Empathy for Pain Involves the Affective But Not Sensory Components of Pain

Abstract

Our ability to have an experience of another's pain is characteristic of empathy. Using functional imaging, we assessed brain
activity while volunteers experienced a painful stimulus and compared it to that elicited when they observed a signal indicating
that their loved one—present in the same room—was receiving a similar pain stimulus. Bilateral anterior insula (AI), rostral
anterior cingulate cortex (ACC), brainstem, and cerebellum were activated when subjects received pain and also by a signal
that a loved one experienced pain. AIand ACC activation correlated with individual empathy scores. Activity in the posterior
insula/secondary somatosensory cortex, the sensorimotor cortex (SI/MI), and the caudal ACC was specific to receiving pain.
Thus, a neural response in AIand rostral ACC, activated in common for “self” and “other” conditions, suggests that the neural
substrate for empathic experience does not involve the entire “pain matrix.” We conclude that only that part of the pain network
associated with its affective qualities, but not its sensory qualities, mediates empathy.

Full text

The time-averaged pulsed flux density of the
pulsars is about 1.8 mJy (1 Jy ⫽10
–26
Wm
–2
Hz
–1
) at 1390 MHz (Table 1), compared with
a total flux density at this frequency of 7 mJy
(3). The ⬃5-mJy unpulsed emission probably
arises in the impact region described above.
We find it remarkable, with much of the
magnetosphere of B blown away by the wind
of A, that B still works as a pulsar. This
suggests that the radio emission is probably
generated close to the neutron star, providing
a direct constraint on the emission height.
Conclusion. We have detected the binary
companion of the millisecond pulsar
J0737⫺3039 as a pulsar, making this the first
known double-pulsar system. This discovery
confirms the neutron-star nature of the compan-
ions to recycled pulsars in eccentric binary sys-
tems and validates the suggested evolutionary
sequences in which a companion star, having
spun up the pulsar, forms a young pulsar in a
supernova explosion (31). The highly relativis-
tic nature of this compact system opens up
opportunities for much more stringent tests of
relativistic gravitation than have been possible
previously. Not only have we already measured
four quantities attributable to, and consistent
with, general relativity, but the mass ratio Ris a
new high-precision constraint that is indepen-
dent of gravitational theories. Within a year or
so, we expect to measure the orbital decay due
to emission of gravitational radiation. If the
intrinsic value due to gravitational-wave damp-
ing can be extracted, it will allow tests of radi-
ative aspects of gravitational theories mixed
with strong-field effects. On somewhat longer
time scales of a few years, we expect to detect
several other relativistic effects, such as geodet-
ic precession of the pulsars’ spin axes, spin-
orbit coupling, and other deviations, making
this a superb test bed for relativity.
The detection of the companion as a pul-
sar also opens up the possibility of using each
pulsar to probe the magnetosphere of the
other. The energy flux from the millisecond
pulsar is strongly affecting the pulse emission
process in the companion, and eclipses of the
millisecond pulsar by the companion are also
seen. Future measurements of orbital varia-
tions in pulse shapes, amplitudes, polariza-
tion, and timing over a range of radio fre-
quencies will give fascinating insights into
magnetospheric processes in pulsars.
References and Notes
1. J. H. Taylor, Philos. Trans. R. Soc. London Ser. A 341,
117 (1992).
2. R. A. Hulse, J. H. Taylor, Astrophys. J. 195, L51 (1975).
3. M. Burgay et al., Nature 426, 531 (2003).
4. V. Kalogera et al.,Astrophys. J., in press; preprint is avail-
able at http://arxiv.org/abs/astro-ph/0312101.
5. Receiver bandwidths used were 64 MHz at 680 MHz,
256 MHz at 1390 MHz, and 576 MHz at 3030 MHz.
6. More information is available at http://pulsar.
princeton.edu/tempo/.
7. T. Damour, J. H. Taylor, Phys. Rev. D 45, 1840 (1992).
8. J. H. Taylor, J. M. Weisberg, Astrophys. J. 345, 434 (1989).
9. I. H. Stairs, S. E. Thorsett, J. H. Taylor, A. Wolszczan,
Astrophys. J. 581, 501 (2002).
10. M. Bailes, S. M. Ord, H. S. Knight, A. W. Hotan,
Astrophys. J. 595, L49 (2003).
11. T. Damour, N. Deruelle, Ann. Inst. Henri Poincare´ 44,
263 (1986).
12. T. Damour, J. H. Taylor, Astrophys. J. 366, 501 (1991).
13. T. Damour, N. Deruelle, Ann. Inst. Henri Poincare´ 43,
107 (1985).
14. This is correct up to so-called first post-Newtonian [(v/c)
2
]
order (7), and any intersection of the PK-parameter lines
must be located on the R line, which will only deviate from
that shown in Fig. 3 by corrections of order (v/c)
4
.
15. This is correct because of the usage of the DD timing
model in our analysis and its definition of mass (11,13).
16. T. Damour, R. Ruffini, C. R. Acad. Sci. Paris Ser. A 279,
971 (1974).
