Neuron, Vol. 32, 537–551, November 8, 2001, Copyright 2001 by Cell Press
Beautiful Faces Have Variable Reward Value:
fMRI and Behavioral Evidence
1994). The strong motivational influence of beauty has
been shown in studies of labor markets suggesting that
there is a ”beauty premium” and “plainness penalty”
(Hamermesh and Biddle, 1994) such that attractive indi-
higher salaries than unattractive individuals (Marlowe et
al., 1996; Frieze et al., 1990, 1991). Darwinian ap-
proaches to the study of facial attractiveness posit that
the features of beautiful faces are important biological
signals of mate value that motivate behavior in others
(Etcoff, 1999; Grammer and Thornhill, 1994; Perrett et
al., 1998; Symons, 1995).
Given the association between beauty and motivated
brain circuitry implicated in reward function underlying
motivated behavior is activated by the social signals
contained in beautiful faces. Research with another so-
cial stimulus, namely money, has implicated an ex-
tended set of brain reward regions with the anticipation
and reception of monetary outcomes (Breiter et al.,
1996b, 2001; Delgado et al., 2000; Elliott et al., 2000;
Knutson et al., 2000, 2001; O’Doherty et al., 2001; Thut
et al., 1997). Although money and beautiful faces can
thetic evaluations. In contrast, it is possible that beauti-
ful faces may stimulate both reward assessments and
aesthetic assessments, each leading to different pat-
terns of brain activity.
Extensive neuroscience research has focused on the
visual processing of faces (e.g., Kanwisher, et al., 1997)
and facial expression (e.g., Breiter et al., 1996a; Morris
et al., 1996; Phillips et al., 1999; Thomas et al., 2001),
while other work has evaluated the visual processing of
symmetry (Grammer and Thornhill, 1994; Perrett et al.,
1999) and attractiveness (Perrett et al., 1998; Bartels
and Zeki, 2000; Nakamura et al., 1998). In this study, we
wished to evaluate faces as potential objects of reward.
Most visual stimuli are not primary reinforcers; indeed,
the sensory representation of an object is different from
its rewarding properties (Rolls, 1999). When animals or
humans respond to rewarding stimuli, they respond to
multiple informational features extracted from distinct
representations of these goal-objects, including the
rate, latency, incidence, intensity, amount, category,
and proximity of the reinforcing stimuli (Breiter and Ro-
sen, 1999; Gallistel, 1990; Shizgal, 1999). The response
of animals and humans to these features appears to be
dependent on their hedonic deficit state regarding such
reinforcers (Cabanac, 1971). In the absence of a defined
deficit state regarding attractive faces, it remains a sa-
lient question whether they could be considered to be
To evaluate this issue, we carried out a study with
three components, each component using the same
categories of faces: beautiful females, average females,
beautiful males, and average males (Figure 1A). One
quality of these images. Another component used a
novel “keypress” task to operationalize the amount of
Itzhak Aharon,1,2,7Nancy Etcoff,3,7Dan Ariely,4,7
Christopher F. Chabris,1,2,5,7Ethan O’Connor,4
and Hans C. Breiter1,2,3,6
1Motivation and Emotion Neuroscience Center
Department of Radiology
Massachusetts General Hospital and
Harvard Medical School
Boston, Massachusetts 02129
2Athinoula A. Martinos Center for Biomedical
Massachusetts General Hospital
Massachusetts Institute of Technology and
Harvard Medical School
Boston, Massachusetts 02129
3Department of Psychiatry
Massachusetts General Hospital and
Harvard Medical School
Building 149, 13thStreet
Charlestown, Massachusetts 02129
4Massachusetts Institute of Technology
Cambridge, Massachusetts 02139
5Department of Psychology
Cambridge, Massachusetts 02138
The brain circuitry processing rewarding and aversive
stimuli is hypothesized to be at the core of motivated
behavior. In this study,discrete categories of beautiful
faces are shown to have differing reward values and
tures of beautiful males and females as attractive, but
exert effort via a keypress procedure only to view pic-
tures of attractive females. Functional magnetic reso-
nance imaging at 3 T shows that passive viewing of
beautiful female faces activates reward circuitry, in
particular the nucleus accumbens. An extended set of
procedures, suggesting that reward circuitry function
does not include aesthetic assessment.
Beauty in human faces has long been considered within
the general category of aesthetic theory (Ruskin, re-
printed 1997; Kant, reprinted 1960) and only recently
within the domain of biology and neuroscience. Recent
research on facial beauty suggests that the perception
of beauty is innate (Slater et al., 1998; Langlois et al.,
and Hill, 1993; Cunningham et al., 1995; Perrett et al.,
7These authors contributed equally to this work.
Knutson et al., 2000, 2001; Rogers et al., 1999; Small et
al., 2001; Stein et al., 1998; Thut et al., 1997), including
the nucleus accumbens (NAc), sublenticular extended
amygdala (SLEA) of the basal forebrain, amygdala, hy-
pothalamus, orbitofrontal cortex (GOb), and ventral teg-
mentum (VT) of the midbrain. To control for variable
attention to stimuli, BOLD signal in a control region
known to be modulated by attention, namely the fusi-
form gyrus (GF) (Wojciulik et al., 1998), was also eval-
Data from Behavioral Measures
Two different groups of heterosexual male subjects
were exposed to two distinct behavioral tasks. One
group of subjects rated facial attractiveness, and an-
other group used the keypress procedure to control the
duration of their exposure to these faces.
Rating Face Attractiveness
Eight young heterosexual males viewed the stimuli se-
quentially, rating each face’s attractiveness on a scale
of 1 (“very unattractive”) to 7 (“very attractive”). The
results (Figure 1B) showed a general effect of beauty
[F(1,7) ? 569.9, p ? 0.0001], a general effect of gender
[F(1,7) ? 23.4, p ? 0.0001], and an interaction [F(1,7) ?
5.8, p ? 0.05]. Within the male and female sets, the
differences between beautiful and average faces were
significant (p ? 0.000001) for each comparison. Most
importantly, for our purpose, the difference between the
[t(7) ? 0.99, p ? 0.36].
The ratings varied with exposure, such that the differ-
ence in ratings between average and beautiful female
faces increased from 2.96 on the first exposure to 3.39
and 3.54 on the second and third exposures; this trend
had a significant linear component, p ? 0.003. This oc-
curred because the ratings for the average faces de-
creased (2.33, 2.00, and 1.91 across exposures), while
the ratings for the beautiful faces increased slightly
(5.29, 5.39, and 5.45). By contrast, for the male faces,
ratings of beautiful and average faces both increased
slightly with exposure, with the difference remaining
fairly steady (2.72, 2.61, and 2.70).
A separate cohort of 15 young heterosexual males com-
pleted the keypress task with a mean number of key-
presses per subject of 6726 (SD 4287), which translated
into large effects on image viewing times. The results
(Figure 1C) showed that subjects expended effort only
to increase the viewing time of beautiful female faces.
For all other categories, they keypressed to make the
faces disappear faster. There was a significant effect of
beauty [F(1,14) ? 21.7, p ? 0.001], a significant effect
of gender [F(1,14) ? 22.4, p ? 0.001], and a significant
interaction [F(1,14) ? 24.3, p ? 0.001]. Most importantly
difference between the beautiful males and beautiful
females [t(14) ? 5.1, p ? 0.0002]. Because the overall
duration of the study was fixed (40 min—allowing sub-
jects to control only the allocation of time between the
different categories), the number of exposures was not
Figure 1. Rating and Keypress Results
(A) A sample of the four picture types used in these tasks (from
left to right): beautiful female, average female, beautiful male, and
(B) Eight heterosexual males rated picture attractiveness on a 1–7
beautiful male 5.46 (0.53), average female 2.08 (0.42), average male
(C) A separate cohort of 15 heterosexual males performed a task
where picture viewing time was a function of the number of their
keypresses. Within each gender, the 80 faces were always pre-
sented in a new random order, with beautiful and average faces
intermixed(Ariely etal.,2001). Themeanviewingtimes were:beauti-
ful female 8.67 s (SD 2.1 s), beautiful male faces 5.9 (1.63), average
female 5.25 (1.35), average male 5.33 (1.43).
