Content uploaded by Paolo Riva
Author content
All content in this area was uploaded by Paolo Riva on Oct 23, 2017
Content may be subject to copyright.
Gender effects in pain detection: Speed and accuracy in decoding female
and male pain expressions
Paolo Riva
⇑
, Simona Sacchi, Lorenzo Montali, Alessandra Frigerio
University of Milano-Bicocca, Italy
article info
Article history:
Received 17 July 2010
Received in revised form 4 February 2011
Accepted 24 February 2011
Available online 23 March 2011
Keywords:
Pain expressions
Emotional expressions
Gender differences
Face perception
Implicit measures
abstract
The ability to detect facial expressions of pain is crucial in eliciting prosocial behaviors towards the
individual experiencing pain. Previous studies have shown that the sufferers’ gender can affect the
observers’ explicit judgment of the pain face, thus suggesting its possible influence on pain decoding.
The present study investigates whether the sufferer’s gender affects the observer’s reflexive or implicit
detection of facial expression of pain. More specifically, we used implicit measures to test whether
observers detect pained expression more quickly or accurately on male or female faces. In three
experimental studies, we devised a set of stimuli using computer-generated faces. In Experiment 1,
prototypical female and male avatars with different facial expressions (pain, anger, disgust, and neu-
tral) were displayed, while subjects’ (N= 34) accuracy and speed at identifying the expressions were
recorded. In Experiment 2, participants (N= 56) watched videos of the avatars displaying dynamic
expressions and had to quickly and accurately identify each expression. In Experiment 3, participants
(N= 38) were shown an androgynous avatar face showing different expressions and were asked to
identify the face as either female or male. Overall, we found that the target’s gender affected the
observer’s reflexive decoding of the facial expression of pain. Specifically, the results showed that par-
ticipants, regardless of their gender, were slower and less accurate in recognizing pain expressions
(but not other expressions) on female faces. Furthermore, androgynous faces displaying pained
expressions were more likely to be categorized as male than female. Several potential explanations
are discussed.
Ó2011 European Federation of International Association for the Study of Pain Chapters. Published by
Elsevier Ltd. All rights reserved.
1. Introduction
Since Darwin (1872), the pain face has been considered primar-
ily for its communicative function (Craig, 1992; Hadjistavropoulos
and Craig, 2004; Prkachin, 2009). Indeed, it often seems advanta-
geous for the sufferer to have the chance to translate a distressing
internal state into a message that can be perceived by someone in
the environment (Prkachin et al., 1994; Simon et al., 2008). Re-
search has found that the specific pain-related facial movements
(Prkachin and Solomon, 2008) are consistent across genders (Kunz
et al., 2006), ages (Hadjistavropoulos, 2005) and cognitive abilities
(Kunz et al., 2007).
To be effective in its communicative function, the pain face,
once encoded, needs to be detected by someone in the environ-
ment. The facial expression of pain is highly salient to observers
(Craig, 1992); it prompts them to provide support for the suf-
ferer or to protect themselves from a threat (Williams, 2002;
Yamada and Decety, 2009). Furthermore, observers need to
identify the pain face correctly and quickly in order to provide
rapid support or to receive immediate signals of a possible
alarm. Researchers have examined observers’ ability to discrim-
inate the facial expression of pain from expressions of other ba-
sic emotions (Simon et al., 2008). However, past research has
also shown variance and errors in the observer’s ability to de-
tect pain (Kappesser and Williams, 2002; Prkachin and Craig,
1994), suggesting that further investigation of the variables that
can bias the ability to detect the facial pain expression is
needed.
Several studies indicate that the sufferer’s social characteristics,
including gender, ethnicity and age, can affect the observers’
perception of pain. For instance, pain may be under-recognized
in older people and members of ethnically disadvantaged groups
(Horgas and Elliott, 2004; Green et al., 2003). Gender may also af-
fect the process. For instance, Robinson and Wise (2004) found that
observers judged that women undergoing a cold pressor task expe-
rienced more intense pain than men. Hirsh et al. (2008) showed
that computer representations of female faces were rated as suffer-
1090-3801/$36.00 Ó2011 European Federation of International Association for the Study of Pain Chapters. Published by Elsevier Ltd. All rights reserved.
doi:10.1016/j.ejpain.2011.02.006
⇑
Corresponding author. Address: Department of Psychology, University of
Milano-Bicocca, 1, Piazza Ateneo Nuovo, 20126 Milano, Italy. Tel.: +39 02 6448
3775; fax: +39 02 6448 3706.
E-mail address: paolo.riva1@unimib.it (P. Riva).
European Journal of Pain 15 (2011) 985.e1–985.e11
Contents lists available at ScienceDirect
European Journal of Pain
journal homepage: www.EuropeanJournalPain.com
ing more intense pain than representations of male faces. These
findings related to observers’ social judgment are consistent with
pain-related gender role stereotypes (Hoffmann and Tarzian,
2001), and they suggest that gender can bias the observer’s decod-
ing of a target face.
However, studies have shown that explicit and reflective social
judgment might diverge from implicit, or reflexive, cognitive pro-
cesses and associations (Greenwald and Nosek, 2008; Sloman,
1996). These latter are faster, they take place automatically and
unintentionally, and they are less influenced by social desirability.
In real life, facial pain cues are often received accidentally, and
behavioral responses can occur under uncertainty or time pres-
sure; under these conditions, observers are more likely to rely on
automatic and implicit processes than controlled reasoning (Tait
et al., 2009). Therefore, implicit methods can assess the involun-
tary effects of the target’s gender on pain detection while bypass-
ing explicit reasoning and social desirability concerns. Consistent
with these premises, the present study aimed to examine, using
implicit measures, whether the sufferer’s gender affects the obser-
ver’s speed and accuracy in detecting otherwise identical facial
expressions of pain.
2. Experiment 1
In Experiment 1, we investigated whether participants de-
tected facial expression more quickly and/or accurately in female
or male faces. Furthermore, we tested the distinctness of the
interaction between gender and the pain face versus other
threat-relevant facial expressions. Anger and disgust were cho-
sen as control stimuli because of their negative valence and their
threat-relevant nature (Kappesser and Williams, 2002; Williams,
2002).
