Ò

PAIN 151 (2010) 687–693

www.elsevier.com/locate/pain

Knowing you care: Effects of perceived empathy and attachment style on

pain perception
Chiara F. Sambo a, Matthew Howard a, Michael Kopelman a, Steven Williams a, Aikaterini Fotopoulou a,b,⇑
a
King’s College London, Institute of Psychiatry, UK
b
Institute of Cognitive Neuroscience, University College London, UK

a r t i c l e i n f o a b s t r a c t

Article history: Other people can have a significant impact on one’s pain. Although correlational data abound, causal
Received 19 April 2010 relationships between one’s pain experience, individual traits of social relating (e.g. attachment style),
Received in revised form 4 August 2010 and social factors (e.g. empathy) have not been investigated. Here, we studied whether the presence of
Accepted 20 August 2010
others and ‘perceived empathy’ (defined as participants’ knowledge of the extent to which observers felt
they understood and shared their pain) can modulate subjective and autonomic responses to pain; and
whether these influences can be explained by individual traits of pain coping and social attachment. Par-
Keywords:
ticipants received noxious thermal stimuli via a thermode attached to their forearm and were asked to
Empathy
Social support
rate their pain. In separate blocks they were witnessed by (a) high-empathic and (b) low-empathic unfa-
Attachment miliar observers, and in a third condition (c) no observer was present (alone condition). We found that
Coping the effects of social presence and empathy on pain ratings depended on individual differences in attach-
ment style. Higher scores on attachment anxiety predicted higher pain ratings in the low-empathy than
in the high-empathy condition; and higher scores on attachment avoidance predicted lower pain ratings
in the alone condition than with social presence. In addition, social presence decreased autonomic
responses to pain irrespective of individual personality traits. To our knowledge this is the first time that
adult attachment style has been shown to modulate the effects of social presence and ‘perceived empa-
thy’ on experimentally induced pain. The results are discussed in relation to recent cognitive models of
pain coping and attachment theory.
Ó 2010 International Association for the Study of Pain. Published by Elsevier B.V. All rights reserved.

1. Introduction noxious stimuli; [34]) is a way of eliciting empathic responses from

There has been a rapid progress recently in elucidating the
psychological and neural mechanisms of empathy, including influ-
ential studies on empathy for pain (i.e. how an ‘observer’ under-
stands the pain of a ‘sufferer’; e.g. [42]). On the contrary, few
experimental investigations explore the role of empathy on pain
(i.e. how an observer’s empathy can affect the pain of another indi-
vidual). This is unfortunate because an abundance of clinical, cor-
relational studies have concluded that psychosocial factors,
including empathy, influence pain [1,48]. For instance, the behav-
iour of spouses can determine pain-related behaviours in chronic
pain sufferers [15]. Recent reviews have argued that initial behav-
ioural [16] and social modelling [12] theories were insufficient to
capture the complex psychosocial variables that influence pain
[7,37,46]. Alternative, cognitive-behavioural models have recently
emerged. For example, the communal coping model [46] maintains
that ‘catastrophizing’ (the tendency to focus on the threat value of
⇑ Corresponding author at: Psychology Department, Institute of Psychiatry, King’s
College London, 5th Floor, Bermondsey Wing, Guy’s Campus, London SE1 9RT, UK.
Tel.: +44 20 7188 0183; fax: +44 20 7188 0184.
E-mail address: a.fotopoulou@kcl.ac.uk (A. Fotopoulou).
others and achieving relational goals, beyond pain reduction.
A related theoretical development is the suggestion that attach-
ment theory [4] may be a useful model for understanding individ-
ual differences in pain [17,33]. Attachment theory supports that
children develop mental schemas of themselves based on their
caregivers’ responses, particularly during encounters with threat.
These ‘attachment styles’ tend to work as life-long templates for
social interactions. It has been consistently shown that insecure
attachment styles (characterised by increased worry over the
responsiveness of others, or alternatively by increased discomfort
with interdependence) are important predictors of pain report
[39]. However, no study has examined the interaction between
attachment style and the ‘on-line’ social modulation of pain.
More generally, in the few existing experimental investigations
on the social modulation of pain, methodological limitations and
differences in operationalising rich psychosocial concepts have
produced conflicting results (e.g. [5,30]). For instance, whereas
some studies have operationalised ‘social support’ as the simple
presence of familiar and unfamiliar others (e.g. [30,45,50,51]),
other studies have collapsed different behaviours (e.g. empathic
comments, reassurance, criticism) into single categories, such as
‘active support’ [5] and ‘pain-promoting’ [6]. In fact, social support

0304-3959/$36.00 Ó 2010 International Association for the Study of Pain. Published by Elsevier B.V. All rights reserved.
doi:10.1016/j.pain.2010.08.035
 Ò
688 C.F. Sambo et al. / PAIN 151 (2010) 687–693
is a multifaceted construct, comprising several dimensions [3],
including received support (tangible help actually provided) and
perceived support (subjective evaluations of support). These facets
can have different effects in clinical settings (e.g. patients’ pain is
more closely related with perceived than with received spouse re-
sponses [15,38]). However, the effects of perceived social attitudes
on pain have not been examined in experimental studies.
To overcome the aforementioned limitations, we conducted a
study on healthy volunteers that manipulated ‘social presence’ (de-
fined as observation by unfamiliar others) and ‘perceived empathy’
(defined as knowledge of the extent to which ‘observers’ felt they
understood and shared participants’ pain) during painful thermal
stimulation. We predicted that (1) the presence versus absence
of ‘observers’, and (2) their ‘high’ versus ‘low’ empathy would lead
to decreased pain ratings and physiological responses; and (3) that
individual differences in attachment style and ‘catastrophizing’
would modulate these effects.