17. B. M. Barker, R. F. O’Connell, Astrophys. J. 199, L25 (1975).
18. J. M. Weisberg, R. W. Romani, J. H. Taylor, Astrophys.
J. 347, 1030 (1989).
19. J. M. Weisberg, J. H. Taylor, Astrophys. J. 576, 942 (2002).
20. M. Kramer, Astrophys. J. 509, 856 (1998).
21. T. Damour, G. Scha¨fer, Nuovo Cimento 101, 127 (1988).
22. N. Wex, Class. Quantum Gravity 12, 983 (1995).
23. D. Bhattacharya, E. P. J. van den Heuvel, Phys. Rep.
203, 1 (1991).
24. G. S. Bisnovatyi-Kogan, B. V. Komberg, Sov. Astron.
18, 217 (1974).
25. N. Shibazaki, T. Murakami, J. Shaham, K. Nomoto,
Nature 342, 656 (1989).
26. Z. Arzoumanian, J. M. Cordes, I. Wasserman, Astro-
phys. J. 520, 696 (1999).
27. E. P. J. van den Heuvel, C. de Loore, Astron. Astrophys.
25, 387 (1973).
28. B. P. Flannery, E. P. J. van den Heuvel, Astron. Astro-
phys. 39, 61 (1975).
29. S. E. Thorsett, D. Chakrabarty, Astrophys. J. 512, 288
(1999).
30. R. N. Manchester, J. H. Taylor, Pulsars (Freeman, San
Francisco, 1977).
31. G. Srinivasan, E. P. J. van den Heuvel, Astron. Astro-
phys. 108, 143 (1982).
32. J. H. Taylor, J. M. Cordes, Astrophys. J. 411, 674
(1993).
33. We would like to thank J. Sarkissian and other members
of the Parkes multibeam team for their kind help with
making the observations described in this paper and N.
Wex and G. Scha¨fer for useful discussions. Extensive use
was made of the PSRCHIVE pulsar analysis system
developed by A. Hotan and colleagues (see http://
astronomy.swin.edu.au/pulsar). The Parkes radio tele-
scope is part of the Australia Telescope, which is funded
by the Commonwealth of Australia for operation as a
National Facility managed by CSIRO. M.B., A.P., and N.D.
acknowledge financial support from the Italian Ministry
of University and Research (MIUR) under the national
program Cofin 2001. F.C. is supported by NSF, NASA,
and National Radio Astronomy Observatory. D.R.L. is a
University Research Fellow funded by the Royal Society.
12 December 2003; accepted 30 December 2003
Published online 8 January 2004;
10.1126/science.1094645
Include this information when citing this paper.
Empathy for Pain Involves the
Affective but not Sensory
Components of Pain
Tania Singer,
1
* Ben Seymour,
1
John O’Doherty,
1
Holger Kaube,
2
Raymond J. Dolan,
1
Chris D. Frith
1
Our ability to have an experience of another’s pain is characteristic of empathy. Using
functional imaging, we assessed brain activity while volunteers experienced a painful
stimulus and compared it to that elicited when they observed a signal indicating that
their loved one—present in the same room—was receiving a similar pain stimulus.
Bilateral anterior insula (AI), rostral anterior cingulate cortex (ACC), brainstem, and
cerebellum were activated when subjects received pain and also by a signal that a loved
one experienced pain. AI and ACC activation correlated with individual empathy scores.
Activity in the posterior insula/secondary somatosensory cortex, the sensorimotor
cortex (SI/MI), and the caudal ACC was specific to receiving pain. Thus, a neural response
in AI and rostral ACC, activated in common for “self” and “other” conditions, suggests
that the neural substrate for empathic experience does not involve the entire “pain
matrix.” We conclude that only that part of the pain network associated with its
affective qualities, but not its sensory qualities, mediates empathy.
Human survival depends on the ability to
function effectively within a social context.
Central to successful social interaction is
the ability to understand others intentions
and beliefs. This capacity to represent men-
tal states is referred to as “theory of mind”
(1) or the ability to “mentalize” (2). Empa-
thy, by contrast, broadly refers to being
able to understand what others feel, be it an
emotion or a sensory state. Accordingly,
empathic experience enables us to under-
stand what it feels like when someone else
experiences sadness or happiness, and also
pain, touch, or tickling (3).
Even though empathy has been exten-
sively discussed and investigated by philos-
ophers and social scientists, only recently
has it become a focus for neuroscience
(3–8). Influenced by perception-action
models of motor behavior and imitation (9),
1
Wellcome Department of Imaging Neuroscience, In-
stitute of Neurology, University College of London, 12
Queen Square, WC1N 3AR London, UK.
2
Headache
Group, Institute of Neurology, University College of
London, Queen Square, WC1N 3BG London, UK.