As already mentioned, implicit methodology was used in this
and the following experiments. Implicit methods are appropriate
in cases in which some sort of knowledge or schema influences
an observer’s cognition or behavior, but the observer might not
be aware of that knowledge and its influence. This methodological
choice was preferred because of the known role of automatic cog-
nitive processes in pain decoding. Tait and colleagues (2009) ar-
gued that observers often perceive sufferers’ pain under
conditions of uncertainty. This phenomena fosters the use of an
intuitive/heuristic cognitive system (also known as ‘‘System 1’’),
which is fast, effortless and automatic, and which can ultimately
bias the observer’s ability to decode pain expressions. Considering
the role of automatic cognitive processes in pain judgment, we re-
corded observers’ reaction times and accuracy to determine
whether an association existed between the targets’ gender and
the observers’ perceptions of pain expressions. Through the accu-
racy scores, we examined the interaction between gender and
the observer’s confusion of the pain face with the other negative
emotional expressions (i.e., misinterpreting pain as disgust or dis-
gust as pain).
2.1. Method
2.1.1. Participants
Participants were 34 undergraduate students (24 females, 10
males; mean age = 25 ± 5.6) at the University of Milano-Bicocca.
They received class credit for participating. All participants were
Caucasian with normal or corrected-to-normal vision.
2.1.2. Stimuli and design
We adopted computer-generated avatar faces to present iden-
tical pain expressions and precisely control the expressive inten-
sity of each face, in line with previous studies (Hirsh et al., 2008,
2009). A series of 400 400 pixel computer avatars was gener-
ated. We used FaceGen 3.1, a computer software program that
allows the user to manipulate the facial movements of an avatar
along with other features, such as its gender, age and ethnic
group. Using this tool, we generated Caucasian faces that were
identical to each other except for their gender and their facial
expressions, thus controlling all the other possible confounding
variables (e.g., the degree of facial attractiveness, the gaze direc-
tion, facial asymmetry, ethnicity and general facial morphology).
Moreover, using computer-generated faces provided us with a
unique way to objectively match the faces’ intensity of expres-
sion; the exact same expression (in terms of the actions involved
and intensity of each movement) could be displayed on a female
or a male avatar face. FaceGen 3.1 includes a series of pre-pro-
grammed slides that allow the user to generate expressions of
anger, disgust, fear, happiness, sadness and surprise. A computer
programmer was hired to add a pain slide (see Fig. S1, see the
online version at 10.1016/j.ejpain.2011.02.006) that included
the facial movements, or action unities (AUs) involved in the
pain face (AUs: 4, 6, 7, 10, and 43). The exclusive presence of
these AUs and the specificity of the resulting pain face compared
with other basic emotions were confirmed by three trained
judges, who were blind to the purpose of the research, using
the Facial Action Coding System (FACS) (Ekman and Friesen,
1978). A subsequent study was conducted to determine whether
naïve observers perceived the new slide’s expression as pained.
Thirty-one participants (58% females, mean age = 23 ± 2.43) were
presented with a series of eight avatar faces expressing pain, an-
ger, disgust and neutrality. Participants had to identify each
expression by choosing one out of eight options: fear, anger, dis-
gust, sadness, shame, happiness, pain and neutrality. The results
showed that participants correctly identified the pain expression
generated with FaceGen 3.1 in 80.6% of the cases (50 times out
of 62 cases). This pattern was similar to that for disgust (79% of
correct hits). Anger was correctly identified in 87% of cases,
whereas the neutral expression was correctly identified in the
93.5% of the total cases.
The resulting experimental design considered the following fac-
tors: 2 (face gender: male vs. female) 4 (expression: anger, dis-
gust, neutral or pain) 2 (participant’s gender: male vs. female),
with the first two factors varied within-subjects and the last one
between subjects.
2.1.3. Procedure
We employed a speeded dichotomous decision task in which
individuals were presented a series of avatar faces on a com-
puter monitor one at a time. Stimuli presentation and response
measurements in this and the subsequent experimental proce-
dures were controlled by the software package e-Prime™
(Schneider et al., 2002). The images were displayed at the center
of a 17-in. color monitor screen on a uniform, black background
with a 75-Hz refresh rate. The level of brightness was consistent
for every image. Participants observed the monitor at a viewing
distance of about 50 cm without head restraint. They were in-
structed that for each trial, they should quickly and accurately
classify avatar facial expressions as pained or ‘‘other’’ by pressing
the ‘‘D’’ or ‘‘K’’ key on the keyboard with their left and right
forefinger, respectively. During each trial, a fixation cross ap-
peared on the screen; approximately 500 ms later, an avatar face
was presented. As soon as the participant provided his or her an-
swer, the avatar face disappeared and another fixation cross ap-
peared. Participants completed a training phase composed of 25
trials and then judged 80 trials (10 stimuli for each condition).
After each trial, participants received feedback regarding the cor-
rectness of their responses. The stimuli were presented
randomly.
985.e2 P. Riva et al. / European Journal of Pain 15 (2011) 985.e1–985.e11
2.2. Results
2.2.1. Accuracy
Participants’ answers were classified as either correct (1) or
incorrect (0). We assigned a score of 1 when the pain expression
was correctly detected and when expressions of anger, disgust or
neutrality were perceived as ‘‘other’’; vice versa, we assigned 0
scores when other expressions were identified as pain and pain
as other expressions. Overall, participants made 289 errors
(11.89%) (M= 8.50, SD = 4.51). An accuracy index was computed
as the sum of the correct answers.
Crucially, the 2 (face gender: male vs. female) 4 (expression:
anger, disgust, neutral or pain) 2 (participant’s gender: male vs.
female) ANOVA revealed an interaction effect between facial
expression and gender, F(1, 32) = 24.15, p< .001,
g
2
p
¼:42. More
specifically, the results showed that accuracy in detecting pain
expressions significantly increased when pain was showed by a
male face compared with a female one; there was no difference be-
tween male and female faces and observers’ accuracy of identifying
the other three expressions (see Table 1).
The analysis also yielded a main effect for expression,
F(1, 32) = 38.26, p< .001,
g
2
p
¼:54. As the post hoc analyses
showed, the expression of anger was misperceived (M= 9.65,
SD = 0.74) to the same extent as the neutral expression (M= 9.85,
SD = 0.55). The participants were less accurate in identifying dis-
gust (M= 8.49, SD = 1.55), which is more similar to pain than the
other two expression (Simon et al., 2008). Participants scored low-
est on identifying the expression of pain (M= 7.76, SD = 1.46). At
this regard, data on the type of errors participants made revealed
that 29.61% of pain expressions were classified as non-pain expres-
sions, while 8.47% of non-pain expressions as pain expression.