work subjects performed in order to change the relative
duration they viewed the different images. The keypress
task evaluated whether these categories of faces had
reward values that distinguished them (i.e., along the
dimension of reward intensity; Shizgal, 1999). For the
neuroimaging component of the study, reward region
activity was evaluated using fMRI at 3 T to determine if
signal changes followed the results of the behavioral
tasks. Six brain regions were targeted that have been
associated with reward function in animals (Everitt and
Shizgal, 1999) and humans (Berns et al., 2001; Breiter
Beauty and Reward
Table 1A. Random Effects Analysis in A Priori Regions
AnatomyROIR/LA/P S/I ANOVA Paired Contrasts
?3p ? 0.0060 FBvsFA (p ? 0.07)
FBvsMB (p ? 0.03)
p ? 0. 0901
p ? 0.0643
p ? 0.9820
p ? 0.9933
p ? 0.7860
1B. Random Effects Analysis in Control Region
Anatomy ROIR/LA/PS/I ANOVA Paired Contrasts
p ? 0.6117
p ? 0.2991
Region of interest (ROI) based analysis of fMRI data from targeted brain regions. The Talairach coordinates (Talairach and Tournoux, 1988)
of maxima for the six a priori and two control ROIs are expressed in mm from the anterior commissure: x, right (?)/left (?); y, anterior (?)/
posterior (?); z, superior (?)/inferior (?). ANOVA results from the interaction of Gender and Attractiveness are tabulated; paired contrasts are
listed for ANOVA results with p ? 0.05/6 clusters in hypothesized reward regions ? 0.0083. GF clusters were considered separately and had
to meet the same threshold. FB stands for beautiful female, FA for average female, and MB for beautiful male.
constant across subjects. On average, subjects viewed
the entire set of 80 images three to four times, with the
pattern of results not changing with repeated viewing.
ROI-Based Random Effects Analysis
Foci of Signal Change in Targeted Anatomic Regions.
Data obtained at all time points during the face and
fixation point phases of the paradigm were evaluated by
statistical parametric analysis. Six ROIs were identified,
one in the left GOb, two in the left SLEA, one in the
right SLEA, one in the left NAc, and one in the right VT.
Furthermore, two foci in opposite hemispheres were
found in the control region of the fusiform gyrus (GF).
Hypothesis Testing via ANOVAand Contrasts: A Priori
Reward Regions. Gender and beauty served as the pre-
sampled from each ROI. The ANOVA results from the
interaction of gender and beauty are described in Table
1A. Only the ROI in the left NAc (see Figures 2 and 4)
exhibited a significant general effect [F(1,5) ? 20.9, p ?
0.0060], although two other ROIs suggested trends to-
contrasts of the NAc time courses were significant for
the beautiful female versus average female conditions
(Table 1A). The bar graphs of average percent signal
change across subjects for these conditions (Figure 2)
supports the results of the paired contrasts in the NAc.
observations. For instance, with regard to the NAc, the
signal change in the average male condition is second
only to that of the beautiful female faces and is not
significantly different than it. It is also notable that signal
profiles in the VT and one focus in the SLEA(C) resemble
the signal profiles between the beautiful female condi-
tion and the average female plus beautiful male condi-
tions in the NAc, even if not significant by the analysis
of variance (Table 1A). One other focus of signal change
in the SLEA(D) also qualitatively resembles those in the
NAc [and VT plus SLEA(C)], but differs with regard to
signal change in the average male condition. In general,
thereis aqualitativeresemblancebetween thekeypress
results for the beautiful female, average female, and
Data from Neuroimaging
Initial data analysis focused on six brain regions: the
nucleus accumbens (NAc), sublenticular extended
amygdala (SLEA) of the basal forebrain, amygdala, hy-
pothalamus, orbitofrontal cortex (GOb), and ventral teg-
a region-of-interest based (ROI-based) analysis was ini-
tially performed using individual data (henceforth re-
ferred to as “random effects analysis”), followed by a
voxel-by-voxel post-hoc analysis of data from the aver-
aged cohort. For the ROI-based random effects analy-
sis, we selected all clusters of activated voxels in the
targeted regions that were significant in the comparison
between the four face categories and the fixation point
baseline. These clusters were then used as regions of
interest (ROIs) to sample individual data for an analysis
of variance. For the ROI-based random effects analysis,
type II errors (false negative) might be expected in the
case of (1) opponent responses to different face condi-
tions, which would tend to cancel as a result of the
averaging of all face categories, or (2) responses con-
fined to a small proportion of face conditions, which
would tend to be diluted by such averaging. Thus, a
post-hoc parametric analysis (i.e., using t statistics) was
performed on a voxel-by-voxel basis using data aver-
aged across the cohort (henceforth referred to as “fixed
effects analysis”) to determine if differential patterns of
reward circuitry activation paralleled results from the
two behavioral studies. General effects for beauty and
gender were each evaluated, followed by their corre-
sponding specific effects: beautiful female versus aver-
age female and beautiful male versus average male (i.e.,
tiful male and average female versus average male (i.e.,
Figure 2. Bar Graphs of Mean Percent Signal
Change and Standard Deviation from the Six
ROIs Selected for the Random Effects
Percent signal change is presented, relative
to the experimental baseline, for each of the
stimulus conditions: beautiful female, aver-
age female, beautiful male, and average male
faces. Note the similar signal profiles for the
first three bars in each graph for the NAc(A),
VT(B), SLEA(C), and SLEA(D).
beautiful male conditions, and signal changes for these
three conditions in the NAc, SLEA(C), SLEA(D), and VT.
Region. The ANOVA results from the interaction of gen-
der and beauty are described in Table 1B, and no ROIs
exhibited a significant general effect.
Fixed Effects (Post-Hoc) Analysis of Average Data
A number of reward regions produced significant signal
change (p ? 0.00011) for the general effects of beauty
and gender and the specific effects of the interactions
between beauty and gender. The following sections are
organized around the general effects with specific ef-
fects (interactions) following.
General Effect of Beauty: All Beautiful versus All Aver-
age Face Comparison. The general comparison of all
beautiful versus average faces revealed four clusters of
significant positive signal change in the right GOb, right
SLEA, and bilateral VT. One focus of subthreshold posi-
tive signal change (p ? 0.001) was further noted in the
left SLEA. Two foci of significant negative signal change
were observed in the bilateral SLEA, one of which (row
9 of Table 2) contains ROI C (Table 1A) of the random
effects analysis. All of these foci of signal change, with
the exception of the right SLEA focus of negative signal
change (Table 2, #A8), coincided with foci of signal
effects that follow.
Specific Effect for Beauty: Beautiful versus Average
Female Comparison. For this contrast, four clusters
were observed with significant positive signal changes:
two in the right GOb, one in the left GOb, and one in
signal change were further observed in the left NAc,
bilateral SLEA, and right VT. Three foci of significant
negative signal change were observed: one in the right
GOb, and two in the left GOb.
Of these activation foci, four directly correspond to
foci observed in the general contrast for beauty. In par-
ticular, activation in the VT, SLEA (see Figure 3), and
GOb (Table 2, #B5, B7, B9, B11) is common to both,
although one SLEA focus is countered by an activation
of opposite valence from thebeautiful male versus aver-
Beauty and Reward
Table 2. Post-Hoc Fixed Effects Analysis
NAc and other VS
Activation clusters are identified by post-hoc voxel-by-voxel analysis. “Anatomy” lists the six targeted reward regions, including putative Brodman Area (BA) where appropriate. Activations are identified by row number
(Number) and specific subtraction (A–F). These subtractions include the general comparisons for beauty and gender and their associated specific contrasts: beautiful female versus average female (BF–AF), beautiful
male versus average male (BM–AM), beautiful female versus beautiful male (BF–BM), and average female versus average male (AF–AM). Talairach coordinates (Talairach and Tournoux, 1988) of these activation foci are
listed, and maxima p values are tabulated, where “E” stands for exponent, such that 1.0E?4 means 1.0 ? 10?4. Activation clusters are listed that produced significant post-hoc effects per a p value threshold corrected
for the volume of voxels sampled in a priori targeted regions. Activations with asterisks did not meet this corrected threshold but had maxima with 1.1 ? 10?4? p ? 9.9 ? 10?4. Activations are recorded in the same
row if their foci are within 2 cm of each other; some contrasts have multiple foci proximate to foci in other contrasts, in which case, the focus with the closest maximum to the others is placed in the row with them.