2. Materials and methods
2.1. Participants
Thirty healthy participants (20 females and 10 males) aged be-
tween 21 and 50 years (mean = 29.1; SD = 7.3) participated in the
study. The study conformed to the principles of the Declaration
of Helsinki and the IASP’s (International Association for the Study
of Pain) guidelines and was approved by the University’s Ethics
Committee. All participants gave written informed consent. Partic-
ipants were excluded if they had previous psychiatric or neurolog-
ical history, suffered from chronic pain disorders, had a history of
substance misuse, or drank more than 28 units of alcohol per week.
Participants were required not to use painkillers on the day of the
testing, to have no more than one-caffeinated drink and no more
than one cigarette in the 6 h previous to the testing, and to refrain
from alcohol in the 24 h prior to the testing.
2.2. Study design
Social context (‘High-Empathy’, ‘Low-Empathy’, and ‘Alone’
conditions) and thermal stimulus intensity (‘Low’, ‘Medium’, and
‘High’) were manipulated in a 3  3 repeated-measures design.
Social context was manipulated by the presence of an unfamiliar
observer and false feedback of the observers’ empathy levels for
the two ‘perceived empathy’ conditions (‘High-Empathy’ and
‘Low-Empathy’) and by the absence of the unfamiliar observer
for the ‘Alone’ condition (see Section 2.4).
Participants received noxious thermal stimuli while in the pres-
ence of a ‘High-Empathy’ observer, a ‘Low-Empathy’ observer, and
in an ‘Alone’ condition (i.e. while they were alone in the testing
room). Based on pilot testing, thermal stimulus intensity was
manipulated in low, medium and high levels (tailored individually;
see Section 2.4).
2.3. Thermal noxious stimuli
Thermal stimuli were administered using a contact thermode
probe (Contact Heat Evoked Potentials Stimulator, CHEPS; Medoc
Ltd., Ramat Yishai, Israel) (thermode area of 573 mm2; 27 mm
diameter). The thermode probe was attached to the volar surface
of the participants’ forearm by Velcro straps. As previous studies
have shown that thermal pain thresholds and tolerance do not dif-
fer between forearm sites (e.g. [41]), the thermode was moved sys-
tematically along the arm by approximately one thermode length
(3 cm; so that it was always applied to the same two dermatomes,
T1 and C5; [36]) between experimental runs to minimize sensitiza-
tion and habituation effects. Each stimulus had a plateau time of
1.5 s, and a heating/cooling rate of 25 °C/s, leading to a total stim-
ulus duration of approximately 3 s. The temperatures of thermal
stimuli in the experimental blocks varied depending on the partic-
ipants’ highest pain ratings in pre-experimental procedures (see
below). The baseline temperature of the thermode was maintained
constant at 32 °C.
2.4. Procedures
The study took place in a mock scanner environment, to facili-
tate future functional neuroimaging studies. Upon their arrival,
participants were familiarized with the environment, equipment,
procedures and thermal stimulation, receiving two 10-s example
trials of thermal stimulation at 38 and 40 °C.
2.4.1. Pre-experimental procedures
Individual pain levels estimation (‘Thresholding’): Three ascending
series of successive thermal stimuli (1.5 s stimulus duration) were
administered to each participant. Participants rated their pain
intensity after each stimulus using a 100-point visual analogue
scale (VAS) anchored with ‘‘no pain” at the left end and ‘‘worst pain
imaginable” at the right end. Each series was discontinued at 50 °C,
or when the participants provided a rating corresponding to 80 or
above. The three series were delivered alternatively on both the
forearms of the participants (starting on the right forearm) (e.g.
[53]). The first series started at 41 °C, and the remaining series
started 4 °C below the highest temperature of their previous series.
Successive stimuli increased by 1 °C (first and second series) and
0.5 °C (third series), and the inter-stimulus interval was 7 s. The
highest temperature of the third series was used to determine
the individual temperatures (low, medium, and high) in the
experimental blocks, in that for each participant temperatures
corresponding to 60%, 75%, and 90% of this highest temperature
were used for the ‘Low’, ‘Medium’, and ‘High’ levels, respectively.
Participants’ mean ‘highest temperature’ measure was 48.5 °C
(SD = 1.2 °C, range = 46–50 °C). A one-way ANOVA including the
factor Gender showed that these did not differ between women
and men (F[1,29] = 0.34, p = 0.56).
False feedback of perceived empathy: Prior to ‘thresholding’, two
‘observers’ of the opposite sex to each participant were introduced
as research collaborators. During ‘thresholding’, they were standing
on the right of the scanner bed where the participants were lying, at
a distance of one meter and outside of the participants’ view. The
following ‘experimental script’ was used to support the false feed-
back that participants were given later about the observers’ empa-
thy levels (see Section 2.4.2 below): Participants were told that
one aim of the study was to examine the nature of empathy and that
observers were asked to try ‘‘to understand and share the partici-
pants’ pain” (i.e. empathize) and to rate their level of empathy after
each of the three series of thermal stimuli using an 11-point scale
(0 = No empathy to 10 = Maximum empathy). Observers could see
the temperatures of thermal stimuli on a computer monitor, but
not the participants’ ratings. To minimize social desirability bias this
fact was especially stressed to the participants. Eight observers in
total were used, four men and four women, matched for age to the
participants. All observers acted as both ‘‘High-Empathy” and
‘‘Low-Empathy” observers across participants with equal fre-
quency. After the experiment, relevant traits of each observer were
rated by each participant (see Section 2.5.4 below).
2.4.2. Main task procedures
Following ‘thresholding’ and immediately prior to each of the
two ‘Empathy’ experimental conditions while the observers were
not present in the room the experimenter informed the partici-
pants that the observer in the forthcoming condition had reported
‘high’ or ‘low’ empathy for them during ‘thresholding’ (‘false feed-
 Ò
C.F. Sambo et al. / PAIN 151 (2010) 687–693
back’). During these conditions, the corresponding observer was
standing next to the scanner bed, as during ‘thresholding’. Each
observer was again able to see the stimulation levels, but not the
participants’ ratings; this fact was again stressed to participants.