These results might arise from the nature of the task, which re-
quired that pain be identified specifically, while the expressions
of anger, disgust and neutrality could be classified roughly into a
‘‘not-pain’’ category.
The ANOVA did not reveal a main effect for participant’s gender,
F(1, 32) = 1.11, p= .30 or interaction effects between participants’
gender and the other factors, Fs(1, 32) < 2.50, ps > .12.
2.2.2. Latency
After the analysis on the level of answer accuracy, we consid-
ered the latency time between the stimuli presentation and the
participants’ answer, for correct answers only. Reaction times
(RT) were filtered at 2 standard deviations (calculated for each con-
dition). This procedure eliminated 5.51 % of the total responses.
Consistent with the analysis on accuracy, statistics revealed a
significant interaction effect between expression and face gender,
F(1, 26) = 3.29, p< .03,
g
2
p
¼:11. As Table 2 shows, the face gender
did not affect the participants’ RT in classifying anger and disgust;
however, when presented with a neutral expression, participants
were faster in answering when the avatar had a female appearance
than when it was male; conversely, when presented with a pain
expression, participants were faster in identifying the stimulus
when the avatar had a male than female appearance (LSD test:
ps < .05).
The statistical analysis did not reveal a main effect of partici-
pants’ gender, F(1, 26) = 3.26, p= .08: in decoding facial expres-
sions, female participants (M= 871.01, SD = 208.09) were as fast
as males (M= 1004.04, SD = 269.39). Moreover participants’ gender
interacted with face gender, F(1, 26) = 5.72, p= .02,
g
2
p
¼:18: fe-
male participants (M = 899.32, SD = 225.36) were as fast as male
participants (M= 978.83, SD = 224.15) when they had to identify
male expressions but male participants (M = 1029.25,
SD = 283.67) were slower than female participants (M = 842.69,
SD = 191.33) when they were asked to decode expressions showed
by female faces. The type of expression showed by the faces signif-
icantly affected the participants RT, F(1, 26) = 14.16, p< .001,
g
2
p
¼:35: participants were relatively faster in classifying neutral
expressions (M= 849.21, SD = 202.69) and relatively slower in clas-
sifying disgust expressions (M= 1054.93, SD = 285.15). Analysis did
not yield the avatar face gender main effect, F(1, 26) = .02, p= .89.
The three-way interaction effect among participants’ gender,
avatar gender and expression was not significant, F(1, 26) = 1.85,
p> .18.
Overall, Experiment 1 provided some preliminary evidence that
pain was more salient in male faces at the reflexive, implicit level.
Participants detected pain on male faces more accurately than on
female faces. Furthermore, they identifying the pain face more
quickly when the avatar had a male face than when it had a female
one. These results were specific to the pain face, as shown by the
analyses of the other expressions.
3. Experiment 2
In Experiment 2, we aimed to replicate the findings of Experi-
ment 1 using dynamic stimuli instead of static facial displays. Spe-
cifically, Experiment 2 was meant to investigate the threshold at
which observers could correctly identify a pain expression as a
function of the gender of the face displaying it. Based on the find-
ings of Experiment 1, we expected that female faces needed to dis-
play higher magnitudes of pain to be detected by observers.
3.1. Method
3.1.1. Participants
Fifty-six Milano-Bicocca university students (33 females; mean
age = 23 ± 2.8) were recruited to participate in a study on the
perception of facial expressions. They received class credit for
participating. All subjects were Caucasian and had normal or cor-
rected-to-normal vision.
3.1.2. Stimuli and design
The participants were shown series of 16-s dynamic visual
stimuli (video) of male and female avatar faces displaying anger,
Table 1
Mean (standard deviations) of accuracy in decoding expression when showed by
female or by male faces. Experiment 1.
Expression Face gender TOT
Female Male
Neutral 9.91
a
(0.38) 9.79
a
(0.73) 9.85 (0.55)
Anger 9.74
a
(0.67) 9.56
a
(0.82) 9.65 (0.74)
Disgust 8.79
a
(1.47) 8.18
a
(1.62) 8.49 (1.55)
Pain 6.88
a
(1.49) 8.65
b
(1.43) 7.76 (1.46)
Note: Different letters indicate in each row statistical differences between male and
female condition, according to the LSD test (p< .001).
Table 2
Mean (standard deviations) of participants’ latency time (in milliseconds) between
the expression presentation and the answer when correct. Experiment 1.
Expression Face gender TOT
Female Male
Neutral 785.18
a
(211.02) 848.69
b
(194.36) 816.94 (202.69)
Anger 925.22
a
(297.04) 905.91
a
(205.50) 915.57 (251.27)
Disgust 969.88
a
(281.03) 1068.23
a
(289.27) 1019.06 (285.15)
Pain 937.81
a
(156.44) 866.49
b
(189.27) 902.15 (212.86)
Note: Different letters indicate in each row statistical differences between male and
female condition, according to the LSD test (p< .05).
P. Riva et al. / European Journal of Pain 15 (2011) 985.e1–985.e11 985.e3
disgust or pain. Each video was created by pooling 100 frames rep-
resenting facial expressions with increasing intensity (the degree
of increase was controlled and matched for each expression using
the FaceGen 3.1 controls). Each video started with a neutral face
and moved progressively through the sequence of 100 frames,
reaching the maximum amount of the three expressions after 16 s.
The experimental design considered the following factors: 2 (face
gender: male vs. female) 3 (expression: anger, disgust or pain) 2
(participant’s gender: male vs. female), with the first two factors var-
ied within-subjects and the last one between subjects.
3.1.3. Procedure
The 16-s dynamic visual stimuli were presented on a computer
monitor one at a time in a randomized order. The images were dis-
played at the center of the screen of a 17-in. monitor, with a uni-
form black background and a 75-Hz refresh rate. The level of
brightness was consistent for every image. Participants viewed
the monitor at a distance of about 50 cm without head restraint.
Participants were asked to watch each video and identify the facial
expressions as quickly and accurately as possible. They were in-
structed to press the space bar on the computer keyboard as soon
as they thought they had identified the expression displayed by the
avatar face. The face disappeared when the space bar was pressed,
and the participant had to indicate which of the three expressions
s/he had identified by pressing a designated key. Participants com-
pleted a training phase composed of two trials, then judged 12 tri-
als (two stimuli for each condition). Participants were given
feedback about the correctness of their response after each trial.