Figure 3. Post-Hoc Analysis (for General Ef-
fects of Beauty, along with Its Specific Con-
Bilateral activation of the SLEA in the general
(rows 2–5). Rows 2 and 3 are the positive
and negative results of the general effect of
activation for the comparison of beautiful fe-
male faces versus average female faces (BF
versus AF), and the negative activation (row
five) for the comparison of beautiful male
faces versus average male faces (BM versus
AM). The SLEA activation (rectangles) in the
contrast of all faces versus baseline and the
general effect B versus A, coincided with
the foci of signal changes observed in the
AF, and BM versus AM). Note that activations
in the specific contrasts are additive and are
thus observed in the general contrast.
age male conditions. Also, clusters in rows 2, 5, and 9
(Table 2) contain ROIs A–C (Table 1A) of the random
Specific Effects for Beauty: Beautiful versus Average
Male Comparison. For this comparison, one focus of
significant positive signal change was identified in the
left VT (Table 2). Four clusters of significant negative
signal change were further observed in the left SLEA
(n ? 2), and the bilateral NAc/ventral striatum. The NAc/
ventral striatum foci had maxima in the ventral striatum
with their clusters covering the NAc (e.g., see Figure 3,
bottom row images for one such example).
One focus of positive signal change and one focus of
negative signal change directly correspond to foci of
positive and negative signal change observed in the
general contrast for beauty. In particular, activation in
the VT and SLEA (see Figure 3 and Table 2, #C4 and
C9) is common to both. Also, clusters in rows 2 and 9
(Table 2) contain ROIs A and C (Table 1A) of the random
General Effect of Gender: All Female versus All Male
Face Comparison. The general comparison of all female
versus all male faces revealed no foci of significant acti-
vation. Subthreshold foci of positive signal change were
Beauty and Reward
observed in the left amygdala, right hypothalamus, and
left VT. A subthreshold focus of negative signal change
was also observed in the left amygdala (Table 2). All of
these foci of signal change coincided with foci of signal
change observed in the comparisons for the specific
effects that follow.
Specific Effect for Gender: Beautiful Female versus
Beautiful Male Comparison. For this contrast, seven
clusters were observed with significant positive signal
change in the left GOb, bilateral SLEA, left NAc/ventral
striatum, left NAc, and right VT (n ? 2) (Table 2). One
focus of subthreshold signal change was observed in
the right NAc. Three foci of significant negative signal
change were observed in the bilateral GOb and left
SLEA, while one focus of subthreshold signal change
was noted in the amygdala.
Although none of these activation clusters account
of them directly overlap foci of opposite signal change
observed in the contrast of average female versus aver-
age male faces described below. An example of this
countervalanced signal change in the left NAc/ventral
striatum can be observed in Figure 4. Also, clusters in
rows 2, 5, and 9 (Table 2) contain ROIs A–C (Table 1A)
of the random effects analysis.
Specific Effects for Gender: Average Females versus
Average Males. For this comparison, six foci of signifi-
cant positive signal change were identified in the right
GOb (n ? 2), left GOb (n ? 2), right hypothalamus, and
left VT (Table 2). Subthreshold positive signal change is
ing significant foci of negative signal change, three foci
are observed: in the right NAc, left NAc/ventral striatum,
change is noted in the left amygdala.
Of these activation foci, three foci of positive signal
spond to foci of subthreshold signal change observed
in the general contrast for gender. In particular, activa-
tion in the amygdala, hypothalamus, and VT (Table 2,
#F4, F17, F18, F21) is common to both. The negative
signal change in the right NAc and left NAc/ventral stria-
tum (Table 2, #F1 and F2; see Figure 4), along with three
other foci of negative and positive signal change (Table
2, #F9, F11, F13), directly overlay oppositely valanced
foci of signal change in the contrast of attractive female
versus attractive male faces described above. It is also
notable that clusters in rows 2, 5, and 9 (Table 2) contain
ROIs A–C (Table 1A) of the random effects analysis.
tified measures of reward valuation, particularly for
beautiful male faces. This dissociation between assess-
ments of attractiveness and quantified measures of re-
ward resembles that described by Berridge (1996, 2000)
regarding the processes of “liking” and “wanting” (also
see Ariely and Loewenstein, 2000). Second, the key-
press procedure revealed that a category of visual stim-
ulus that is not linked to any post experiment asset or
without being associated with any clear hedonic deficit
Third, the fMRI experiment revealed a significant re-
sponse in the nucleus accumbens to the beautiful fe-
male faces by both random effects and fixed effects
analyses. The fixed effects analysis further indicated
NAc, along with the VT, SLEA, and GOb, also produced
significant activation with regard to the same stimulus
conditions. Both the random effects (Figure 2) and fixed
effects analyses (Table 2) revealed relative positive sig-
nal changes in the NAc for heterosexual males viewing
rewarding faces (i.e., beautiful female versus average
female conditions) and relative negative signal changes
for nonrewarding faces (i.e., beautiful male versus aver-
age male conditions). These results have some analogy
to the negative deflection in BOLD signal reported for
versus a nonaversive baseline (Becerra et al., 2001).
In conjunction with results of human neuroimaging
work using drug, gustatory, tactile, and monetary stim-
uli, these beauty results strongly support a fourth point,
that at the spatial scale of subcortical nuclei and their
cortical projection fields, there appears to be a general-
ized circuitry processing rewarding stimuli.
Lastly, the fMRI data suggest a hypothesis for further
study regarding the potential effect of “liking” and
2000). Specifically, we observed patterns of activation
from four of six ROIs in the NAc, SLEA, and VT that
generally follow the keypress results for the beautiful
female, average female, and beautiful male conditions
(Figures 1C and 2). The observation of similar signal
profiles across multiple reward regions resembles re-
ports of similar signal changes across a distributed set
of reward regions with money (Breiter et al., 2001). Al-
though the outcomes for the behavioral tasks and fMRI
procedure were not perfectly symmetric, no reward re-
gion per se mirrored the outcomes of the rating experi-
ment (Figure 5). The fixed effects analysis supported the
random effects analysis by showing that components of
theNAc, SLEA,and VTparalleled thekeypress datawith
regard to subtractions between the beautiful female,
average female, and beautiful male conditions (Figure
5). These results suggest that reward regions respond
of the faces rather than to their aesthetic components.
The goal of this study was to provide evidence for sepa-
rable measures of aesthetic versus rewarding qualities
in beautiful faces and to provide an initial understanding
of the neural mechanisms that underlie these qualities.
To reach this goal, we carried out both behavioral and
fMRI experiments involving the visual assessment of
beautiful and average faces. The results of these experi-
ments produced five primary points.
ation betweenassessments of attractivenessand quan-
Dissociation of Rewarding and Aesthetic
Heterosexual male subjects rated beautiful male faces
as very attractive but did not expend effort to increase
the viewing times of these faces. The most reasonable
inference is that they found them aesthetically pleasing
Figure 4. Post-Hoc Analysis (for General Effects of Gender, along with Its Specific Contrasts)
Activation of the left NAc in at least two of three contiguous brain slices for the contrast of all faces versus baseline (row 1) but not the general
effect of gender (rows 2 and 3 for positive and negative signal changes, respectively). Row 4 shows the positive signal changes from the
comparison of beautiful female faces versus beautiful male faces (BF versus BM), while row 5 shows the negative signal changes of the
comparison of average female faces versus average male faces (AF versus AM). Activation in the left NAc (in boxes) can be observed as foci
of opposite signal change (positive BF versus BM, and negative AF versus AM). This explains the absence of this activation in the general
effect of gender.
suggest that the ratings procedure and the keypress
procedure measure aesthetic value and reward value
respectively. The keypress results further indicate that
visual stimuli can have reward value by an objective
measure, namely work performed to increase or de-
crease exposure time to different images. In general,
the reward value associated with a stimulus is not a
static, intrinsic property, but is rather a function of the
animal’s internal state at the time and of its past experi-
ence with the stimulus. Primary homeostatic functions
logical conditions. In the current study, the beautiful
cific deficit state, post experiment asset position, or
object that could potentially impact on internal physio-
Beauty and Reward
men like to watch beautiful women’s faces; however, it
was not obvious that this class of stimuli would activate
the classical reward circuitry that has previously been
associated with drug rewards, homeostatic rewards,
and monetary rewards, all of which have direct physio-
logical implications or can be readily transformed into
goal-objects with physiological effects. This is a critical
issue since many literatures outside of neuroscience
consider facial attractiveness as a social construct that
is not necessarily tied into the function of fundamental
reward circuitry (reviewed in Etcoff, 1999).