Observers were now asked to rate their overall empathy for each
level of thermal stimuli at the end of each experimental block. Par-
ticipants and observers were not allowed to directly interact in any
way during the task. The experimenter remained outside the test-
ing room during each experimental block, but she ensured that the
procedures were followed via the available visual and audio obser-
vation equipment.
Experimental blocks: The order of the experimental conditions
(‘High-Empathy’, ‘Low-Empathy’, and ‘Alone’) was counterbal-
anced across participants. Each experimental condition consisted
of one block of 45 trials, including three ‘‘buffer” trials at the start
of each block (not included in the analysis); 12 trials per each stim-
ulus intensity (‘Low’, ‘Medium’, and ‘High’) (i.e. 36 trials in total);
and six ‘catch’ trials with ‘‘null events” in which no stimulus was
delivered (to control for any response bias). The order of trials
was pseudo-randomized, in that stimuli of the same temperature
level were not delivered in more than two consecutive trials.
Trial sequence (see Fig. 1 for a schematic representation): Each
trial started with a white fixation cross presented at the centre of
the screen for 1–4 s, followed by a green cross (3 s). Participants
had to maintain fixation on the cross. The onset of the thermal
stimulus was synchronized with the onset of green cross. After
each thermal stimulus participants had to rate their pain intensity
(50% of the trials) and unpleasantness (50% of the trials) on a VAS
(see below), which stayed on the screen for 5 s.
2.5. Measures
2.5.1. Pain intensity and unpleasantness
Participants rated their pain intensity and unpleasantness using
separate visual analogue scales (VAS). The VAS for pain intensity
and unpleasantness were anchored with ‘‘no pain” (or ‘‘not
unpleasant”, respectively) at the left end and ‘‘worst pain imagin-
able” (or ‘‘most unpleasant imaginable”, respectively) at the right
Fig. 1. Schematic representation of the sequence of events in each trial. At the beginning of each trial, a white fixation cross appeared at the centre of the screen for 1–4 s. This
was replaced with a green fixation cross of the same size for 3 s. The onset of the thermal stimulus was synchronized with the onset of the green cross. Thermal stimuli had a
plateau of 1.5 s after reaching the target temperature. Following that, a VAS replaced the fixation cross and stayed on the screen for 5 s. The inter-stimulus interval (ISI) was
6–9 s. The total duration of one trial was 9–12 s.
689
end. Participants were instructed to move the joystick-controlled
cursor on the VAS to indicate the amount of pain that they experi-
enced. The VAS ratings were then computed as discrete values
between 0 (left end of the scale) and 100 (right end).
2.5.2. Psychophysiological measures
Skin conductance (SC) activity was recorded continuously,
using two silver chloride electrodes attached to the palmar surface
of the index and the middle fingers of the participant’s right hand.
The electrodes were connected to a PSYLAB SC5 25-bit digital
amplifier. Data were analyzed using locally developed software.
SC responses (SCR) to heat stimuli were computed as the mean
SC activity in the 6-s window following stimulus onset. The inter-
val between successive measurement windows was 3–7 s. All SCR
were square-root transformed prior to statistical analyses to nor-
malize the distribution of the data [26].
Heart rate (HR) was recorded continuously, using a pulse oxim-
eter (Nonin 8600FO, Nonin Medical, Plymouth, MN, USA), with the
pulse probe attached to the ring finger of the participant’s right
hand. Interbeat intervals were transformed off-line in second-by-
second beats per minute (bpm), using locally developed software.
Mean HR responses (HRR) to heat stimuli were computed by aver-
aging the bpm values in the 6-s window following stimulus onset.
The interval between successive measurement windows was 3–7 s.
In addition, the average of bpm values in the 1-s interval preceding
stimulus onset was calculated as a measure of pain anticipation. All
HRR were square-root transformed prior to statistical analyses to
achieve statistical normality.
2.5.3. Self-report psychometric measures
All participants completed the following relevant psychometric
questionnaires: The 40-item State-Trait Anxiety Inventory (STAI)
[44], measuring ‘‘state” and ‘‘trait” anxiety in adults, was com-
pleted before the ‘thresholding’ procedure. All other questionnaires
described below were completed after the experimental conditions:
The 30-item Relationship Scale Questionnaire (RSQ) [18] assesses
adult attachment styles. A recent influential meta-analysis recom-
mends that the scores are used to obtain measures of two major
 Ò
690 C.F. Sambo et al. / PAIN 151 (2010) 687–693
dimensions of attachment style, namely anxiety and avoidance [23].
The Pain Vigilance and Awareness Questionnaire (PVAQ) [31] includes
16 items assessing sensitivity to pain and attention to changes in
pain. The Coping Strategies Questionnaire (CSQ) [40] contains 50
statements about behaviours and thoughts that people may use
to deal with pain including coping strategies (44 items) and
catastrophizing (CSQ-Cat; 6 items).
2.5.4. Post-task ratings
We designed a questionnaire with two sets of questions, one
investigating the participant’s experience of each experimental
block (i.e. five items per block) and one investigating their experi-
ence of the two observers (i.e. three items for each observer). The
items were: ‘‘How much pain did you experience overall?”;
‘‘How much distraction did you experience from (a) the experi-
mental environment, (b) your own thoughts, and (c) the presence
of a person in the room (if any)?”; ‘‘How much distress did you
experience?”; ‘‘How much reassurance did you experience from
each of the research collaborators?”; ‘‘How empathic do you be-
lieve the research collaborators are?”; ‘‘Please rate attractiveness
of the research collaborators”. Participants were required to rate
each item on an 11-point scale (0 = Not at all to 10 = Maximum
extent).
2.5.5. Statistical analyses
Two main types of analyses were performed, each separately for
VAS ratings, post-stimulus SCR, and pre- and post-stimulus HRR:
(1) repeated-measures ANOVAs and (2) multiple regression analyses.
(1) A repeated-measures ANOVA for VAS ratings included the
within-subject factors Social Context (‘High-Empathy’, ‘Low-
Empathy’, and ‘Alone’), Stimulus intensity (‘High’, ‘Medium’,
and ‘Low’) and VAS Type (‘Intensity’ and ‘Unpleasantness’).