3.2. Results
The scores for reaction times and accuracy were subjected to a 2
(face gender: male vs. female) 3 (expression: anger, disgust or
pain) 2 (participant’s gender: male vs. female) ANOVA. Reaction
time was the crucial dependent variable (because it directly re-
flects the detection threshold), and accuracy was not independent
of reaction times (e.g., higher latency corresponds with higher
expression intensity and, consequently, an easier task); however,
we report the analyses of reaction times and accuracy separately
for improved clarity.
3.2.1. Latency
We recorded the time between the beginning of the video and
when the participants pressed the space bar and selected cases
in which participants correctly identified each expression. Reaction
times (RT) were filtered at 2 standard deviations (calculated for
each condition). This procedure eliminated 2.73% of the total
responses.
Facial gender interacted with the type of expression,
F(1, 60) = 10.058; p< .001;
g
2
p
¼:251. As shown in Table 3, three
patterns of results were found. The face’s gender did not affect par-
ticipants’ RT for accurately decoding an expression of anger. Partic-
ipants decoded disgust more quickly when it was displayed by a
female face compared with a male face. In contrast, and consistent
with the findings on static facial displays in Experiment 1, partici-
pants identified dynamic pain expressions more quickly on male
avatar faces than female ones.
Neither the face’s gender (F(1, 30) = 1.68, p= .204) or of the type
of expression (F(1, 60) = .27, p= .762) had a significant main effect
on participants’ RT.
The statistical analysis did not reveal a main effect for partici-
pants’ gender, F(1, 30) = 1.16, p= .289. No interaction effects were
found between participants’ gender and the other factors,
Fs(1, 30) < 1.52, ps > .22.
3.2.2. Accuracy
Participants’ answers were then classified as either correct (1)
or incorrect (0). Overall, participants made 222 errors (M= 3.96,
SD = 2.41; 33.03% of the total; Table S1, see the online version at
10.1016/j.ejpain.2011.02.006 for the exact type of errors made).
An accuracy index was computed as the sum of the correct
answers.
In line with the results of Experiment 1, a significant interaction
between face gender and expression was present, F(2, 108) = 7.69,
p< .001,
g
2
p
¼:12. As shown in Table 4, participants were more
accurate in decoding the expression of anger and disgust on female
rather than male faces. However, the higher accuracy of detecting
facial expression on a female face dropped when avatars displayed
a pain expression.
There was a significant main effect of the face gender on accu-
racy level, F(1, 54) = 25.04, p< .001,
g
2
p
¼:32. Post hoc analyses re-
vealed that accuracy was greater when the expression was
displayed by a female (M= 1.47, SD = 0.57) than a male target
(M= 1.17, SD = 0.67).
The analysis also yielded a main effect of expression,
F(2, 108) = 5.14, p= .008,
g
2
p
¼:08. More specifically, participants
were less accurate in decoding the expression of pain (M= 1.18,
SD = 0.70) than anger (M= 1.39, SD = 0.65) and disgust (M= 1.41,
SD = 0.79). No significant differences emerged between the two lat-
ter factors.
The ANOVA revealed neither a main effect of participant’s gen-
der, F(1, 54) = 1.49, p= .23, nor interaction effects between partici-
pants’ gender and the other factors, Fs(1, 54) < .58, ps > .55.
In sum, Experiment 2 confirmed Experiment 1’s findings that
the male pain face is more salient than the female pain face at
the implicit level. Results indicated that dynamic expressions of
pain – unlike other negative emotions – were judged as pained
more quickly if they were displayed by a male face. The null results
for the accuracy score do not contradict the findings for reaction
time. Indeed, at the moment participants identified an emotion
in the dynamic display, the mean pain expression intensity dis-
played by female avatars was likely to be higher than that of male
avatars. Thus, we obtained the same accuracy scores for male and
female faces, but the female pain expressions were easier to detect
because they were watched for a longer time and were therefore
more intense.
Table 3
Mean (standard deviations) of participants’ latency time (in milliseconds) between
the expression presentation and the answer when correct. Experiment 2.
Expression Face gender TOT
Female Male
Anger 10003.98
a
(500.99) 10724.702
a
(470.86) 10523.10 (2767.14)
Disgust 9199.50
a
(428.03) 11458.09
b
(500.112) 10361.98 (2913.40)
Pain 11318.68
a
(436.14) 9829.29
b
(580.06) 10771.91 (2637.84)
Note: Different letters indicate in each row statistical differences between male and
female condition, according to the LSD test (p< .001).
Table 4
Mean (standard deviations) of accuracy in decoding expression when showed by
female or by male faces. Experiment 2.
Expression Face gender TOT
Female Male
Anger 1.59
a
(0.72) 1.18
b
(0.95) 1.40 (0.65)
Disgust 1.68
a
(0.74) 1.14
b
(0.69) 1.41 (0.71)
Pain 1.15
a
(0.93) 1.20
a
(0.93) 1.19 (0.68)
Note: Different letters indicate in each row statistical differences between male and
female condition, according to the LSD test (p< .001).
985.e4 P. Riva et al. / European Journal of Pain 15 (2011) 985.e1–985.e11
4. Experiment 3
Experiment 3 was designed to further test the prediction that
the facial expression of pain and the target’s gender interact at
the implicit level. The findings from Experiments 1 and 2 revealed
that perceivers detect pain on male faces more quickly and accu-
rately than on female faces. Reversing the perspective of these pre-
vious experiments, we predicted that detecting a facial expression
of pain would lead observers to perceive an androgynous target as
more masculine.
4.1. Method
4.1.1. Participants
Participants were 39 undergraduate students (29 females, 10
males; mean age = 25 ± 5.31) at the University of Milano-Bicocca.
They received class credit for participating. All participants were
Caucasian with normal or corrected-to-normal vision.
4.1.2. Stimuli and design
Ten versions of an androgynous avatar face were generated using
FaceGen 3.1. The ambiguous stimuli were created by setting the
gender bar at the midpoint between the male and female poles. A pi-
lot study with 21 participants (56% females, mean age = 22 ± 2.10)
was conducted to test that the adopted avatar face conveyed similar
degrees of perceived masculinity and femininity. Participants rated
slightly different versions of the androgynous face on a scale from 0
(completely masculine) to 10 (completely feminine). For the study,
we selected the avatar faces whose mean scores were not statisti-
cally different from the middle point of the scale, t(1, 20) = 1.60,
p= .13.