In therandom effectsanalysis of ROIdata fromthe six
statistical threshold and produced subsequent pair-
wise comparisons for contrasts involving signal col-
lected during the observation of beautiful female faces.
The NAc, like the fusiform gyrus (GF), was one of the
few brain regions to even meet a Bonferroni threshold
(p ? 7.1 ? 10?6) for the general contrast of all face
conditions versus baseline used in the ROI-selection
procedure (Breiter et al., 1997, 2001). The NAc achieved
this despite having counterposed activation for condi-
tions that were nonrewarding versus conditions that
were rewarding (see Figure 4). In contrast to the NAc,
the fusiform gyrus (GF) evidenced two foci of BOLD
signal change in the comparison of all face conditions
versus baseline; however, the sex by beauty interaction
of variance. BOLD signal change in the GF in response
to faces has been reported to be affected by alterations
icant contrasts between conditions in the GF for this
sample of subjects argues that the significant effects
observed in the NAc were not strictly due to differences
in subject attention to stimuli.
Given the keypress results, the observation of NAc
activity in response to beautiful female faces implicates
it in the assessment of reward value. Since the classic
formulation of a dedicated reward circuitry (Wise et al.,
1978; Louilot et al., 1989), the meso-accumbens dopa-
of circuitry processing the reinforcing effects of most
has complex reward functions, including the assess-
ment of reward expectancy (Schultz et al., 1997; Berns
et al., 1997; Breiter and Rosen, 1999; Breiter et al., 2001;
Schultz and Dickinson, 2000; Knutson et al., 2001). The
current work extends the set of categories of stimuli
producing NAc activity beyond homeostatic, monetary,
and drug rewards, to include rewards with direct social
relevance, such as beautiful faces.
Fixed effects analysis of averaged data was further
performed to determine if other reward regions pro-
duced differential patterns of activation that paralleled
the dissociation of effects for the beautiful face condi-
tions in the behavioral studies. Indeed, distinct profiles
regarding beauty and gender in other reward regions
besides the NAc and proximate ventral striatum (i.e.,
NAc/ventral striatum), namely the SLEA, VT, and GOb.
Of these regions, no focus of signal change met our
post-hoc corrected statistical threshold for the general
effect of gender (Table 2). This appears to be partly due
to the counterposing of activations (see Figure 4) in the
Figure 5. A Summary of the Behavioral and BOLD Signal Results
for the Specific Effects of Beauty and Gender
The left table in the figure shows “p ? 0.05” when the difference in
the results from the rating and keypress tasks is statistically signifi-
cant and shows “NS” when it is not. The table on the right describes
the BOLD signal changes in the following reward regions: (1) right
NAc (activation clusters listed in row 1 of Table 2), (2) region of the
left ventral striatum proximate to the NAc, including the NAc (NAc/
VS) (this includes the maxima that was localized in the NAc via the
ROI selection procedure for the random effects analysis as ROI A
in Table 1A, and the cluster localized over a larger area from the
fixed effects analysis that is listed in row 2 of Table 2), (3) left SLEA
(this is ROI C in Table 1A that was localized in the SLEA via the
ROI selection procedure for the random effects analysis, and the
activation clusters listed in row 9 of Table 2 from the fixed effects
analysis), and (4) right and left VT (this is ROI B in Table 1A that
was localized in the right VT via the ROI selection procedure for the
random effects analysis, and the activation clusters in rows 5 and
4 of Table 2 from the fixed effects analysis for the right and left VT,
respectively). Activation in brackets were subthreshold per Table
2. In viewing this summary table, it is important to consider the
comparisons as tentative, since there is not perfect parallelism be-
tween the behavioral data and imaging data (see Discussion). With
this in mind, the BOLD signal in the NAc, NAc/ventral striatum (NAc/
VS), SLEA, and VT produce positive signal changes when p ? 0.05
for the keypress task (emphasized with text in bold tone). These
regions also alternate signal valence across conditions, while VT
activation switches laterality; these observations suggest a need
for greater dynamic range in the behavioral tasks to fully determine
their correspondence with reward circuitry signal change.
logical signals. Further work is clearly needed to evalu-
ate alliesthesia effects (Cabanac, 1971, 1979) for this
category of stimuli. Further behavioral tasks are also
needed to evaluate informational features not assessed
in the current study. For instance, it is possible that
young heterosexual males experience the perception
of beautiful males in a more aversive fashion than the
spondence of NAc signal change between beautiful
male and average male faces, to NAc signal change
between aversive and nonaversive thermal stimuli (Be-
cerra et al., 2001).
Reward Circuitry Activation to Facial Stimuli
One goal of this study was to determine if beautiful
faces, or a subgroup thereof, motivated behavior on
one hand, while on the other producing reward circuitry
specific contrasts of beautiful women versus beautiful
men and average women versus average men (Table 2),
thus reducing their additive effect for the general effect
of gender. In the case of beauty, the general effect
largely reflected additive inputs (see Figure 3) between
responses for beautiful women versus average women
and beautiful men versus average men (Table 2).
Alongside the observations in the NAc, NAc/ventral
striatum, and VT, the post-hoc observations of signal
change in the GOb and SLEA is consistent with exten-
sive animal experimentation. GOb neurons in the rat
(Schoenbaum et al., 1999) and monkey (Rolls, 1999;
Schultz and Dickinson, 2000) fire during the anticipation
and experience of positive and negative outcomes. Fur-
thermore, the responses of GOb neurons may represent
relative reward preferences (Schultz and Dickinson,
2000). This latter observation has particular relevance
to the current data showing GOb activation for subtrac-
tionsinvolving beautifulfemalefacescompared toother
face conditions. The presence of GOb activity in the
average female versus average male subtraction may
also be similarly interpreted if one assumes the behav-
ioral tasks did not have enough dynamic range to distin-
guish a preference for average females over average
males. Alternately, the GOb activation associated with
this contrast could reflect the assessment of conflict
between options connected to the stimuli, as has been
shown to be important for choices between gambling
positions (Bechara et al., 1998).
Neurons in the SLEA are activated by rewarding brain
stimulation in the rat (Arvanitogiannis et al., 1996, 2000;
Flores et al., 1997; Nakahara et al., 1999; Shizgal, 1997).
Furthermore, lesions of the SLEA reduce the rewarding
effect of stimulation to the medial forebrain bundle (Ar-
vanitogiannis et al., 1996), reduce cocaine self-adminis-
tration (Robledo and Koob, 1993), and disrupt operant
performance for sucrose pellets in rats (Brown et al.,
the SLEA, GOb, and VT may respond to nonrewarding
stimuli: transient activation in these regions has been
reported following painful thermal stimulation (Becerra
et al., 2001).
The patterns of signal change observed in reward
regions with the random effects and the fixed effects
analyses suggest some potential parallels with the be-
havioral tasks (Figures 1C, 2, and 5). Such comparisons
must be considered tentative since there is not perfect
parallelism between the behavioral data and imaging
data in that the imaging data has a broader scale of
measurement and thus greater sensitivity (suggesting a
need for greater sensitivity of measurement in future
behavioral tasks). Signal change from some of the ROIs
(e.g., ROIs A–D from Table 1A and Figure 2) used in the
random effects analysis suggest a qualitative similarity
to a subset of results from the keypress procedure (i.e.,
the beautiful female, average female, and beautiful male
conditions) (Figure 1C). The results of the fixed effects
analysis support this observation, in that clusters of
activation in the right NAc (row 1 of Table 2), left NAc/
ventral striatum (row 2 of Table 2), left SLEA (row 9 of
Table 2), and right VT (row 5 of Table 2) produce positive
signal changes for comparisons between conditions
that are significant for the keypress task (Figure 5).
These results support the observations of the random
effects analysis in that three of these clusters overlap
ROIs in the NAc(A), VT(B), and SLEA(C) used in the ran-
dom effects analysis (Table 1A and Figure 2). If one
looks at all of the contrasts for the specific effects of
beauty and gender, one notes that the BOLD signal in
valence across conditions, while VT activation switches
laterality (Table 2 and Figure 5). Suggestions that these
activations follow the outcomes of the keypress task
are not warranted, though, since there is a potential
SLEA, and VT (e.g., ROIs A–C in Figure 2) during the
average malecondition and thekeypress resultsfor that
experimental condition (Figure 1C). Behavioral work to
characterize responses to the average male faces may
be helpful in this regard. Furthermore, future studies
with more sensitive behavioral experiments might con-
sider use of the same subjects for both behavioral and
imaging tasks to allow quantitative evaluation via multi-
variate analysis of these observations.