*To whom correspondence should be addressed. E-
mail: t.singer@fil.ion.ucl.ac.uk
RESEARCH ARTICLES
www.sciencemag.org SCIENCE VOL 303 20 FEBRUARY 2004 1157

Preston and de Waal (8) proposed a model
of empathy that incorporates most theoret-
ical accounts of, as well as empirical find-
ings on, empathy. The key suggestion is
that observation or imagination of another
person in a particular emotional state auto-
matically activates a representation of that
state in the observer, with its associated
autonomic and somatic responses (“auto-
matic”refers to a process that does not
require conscious and effortful processing
but can nevertheless be inhibited or con-
trolled). The philosopher Susanne Langer
has described it as an involuntary breach of
individual separateness (10).
Recent neuroimaging studies have ex-
plored the neural correlates of empathic
experience by measuring brain activity
while subjects watched video clips showing
actors telling personal stories with neutral
or sad content accompanied by sad and
happy facial expressions (11), by comparing
brain activity associated with the imitation
of and the mere observation of pictures
showing different emotional facial expres-
sions (12), and by comparing neural re-
sponses elicited by watching videos of
faces with emotional expressions of disgust
and pleasure with responses induced by
smelling aversive and pleasant odors (13).
Whereas the first study did not permit the
identification of shared networks underly-
ing emotions in self and others, the latter
studies report activation in areas previously
identified in the perception and production
of facial emotional expressions (action rep-
resentation and emotional systems) as well
as during the smelling of aversive odors
(insula). These results suggest that regions
associated with feeling an emotion can be
activated by seeing the facial expression of
the same emotion, a phenomenon described
as emotional contagion.
A paradigm to study empathy in
vivo. The present study extends previous
findings in a number of ways. First, we
tested for shared and unique networks for
experienced and empathic pain. More spe-
cifically, we provide evidence for pain-
related empathic responses and demon-
strate that empathic experience does not
involve activation of an entire pain matrix,
but only of that component associated with
the affective dimension of pain experience.
Moreover, we show a relation between em-
pathy-related brain activity and individual
difference in empathy as assessed by com-
monly used empathy scales. We also
demonstrate—in contrast to accounts of
emotional contagion—that empathic re-
sponses can be elicited automatically in the
absence of an emotional cue (such as facial
emotional expressions) through mere pre-
sentation of an arbitrary cue that signals the
feeling state of another person.
We used functional magnetic resonance
imaging (fMRI) to explore both common
and unique brain circuitry involved when
we experience pain in ourselves, as well as
when we observe someone else feeling
pain. Previous studies on pain have re-
vealed that noxious stimuli consistently ac-
tivate the secondary somatosensory cortex
(SII), insular regions, the anterior cingulate
cortex (ACC), the movement-related areas
such as the cerebellum and supplementary
motor areas and, less robustly, the thalamus
and the primary somatosensory cortex (SI)
(14,15). This pain-related network, which
is commonly referred to as the pain matrix,
served to define regions of interest in the
present study.
To investigate pain-related empathy, we
developed a paradigm that allows the in-
vestigation of empathic experience in vivo,
with the object of empathy being a real
person present in the same room. More
specifically, we investigated pain-related
empathy in 16 couples, under an assump-
tion that couples are likely to feel empathy
for each other. We assessed brain activity
in the female partner while painful stimu-
lation was applied to her or to her partner’s
right hand through an electrode attached to
the back of the hand. The partner was
seated next to the MRI scanner and the
right hand of each subject was placed on a
tilted board, allowing the female partner
with help of a mirror system to see her and
her partner’s right hand. On a large screen
situated behind the board, cues were pre-
sented in random order indicating whether
she (self) or her partner (other) would get
low (no pain condition) or high (pain con-
dition) stimulation. We were especially in-
terested in comparing pain-related brain ac-
tivity (assessed by the difference between
trials involving painful and nonpainful
stimulation) in the context of “self”and
“other.”Questionnaires administered after
scanning served to validate measurements
of individual pain threshold made before
scanning, to obtain subjective evidence for
empathic experience during scanning, and
to assess stable individual differences in
empathy in order to determine whether
these scores predict the amplitude of
empathy-related brain activity.
Analysis of pain intensity ratings after
Fig. 1. Pain-related acti-
vation associated with
either experiencing pain
in oneself or observing
one’s partner feeling
pain. Areas in green rep-
resent significant activa-
tion (P⬍0.001) for the
contrast pain–no pain in
the “self” condition and
areas in red for the con-
trast pain–no pain in the
“other” condition. The
results are superimposed
on a mean structural
scan of the 16 subjects.
Activations are shown
on sagittal (A and B) and
axial (C and D) slices. (A)
Activation in ACC and
cerebellum. (B) Bilateral
insula cortex extending
into lateral prefrontal
cortex, left posterior in-
sula extending into sec-
ondary somatosensory
cortex (SII), bilateral oc-
cipital cortex, and fusi-
form cortex. (C) Bilateral
insula and mediodorsal
thalamus. (D) Middle
and lateral cerebellum/
fusiform gyrus. For coor-
dinates of peak activa-
tions from “self” and
“other” conditions, see
tables S1 and S2.