The avatars’ facial movements were then manipulated to gener-
ate four different facial expressions: neutrality, anger, disgust and
pain (see Fig. S2, see the online version at 10.1016/j.ej-
pain.2011.02.006). For an in-depth investigation of the hypothesis,
we also manipulated the intensity of each facial expression. The
stimuli differed according to the intensity level displayed: intensity
was categorized as low (corresponding levels of the built-in Face-
Gen 3.1 controls: 0.1, 0.2, 0.3), medium (0.4, 0.5, 0.6, 0.7) or high
(0.8, 0.9, 1). The crucial stimuli set consisted of three images per
condition, resulting in a total of 36 trials. We also included fillers
(12 male avatars and 12 female avatars, varied by expression and
intensity level) to increase the variability of the stimuli shown to
participants.
The number of ‘‘female’’ answers and the latency time were
subjected to a 3 (expression: pain, anger or disgust) 3 (intensity
level: low, medium or high) 2 (participant’s gender: male vs. fe-
male) ANOVA, with the first two factors varied within-subjects and
the last one between subjects. The factorial design was incomplete
because the neutral expression’s intensity was not manipulable
(the nine neutral faces did not differ by degree of expression
intensity).
4.1.3. Procedure
We employed a speeded dichotomous decision task. Respon-
dents were presented with a series of computer-generated avatar
faces displayed at the center of the screen of a 17-in. monitor, on
a uniform, black background with a 75-Hz refresh rate. The level
of brightness was consistent for every image. Participants viewed
the monitor from a distance of about 50 cm without head restraint.
They were asked to categorize each target as female or male by
pressing two corresponding keys on the computer keyboard using
their left and right forefingers, respectively. The stimuli were pre-
sented randomly, and participants were instructed to respond as
quickly and accurately as possible.
4.2. Results
4.2.1. Gender classification
The gender categorization of androgynous stimuli was ana-
lyzed. We assigned a score of 0 to male answers and 1 to female
answers.
The analysis yielded a significant main effect for expression,
F(1, 37) = 58.77, p< .001,
g
2
p
¼:61 (see Table 5). Supporting our pre-
diction, the lowest number of female answers was associated with
pain expressions (M= 0.37, SD = 0.71), followed by disgust
(M= 0.99, SD = 1.05) and anger expressions (M= 1.84, SD = 1.07).
The analysis also yielded a main effect for expression level of inten-
sity, F(1, 37) = 9.54, p< .01,
g
2
p
¼:20. Participants classified the
ambiguous stimuli as female more frequently at low levels of inten-
sity (M= 1.30, SD = 1.02) than at medium (M= 1.02, SD = 0.94) and
high levels of intensity (M= 0.89, SD = 0.88).
In the previous analysis, we did not consider the neutral expres-
sions because their intensity could not be manipulated. However,
we compared the total number of ‘‘female’’ answers by expression
type independent of intensity level using a 4-level within-subject
ANOVA (expression: neutral, anger, disgust or pain). The analysis
confirmed a significant effect for expression type, F(1, 38) = 88.99,
p< .001,
g
2
p
¼:70. The ambiguous stimuli showing neutral expres-
sions obtained the highest number of ‘‘female’’ responses
(M= 7.22, SD = 2.45) compared with the other three expressions,
p< .001 (see Table 5). The ‘‘female’’ responses to the other three
expressions remained significantly different.
The ANOVA did not reveal a main effect for participant gender,
F(1, 37) = 0.02, p= .87 or interaction effects between participant
gender and the other factors, Fs(1, 37) < 2.58, ps > .12.
These results replicate those of Experiments 1 and 2 and show
that decisions about a face’s gender and its emotional expressions
of pain are not independent. Therefore, we found that androgynous
faces displaying pain were considered more masculine than femi-
nine when the images were controlled for other negative emotions.
5. General discussion
The detection of pain in others is a highly salient event for
onlookers (Craig, 1992; Williams, 2002). However, this strong sal-
ience can be moderated by the target’s social characteristics. In-
deed, consistent with pain-related gender role stereotypes
(Hoffmann and Tarzian, 2001) early studies suggested that gender
can actually affect the observers’ explicit judgment of pain on fe-
male and male target (see Hirsh et al., 2008). Yet, past research
has also shown that explicit social judgment might diverge from
implicit, or reflexive, cognitive processes and associations (Green-
wald and Nosek, 2008; Sloman, 1996). Whereas explicit judgments
can be biased by a wide range of factors, such as response biases,
faking, motivation and opportunity (e.g., the motivation to control
prejudice) implicit processes tend to be more free of response fac-
tors such as social desirability and faking tendencies (Greenwald
et al., 2002). Moving from these premises, in the present study,
Table 5
Mean (standard deviations) of number of female categorizations by type of expression
and intensity level. Experiment 3.
Expression Intensity
SUM Low Medium High
Anger 5.52
a
(3.04) 1.95
a
(1.05) 1.75
a
(1.08) 1.82
a
(1.09)
Disgust 2.98
b
(2.68) 1.44
b
(1.16) 0.98
b
(1.00) 0.56
b
(0.99)
Pain 1.12
c
(1.93) 0.51
c
(0.85) 0.33
c
(0.72) 0.28
c
(0.56)
Neutral 7.22
d
(2.45)
Note: Different letters indicate in each column statistical differences among each
expression, according to the LSD test (p< .05).
P. Riva et al. / European Journal of Pain 15 (2011) 985.e1–985.e11 985.e5
we found that the target’s gender affected the observer’s reflexive
decoding of the facial expression of pain. More specifically, we
found that observers’ ability to detect pain in a female face was
lower than their ability to detect pain in male faces. This tendency
was consistent across three experiments and represented a unique
response to pain expressions compared with a baseline (e.g., neu-
tral expression) or other negative facial expressions (e.g., anger and
disgust).
Overall, the results showed that male pain faces were more eas-
ily processed at the reflexive level. In Experiment 1, participants
confused the pain expression with other expressions to a higher
degree in female faces than male faces. Experiment 2 showed that
participants took longer to correctly identify a dynamic pain
expression on a female face compared with a male face. Because
the dynamic stimuli showed facial expressions of increasing inten-
sity, a more intense pain expression was needed for participants to
correctly identify pain on a female face, whereas lower-intensity
pain expressions were sufficient to detect suffering on male faces.