(2) Multiple linear regression was used to examine the extent to
which demographic (age, gender) and psychometric variables
predicted the effects of social context on pain. Analyses were
conducted on differential measures of social context, sepa-
rately for VAS, SCR, and HRR. To assess the effect of ‘‘Social
Presence” on VAS ratings, SCR, and HRR, we averaged individ-
ual scores in the ‘Low-Empathy’ and the ‘High-Empathy’ con-
ditions (i.e. ‘presence’ conditions) and we subtracted these
from the ‘alone’ condition. Similarly, to assess the effect of
‘‘Perceived Empathy” on VAS, SCR, and HRR, we subtracted
the ‘High-Empathy’ from the ‘Low-Empathy’ condition.
Demographic and psychometric variables were used as
regressors. Because three measures of anxiety were initially
used, we assessed their reciprocal correlation. Trait anxiety
was positively correlated with state anxiety (r = 0.37,
p < 0.05) and attachment anxiety (r = 0.45, p < 0.05), and
Fig. 2. Mean VAS ratings (a), SCR (b), and HRR (c), and their respective standard errors are shown for the three experimental conditions (HE, ‘High-Empathy’; LE, ‘Low-
Empathy; A, ‘Alone’). The respective units of measure are indicated in brackets.
hence trait anxiety scores were not included in the regression
analyses. In all regression analyses, we used a stepwise proce-
dure to identify the variables that significantly contributed to
the effects.
3. Results
3.1. Repeated-measures ANOVAs
Participants’ average scores for VAS, post-stimulus SCR, and
pre- and post-stimulus HRR are shown in Fig. 2. The SCR and
HRR data of two participants were not obtained due to a recording
failure. As expected, ‘‘null events” in the experimental conditions
were rated very low on the VAS (M = 2.3/100, SD = 1.6, range
= 0–4). ‘‘Null events” data were not analyzed further.
VAS ratings: Mean VAS ratings (and standard deviations), aver-
aged across stimulus intensity and VAS type, were as follows:
M = 48.61/100 (SD = 15.73) in the ‘High-Empathy’ condition;
M = 47.63/100 (SD = 14.77) in the ‘Low-Empathy’ condition; and
M = 47.77/100 (SD = 14.76) in the ‘Alone’ condition. In the re-
peated-measures ANOVA, the main effects of Social Context and
VAS Type were not significant (F[2,58] = 0.11, p = 0.89, and
F[1,29] = 0.031, p = 0.86, respectively). The main effect of Stimulus
Intensity was significant (F[1.2,35.3] = 101.30, p < 0.01), with post-
hoc tests (a = 0.017, Bonferroni corrected) indicating significantly
lower pain ratings for the ‘Low’ vs. ‘Medium’ (t[29] = 9.62,
p < 0.017 Bonferroni corrected), ‘Low’ vs. ‘High’ (t[29] = 10.72,
p < 0.017), and ‘Medium’ vs. ‘High’ (t[29] = 8.82, p < 0.017) inten-
sity, suggesting that participants’ pain levels were higher for higher
stimulus temperatures, as we would expect. The interactions be-
tween the factor Social Context and the factors Stimulus Intensity
and VAS Type, respectively, were not significant (F[4, 116] = 0.72,
p = 0.57, and F[1.6,48.7] = 2.21, p = 0.12, respectively). The addition
of the between-factor Gender in the ANOVA did not lead to any sig-
nificant interactions between Gender and any of the factors (all
F 6 2.24, all p P 0.13).
Post-stimulus SCR: A repeated-measures ANOVA conducted on
mean SCR revealed a significant main effect of Social Context
(F[1.6,43.8] = 4.46, p < 0.05), with post-hoc tests (a = 0.017, Bonfer-
roni corrected) showing that mean SCR were lower in the
‘High-Empathy’ and ‘Low-Empathy’ conditions than in the ‘Alone’
condition (t[27] = 2.14, p < 0.017, and t[27] = 2.45, p < 0.017,
respectively), while mean SCR in the ‘High-Empathy’ and ‘Low-
Empathy’ conditions did not differ statistically (t[27] = 0.52,
p = 0.60). The main effect of Stimulus Intensity and the interaction
between Social Context and Stimulus Intensity were not significant
(F[1,28.2] = 0.23, p = 0.64, and F[1.5,40.62] = 0.57, p = 0.52, respectively).
Post-stimulus HRR: In a repeated-measures ANOVA performed
on mean HRR, a main effect of Social Context was found (F[1.3,35.4]
 Ò
C.F. Sambo et al. / PAIN 151 (2010) 687–693
= 6.26, p < 0.05). Post-hoc tests (a = 0.017, Bonferroni corrected)
revealed that mean HRR were significantly lower in the ‘High-
Empathy’ and ‘Low-Empathy’ conditions than in the ‘Alone’ condi-
tion (t[27] = 2.83, p < 0.017 for the ‘High-Empathy’ to ‘Alone’
comparison, and t[27] = 2.40, p < 0.017, for the ‘Low-Empathy’ to
‘Alone’ comparison), while the difference between mean HRR in
the ‘High-Empathy’ and the ‘Low-Empathy’ conditions was not sig-
nificant (t[27] = 1.4, p = 0.16). A main effect of Stimulus Intensity
was also found (F[2,52] = 8.71, p < 0.05), and post-hoc tests
(a = 0.017, Bonferroni corrected) showed a significant difference
between mean HRR for ‘Low’ and ‘High’ (t[27] = 2.88, p < 0.017),
and ‘Medium’ compared to ‘High’ intensity (t[27] = 2.67,
p < 0.017), but not for ‘Low’ compared to ‘Medium’ intensity
(t[27] = 0.56, p = 0.57).