RESEARCH ARTICLES
20 FEBRUARY 2004 VOL 303 SCIENCE www.sciencemag.org1158

scanning confirmed the individual thresh-
olds for nonpainful and painful stimulation
determined before scanning (fig. S1) (16 ).
In addition, the unpleasantness ratings in-
dicated empathic involvement of the sub-
jects. Nonpainful trials were rated as being
significantly less unpleasant than painful
trials, irrespective of whether the pain was
applied to themselves or to the partner
[main effect of pain: F (1, 15) ⫽19.93, P⬍
0.001; main effect of self/other: F (1, 15) ⫽
0.12, P⫽0.73)].
Does empathizing with pain in others
activate the entire pain matrix? Compar-
ison of brain activity associated with painful
and nonpainful trials in the “self”condition
revealed increased activity in contralateral
SI/MI, in bilateral SII with a peak activa-
tion in contralateral posterior insula ex-
tending into SII, in bilateral mid and ante-
rior insula, in ACC [caudal and posterior
rostral zones, using Picard and Strick’s ter-
minology, (17 )], in right ventrolateral and
mediodorsal thalamus, brainstem, and mid
and right lateral cerebellum (Fig. 1, A to D,
green). These regions have all been identi-
fied as responding to painful stimuli in
previous imaging studies (14,15). Many
structures in this pain matrix were also
activated when pain was applied to the
partner, that is, in the absence of somato-
sensory stimulation (Fig. 1, A to D, red).
When comparing painful with nonpainful
trials in the context of “other,”increases in
brain activity were observed in the ACC
(anterior and posterior rostral zones), the
anterior insula (AI) bilaterally with an ex-
tension into inferior prefrontal cortex, the
cerebellum, and the brainstem. In addition,
significant activations were observed in the
ventral and dorsal visual stream, including
bilateral fusiform cortex, lateral occipital
and right posterior superior temporal sul-
cus, the left inferior parietal cortex, and the
left superior frontal cortex.
To test for common networks activated
by pain in the self and other conditions, we
performed a conjunction analysis (Fig. 2, A
and B), as well as a more conservative
masking procedure in which we masked the
simple contrast pain–no pain in others by
the simple contrast pain–no pain in self.
Both analyses revealed a network common
to pain in self and other conditions that
comprised ACC (caudal and posterior ros-
tral zones), bilateral middle insula and AI
(with a peak activation in the right AI),
brainstem, and lateral cerebellum.
We also tested for an interaction in or-
der to identify regions where there were
stronger pain-related activations in the self
as compared to the other condition. These
analyses confirmed that the contralateral
activations in SI/MI and SII/posterior insu-
la, as well as a region in caudal ACC, were
specific to the pain experience in the self
condition (Fig. 3). In contrast, extrastriate
visual related activations were specific to
the empathy condition.
Inspection of the time courses of pain-
related activity for self and others suggest
two peaks, probably as the result of an instant
response to the anticipation cue followed by
another response to the delivery of the pain
stimulus 3.5 s later. Accordingly, pain-related
activations for self in ACC and AI peak first
early, around 2 to 4 s, and again around 8 to
12 s (Fig. 2C). Activation specific to pain in
self in SI, SII, and caudal ACC seems only to
peak late, around 8 to 12 s, probably regis-
tering the actual receipt of the painful stimu-
lus (Fig. 3, A to C). Although the present
study was not designed to test differences
between anticipation and receipt of pain in
self and others, the present data suggest that
both anticipation and experience of pain in-
volve ACC and AI and that activation of
these networks is involved in understanding
the pain of others.
Do people scoring higher in empathy
show higher pain-related brain activity? To
explore further the role of anterior insular
cortex and ACC—the two major regions of
the pain matrix identified as being also
involved when empathizing with the pain
of others—we determined whether individ-
ual differences in empathy, assessed by two
questionnaires, covary with brain activity
elicited in the empathy conditions (pain–no
pain in the “other”condition). As Fig. 4
illustrates, individual differences in empa-
thy as measured by two empathy scales, the
Balanced Emotional Empathy Scale (18,
19) and the Empathic Concern Scale (a
subscale of the Interpersonal Reactivity In-
dex) (20), showed significant covariation
with activity in ACC (posterior rostral
zone) and left AI, but not right AI. Thus,
subjects who scored higher on general em-
pathy scales showed stronger activations in
areas significantly activated when the sub-
jects perceived their partner as being in
pain. In addition, an anterior part of ACC
(anterior rostral zone adjacent to paracin-
gulate sulcus) and lateral right cerebellum
showed significant covariation with self-
rated individual differences in empathy.