Finally, Experiment 3 showed that participants tended to perceive
an androgynous face as less feminine when it displayed pain than
other negative emotions.
Although our experimental design did not allow us to identify
the underlying mechanism, several parallel and convergent expla-
nations might be pointed out. Further research should directly
investigate the processes that determine the higher salience of
male pain face at the implicit level.
First, we can speculate that the observers’ ability to decode pain
faces might depend upon gender differences in pain behavior,
which implies different exposures to male and female pain faces.
Indeed, research has reported significant gender differences in
the perception, experience and expression of pain. A large amount
of clinical literature indicates that women more frequently experi-
ence and are more intensely affected by a large number of pain
syndromes – both chronic and acute – and they describe the pain
as more intense, widespread, and of a greater duration than men
do (Dao and LeResche, 2000; Heitkemper and Jarrett, 2001; Morin
et al., 2000; Robinson et al., 1998; Rollman and Lautenbacher,
2001; Unruh, 1996). Experimental research has found that women
exhibit significantly greater sensitivity to pain and lower pain tol-
erance (Fillingim and Maixner, 1995; Wise et al., 2002). For that
reason, even though men and women do not seem to differ in their
facial configurations in response to pain stimuli (Kunz et al., 2006),
it might be that onlookers are more frequently exposed to facial
pain expressions on female rather than male faces precisely be-
cause women suffer more from a larger variety of pain syndromes.
Studies on pain judgment have reported the systematic tendency
of observers who are more exposed to patients in pain to underes-
timate the pain those patients experience (Marks and Sachar,
1973; Marquié et al., 2003). The commonly referenced socio-cogni-
tive process is thought to result from habituation; that is, the con-
sequence of being repeatedly exposed to others’ suffering might be
a diminished sensitivity to pain in others (Prkachin et al., 2007).
Habituation has received empirical support (Prkachin and Ro-
cha, 2010; Prkachin et al., 2004), although boundary conditions
have been pointed out (i.e., the observer’s accuracy can increase
after spending more time with the person in pain) (Miaskowski
et al., 1997). Taking into account the habituation bias, a greater
exposure to females’ pain expressions might lead to a decreased
sensitivity to pain displayed by women. Indeed, Experiments 1
and 2 showed that participants took longer and needed more in-
tense pain expressions to correctly identify pain on female com-
pared with male faces. Future studies should investigate whether
habituation underpins observers’ impaired ability to detect female
pain compared with male pain. Observers could be experimentally
exposed either to a male or a female face for a specific length of
time, similar to the design used by Prkachin and Rocha (2010),to
determine whether length of exposure produces different out-
comes. However, this possible explanation is inconsistent with
our findings related to anger detection under the control condition.
It is well known that people are more exposed to male expressions
of anger than female; according to the habituation hypothesis, we
would expect greater speed and accuracy in identifying anger on a
female face.
As a potential parallel to the habituation bias (e.g., decreased
sensitivity to female pain faces resulting from greater exposure
to them), top-down inhibition processes might have contributed
to observers’ impaired ability to decode female pain faces. Stereo-
types are known to affect the way observers perceive a target face
(Hugenberg and Bodenhausen, 2004; Hugenberg and Sacco, 2008).
The psychology literature has widely shown that stereotypes bias
the processing of potentially ambiguous information in a stereo-
type-consistent manner across multiple domains (Darley and
Gross, 1983; Duncan, 1976). Consistent with the notion that ex-
pected gender differences in facial expressivity can be prompted
by gender stereotypes, a recent theory of pain underestimation
posits that female gender is an invalidating factor in pain judg-
ment: given the stereotypical views of women as dramatizing,
observers’ certainty about a women’s pain might be reduced (Tait
et al., 2009). Accordingly, our study shows higher error rates in
perceiving pain expression in females than in males.
Further evidence of gender influences in observers’ ability to de-
code pain expressions comes from brain research. Simon et al.
(2006) found the neurological activation of the observer’s brain de-
pended upon the gender of the person expressing pain, but not
upon the observer’s gender. In their study, the authors provided
fMRI support for their hypothesis that the ‘‘implicit processing of
male pain expression triggers an emotional reaction characterized
by a threat-related response’’ (p. 309). Indeed, several activations
induced by male facial displays were significantly decreased when
observers watched female facial displays of pain. Observing male
actors expressing pain activated several areas known for a
threat-related response, like the ventromedial prefrontal cortex,
SII/posterior insula and anterior insula. However, several areas
activated by male facial displays – including somatosensory areas,
the amygdala and the perigenual ACC – registered a significant de-
crease when observers watched female facial displays of pain. The
authors argued that the interaction between the pattern of neural
activation and actor’s gender might be due to the social communi-
cative value of pain conveyed by male and female faces. A male
pain expression may be more strongly linked with potentially
threatening situations than a female pain expression (LeDoux,
2000), activating the fight-flight system in the brain of the obser-
ver. Indeed, previous scholars suggested that pain can foment a
disposition towards aggression (Berkowitz, 1993), and recent
empirical data confirmed that pain increases aggressive tempta-
tions in humans (Riva et al., 2010). Because men are physically lar-
ger and stronger and therefore more dangerous on average than
women, detecting a male in pain might pose a threat particularly
salient. Simon et al.’s (2008) results are consistent with our find-
ings that observers are selectively (compared to neutral and anger
expression) more accurate and faster at decoding pain displayed by
male faces. Moreover, Experiment 3 showed that under conditions
of ambiguity (e.g., androgynous faces), participants perceived the
pain expressions as more masculine. Again, this might be the con-
sequence of a pain detection system that is selectively biased to
detect any possibility of a relevant threat in the environment. Thus,
the tendency to perceive a gender-ambiguous pain expression as
masculine may rely upon the higher cost of failing to detect pain
in a male compared with a female. In other words, because the cost
of overdetecting pain in a male face might be lower than the cost of
a missing it, the observer’s judgment may be biased toward deci-
phering the expression of pain on an androgynous face as mascu-
985.e6 P. Riva et al. / European Journal of Pain 15 (2011) 985.e1–985.e11
line (for a discussion of the error management system that might
favor a bias for false alarms over misses, see Haselton and Buss,
2000).