Pre-stimulus HRR: An additional repeated-measures ANOVA was
conducted in the one-second interval preceding heat stimulus on-
set (i.e. associated to anticipation of pain) with the factor Social
context. The effect of Social Context was significant
(F[1.5,40.7] = 9.85, p < 0.01), and post-hoc tests (a = 0.017, Bonferroni
corrected) indicated that in this pre-stimulus interval HRR were
lower when participants experienced pain in the ‘High-Empathy’
and in the ‘Low-Empathy’ conditions than in the ‘Alone’ condition
(t[27] = 3.53, p < 0.017, and t[27] = 2.84, p < 0.017, respectively),
while no significant difference was found between the ‘High-
Empathy’ and the ‘Low-Empathy’ conditions (t[27] = 2.26,
p = 0.031, ns at a = 0.017).
3.2. Multiple regression analysis
Separate multiple regression analyses were performed to assess
the effect of Social Presence (scores of the ‘Alone’ condition minus
the mean of the ‘Presence’ conditions) and Empathy (scores of
‘Low’ minus ‘High’ empathy conditions) on VAS ratings, SCR, and
HRR. Demographic (age, gender) and psychometric variables (state
anxiety, attachment anxiety, attachment avoidance, pain vigilance,
coping, and catastrophizing) were entered as regressors in the
model.
VAS ratings: Descriptive scores of attachment style and other
psychometric measures are shown in Table 1. The only predictor
of effects of Social Presence on VAS ratings was attachment avoid-
ance (F[1,28] = 5.21, p < 0.05, R2 = .16; B = .40, t = 2.28, p < 0.05).
Higher scores on attachment avoidance predicted lower pain rat-
ings in the ‘Alone’ condition than in the presence of an observer.
On the other hand, attachment anxiety was the only predictor for
the effect of Empathy on pain ratings (F[1,28] = 9.73, p < 0.01,
R2 = .26; B = .51, t = 3.12, p < 0.01). Here, higher scores on attach-
ment anxiety predicted higher pain ratings in the ‘Low-Empathy’
than in the ‘High-Empathy’ condition. None of the other variables,
including catastrophizing, predicted the effects of Social Presence or
Empathy on VAS ratings.
SCR and HRR: The same regression analyses on SCR and HRR
(including pre- and post-stimulus HRR) proved not to be signifi-
cant (all p P 0.13).
Table 1
Mean values, standard deviations (SD), and range of participants’ scores on
psychometric measures.
Mean SD Range
State anxiety (STAI) 32.53 7.41 21–52
Trait anxiety (STAI) 36.40 6.88 26–52
Attachment anxiety (RSQ) 32 9.75 13–60
Attachment avoidance (RSQ) 29.93 5.13 19.42
Pain vigilance and awareness (PVAQ) 43.20 12.44 25–94
Coping (CSQ) 105.53 34.87 45–185
Catastrophizing (CSQ) 8.46 5.59 0–17
691
Table 2
Mean values (and standard deviations) of post-test ratings.
High- Low- Alone
empathy empathy
Overall pain 6.7 (1.4) 6.7 (1.6) 6.9 (1.6)
Distraction (environment) 2.17 (2.40) 2.27 (2.48) 2.31 (2.73)
Distraction (thoughts) 3.50 (2.63) 3.55 (2.91) 4.24 (2.71)
Distraction (person) 1.70 (2.03) 1.93 (2.31) n/a
Distress 4.1 (2.6) 3.9 (2.6) 3.9 (2.9)
Reassurance from observers 2.0 (2.7) 1.8 (2.5) n/a
Perceived observers’ empathy 5.0 (2.9) 3.3 (2.4) n/a
Attractiveness 4.5 (2.5) 2.5 (1.9) n/a
3.3. Post-task ratings
The mean ratings of the post-task questions are shown in
Table 2. Participants’ ratings between conditions (‘High-Empathy’,
‘Low-Empathy’, and ‘Alone’) were compared using repeated-mea-
sures ANOVAs. No significant differences between experimental
conditions were found for ratings of: (a) overall pain (F[2,56]
= 0.30, p = 0.73); (b) distraction by the experimental environment,
the participant’s own thoughts, and the presence of observers (all
F 6 2.2, all p P 0.12); (c) levels of distress (F[1.1,41.2] = 0.41,
p = 0.60). In addition, ratings of reassurance from the ‘High-Empa-
thy’ and the ‘Low-Empathy’ observers did not differ significantly
(t[29] = 0.92, p = 0.36). Ratings of the observers’ perceived empathy
were significantly higher for the ‘High-Empathy’ than for the ‘Low-
Empathy’ observer (t[29] = 3.22, p < 0.05), indicating that partici-
pants generally trusted the feedback of the experimenter on the
observers’ empathy levels. Ratings of the observers’ attractiveness
were also significantly higher for the ‘High-Empathy’ than for the
‘Low-Empathy’ observer (t[29] = 3.26, p < 0.05). In addition, ratings
of the observers’ attractiveness and perceived empathy levels were
positively correlated for both the ‘High-Empathy’ and the ‘Low-
Empathy’ observers (r = 0.40, p < 0.05, and r = 0.66, p < 0.001,
respectively). In the light of the results on ratings of observers’
attractiveness, as well as previous experimental evidence that pain
reports may be influenced by the attractiveness of observers [27],
one additional VAS ANOVA was computed that specifically com-
pared the ‘High-Empathy’ and ‘Low-Empathy’ conditions including
the ratings of observers’ attractiveness as covariates. However, this
analysis revealed no significant main effect or interactions with the
covariates (all F[1,27] 6 2.33, all p P 0.14).