Fig. 2. Shared networks observed when pain was applied to self or to the partner. (A) and (B)
illustrate results of a conjunction analysis between the contrasts pain–no pain in the context
of self and other at P⬍0.001. Results are shown on sagittal (A) and coronal (B) sections of
the mean structural scan. Coordinates refer to peak activations and are in mm. Increased
pain-related activation was observed in ACC [(–9, 6, 42); (0, 24, 33)], left insula [(–30, 9, 6);
(–39, 0, 18); (–48, 12, –3)], right anterior insula [(60, 15, 3); (39, 12, 3); (42, 27, –6)],
cerebellum [(–33, –66, –24); (27, –66, –18)] and brainstem (–6, –39, –30). (C) The time
courses of pain-related activation (pain–no pain) for peak activations in ACC and right anterior
insula cortex for self (green lines) and others (red lines). A complete list of activated areas is
in table S3. The anticipation cue was presented at time 0, and the pain was applied between
3 and 4 s into the trial (mean 3.5 s).
RESEARCH ARTICLES
www.sciencemag.org SCIENCE VOL 303 20 FEBRUARY 2004 1159

These findings underscore the crucial role
of rostral ACC and anterior insula cortices
for empathic experience related to pain.
They are also in line with a report based on
single-neuron recordings in a precingu-
lotomy patient that provided evidence that
neurons in the ACC can respond not only
when a person receives a painful stimulus
but also when a person observes or antici-
pates a potentially painful stimulus deliv-
ered to an experimenter (21). Note that these
regions are fundamentally different from sites
subserving mirror neurons that respond when
performing or watching a conspecific perform-
ing particular movements (22).
Numerous imaging studies have investi-
gated the neural correlates of pain experience
(14,15), but none have explored the empathic
experience of pain. Our results confirm pre-
vious findings of pain-related activation in
SI, SII, bilateral insula cortex, ACC, thala-
mus, brainstem, and cerebellum. In addition,
our data show that only part of the network
mediating pain experience is shared when
empathizing with pain in others. Empathizing
with someone else’s pain elicited activity
principally in left and right AI, ACC, lateral
cerebellum, and brainstem. This activity was
elicited without an explicit focus on empathy
insofar as subjects were not aware of the
study aims, nor were they required to make
an explicit judgement of what their loved one
was feeling at the time of the scan. The
finding of empathy-related activation sug-
gests an automatic engagement of empathic
processes when perceiving pain in others.
Moreover, our analysis demonstrates that
pain-related activation in contralateral SI,
SII/posterior insula, and caudal ACC are spe-
cific to self-experienced pain, as opposed to
perceived pain in others.
Recent neuroimaging studies on pain
have demonstrated contralaterally biased
representations of painful stimulus in SI,
distinct parts of SII, and posterior insula, as
well as in lateral thalamus, which suggests
that these structures provide sensory-
discriminative information concerning no-
ciceptive stimuli such as location, quality,
and intensity (23,24 ). In contrast, AI and
ACC do not show such a contralateral bias,
supporting a role in coding the autonomic
and affective dimension of pain such as the
subjectively perceived unpleasantness (15,
25–27 ). Using hypnosis as a tool to disas-
sociate sensory-discriminative from affec-
tive pain components, activation in ACC
(posterior rostral zone) was shown to be
modulated by perceived unpleasantness,
whereas activation in SI and SII was unaf-
fected (28). Similarly, attentional manipu-
lations modulate ACC (posterior rostral
zone) and right AI, but not posterior insular/
SII cortices (29). The role of ACC (poste-
rior rostral zone) and right AI in coding the
subjective affective dimension of pain is
also highlighted by findings showing that
activation of right AI was correlated with
subjective intensity ratings of thermal sen-
sation in a manner that is distinct from a
linear representation of stimulus temp-
erature in posterior insula (30). Indeed,
anticipation of pain activates more anterior
insular regions, whereas the actual experi-
ence of pain activates more posterior insu-
la, which suggests that the latter is associ-
ated with the actual sensory experience of
pain and the former with affective dimen-
sions such as the anticipatory arousal and
anxiety of pain (31). Furthermore, subjec-
tive reduction of pain associated with pla-
cebo and opioid analgesia is associated
with increased activity in rostral ACC and
right AI (32).
Conclusion. The strong anatomical con-
nections between regions constituting the
pain matrix suggest that these regions do not
function independently in encoding different
aspects of pain but are highly interactive.
Nevertheless, the results of the present study
and previous investigations suggest a segre-
gation of sensory-discriminative and auto-
nomic-affective attributes of the pain experi-
ence. Rostral ACC and AI appear to reflect
the emotional experience that evokes our re-
actions to pain and constitutes the neural
basis for our understanding of the feelings of
others and ourselves.
The above proposal is consistent with a
recent model by Craig based on detailed
anatomical observations that conceives of
pain as one of the homeostatic emotions
that reflects the internal (interoceptive)
Fig. 3. Brain activity specific to the experience of pain in oneself. (Ato C) illustrate results of
the interaction between the two factors (pain–no pain and self/other). Statistical maps
(threshold set at P⬍0.001) are overlayed on axial (A), coronal (B), and sagittal (C) sections
from the mean structural scan. Coordinates refer to peak activations and are in mm. (A)
Increased pain-related activity [(–21, –33, 75); (–30, –24, 72); (–27, –36, 60)] in SI/MI. (B)
Activity in left posterior insula/SII [–39, –27, 24]. (C) Activity in a part of caudal ACC (6, 6, 42).