The keypress task indicated a significant difference
between categories of faces when one of the categories
was that of the beautiful female faces (which was the
only condition to be positively reinforcing). The other
indicating that they were mildly aversive). The presence
ing reinforcing rewards versus contrasts involving non-
lel to data indicating positive BOLD signal changes to
rewards (such as cocaine and morphine; Breiter et al.,
1997, 2000), and the report of negative BOLD signal
changes to painful stimuli (Becerra et al., 2001). Given
that the changes in signal observed for rewarding and
nonrewarding outcomes is dependent on expectancy
conditions (Breiter et al., 2001), such differences in sig-
nalvalence intheNAc andotherrewardregions maynot
be absolute but represent a relative ordering of signal
associated with the experimental conditions.
General Circuitry of Reward and Its Activity
during Aesthetic Assessment
A similar set of brain reward regions appears to respond
in common to very distinct categories of reward. At the
ogy, at least five brain regions (plus or minus the hypo-
have been implicated in the perceptual assessment of
rewarding stimuli. In particular, the NAc, SLEA, amyg-
dala, GOb, and VT have all been implicated in expec-
tancy functions (Breiter et al., 2001) and to various de-
grees in the processing of positive and negative
outcomes. Inhuman neuroimagingstudies, stimulilead-
ing to significant signal changes in these regions have
and morphine, pleasant or aversive tastants, pleasant
tactile stimuli, and monetary rewards (Berns et al., 2001;
Breiter et al., 1996b, 1997, 1998, 2000, 2001; Breiter and
Rosen, 1999; Delgado et al., 2000; Drevets et al., 2001;
Elliott et al., 2000; Francis et al., 1999; Knutson et al.,
2000, 2001; O’Doherty et al., 2001; Rogers et al., 1999;
Small et al., 2001; Stein et al., 1998; Thut et al., 1997;
Zald et al., 1998). Such congruence of findings between
Beauty and Reward
brain regions producing responses at the spatial scale
of cubic millimeters of tissue and reward stimuli that
of a common generalized circuitry that processes re-
ward information across category and dissects discrete
features of such stimuli for the planning of behavior
(Breiter and Rosen, 1999; Breiter et al., 2001). Future
tributions of these distributed brain regions to the as-
sessment of reward and aversion, and the organization
of behavior around these assessments.
Together with the contrasting results of the rating and
keypress procedures, our fMRI results suggest a hy-
pothesisthat aestheticevaluationmaybe separatefrom
reward assessment; although both might follow a com-
inspection of results from the fixed effects analysis sug-
gests that the VT, SLEA, and GOb may be associated
with the general effect of beauty. Of these regions, one
focus in the SLEA (Table 2, #A8), in particular, is unique
to the general effect and cannot be accounted for by
the subsequent specific effects. Such an interpretation
would be supported by the fact that the SLEA and VT
share connections through the medial forebrain bundle,
which is an important site of brain stimulation reward
effects (Olds and Milner, 1954) and has been theorized
gal, 1999). VT dopamine neurons have also been inter-
preted to respond to salient sensory events, regardless
of their relationship to reward (Breiter et al., 2001; Hor-
Arguing against a common basis for aesthetic and
reward assessments, though, is the observation that of
the two behavioral tasks, only the keypress procedure
produced responses that paralleled to some degree re-
sponses in reward regions such as the NAc, SLEA, and
VT (Figures 1C, 2, and 5). No reward region responded
in a similar fashion to the outcomes of the rating proce-
dure. These results are consistent with predictions of
by Berridge and colleagues (Berridge, 1996, 2000),
which discusses “wanting,” but not “liking,” in terms of
meso-accumbens dopaminergic function (Wyvell and
A longstanding debate in aesthetics concerns the
question of whether perception of beauty can be “disin-
The post-hoc analysis of our data did not show com-
pletely separated patterns of activation in reward re-
gions for aesthetic assessments and reward assess-
ments, but instead demonstrated a set of regions that
resembled aspects of the keypress results alone. This
suggests that beautiful faces produce a valuation signal
(potentially involving the SLEA), which is processed in
different brain regions for reward functions and aes-
thetic judgements in young heterosexual male subjects.
Given that it may be particularly difficult to dissociate
rewarding fromaesthetic effects with faces,future stud-
ies with non-face stimuli and nonbiological categories
of goal-objects may help determine if these effects are
A number of limitations apply to this fMRI study (see
Breiter et al., 1996a, 1996c, 1997, 2001). These issues
include the limited signal-to-noise ratios of BOLD sig-
nals recorded from small subcortical structures, spatial
resolution after data analysis, and magnetic suscepti-
Particular to this study, it is important to note that
participants in all three experiments were young hetero-
sexual males. Recent evidence suggests that women’s
facial preferences may change across the menstrual
cycle. Penton-Voak et al. (1999) have found that women
in the follicular phase of the menstrual cycle are more
petence and therefore prefer stereotypically “mascu-
line” faces more than they do at other times of their
cycle. Given the more complex results of human female
tion with male subjects only. Thus, these results should
not be generalized to women, nor to individuals with a
different sexual orientation.
Finally, it is possible that the difference between the
two behavioral tasks was not caused by a difference
was a result of scaling changes in the ratings task. It
is possible that when using ratings as the response
measure, our subjects rated males relative to other
males and females relative to other females (Ariely and
Loewenstein, 2000). The keypress method does not suf-
fer from this changing scale problem because effort is
a scale that is constant across all images presented.
Further work is needed in order to better understand if
the rating task results are a response language artifact
or a real effect.
This study sought to determine if social stimuli distinct
from money, namely categories of faces segregated by
gender and attractiveness, had different reward values,
and in turn induced reward circuitry activity when ob-
served. The keypress procedure demonstrated that a
social stimulus that was not linked to any post experi-
ment asset or reward, had a distinct valuation that could
be objectively quantified. fMRI of subjects passively
viewing these faces demonstrated a significant effect
by analysis of variance across individual subjects in the
NAc, particularly in response to the beautiful female
faces. By the post-hoc analysis, no one region re-
sponded in a similar fashion to the rating study alone,
but anumber of reward regionsproduced signal change
with approximate similarity to the keypress data. These
results suggest that aesthetic judgement may not be
mediated by reward systems.
The observations of this study build upon prior neuro-
imaging work with multiple categories of reward stimuli
to argue that, at the spatial resolution of current imaging
techniques, an extended set of deep gray matter and
paralimbic regions serves as a generalized circuitry to
process goal-related stimuli. In a previous study using
monetary reward and a game of chance, our group ob-
served this extended set of reward regions to display
differential activation to expectancy conditions and to
the monetary gains and losses embedded in them
selected from print media. Each set consisted of 20 male and 20
(Breiter et al., 2001). The segregation of expectancy and
outcome responses demonstrated that reward/motiva-
tion systems could be dissected into their constituent
functional subsystems, but left open the issue of their
involvement in valuation processes. The current study
suggests that the same set of regions implicated in the
processing of expectancies and outcomes, may also be
involved with valuation functions. A challenge for future
work is to determine the varied contributions of this
distributed set of reward regions to the processes of
expectancy, valuation, and their combination for moti-
by the design and focus of the current work. For in-
stance, there is the issue that differences in the gender
might lead to different behavioral and imaging results.
The signals from reward regions in response to average
male faces did not resemble the results of the two be-
havioral tasks, suggesting the need for assessing other
subjective dimensions. Another question arises about
ize to those produced by bodies, or nonanimate stimuli.
Lastly, there is the issue of what brain regions might
process information regarding “aesthetic” beauty (i.e.,
follow the pattern of the rating task results), as opposed
to processing information regarding “desirable” beauty.
In pursuing these issues regarding brain function, it is
possible we may begin to approach an understanding
of Ruskin’s concept of the “sublime” (reprinted 1997)
with regard to the grandeur imbedded in the function
Rating Procedure. The rating task measured the attractiveness of
the female and male faces for the experimental subjects. Subjects
viewed the 80 faces three times each, on each occasion rating each
face’s “attractiveness” on a scale of 1–7, with 1 representing “very
unattractive” and 7 representing “very attractive.” Faces were orga-
nized in blocks by gender, and the order of gender was counterbal-
ful and average faces were always presented in a new randomly
Keypress Procedure. The keypress task examined the reward
value of average and beautiful faces. Subjects were told that they
would be exposed to a series of pictures that if not interfered with,
would change every eight seconds. However, if they wanted a pic-
ture to disappear faster, they could alternate pressing the “z” and
“x” keys, whereas if they wanted a picture to stay longer on the
screen, they could alternate pressing the “n” and “m” keys. The
dependent measures of interest were the amount of work in units
of key press that subjects exerted in response to the different cate-
gories of stimuli, and their resulting viewing durations.
