - 10
0 - 10
minutes
Figure 3. Mean values of AU 4 (brow lowering) (product scores) during tonic heat stimulation at nonpainful and painful intensities.
AU 4 Self-Report (Table 2)
The association between AU 4 and self-report differed
substantially between men and women. Whereas men
showed no relationship at all between AU 4 and selfreported pain ratings, moderate correlations could be
found for women. However, given our stricter value of
ORIGINAL REPORT/Kunz et al
923
Male
Female
18
16
Non-Painful Heat
14
12
= 4,96
= 7,70
= 5,66
= 6,30
= 9,41
= 2,91
10
8
6
4
2
0
18
16
Painful Heat
= 20,21
= 21,47
= 13,14
= 12,90
= 14,05
= 18,43
14
12
10
8
6
4
2
0
0 - 10
0 - 10
0 - 10
minutes
Figure 4. Mean values of AUs 6/7 (eye tightening) (product scores) during tonic heat stimulation at nonpainful and painful
intensities.
tween AUs 6/7 and self-report, women showed moderate correlations. Intriguingly, 7 correlations reached the
level of significance in women despite our stricter value
of .001. (Applying the conventional value of .05 would
have yielded 10 significant correlations in women vs 3 in
men). Computing correlations for both sexes combined
yielded only small correlation coefficients.
In 2 parameters (intensity and duration of aggregated
responses to painful heat [type 3]), the difference in correlation coefficients between men and women was
larger than .40. Therefore, multiple regressions were
conducted for these 2 parameters to test statistically
whether sex moderates the relationship between AUs
6/7 and self-report. The results are displayed in Table 6.
No significant change in R2 was found.
value of .05 would have yielded 4 significant correlations in women versus 1 in men). Nevertheless, correlation coefficients were continuously higher in females
than in males. Computing correlations for both sexes
combined yielded no noteworthy correlation coefficients. Because the difference in correlation coefficients
between men and women was always smaller than .40,
no multiple regressions were computed.
AU 43 Self-Report (Table 5)
No substantial sex differences in the relationship between AU 43 and self-report ratings could be found.
However, we still found slightly higher correlation coefficients in women than in men, and 2 of the correlations
reached the level of significance in women. (Applying
the conventional value of .05 would have yielded 4
significant correlations in women versus 1 in men). Computing correlation for both sexes combined yielded no
noteworthy correlation coefficients. Because the difference in correlation coefficients between men and
women was always smaller than .40, no multiple regressions were computed.
924
Male
Female
16
Non-Painful Heat
14
12
10
8
6
= 0,23
= 0,39
= 0,52
= 2,74
= 0,93
= 0,77
= 0,16
= 5,05
= 0,29
= 7,69
4
2
0
18
16
Painful H eat
14
12
10
= 0,74
= 11,53
8
6
4
2
0
0 - 10
0 - 10
0 - 10
m inutes
Figure 5. Mean values of AUs 9/10 (nose wrinkling) (product scores) during tonic heat stimulation at nonpainful and painful
intensities.
Discussion
The major findings of the present study are that males
and females differed neither in their subjective responses to pain nor in their facial expressiveness of pain,
but sex had a notable influence on the relationship between self-report and facial expression of pain, at least in
some AUs, with women showing a stronger relationship
between the two domains. We discuss these findings in
this order.
ORIGINAL REPORT/Kunz et al
925
18
M ale
Female
16
AU 43 (product score)
Non-Painful Heat
14
12
= 043
= 1,62
10
= 2,46
= 4,40
= 3,49
= 1,70
= 9,17
= 1,91
= 5,45
= 1,91
8
6
4
2
0
18
AU 43 (product score)
16
14
P a in fu l H e a t
12
10
= 1,60
= 5,21
8
6
4
2
0
0 - 10
0 - 10
0 - 10
m in u te s
Figure 6. Mean values of AU 43 (eye closing) (product scores) during tonic heat stimulation at nonpainful and painful intensities.
mean of standard deviations for each block of stimulation.
Testing the Moderating Effect of Sex on the Relationship Between Self-Report and
Facial Expression of Pain
Table 6.
STEP
1
2
1
2
1
2
1
2
VARIABLES (DEPENDANT
SELF-REPORT)
VARIABLE:
bETA
.10
.04
1.16
.01
.04
1.14
.61
.25
2.48
.05
.26
2.19
.26
.07
.51
.33
.07
.57
1.60
.46
.87
2.15
.46
1.05
R2
Total R2
.012
.012
.23
.144
.002
.156
.002
2.23
.04
.117
.119
1.62
.067
.067
1.32
.019
.113
.086
.113
1.13
2.35
.026
.139
1.94
926
sadness, happiness), for which women have been found
to be facially more expressive than men.5,19 It is likely
that the predictable and controllable way of experimental pain induction produces a rather low level of affective
responding. Therefore, the findings derived from other
emotions (such as psychologic stressors as sadness and
fear) may not be applicable to more physical stressors
such as pain.
ORIGINAL REPORT/Kunz et al
927
these arguments, it may seem surprising that the correlation coefficients obtained by the three different types
of correlation analyses (Fig 1) were not very different.
However, our effort was rewarded by the certainty that
we did not miss certain aspects of the relationship between self-report and facial expression of pain. The analysis including more than one stimulus intensity produced
tentatively the highest correlation coefficients. This finding corroborates our hypothesis that approaches with
multiple stimulus intensities are to be preferred to approaches with only one intensity, which we derived from
data obtained by phasic pain stimulation.20
Regardless of the type of correlation analyses conducted we found substantial differences between the 4
AUs (at least for women and for both sexes combined).
Whereas AUs 4 and 6/7 produced small to moderate correlations with self-report ratings, AUs 9/10 and 43 produced only negligible correlations. Therefore, AUs 4 and
6/7 seem to be the most qualified AUs for encoding the
subjective pain experience, at least when tonic heat pain
of moderate intensity is applied.
As a limitation of the present study, the small sample
size has to be acknowledged. When looking for moderator effects, much larger sample sizes are recommended
to guarantee adequate power. However, owing to the
time- and effort-consuming procedure of FACS coding in
References
13. Fillingim RB, Keefe FJ, Light KC, Booker DK, Maixner W:
The influence of gender and psychological factors on pain
perception, J Gen Cult Health 1:21-36, 1996
928
22. LaFrance M, Hecht MA, Paluck EL: The contingent smile:
A meta-analysis of sex differences in smiling. Psychol Bulletin 129:305-334, 2003
23. Lang PJ, Greenwald MK, Bradley MM, Hamm AO: Looking at pictures: Affective, facial, visceral, and behavioral reactions. Psychophys 30:261-273, 1993
24. Lautenbacher S, Nielsen J, Andersen T, Arendt-Nielsen L:
Spatial summation of heat pain in males and females. Somatosens Mot Res 18:101-105, 2001
25. Lautenbacher S, Rollman GB: Sex differences in responsiveness to painful and nonpainful stimuli are dependent
upon the stimulation method. Pain 53:255-264, 1993
26. Lautenbacher S, Roscher S, Strian F: Tonic pain evoked
by pulsating heat: Temporal summation mechanisms and
perceptual qualities. Somatosens Mot Res 12:59-70, 1995
35. Rollman GB, Lautenbacher S: Sex differences in musculoskeletal pain. Clin J Pain 17:20-24, 2001