On the bottom, time courses are displayed for peak voxels of contralateral SI, SII, and caudal
ACC. Green lines reflect parameter estimates for pain-related activity (pain–no pain) in self;
red lines reflect pain-related activity in others.
RESEARCH ARTICLES
20 FEBRUARY 2004 VOL 303 SCIENCE www.sciencemag.org1160

condition of the body, similar to tempera-
ture, sensual touch, itch, hunger, or thirst
(25,33). More specifically, it is assumed
that an image of the body’s internal state is
mapped to the brain by afferents that pro-
vide input by way of the ventromedial tha-
lamic nucleus to area 3a (sensorimotor cor-
tex), as well as to the mid/posterior dorsal
insula. In humans, this high-resolution,
modality-specific sensory representation of
the physiological condition of the body in
the posterior insula is initially re-represent-
ed in the AI on the same side of the brain,
and then, by way of a callosal pathway,
remapped to the other side of the brain in
the right AI. Such a second-order re-
representation in the right AI is assumed to
subserve subjective feelings and the aware-
ness of a physical self as a feeling entity. At
the same time, afferents also project by way
of the medial dorsal thalamic nucleus to
produce behavioral drive in ACC. Thus,
direct activation of both the insula (limbic
sensory cortex) and the ACC (the limbic
motor cortex) may correspond to a simul-
taneous generation of both a feeling and an
affective motivation with its attendant au-
tonomic effects (25,33). Indeed, imaging
studies focusing on the relation between
peripheral measures of arousal and brain
activity give robust evidence for the crucial
role of rostral ACC and AI cortices in the
representation of internal bodily states of
arousal, as well as emotional awareness
(26,27 ). Furthermore, activation in these
regions has been observed in a wide range
of imaging studies associated with positive
and negative subjective feelings expressed
by subjective ratings of facial trustworthi-
ness (34), musical enjoyment (35), sensual
touch (36 ), and distress resulting from
social exclusion (37 ). More generally,
these regions may play a critical role in
second-order representations of bodily ho-
meostatic states that underpin core repre-
sentations of self (38,39).
Our data suggest that empathizing with
the pain of others does not involve the acti-
vation of the whole pain matrix, but is based
on activation of those second-order re-
representations containing the subjective
affective dimension of pain. Accordingly, we
propose that these cortical re-representations
have a dual function. First, they form the
basis for our ability to form subjective
representation of feelings that allow us to
predict the effects of emotional stimuli with
respect to the self. Second, they serve as the
neural basis for our ability to understand
the emotional importance of a particular
stimulus for another person and to predict
its likely associated consequences. From a
functional and evolutionary perspective, a
detailed representation of the source and
nature of a noxious stimulus (i.e., intensity,
location) is of functional relevance when it
concerns our own body, allowing effective
engagement of executive action systems
(i.e., removing the noxious source). In
contrast, the understanding of someone
else’s emotional reaction to pain does not
necessitate such a detailed sensory-
discriminative representation of the nox-
ious stimulus but rather a representation of
the subjective relevance of the stimulus as
reflected in the subjective unpleasantness
that the other person feels. Such decoupled
representations—which are independent of
the sensory inputs of the outside world—
have been postulated to be necessary for
our ability to mentalize, that is, to under-
stand the thoughts, beliefs, and intentions
of others (2). Our data suggest that we use
similar decoupled representations to under-
stand the feelings of others and that our
ability to empathize has evolved from a
system for representing our internal bodily
states and subjective feeling states.
References and Notes
1. D. Premack, G. Woodruff, Behav. Brain Sci. 1, 515
(1978).
2. U. Frith, C. D. Frith, Philos. Trans. R. Soc. London Ser.
B358, 459 (2003).
3. V. Gallese, Philos. Trans. R. Soc. London Ser. B 358,
517 (2003).
4. R. Adolphs, Trends Cognit. Sci. 3, 469 (1999).
5. R. Adolphs, Nature Rev. Neurosci. 4, 165 (2003).
6. L. Brothers, Concepts Neurosci. 1, 27 (1990).
7. R. W. Levenson, A. M. Ruef, J. Pers. Soc. Psychol. 63,
234 (1992).
8. S. D. Preston, F. B. M. de-Waal, Behav. Brain Sci. 25,1
(2002).
9. W. Prinz, Eur. J. Cognit. Psychol. 9, 129 (1997).
10. S. K. Langer, An Essay on Human Feelings (John Hop-
kins Press, Baltimore, MD, 1988).
11. J. Decety, T. Chaminade, Neuropsychologia 41, 127
(2003).
12. L. Carr, M. Iacoboni, M. C. Dubeau, J. C. Mazziotta,
G. L. Lenzi, Proc. Natl. Acad. Sci. U.S.A. 100, 5497
(2003).