Each pair of keypresses increased or decreased the total viewing
time according to the following formula:
NewTotalTime ? OldTotalTime ? (ExtremeTime ?
where ExtremeTime was 0 s for keypresses reducing the viewing
time, ExtremeTime was 16 s for keypresses increasing the viewing
time, and K was a scaling constant set to 40. If the elapsed time
for the picture surpassed the total time determined by keypressing,
the picture was removed and the next trial began. A “slider” was
displayed left of each picture indicating total viewing time at any
moment, and changing with every keypress (Ariely et al., 2001). This
procedure was controlled by Authorware (Macromedia).
Subjects were informed that the task would last 40 min and that
this length was independent of their behavior during the task, as
was their overall payment of $14.
Subject Instructions Prior to Scanning. Subjects were told that they
would see rapid presentations of faces intermixed with fixation
points, that sometimes the faces would change, and that they were
to focus on the fixation points while getting an overall sense of
the faces. One to two minutes prior to imaging, subjects viewed a
separate set of average faces with neutral expressions for image
focusing and centering.
Functional Imaging. Subjects were scanned on an Instascan de-
vice (3 T General Electric Signa, modified by Advanced NMR Sys-
tems) using a head coil, bite-bar, and standard protocol (Breiter
et al., 2001). This protocol included: (1) a sagittal localizer scan
[conventional T1-weighted spoiled gradient refocused gradient
echo (SPGR) sequence; through-plane resolution ? 2.8 mm; 60
slices] to orient, for subsequent scans, 16 contiguous axial oblique
with second order shims to optimize magnetic field homogeneity
(mean FWHM ? 26.6Hz, SD 2.6), (3) an SPGR T1-weighted flow-
compensated scan (resolution ? 1.6 mm ? 1.6 mm ? 7 mm), (4) a
T1-weighted echo planar inversion recovery sequence (TI ? 1200
ms, in-plane resolution ? 1.6 mm), and (5) a gradient echo, T2*-
weighted functional sequence (TR ? 2000 ms, TE ? 30 ms; flip ?
60?; FOV ? 40 ? 20 cm; in-plane resolution ? 3.125 mm; 110
images per slice; disdaq ? 4).
Experimental Paradigm. Each experimental run included five 28 s
baseline epochs interleaved with four 20 s face epochs (e.g., block
design). To minimize attentional modulation of gaze (Breiter et al.,
1996a), face stimuli were presented in a tachistoscopic fashion for
200 ms (face, or empty oval with a fixation point for baseline), fol-
lowed by a blank screen for 3800 ms. During each run, either male
or female faces were presented, with alternating epochs of average
and beautiful faces. Epochs of average and beautiful faces were
All subjects in the three experiments gave informed consent to
participate in these procedures following the rules of the Subcom-
mittee on Human Studies at the Massachusetts General Hospital,
the Institutional Review Board at the Massachusetts Institute of
Technology, and the Institutional Review Board at Harvard Uni-
A group of eight male subjects, ages 21–35 years (mean 24.0, SD
4.4), all heterosexual and right-handed by self-report, participated
in the rating study, and were paid $15. A separate group of fifteen
male subjects between the ages of 21 and 25 years (mean 23.0, SD
2.3), also all heterosexual and right-handed by self-report, partici-
pated in the “keypress” study and were paid $14 for their partici-
Ten male volunteers, who had not participated in the behavioral
tasks, volunteered for the fMRI experiment. All subjects were medi-
cally, neurologically, and psychologically normal by self-report and
physician-directed medical review of systems. All subjects were
right-handed and heterosexual by self-report. fMRI data collected
from three participants had significant signal spiking, along with
uncorrectable motion, and could not be analyzed; data from one
participant was excluded due to noncompliance with instructions.
The six remaining participants were aged 21–28 (mean 25.2, SD 2.5).
Two sets of 40 nonfamous human faces [digitized at 600 dpi in 8-bit
grayscale, spatially downsampled, and cropped to fit in an oval
“window” sized 310–350 pixels wide by 470 pixels high using Pho-
toshop 4.0 software (Adobe Systems)] classified as “beautiful” and
“average” (according to pilot test results) (see Figure 1A), were
Beauty and Reward
counterbalanced within run, with an *A*B*B*A* or *B*A*A*B* order
(A ? average, B ? beautiful, * ? fixation point baseline). The gender
of the faces shown and block sequence was counterbalanced
across the eight runs administered to each subject. This resulted
in each face being viewed twice. A break of approximately 2–4 min
was taken between runs.
Subject Debriefing. After functional imaging, subjects filled out a
questionnaire about what they had seen and thought, and com-
pleted the Beck Anxiety and Depression Inventories (BAI: mean 2.2,
SD 1.7, range 0–4 out of 63; BDI: mean 3.7, SD 3.8, range 0–12 out
of 63). Three of the six subjects spontaneously reported noticing
that the faces differed in attractiveness.
an overall ? ? 0.05, activation clusters had to meet a corrected p
value threshold (Breiter et al., 1996a, 1996c, 2001) for the volume
of tissue (30.97cc) sampled in the six a priori regions (i.e., p ? 0.05/
453 voxels, or p ? 1.1 ? 10?4) (Makris et al., 1999).
Statistical maps of group-averaged data were superimposed over
high-resolution conventional T1-weighted images that had been
transformed into the Talairach domain and averaged. Primary ana-
tomic localization of activation foci was based on the Talairach
coordinates (Talairach and Tournoux, 1988) of the maximum voxel
from each activation cluster, with secondary confirmation of this via
inspection of the juxtaposition of statistical maps with structural
scans. Localization followed the region of interest conventions de-
scribedpreviously (Breiteret al.,1997)for theNAc, SLEA,amygdala,
hypothalamus, VT, and GOb (BA 11/47) (Breiter et al., 2001), and
the ventral striatum proximate to the NAc (Drevets et al., 2001). All
clusters of activation were checked against the functional image
any areas of susceptibility artifact (Breiter et al., 1997, 2001).
Transformation of fMRI BOLD Data before
Motion Correction and Talairach Transformation. These procedures
followed those detailed elsewhere (Breiter et al., 2001).
Signal Normalization and Filtering. For all eight runs, fMRI data
in the Talairach domain were intensity scaled on a voxel-by-voxel
basis to a standard value of 1000 and detrended. The mean signal
intensity for each voxel over all runs was removed on a time point
by time point basis.
Concatenating and Averaging across Subjects. Pairs of identical
110-time point runs of Talairach-transformed functional data were
averaged within each subject (i.e., given four types of runs—female
faces: *A*B*B*A*, female faces: *B*A*A*B*, male faces: *A*B*B*A*,
enated within each subject. These 440 time point sequences were
Statistical Mapping, ROI-Based Analysis
Statistical Mapping of General Effects as ROIs. Individual data was
evaluated to confirm the presence of a normal signal distribution.
Then, a t statistic map was created from the data averaged across
subjects, using time point (n ? 440) as the random factor, and
contrasting all time points during face events versus all time points
during fixation point events.
Within this map, clusters of activation that had maxima (i.e., mini-
mum p value) in one of the six targeted reward regions were identi-
fied using a cluster-growing algorithm (Bush et al., 1996), and se-
lected as ROIs for analyses of variance with subject (n ? 6) as the
random factor (i.e., “random effects” analysis). In order to maintain
an overall ? ? 0.05, this algorithm specifically localized activations
that met a corrected p value threshold of p ? 0.0083 for the number
of hypothesized brain regions interrogated (0.05/6 ? 0.0083). See
Breiter et al. (2001) for specific cluster constraints.
Hypothesis Testing of Individual Data: ANOVAs and Paired Con-
trasts. ROIs in targeted reward regions were then used to sample
signal (as percentage signal change) from each of the eight runs in
each subject. Separate, repeated-measures analyses of variance
for each ROI were computed, with Gender (male or female) and
Attractiveness (beautiful or average) as the dependent measures.
The resultof primary interestin the ANOVAwas the gender? attrac-
tiveness interaction. Given that ANOVAs were carried out for six
different ROIs in hypothesized reward regions, we used a more
stringent ? level (p ? 0.0083). In cases that met the criterion ? level,
the pair-wise contrasts were computed (Table 1).