4. Discussion (1500 w)
We explored the effects of social presence (observation by unfa-
miliar others) and perceived empathy (observation by unfamiliar
others thought of as having high versus low-empathy for the par-
ticipants) on the perception of experimentally induced pain. Con-
trary to our predictions, the presence of an unfamiliar other and
their high ‘perceived’ empathy were not generally associated with
lower pain ratings. However, consistent with our predictions, we
found that both skin conductance and heart rate responses were
lower when participants experienced pain in the presence of an ob-
server (with either low- or high-empathy) than when they were
alone. Moreover, in line with our hypotheses, we found that the ef-
fect of perceived empathy and social presence on pain ratings was
uniquely predicted by the participants’ attachment styles. First,
higher scores on attachment anxiety significantly predicted lower
pain ratings in the presence of a ‘high-empathic’ than of a ‘low-em-
pathic’ observer; and second, higher scores on attachment avoid-
ance significantly predicted lower pain ratings when participants
experienced pain alone than in the presence of an observer. These
findings are discussed below.
Previous experimental investigations have found conflicting re-
sults regarding the role of social support on pain. Some studies
 Ò
692 C.F. Sambo et al. / PAIN 151 (2010) 687–693
show increased pain perception in conditions of social support (e.g.
[30]), while others find the reverse effect (e.g. [5,24]). The reasons
for these conflicting findings remain unclear, as these studies have
manipulated different dimensions of social support and have used
between-subjects designs. Our results, based on within-subject
comparisons, did not find an overall effect towards either direction.
Interestingly, we found that individual differences in attachment
style may interplay with social factors in affecting pain reports. In-
deed, attachment theory [4,10] has been proposed as a heuristic
model for understanding individual differences in pain [33]. It
has been shown that an insecure attachment style is an important
predictor of experimental and chronic pain [39]. However, no
study had hitherto examined the interaction between individual
attachment styles and the ‘on-line’ social modulation of pain.
In our study, higher scores on attachment anxiety were associ-
ated with lower pain ratings when participants perceived the
observers as having high-empathy than low-empathy. A recent
neuroimaging study has reported increased responses to negative
social feedback in brain regions involved in the control of emotion-
ally significant information (i.e. amygdala) in individuals with
higher attachment anxiety [52]. Our finding adds to this evidence
by showing that low perceived empathy by others also worsens
the pain experience of individuals with higher attachment anxiety.
Importantly, our results are consistent with the more general find-
ing that anxious attachment in healthy volunteers may be associ-
ated with increased perception and maladaptive reactions to and
appraisals of experimentally induced pain [32,34]. For example,
individuals scoring high in attachment anxiety experienced pain
earlier in the cold-pressor task than those scoring low in attach-
ment anxiety [34]. Our results potentially provide additional spec-
ificity to these findings, indicating that although perceived
empathy may not reduce pain perception in all individuals, it is
possible that in individuals scoring high in attachment anxiety,
pain reduction may be noted when empathy from others is clearly
perceived. This implies that their normally increased pain percep-
tion may be linked with their tendency to doubt support and
acceptability from others, irrespective of their explicit behaviour.
Future experimental studies should aim to determine whether in-
creased pain perception in these individuals, possibly aiming to in-
crease empathy and caretaking by others, is one of the reasons why
they show a predisposition to develop persistent pain conditions
(see also [33,39]).
Individuals scoring high in the dimension of anxious attach-
ment have also been found to show increased pain catastrophiz-
ing (e.g. [34]). According to the communal coping model of pain
catastrophizing [46,47], individuals differ in the extent to which
they adopt interpersonal coping strategies [11,47]. When high
pain catastrophizers experience pain in the presence of observers
they engage in displays of their pain-related distress as a means
of coping with pain and eliciting attention or support from others
[21,46]. However, these social cueing effects have been noted
only for pain behaviours (various actions or postural displays that
are enacted during the experience of pain; [13,19]) and not for
pain ratings [45]. Consistently, in the present study, we did not
observe an association between catastrophizing and the modula-
tion of pain ratings by empathy or social presence. Moreover, gi-
ven our aim to reliably measure the effects of empathy on pain
and to limit potential social desirability effects, we minimized ac-
tive interpersonal interactions and ensured that observers were
not aware of the participants’ ratings. Thus, we did not measure
pain behaviours. However, given the theoretical compatibility be-
tween the implications of our findings and the communal copy-
ing model, it would be interesting in future research to
determine whether individual differences in catastrophizing have
an effect on the social modulation of pain behaviours by per-
ceived empathy.
A further finding of our study was that consistently with
traits of attachment avoidance, such as preference for indepen-
dence and self-reliance [29,35], higher scores on attachment
avoidance were associated with lower pain ratings when partic-
ipants experienced pain alone than with an observer. Avoidant
individuals tend not to trust others, to avoid seeking support
and help from caregivers, as well as to adopt ‘‘deactivating’’
strategies, including suppression of negative affect [22,28]. Our
results suggest that these traits also apply to the perception of
pain in social context. Highly avoidant individuals perceive pain
stimulation as more painful in the presence of others, irrespec-
tive of the others’ perceived empathy. The presence of other peo-
ple during pain stimulation may result in increased anxiety for
these individuals who tend to mistrust social relationships (e.g.
[14]), and this may in turn increase pain perception. Interest-
ingly, a recent neuroimaging study [43] showed that pain-related
activation in the amygdala was reduced in ‘‘selfish” rather than
pro-social individuals, following administration of oxytocin (a
neuropeptide linked to social trust and attachment). The authors
concluded that despite contrary appearances the actions of ‘‘self-
ish individuals” may be driven by anxiety. It should also be
stressed that, as aforementioned, the present study was design
as to minimize free interaction and social desirability effects
and thus we did not measure pain-related behaviours. In future
studies, it would be interesting to investigate whether pain-re-
lated behaviours of highly avoidant individuals are modulated
in the same direction by social context as pain ratings (i.e. they
show increased pain-related behaviours in the presence of oth-
ers) or whether pain ratings and pain-related behaviours show
different patterns of social modulation.