13. B. Wicker et al., Neuron 40, 655 (2003).
14. K. D. Davis, Neurol. Res. 22, 313 (2000).
15. R. Peyron, B. Laurent, L. Garcia-Larrea, Neurophysiol.
Clin. 30, 263 (2000).
16. Materials and methods are available as supporting
material on Science Online.
17. N. Picard, P. L. Strick, Cerebral Cortex 6, 342
(1996).
18. A. Mehrabian, N. Epstein, J. Pers. 40, 525 (1972).
19. A. Mehrabian, Aggress. Behav. 23, 433 (1997).
20. M. A. Davis, JSAS Catalog of Selected Documents in
Psychology 10, 85 (1980) (American Psychological
Association, Journal Supplement Abstract Service).
21. W. D. Hutchison, K. D. Davis, A. M. Lozano, R. R.
Tasker, J. O. Dostrovsky, Nature Neurosci. 2, 403
(1999).
22. J. Grezes, J. Decety, Hum. Brain Mapp. 12, 1 (2001).
23. U. Bingel et al., Neuroimage 18, 740 (2003).
24. J. C. Brooks, T. J. Nurmikko, W. E. Bimson, K. D. Singh,
N. Roberts, Neuroimage 15, 293 (2002).
25. A. D. Craig, Nature Rev. Neurosci. 3, 655 (2002).
26. H. D. Critchley, S. Wiens, P. Rotshtein, A. O
¨hman, R. J.
Dolan, Nature Neurosci. 7, 189 (2004).
27. H. D. Critchley et al., Brain 126, 2139 (2003).
28. P. Rainville, G. H. Duncan, D. D. Price, B. Carrier, M. C.
Bushnell, Science 277, 968 (1997).
29. R. Peyron et al., Brain 122, 1765 (1999).
Fig. 4. Activation level (parameter estimates) observed within peaks of the ACC and the left insula
during empathy-related conditions (pain–no pain in other) are significantly correlated with
individual differences in empathy as measured by (A) the Empathic Concern Scale of Davis (20) and
(B) the Balanced Emotional Empathy Scale of Mehrabian (18,19). The lines represent the linear best
fit; rrefers to the correlation coefficient. All correlations are significant on the P⬍0.05 level. Peak
activations lie within regions of ACC and left insula that were activated in the simple contrast
pain–no pain in others. Coordinates refer to peak activations and are in mm.
RESEARCH ARTICLES
www.sciencemag.org SCIENCE VOL 303 20 FEBRUARY 2004 1161

30. A. D. Craig, K. Chen, D. Bandy, E. M. Reiman, Nature
Neurosci. 3, 184 (2000).
31. A. Ploghaus et al., Science 284, 1979 (1999).
32. P. Petrovic, E. Kalso, K. M. Petersson, M. Ingvar, Science
295, 1737 (2002).
33. A. D. Craig, Curr. Opin. Neurobiol. 13, 500 (2003).
34. J. S. Winston, B. A. Strange, J. O’Doherty, R. J. Dolan,
Nature Neurosci. 5, 277 (2002).
35. A. J. Blood, R. J. Zatorre, Proc. Natl. Acad. Sci. U.S.A.
98, 11818 (2001).
36. H. Olausson et al., Nature Neurosci. 5, 900 (2002).
37. N. I. Eisenberger, M. D. Lieberman, K. D. Williams,
Science 302, 290 (2003).
38. P. S. Churchland, Science 296, 308 (2002).
39. A. R. Damasio, Descartes’ Error (Putnam, New York,
1994).
40. This work was supported by a personal grant for T.S.
from the German Academy of Natural Sciences
Leopoldina, Halle, with the grant BMBF-LPD 9901/8-
73 from the Ministry of Education and Science, by the
Wellcome Department of Imaging Neuroscience, and
by the University College of London. R.J.D. and C.D.F.
are in receipt of Wellcome Trust program grants. We
thank S. Kiebel, J. Schultz, K. Wiech, R. Kalisch, P.
Aston, E. Featherstone, and P. Allen for their help.
Supporting Online Material
www.sciencemag.org/cgi/content/full/303/5661/1157/
DC1
Materials and Methods
Fig. S1
Tables S1 to S5
References
11 November 2003; accepted 23 December 2003
Placebo-Induced Changes in fMRI
in the Anticipation and
Experience of Pain
Tor D. Wager,
1
*†James K. Rilling,
2
Edward E. Smith,
1
Alex Sokolik,
3
Kenneth L. Casey,
3
Richard J. Davidson,
4
Stephen M. Kosslyn,
5
Robert M. Rose,
6
Jonathan D. Cohen
2,7
The experience of pain arises from both physiological and psychological factors,
including one’s beliefs and expectations. Thus, placebo treatments that have no
intrinsic pharmacological effects may produce analgesia by altering expecta-
tions. However, controversy exists regarding whether placebos alter sensory
pain transmission, pain affect, or simply produce compliance with the sugges-
tions of investigators. In two functional magnetic resonance imaging (fMRI)
experiments, we found that placebo analgesia was related to decreased brain
activity in pain-sensitive brain regions, including the thalamus, insula, and
anterior cingulate cortex, and was associated with increased activity during
anticipation of pain in the prefrontal cortex, providing evidence that placebos
alter the experience of pain.