Statistical Mapping, Post-Hoc Voxel-by-Voxel Parametric
Analysis on Averaged Data
A post-hoc analysis on the averaged data was undertaken on a
voxel-by-voxel basis to check for potential activations missed in
the ROI-based analysis (see “Data from Neuroimaging” in Results).
Parametric statistical maps (using voxel-by-voxel t tests) were gen-
erated on data that had been averaged by the four types of runs
data were intensity scaled (i.e., normalized to the first run). General
contrasts were computed for all beautiful face conditions versus all
average face conditions (i.e., general effect of beauty) and for all
female faces versus all male faces (i.e., general effect for gender).
The four specific contrasts following off of these general effects are
listed in Table 2.
Clusters of positive and negative signal change were identified
for each contrast in the six targeted regions. In order to maintain
This work was supported by grants to Dr. Breiter from the National
Institutes of Drug Abuse (grants #00265 and #09467), the Office of
National Drug Control Policy and Counterdrug Technology Assess-
ment Center (ONDCP-CTAC), and Drs. Breiter and Aharon from the
National Center for Responsible Gaming (NCRG). Dr. Breiter was
also supported by the National Foundation for Functional Brain Im-
aging. Dr. Etcoff was supported by the Lynn M. Reid Fellowship of
Harvard Medical School, and Dr. Chabris by a postdoctoral fellow-
ship from the National Institutes of Health through the MGH-NMR
Center. Wewould like tothank Alex Pentland andElizabeth Huffman
for their help and assistance and Mark E. Glickman for statistical
Received June 5, 2001; revised October 12, 2001.
Ariely, D., and Loewenstein, G. (2000). The importance of duration
in ratings of, and choices between, sequences of outcomes. J. Exp.
Psychol. 129, 508–523.
Ariely, D., Breiter, H.C., and Aharon, I. (2001). From mice to men: a
continuous measurement of motivation. MIT Working Paper 4158.
Arvanitogiannis, A., Waraczynski, M., and Shizgal, P. (1996). Effects
of excitotoxic lesions of thebasal forebrain on MFB self-stimulation.
Physiol. Behav. 59, 795–806.
Shizgal, P. (2000). Fos expression following self-stimulation of the
medial prefrontal cortex. Behav. Brain Res. 107, 123–132.
Bartels, A., and Zeki, S. (2000). The neural basis of romantic love.
NeuroReport 11, 3829–3834.
Becerra, L., Breiter, H.C., Wise, R., Gonzalez, R.G., and Borsook,
D. (2001). Activation of reward circuitry following noxious thermal
stimuli. Neuron, in press.
Bechara, A., Damasio, H., Tranel, D., and Anderson, S.W. (1998).
Dissociation of working memory from decision making within the
human prefrontal cortex. J. Neurosci. 18, 428–437.
Berns, G.S., Cohen, J.D., and Mintun, M.A. (1997). Brain regions
responsive to novelty in the absence of awareness. Science 276,
Berns, G.S., McClure, S.M., Pagnoni, G., and Montague, P.R. (2001).
Predictability modulates human brain response to reward. J. Neu-
rosci. 21, 2793–2798.
Berridge, K.C. (1996). Food reward: brain substrates of wanting and
liking. Neurosci. Behav. Rev. 20, 1–25.
Berridge, K.C. (2000). Reward learning: reinforcement, incentives
and expectations. In The Psychology of Learning and Motivation,
Vol. 40. D.L. Medin, ed. (New York: Academic Press).
Breiter, H.C., and Rosen, B.R. (1999). Functional magnetic reso-
nance imaging of brain reward circuitry in the human. Ann. NY Acad.
Sci. 877, 523–547.
Breiter, H.C., Etcoff, N.L., Whalen, P.J., Kennedy, W.A., Rauch, S.L.,
Buckner, R.L., Strauss, M.M., Hyman, S.E., and Rose, B.R. (1996a).
Response and habituation of the human amygdala during visual
processing of facial expression. Neuron 17, 875–887.
Activation of striatum and amygdala during reward conditioning: an
fMRI study. Neuroimage 3, S220.
Breiter, H.C., Rauch, S.L., Kwong,K.K., Baker, J.R., Weisskoff, R.M.,
Kennedy, D.N., Kendrick, A.D., Davis, T.L., Jiang, A., Cohen, M.S.,
et al. (1996c). Functional magnetic resonance imaging of symptom
provocation in obsessive-compulsive disorder. Arch. Gen. Psychia-
try 53, 595–606.
Breiter, H.C., Gollub, R.L., Weisskoff, R.M., Kennedy, D.N., Makris,
N., Berke, J.D., Goodman, J.M., Kantor, H.L., Gastfriend, D.R., Rior-
den, J.P., et al. (1997). Acute effects of cocaine on human brain
activity and emotion. Neuron 19, 591–611.
Breiter, H.C., Gollub, R.L., Edmister, W., Talavage, T., Makris, N.,
Melcher, J., Kennedy, D., Kantor, H., Elman, I., Riorden, J., et al.
(1998). Cocaine induced brainstem and subcortical activity ob-
served through fMRI with cardiac gating. Paper presented at: Inter-
national Society for Magn. Reson. Med. (Sydney, Australia).
Breiter, H.C., Becerra, L., Gonzalez, R.G., Jenkins, L., Huffman, E.,
Harter, K., Comite, A., and Borsook, D. (2000). Morphine induced
reward and pain circuitry activation in drug naı ¨ve humans. Paper
presented at: American Pain Society (Atlanta, USA).
Breiter, H.C., Aharon, I., Kahneman, D., Dale, A., and Shizgal, P.
(2001). Functional imaging of neural responses to expectancy and
experience of monetary gains and losses. Neuron 30, 619–639.
Brown, V.J., Latimer, M.P., and Winn, P. (1996). Memory for the
amygdala. Brain Res. Bull. 39, 163–170.
Bush, G., Jiang, A., Talavage, T., and Kennedy, D. (1996). An auto-
matedsystemfor localizationandcharacterizationof functionalMRI
activations in four dimensions. Paper presented at: Second Interna-
tional Conference on Functional Mapping of the Human Brain (Or-
lando, FL, Academic Press).
Cabanac, M. (1971). Physiological role of pleasure. Science 173,
Cabanac, M. (1979). Sensory pleasure. Q. Rev. Biol. 54, 1–29.
Cunningham, M., Roberts, A.R., Barbee, A.P., Cruen, P.B., and Wu,
C.-H. (1995). Consistency and variability in the cross-cultural per-
ception of female physical attractiveness. J. Personality and Social
Psychol. 68, 261–279.
Delgado, M.R., Nystrom, L.E., Fissell, K., Noll, D.C., and Fiez, J.A.
(2000). Tracking the hemodynamic responses for reward and pun-
ishment in the striatum. J. Neurophys. 84, 3072–3077.
Drevets, W.C., Gautier, C., Price, J.C., Kupfer, D.J., Kinahan, P.E.,
Grace, A.A., Price, J.L., and Mathis, C.A. (2001). Amphetamine-
induced dopamine release in human ventral striatum correlates with
euphoria. Biol. Psych. 49, 81–96.
Elliott, R., Friston, K.J., and Dolan, R.J. (2000). Dissociable neural
responses in human reward systems. J. Neurosci. 20, 6159–6165.
Etcoff, N. (1999). Survival of the Prettiest (New York: Doubleday).
Everitt, B.J., and Robbins, T.W. (1992). In The Amygdala: Neurobio-
logical Aspects of Emotion, Memory, and Mental Dysfunction. J.P.
Aggleton, ed. (New York: Wiley-Liss), pp. 401–429.
Flores,C.,Arvanitogiannis, A.,andShizgal,P. (1997).Fos-likeimmu-
noreactivity in forebrain regions following self- stimulation of the
lateral hypothalamus and the ventral tegmental area. Behav. Brain
Res. 87, 239–251.
ning, A., Clare, S., and Smith, E. (1999). The representation of pleas-
ant touch in the brain and its relationship with taste and olfactory
areas. Neuroreport 10, 453–459.
Frieze, I.H., Olson, J.E., and Good, D.C. (1990). Perceived and actual
discrimination in the salaries of male and female managers. J. Appl.
Soc. Psychol. 20, 46–67.