The main advantages of the study were twofold. First, the use of
implicit methods allowed us to assess the involuntary effects of the
target’s gender on pain detection while bypassing social desirabil-
ity issues. Furthermore, in many naturalistic circumstances, pain
cues are unclear and treatment decisions have to be made under
conditions of uncertainty or time pressure. Under these conditions,
decision-makers rely on automatic, heuristic and implicit pro-
cesses rather than on more laborious reasoning (Tait et al., 2009).
Our methodological choice also accounted for seemingly conflict-
ing past results, which were based on explicit methods and indi-
cated that female faces are judged to express greater pain
intensity than male faces (Hirsh et al., 2008). These results are con-
sistent with explicit, stereotypical views that women are more
prone to express their pain (Hoffmann and Tarzian, 2001), yet they
can diverge from the implicit and automatic cognitive processes
involved in pain decoding. Second, the use of computer-generated
avatars allowed us to create stimuli with identical facial expres-
sions but different genders. However, the use of computer-gener-
ated faces might limit the generalizability of the current study’s
findings; although they allowed us to achieve greater experimental
control, their use might have limited the external validity of the
study. Thus, these findings should be interpreted with caution be-
cause we do not yet know the degree to which findings from com-
puterized stimuli are generalizable to human faces. Further
replication with different stimuli and samples is warranted. More-
over, the influence of other social factors, such as age, race and eth-
nicity, should also be investigated.
With regard to observers’ gender, neither main nor interaction
effects have been consistently found during the three experiments.
The only analysis that showed influence of observers’ gender re-
vealed that female participants were faster than males in decoding
anger, disgust and neutral expression. These results are in line with
the notion that female are generally better than males at basic face
perception (McBain et al., 2009; McClure, 2000). However, in line
with previous research (Simon et al., 2006), we did not find any
interaction between observers’ gender and the specific expression
of pain. Nevertheless, failures to find any consistent effect related
to the participants’ sex might be due to a lack of power from an
insufficient number of subjects or to the uneven distribution of
participants’ sex in our samples.
In terms of clinical implications, our findings are in keeping
with evidence that women are less likely to receive treatment for
pain than men (Calderone, 1990; McDonald, 1994; Hoffmann and
Tarzian, 2001). In this sense, observers’ impaired ability to detect
the female pain face could lead them to disregard the sufferer’s
experience and needs and fail to provide adequate care.
6. Conflict of Interest
The authors have no conflict of interest with respect to this
article.
Acknowledgements
We would like to thank Eric D. Wesselmann for comments on
earlier versions of this paper. We also thank Shannon Lapsley
and Shajuan Jackson for proof reading the article.
Appendix A. Supplementary material
Supplementary data associated with this article can be found, in
the online version, at doi:10.1016/j.ejpain.2011.02.006.
References
Berkowitz L. Aggression: its causes, consequences, and control. New York: McGraw-
Hill; 1993.
Calderone JL. The influence of gender on the frequency of pain and sedative
medication administered to postoperative patients. Sex Roles 1990;23:713–25.
Craig KD. The facial expression of pain: better than a thousand words? J Pain
1992;1:153–62.
Dao TT, LeResche L. Gender differences in pain. J Orofacial Pain 2000;14:169–84.
Darley JM, Gross PH. A hypothesis-confirming bias in labeling effects. J Pers Soc
Psychol 1983;44:20–33.
Darwin C. The expression of the emotions in man and animals. New
York: Philosophical Library; 1872.
Duncan BL. Differential social perception and attribution of intergroup violence:
testing the lower limits of stereotyping blacks. J Pers Soc Psychol
1976;34:590–8.
Ekman P, Friesen WV. Facial action coding system: a technique for the
measurement of facial movement. Palo Alto, Calif.: Consulting Psychologists
Press; 1978.
Fillingim RB, Maixner W. Gender differences in the responses to noxious stimuli.
Pain Forum 1995;4:209–21.
Green CR, Anderson KO, Baker TA, Campbell LC, Decker S, Fillingim RB, et al. The
unequal burden of pain: confronting racial and ethnic disparities in pain. Pain
Med 2003;4:277–94.
Greenwald AG, Banaji MR, Rudman LA, Farnham SD, Nosek BA, Mellott DS. A unified
theory of implicit attitudes, stereotypes, self-esteem, and self-concept. Psychol
Rev 2002;109:3–25.
Greenwald AG, Nosek BA. Attitudinal dissociation: what does it mean? In: Petty R,
Fazio RH, Briñol P, editors. Attitudes: insights from the new implicit
measures. Hillsdale, NJ: Lawrence Erlbaum Associates; 2008. p. 65–82.
Hadjistavropoulos T. Assessing pain in older persons with severe limitations in
ability to communicate. In: Gibson S, Weiner D, editors. Pain in older
persons. Seattle: IASP Press; 2005. p. 135–51.
Hadjistavropoulos T, Craig KD. Social influences and the communication of pain. In:
Hadjistavropoulos T, Craig KD, editors. Pain: psychological perspectives. New
York: Erlbaum; 2004. p. 87–112.
Haselton MG, Buss DM. Error management theory: a new perspective on biases in
cross-sex mind reading. J Pers Soc Psychol 2000;78:81–91.
Heitkemper MM, Jarrett M. Gender differences and hormonal modulation in visceral
pain. Curr Pain Headache Rep 2001;5:35–43.
Hirsh AT, Alqudah AF, Stutts LA, Robinson ME. Virtual human technology: capturing
sex, race, and age influences in individual pain decision policies. Pain
2008;140(1):231–8.
Hirsh AT, George SZ, Robinson ME. Pain assessment and treatment disparities: a
virtual human technology investigation. Pain 2009;143(1–2):106–13.
Hoffmann DE, Tarzian AJ. The girl who cried pain: a bias against women in the
treatment of pain. J Law Med Ethics 2001;29:13–27.
Horgas AL, Elliott AF. Pain assessment and management in persons with dementia.
Nurs Clin North Am 2004;39:593–606.
Hugenberg K, Bodenhausen GV. Ambiguity in social categorization. The role of
prejudice and facial affect in race categorization. Psychol Sci 2004;15(5):342–5.
Hugenberg K, Sacco DF. Social categorization and stereotyping: how social
categorization biases person perception and face memory. Soc Pers Psych
Compass 2008;2(2):1052–72.
Kappesser J, Williams ACdC. Pain and negative emotions in the face: judgments by
health care professionals. Pain 2002;99(1–2):197–206.