The idea that sensory experience is shaped
by one’s attitudes and beliefs has gained
currency among psychologists, physicians,
and the general public. Perhaps nowhere is
this more apparent than in our ability to
modulate pain perception. A special case of
this phenomenon is placebo analgesia, in
which the mere belief that one is receiving
an effective analgesic treatment can reduce
pain (1–5). Recently, some researchers
have attributed placebo effects to response
bias and/or to publication biases (6), which
raises the issue of whether placebo treat-
ments actually influence the sensory, affec-
tive, and cognitive processes that mediate
the experience of pain.
One important piece of evidence that
placebo effects are not simply due to re-
sponse or publication bias is that such ef-
fects can be reversed by the mu-opioid
antagonist naloxone (2,3,7), suggesting
that some kinds of placebo effects may be
mediated by the opioid system. However,
naloxone has also been shown to produce
hyperalgesia independent of placebo, in
some cases offsetting rather than blocking
the effects of placebo analgesia (8). Al-
though pharmacological blockade provides
suggestive evidence regarding the neuro-
chemical mechanisms mediating placebo
effects, such data do not illuminate the
nature of the information-processing sys-
tem that gives rise to such effects. Neuro-
imaging data can provide complementary
evidence of how pain processing in the
brain is affected by placebos and about the
time course of pain processing. Identifying
placebo-induced changes in brain activity
in regions associated with sensory, affec-
tive, and cognitive pain processing (9) may
provide insight into which components of
pain processing are affected by placebo. In
addition, identifying changes that occur at par-
ticular times—in anticipation of pain, early or
late during pain processing—may shed light on
how cognitive systems mediating expectancy
interact with pain and opioid systems.
In two functional magnetic resonance
imaging (fMRI) experiments (n⫽24 and
n⫽23), we examined two hypotheses re-
garding the psychological and neural mech-
anisms that underlie placebo analgesia. Our
first hypothesis was that if placebo manip-
ulations reduce the experience of pain,
pain-responsive regions of the brain should
show a reduced fMRI blood oxygen level–
dependent (BOLD) signal (a measure relat-
ed to neural activity) during pain. [Pain-
responsive regions, or the “pain matrix,”
include thalamus, somatosensory cortex,
insula, and anterior cingulate cortex (10–
14).] Our second hypothesis was that pla-
cebo modulates activity of the pain matrix
by creating expectations for pain relief,
which in turn inhibit activity in pain-
processing regions. Converging evidence
suggests that the prefrontal cortex (PFC),
the dorsolateral aspect (DLPFC) in partic-
ular, acts to maintain and appropriately up-
date internal representations of goals and
expectations, which modulate processing in
other brain areas (15,16). Thus, stronger
PFC activation during the anticipation of
pain should correlate with greater placebo-
induced pain relief as reported by participants
and greater placebo-induced reductions in neu-
ral activity within pain regions (17).
Placebo reduces reported pain and
brain activity in Study 1 (shock pain). The
design of Study 1 is illustrated in Fig. 1A
(see the figure legend for a description)
(18). First, to confirm that application of
shock elicited a neural response in pain-
related areas, we compared brain activity in
the intense shock versus no shock condi-
tions. This revealed activation of the classic
pain matrix (11,14,19,20), including thal-
amus, primary somatosensory cortex/
primary motor cortex (S1/M1), secondary
somatosensory cortex (SII), midbrain, an-
terior insula, anterior cingulate cortex
(ACC), ventrolateral prefrontal cortex, and
cerebellum (fig. S1). As expected, activa-
tions in thalamus, S1, SII, and M1 were
larger in the left hemisphere, contralateral
to the wrist where shocks were applied,
whereas cerebellar activation was ipsilater-
al, although some bilateral activation was
observed in each of these areas. We also
1
Department of Psychology, University of Michi-
gan, 525 East University, Ann Arbor, MI 48109–
1109, USA.
2
Center for the Study of Brain, Mind and
Behavior, Princeton University, Princeton, NJ
08544, USA.
3
Department of Neurology, Veterans
Affairs Medical Center, University of Michigan, Ann
Arbor, MI 48109, USA.
4
Department of Psychology,
University of Wisconsin, Madison, WI 53706, USA.
5
Department of Psychology, Harvard University,
Cambridge, MA 02138, USA.
6
Mind Brain Body and
Health Initiative, University of Texas Medical
Branch, Galveston, TX 77555, USA.
7
Department of
Psychology, Princeton University, Princeton, NJ
08544, USA.
*To whom correspondence should be addressed. E-mail:
torw@umich.edu, tor@paradox.psych.columbia.edu
†Present address: Department of Psychology, Colum-
bia University, 1190 Amsterdam Avenue, New York,
NY 10027, USA.
RESEARCH ARTICLES
20 FEBRUARY 2004 VOL 303 SCIENCE www.sciencemag.org1162