A third main finding of this study is that skin conductance re-
sponses and heart rate responses were increased when partici-
pants experienced pain alone compared to when they were in
the presence of a ‘high-’ and ‘low-empathic’ observer. The effect
of social presence was found for both pre- and post-stimulus
time intervals indicating that physiological responses associated
with the anticipation of pain and with the actual pain experience
were similarly modulated by the presence of an observer. In
addition, none of the participants’ individual characteristics con-
tributed to these effects. These findings differ to some extent
from the effects obtained on pain ratings. A similar dissociation
between pain reports and the correspondent physiological activ-
ity has been previously reported in experimental investigations
on the social modulation of pain [2,30]. Our results can be inter-
preted according to previous studies that show that social sup-
port from unfamiliar and close others can reduce cardiovascular
reactivity during acute stressors in the laboratory (see [25] for
a review; but see [2,30]) and more generally social support dur-
ing stressful situations has been linked with potential health ben-
efits [8,9,49]. Thus, our results may indicate that the presence of
observers may attenuate physiological arousal in response to
noxious thermal stimuli. An alternative interpretation is that
the presence of others might reduce attention towards pain (i.e.
it may have a distraction effect) [20] resulting in decreased
pain-related autonomic responses. Future investigations will help
clarify the nature of the relationship between social support and
physiological responses to noxious stimuli.
Although experimental studies of pain have the advantage of
addressing causal relationships, caution should be applied when
interpreting the results of our study. First, caution should be ap-
plied when translating findings from healthy volunteers to clinical
populations, from experimentally induced to clinical pain and from
experimentally controlled to spontaneous social interactions. Sec-
ond, our study focused on the effects of perceived empathy and so-
cial presence on pain ratings, thus our findings may not address
other facets of social support and interaction. Third, individual
 Ò
C.F. Sambo et al. / PAIN 151 (2010) 687–693
traits were measured by self-report questionnaires that are not
shared by all studies and may be subject to social desirability
biases. For instance, self-report questionnaires measuring adult
attachment style may generate different findings than procedures
such as the Adult Attachment Interview (AAI) (see [39] for
discussion).
In conclusion, our results suggest that social presence and per-
ceived empathy may interact with individual differences in attach-
ment style to influence one’s pain perception. This may result in
individuals with an avoidant attachment style experiencing less
pain when alone, and individuals with higher attachment anxiety
experiencing less pain in the presence of an ‘empathic’ other.
Conflict of interest statement
The authors have no financial relationships that might lead to a
conflict of interest.
Acknowledgements
We would like to thank Chris Andrew, Jeff Dalton, and Pallavi
Chhibbar for their assistance with this study. The study was sup-
ported by grants by the Hope for Depression Research Foundation
(HDRF), Institute for the Study of Affective Neuroscience (ISAN)
and Volkswagen Foundation ‘European Platform for Life Sciences,
Mind Sciences and Humanities’ (to A.F.).
References
[1] Andrasik F, Flor H, Turk DC. An expanded view of psychological aspects in head
pain: the biopsychosocial model. Neurol Sci 2005;26:S87–91.
[2] Aslaksen PM, Myrbakk IN, Høifødt RS, Flaten MA. The effect of experimenter
gender on autonomic and subjective responses to pain stimuli. Pain
2007;129:260–8.
[3] Barrera M. Distinctions between social support concepts, measures, and
models. Am J Community Psychol 1986;14:413–45.
[4] Bowlby J. Attachment and loss: vol. I. Attachment. New York: Basic Books;
1969/1997.
[5] Brown JL, Sheffield D, Leary MR, Robinson ME. Social support and experimental
pain. Psychosom Med 2003;65:276–83.
[6] Chambers CT, Craig KD, Bennett SM. The impact of maternal behavior on
children’s pain experiences: an experimental analysis. J Pediatr Psychol
2002;27:293–301.
[7] Cano A, de C Williams AC. Social interaction in pain: reinforcing pain behaviors
or building intimacy? Pain 2010;149:9–11.
[8] Cohen S. Social relationships and health. Am Psychol 2004;59:676–84.
[9] Cohen S, Wills TA. Stress, social support, and the buffering hypothesis. Psychol
Bull 1985;98:310–57.
[10] Collins NL, Freeney BC. An attachment theory perspective on closeness and
intimacy. In: Mashek DJ, Aron A, editors. Handbook of closeness and
intimacy. Mahwah, NJ: Lawrence Erlbaum Associates; 2004. p. 163–88.
[11] Coyne JC, Fiske V. Couples coping with chronic illness. In: Akamatsu TJ, Parris
Stephens MA, Hobfoll SE, Crowther JH, editors. Family health psychology.
Washington, DC: Hemisphere; 1992.
[12] Craig KD. Social disclosure, coactive peer companions and social modeling as
determinants of pain communications. Can J Behav Sci 1978;10:91–104.
[13] Craig KD. Emotions and psychobiology. In: Wall PD, Melzack R, editors.
Textbook of pain. Edinburgh: Churchill-Livingstone; 1999. p. 331–43.
[14] Feeney JA, Noller P. Attachment style as a predictor of adult romantic
relationships. J Pers Soc Psychol 1990;58:281–91.
[15] Flor H, Kerns RD, Turk DC. The role of spouse reinforcement, perceived
pain, and activity levels in chronic pain patients. J Psychosom Res 1987;31:
251–9.
[16] Fordyce WE. Behavioral methods for chronic pain and illness. St. Louis: CV
Mosby; 1976.
[17] Fraley RC, Niedenthal PM, Marks M, Brumbaugh C, Vicary A. Adult attachment
and the perception of emotional expressions: probing the hyperactivating
strategies underlying anxious attachment. J Pers 2006;74:1163–90.
[18] Griffin DW, Bartholomew K. Models of the self and other: fundamental
dimensions underlying measures of adult attachment. J Pers Soc Psychol
1994;67:430–45.
[19] Hadjistavropoulos T, Craig KD. A theoretical framework for understanding self-
report and observational measures of pain: A Communications Model. Behav
Res Ther 2002;40:551–70.
[20] Hodes RL, Howland EW, Lightfoot N, Cleeland CS. The effects of distraction on
responses to cold pressor pain. Pain 1990;41:109–14.