Kunz M, Gruber A, Lautenbacher S. Sex differences in facial encoding of pain. J Pain
2006;7(12):915–28.
Kunz M, Scharmann S, Hemmeter U, Schepelmann K, Lautenbacher S. The facial
expression of pain in patients with dementia. Pain 2007;133(1–3):221–8.
LeDoux J. Cognitive–emotional interactions: listen to the brain. In: Lane RD, Nadel L,
editors. Cognitive neuroscience of emotion. New York: Oxford University Press;
2000. p. 129–55.
Marks RM, Sachar EJ. Undertreatment of medical inpatients with narcotic
analgesics. Ann Intern Med 1973;78:173–81.
Marquié L, Raufaste E, Lauque D, Mariné C, Ecoiffier M, Sorum P. Pain rating by
patients and physicians: evidence of systematic pain miscalibration. Pain
2003;102:289–96.
McBain R, Norton D, Chen Y. Females excel at basic face perception. Acta Psychol
2009;130(2):168–73.
McClure EB. A meta-analytic review of sex differences in facial expression
processing and their development in infants, children, and adolescents.
Psychol Bull 2000;126:424–53.
McDonald DD. Gender and ethnic stereotyping and narcotic analgesic
administration. Res Nurs Health 1994;14:45–9.
Miaskowski C, Zimmer EF, Barrett KM, Dibble SL, Wallhagen M. Differences in
patients’ and family caregivers’ perceptions of the pain experience influence
patient and caregiver outcomes. Pain 1997;72:217–26.
Morin C, Lund JP, Villarroel T, Clokie CML, Feine JS. Differences between the sexes in
post-surgical pain. Pain 2000;85:79–85.
Prkachin KM. Assessing pain by facial expression: facial expression as nexus. Pain
Res Manage 2009;14(1):53–8.
Prkachin KM, Berzins S, Mercer SR. Encoding and decoding of pain expressions: a
judgment study. Pain 1994;58(2):253–9.
Prkachin KM, Craig KD. Expressing pain: the communication and interpretation of
facial pain signals. J Nonverbal Behav 1994;19:191–205.
P. Riva et al. / European Journal of Pain 15 (2011) 985.e1–985.e11 985.e7
Prkachin KM, Mass H, Mercer SR. Effects of exposure on perception of pain
expression. Pain 2004;111(1–2):8–12.
Prkachin KM, Rocha EM. High levels of vicarious exposure bias pain judgments. J
Pain 2010;11(9):904–9.
Prkachin KM, Solomon PE. The structure, reliability and validity of pain
expression: Evidence from patients with shoulder pain. Pain 2008;139(2):
267–74.
Prkachin KM, Solomon PE, Ross J. Underestimation of pain by health-care providers:
towards a model of the process of inferring pain in others. Can J Nurs Res
2007;39:88–106.
Riva P, Wirth JH, Williams KD. The social impact of suffering: physical pain thwarts
social needs. In: Presented at the annual meeting of the midwestern
psychological association, Chicago, IL, May 2010.
Robinson ME, Wise EA. Prior pain experience: influence on the observation of
experimental pain in men and women. J Pain 2004;5:264–9.
Robinson ME, Wise EA, Riley JL, Atchison JA. Sex differences in clinical pain: a multi-
sample study. J Clin Psych Med Settings 1998;5(4):413–24.
Rollman GB, Lautenbacher S. Sex differences in musculoskeletal pain. Clin J Pain
2001;17:20–4.
Simon D, Craig KD, Gosselin F, Belin P, Rainville P. Recognition and
discrimination of prototypical dynamic expressions of pain and emotions.
Pain 2008;135(1–2):55–64.
Simon D, Craig KD, Miltner WHR, Rainville P. Brain responses to dynamic facial
expressions of pain. Pain 2006;126:309–18.
Schneider W, Eschman A, Zuccolotto A. E–Prime reference guide. Pittsburgh,
PA: Psychology Software Tools Inc.; 2002.
Sloman SA. The empirical case for two systems of reasoning. Psychological Bulletin
1996;119:3–22.
Tait RC, Chibnall JT, Kalauokalani D. Provider judgments of patients in pain: seeking
symptom certainty. Pain Med 2009;10:11–34.
Unruh AM. Gender variations in clinical pain experience. Pain 1996;65:123–67.
Williams ACdC. Facial expression of pain: an evolutionary account. Behav Brain Sci
2002;25:439–88.
Wise EA, Price DD, Myers CD, Heft MW, Robinson ME. Gender role expectations of
pain: relationship to experimental pain perception. Pain 2002;96(3):335–42.
Yamada M, Decety J. Unconscious affective processing and empathy: an
investigation of subliminal priming on the detection of painful facial
expressions. Pain 2009;143:71–5.
985.e8 P. Riva et al. / European Journal of Pain 15 (2011) 985.e1–985.e11
Table S1
Frequencies and percentages (in parenthesis) of hits and errors the participants made
in Experiment 2.
Overall Male target Female target
Pain
Total number of responses 220 111 109
Correct answers 133 (60.4%) 68 (61.2%) 65 (59.6%)
Judged as anger 50 (22.7%) 29 (26.1%) 21 (19.2%)
Judged as disgust 37 (16.8%) 14 (12.6%) 23 (21.1%)
Anger
Total number of responses 223 112 111
Correct answers 184 (82.5%) 94 (83.9%) 90 (81.0%)
Judged as pain 20 (8.9%) 9 (8.0%) 11 (9.9%)
Judged as disgust 19 (8.5%) 9 (8.0%) 10 (9.0%)
Disgust
Total number of responses 220 110 110
Correct answers 133 (60.4%) 67 (60.9%) 66 (60%)
Judged as anger 61 (27.7%) 33 (30.0%) 28 (25.4%)
Judged as pain 26 (11.8%) 10 (9.0%) 16 (14.5%)
P. Riva et al. / European Journal of Pain 15 (2011) 985.e1–985.e11 985.e9
Fig. S1. Examples of female and male stimuli used in Experiments 1 and 2 to manipulate each facial expression (from the top: neutral, anger, disgust and pain expression).
985.e10 P. Riva et al. / European Journal of Pain 15 (2011) 985.e1–985.e11
Fig. S2. The four androgynous faces adopted in Experiment 3.
P. Riva et al. / European Journal of Pain 15 (2011) 985.e1–985.e11 985.e11