Frieze, I.H., Olson, J.E., and Russell, J. (1991). Attractiveness and
income for men and women in management. J. Appl. Soc. Psychol.
Gallistel, C.R. (1990). The Organization of Learning (Cambridge, MA:
Grammer, K., and Thornhill, R. (1994). Human (Homo sapiens) facial
attractiveness and sexual selection: the role of symmetry and aver-
ageness. J. Comp. Psychol. 108, 233–242.
Hamermesh, D.S., and Biddle, J.E. (1994). Beauty and the labor
market. American Economic Review 84, 1174–1194.
Horvitz, J.C. (2000). Mesolimbocortical and nigrostriatal dopamine
responses to salient non-reward events. Neuroscience 96, 651–656.
Jones, D.M., and Hill, K. (1993). Criteria of facial attractiveness in
five populations. Human Nat. 4, 271–296.
Kant, I. (1960). Observations on the Feeling of the Beautiful and
Sublime (Trans. Goldthwait, J.T.U. California Press, original
Kanwisher, N., McDermott, J., and Chun, M. (1997). The fusiform
face area: a module in human extrastriate cortex specialized for the
perception of faces. J. Neurosci. 17, 4302–4311.
Knutson, B., Adams, C.M., Fong, G.W., and Hommer, D.J. (2001).
Anticipation of increasing monetary reward selectively recruits nu-
cleus accumbens. J. Neurosci. 15, 1–5.
Knutson, B., Westdorp, A., Kaiser, E., and Hommer, D. (2000). FMRI
visualization of brain activity during a monetary incentive delay task.
Neuroimage 12, 20–27.
Langlois, J.H., Roggman, L.A., Casey, R.J., Ritter, J.M., Rieser-Dan-
ner, L.A., and Jenkins, V.Y. (1987). Infant preferences for attractive
faces: rudiment of a stereotype? Dev. Psychol. 23, 363–369.
Langlois, J.H., Ritter, J.M., Riggman, L.A., and Vaughn, L.S. (1991).
Facial diversity and infant preferences for attractive faces. Dev.
Psychol. 27, 79–84.
Louilot, A., Taghzouti, K., Simon, H., and LeMoal, M. (1989). Limbic
Makris, N., Meyer, J.W., Bates, J.F., Yeterian, E.H., Kennedy, D.N.,
and Caviness, V.S. (1999). MRI-based topographic parcellation of
human cerebral white matter and nuclei II. Rationale and applica-
tions with systematics of cerebral connectivity. Neuroimage 9,
Marlowe, C.M., Schneider, S.L., and Nelson, C.E. (1996). Gender
and attractiveness biases in hiring decisions: are more experienced
managers less biased? J. Appl. Psychol. 81, 11–21.
Morris, J.S., Frith, C.D., Perrett, D.I., Rowland, D., Young, A.W.,
Calder, A.J., and Dolan, R.J. (1996). A differential neural response
in the human amygdala to fearful and happy facial expression. Na-
ture 383, 812–815.
Nakahara, D., Ishida, Y., Nakamura, M., Kuwahara, I., Todaka, K.,
and Nishimori, T. (1999). Regional differences in desensitization of
c-Fos expression following repeated self-stimulation of the medial
forebrain bundle in the rat. Neuroscience 90, 1013–1020.
Nakamura, K., et al. (1998). Neuroanatomical correlates of the as-
sessment of facial attractiveness. NeuroReport 9, 753–757.
O’Doherty, J., Kringelbach, M.L., Rolls, E.T., Hornack, J., and An-
drews, C. (2001). Abstract reward and punishment representations
in the human orbitofrontal cortex. Nat. Neurosci. 4, 95–102.
Olds, J., and Milner, P.M. (1954). Positive reinforcement produced
by electrical stimulation of septal area and other regions of rat brain.
J. Comp. Physiol. Psychol. 47, 419–427.
Penton-Voak, I.S., Perrett, D.I., Castles, D.I., Kobayashi, T., Burt,
D.M., Murray, L.K., and Minamisawa, R. (1999). Menstrual cycle
alters face preference. Nature 399, 741–742.
Perrett, D.I., et al. (1998). Effects of sexual dimorphism on facial
attractiveness. Nature 394, 884–887.
Perrett, D.I., May, K.A., and Yoshikawa, S. (1994). Facial shape and
judgments of female attractiveness. Nature 368, 239–242.
Perrett, D.I., Burt, M.D., Penton-Voak, I.S., Lee, K.J., Rowland, D.A.,
Beauty and Reward Download full-text
et al. (1999). Symmetry and human facial attractiveness. Evol. Hu-
man Behav. 20, 295–307.
Phillips, M.L., Williams, L., Senior,C., Bullmore, E.T., Brammer, M.J.,
Andrew, C., Williams, S.C.R., and David, A.S. (1999). A differential
neural response to threatening and non-threatening negative facial
expressions in paranoid and non-paranoid schizophrenics. Psych.
Res. Neuroimag. 92, 11–31.
Robledo, P., and Koob, G.F. (1993). Two discrete nucleus accum-
tion in the rat. Behav. Brain Res. 55, 159–166.
Rogers, R.D., et al. (1999). Choosing between small, likely rewards
and large, unlikely rewards activates inferior and orbital prefrontal
cortex. J. Neurosci. 20, 9029–9038.
Rolls, E.T. (1999). The Brain and Emotion (Oxford: University of Ox-
Ruskin, J. (1997). Lectures on Art. Reprint edition (Allsworth Press).
Schoenbaum, G., Chiba, A.A., and Gallagher, M. (1999). Neural en-
coding in orbitofrontal cortex and basolateral amygdala during ol-
factory discrimination learning. J. Neurosci. 19, 1876–1884.
Schultz, W., Dayan, P., and Montague, P.R. (1997). A neural sub-
strate of prediction and reward. Science 275, 1593–1599.
Schultz, W., and Dickinson, A. (2000). Neuronal coding of prediction
errors. Annu. Rev. Neurosci. 23, 473–500.
Shizgal, P. (1997). Neural basis of utility estimation. Curr. Opin. Neu-
robiol. 7, 198–208.
Shizgal, P. (1999). In Well Being: The Foundations of Hedonic Psy-
chology. D. Kahneman, E. Diener, and N. Schwarz, eds. (New York:
Russell Sage Foundation), pp. 502–526.
worth, G., Parsons, S., and Samuels, C. (1998). Newborn infants
prefer attractive faces. Infant Behav. Devel. 21, 345–354.
Small, D.M., Zatorre, R.J., Dagher, A., Evans, A.C., Jones-Gotman,
M. (2001). Changes in brain activity related to eating chocolate: from
pleasure to aversion. Brain 124, 1720–1733.
Stein, E.A., et al. (1998). Nicotine-induced limbic cortical activation
in the human brain: a functional MRI study. Am. J. Psychiatry 155,
Symons, D. (1995). Beauty is in the adaptations of the beholder: the
evolutionary psychology of human female sexual attractiveness. In
Sexual Nature, Sexual Culture. P.R. Abramson and S.D. Pinkerton,
eds. (Chicago: University of Chicago Press).
Talairach, J., and Tournoux, P. (1988). Co-planar stereotaxic atlas
of the human brain. (New York: Thieme Medical Publishers).
Thomas, K.M., Drevets, W.C., Whalen, P.J., Eccard, C.H., Dahl, R.E.,
Ryan, N.D., and Casey, B.J. (2001). Amygdala response to facial
expressions in children and adults. Biol. Psychiatry 49, 309–316.
Thut, G., et al. (1997). Activation of the human brain by monetary
reward. Neuroreport 8, 1225–1228.
Wise, R., Spinder, J., DeWit, H., and Gerber, G. (1978). Neuroleptic-
induced “anhedonia” in rats: pimozide blocks the reward quality of
food. Science 201, 262–264.
Wojciulik, E., Kanwisher, N., and Driver, J. (1998). Covert visual
attention modulates face-specific activity in the human fusiform
gyrus: fMRI study. J. Neurophysiol. 79, 1574–1578.
Wyvell, C.L., and Berridge, K.C. (2000). Intra-accumbens amphet-
amine increases the conditioned incentive salience of sucrose re-
ward: enhancement of reward “wanting” without enhanced “liking”
or response reinforcement. J. Neurosci. 20, 8122–8130.
Zald, D.H., Lee, J.T., Fluegel, K.W., and Pardo, J.V. (1998). Aversive
gustatory stimulation activates limbic circuits in humans. Brain 121,