693
[21] Keefe FJ, Lefebvre JC, Egert JR, Affleck G, Sullivan MJ, Caldwell DS. The
relationship of gender to pain, pain behavior, and disability in osteoarthritis
patients: the role of catastrophizing. Pain 2000;87:325–34.
[22] Kumin I. Pre-object relatedness. New York: Guilford; 1996.
[23] Kurdek LA. On being insecure about the assessment of attachment styles. J Soc
Pers Rel 2002;19:811–34.
[24] Jackson T, Iezzi T, Chen H, Ebnet S, Eglitis K. Gender, interpersonal transactions,
and the perception of pain: an experimental analysis. J Pain 2005;6:228–36.
[25] Lepore SJ. Problems and prospects for the social support-reactivity hypothesis.
Ann Behav Med 1998;20:257–69.
[26] Levey AB. Measurement units in psychophysiology. In: Martin I, Venables PH,
editors. Techniques in psychophysiology. New York: Wiley; 1980.
[27] Levine FM, De Simone LL. The effects of experimenter gender on pain report in
male and female subjects. Pain 1991;44:69–72.
[28] Main M. The organized categories of infant, child, and adult attachment:
flexible vs. inflexible attention under attachment-related stress. J Am
Psychoanal Assoc 2000;48:1055–96.
[29] Maunder RG, Hunter JJ. Attachment and psychosomatic medicine: develop-
mental contributions to stress and disease. Psychosom Med 2001;63:556–67.
[30] McClelland LE, McCubbin JA. Social influence and pain response in women and
men. J Behav Med 2008;31:413–20.
[31] McCracken LM. ‘Attention’ to pain in persons with chronic pain: a behavioural
approach. Behav Ther 1997;28:271–84.
[32] McWilliams LA, Asmundson GJ. The relationship of adult attachment
dimensions to pain-related fear, hypervigilance, and catastrophizing. Pain
2007;127:27–34.
[33] Meredith PJ, Ownsworth T, Strong J. A review of the evidence linking adult
attachment theory and chronic pain: presenting a conceptual model. Clin
Psychol Rev 2008;28:407–29.
[34] Meredith PJ, Strong J, Feeney JA. The relationship of adult attachment to
emotion, catastrophizing, control, threshold and tolerance, in experimentally-
induced pain. Pain 2006;120:44–52.
[35] Mikulincer M, Shaver PR. The attachment behavioral system in adulthood:
activation, psychodynamics, and interpersonal processes. In: Zanna MP, editor.
Advances in experimental social psychology, Vol. 35. San Diego, CA: Academic
Press; 2003. p. 53–152.
[36] Netter FH. Atlas of Human Anatomy. 3rd ed. Teterboro, NJ: Icon Learning
Systems; 2003.
[37] Newton-John TRO. Solicitousness and chronic pain: a critical review. Pain Rev
2002;9:7–27.
[38] Paulsen JS, Altmaier EM. The effects of perceived versus enacted social support
on the discriminative cue function of spouses for pain behaviors. Pain
1995;60:103–10.
[39] Porter LS, Davis D, Keefe FJ. Attachment and pain: recent findings and future
directions. Pain 2007;128:195–8.
[40] Rosenstiel AK, Keefe FJ. The use of coping strategies in chronic low back pain
patients: relationship to patient characteristics and current adjustment. Pain
1983;17:33–44.
[41] Shaffnel N, Wittwer A, Kut E, Folkers G, Benninger DH, Candia V. Heat pain
threshold and tolerance show no left–right perceptual differences at
complementary sites of the human forearm. Neurosci Lett 2008;440:309–13.
[42] Singer T, Seymour B, O’Doherty J, Kaube H, Dolan RJ, Frith CD. Empathy for pain
involves the affective but not sensory components of pain. Science
2004;303:1157–62.
[43] Singer T, Snozzi R, Bird G, Petrovic P, Silani G, Heinrichs M, Dolan RJ. Effects of
oxytocin and prosocial behavior on brain responses to direct and vicariously
experienced pain. Emotion 2008;8:781–91.
[44] Spielberg CD, Gorsuch RL, Lushene RE. Manual for the state-trait anxiety
inventory. Paolo Alto: Calif. Consulting Psychologists Press; 1970.
[45] Sullivan MJL, Adams H, Sullivan ME. Communicative dimensions of pain
catastrophizing: social cueing effects on pain behaviour and coping. Pain
2004;107:220–6.
[46] Sullivan MJL, Thorn B, Haythornthwaite JA, Keefe F, Martin M, Bradley LA,
Lefebvre JC. Theoretical perspectives on the relation between catastrophising
and pain. Clin J Pain 2001;17:52–64.
[47] Thorn BE, Ward LC, Sullivan MJL, Boothby JL. Communal coping model of
catastrophizing: conceptual model building. Pain 2003;106:1–2.
[48] Tait RC. Empathy: necessary for effective pain management? Curr Pain
Headache Rep 2008;12:108–12.
[49] Uchino BN, Cacioppo JT, Kiecolt-Glaser JK. The relationship between social
support and physiological processes. A review with emphasis on underlying
mechanisms and implications for health. Psychol Bull 1996;119:488–531.
[50] Vervoort T, Goubert L, Eccleston C, Verhoeven K, De Clercq A, Buysse A,
Crombez G. The effects of parental presence upon the facial expression of
pain: the moderating role of child pain catastrophizing. Pain 2008;138:
277–85.
[51] Vlaeyen JWS, Hanssen M, Goubert L, Vervoort T, Peters M, van Breukelen G,
Sullivan MJL, Morley S. Threat of pain influences social context effects on
verbal pain report and facial expression. Behav Res Ther 2009;47:774–82.
[52] Vrtička P, Andersson F, Grandjean D, Sander D, Vuilleumier P. Individual
attachment style modulates human amygdala and striatum activation during
social appraisal. PLoS One 2008;3:1–11.
[53] Wunsch A, Philippot P, Plaghki L. Affective associative learning modifies the
sensory perception of nociceptive stimuli without participant’s awareness.
Pain 2003;102:27–38.