!

SELECTED'READINGS'IN''
CONSUMER)NEUROSCIENCE)&)
NEUROMARKETING)
!
2nd$edition$

!
Compiled$by$
Thomas$Zoëga$Ramsøy$$
2014

 CONSUMER PREFERENCE: WANTING & LIKING

A"recurring"theme"in"this"compendium"is"whether"choices"are"conscious"or"unconscious."Both"
from"psychology"and"neuroscience,"we"know"that"we"have"(at"least)"two"different"motivation"
systems."Some"scholars"have"distinguished"between"a"wanting"system"and"a"liking"system."The"
wanting'system"operates"unconsciously,"and"is"known"to"reflect"the"operations"of"a"separate,"
deep"brain"system,"such"as"the"basal"ganglia."

Conversely,"the"liking'system"is"related"to"our"overt,"conscious"hedonic"experience,"and"is"
reflected"in"our"verbal"utterances."This"system"is"thought"to"be"based"on"the"brain’s"prefrontal"
regions,"such"as"the"orbitofrontal"cortex,"and"possibly"the"anterior"insula.

In"most"cases,"what"you"want"is"what"you"like."But"in"some"instances,"those"processes"break"
down."Consumer"choice"is"packed"with"such"examples."Think"of"the"choice"of"eating"a"
chocolate"while"on"a"diet."You"feel"the"urge"for"a"snack"and"the"sugar"boost,"but"consciously"
think"it’s"a"bad"thing"to"do."Which"of"the"two"will"win?"

This"is"a"tell"tale"sign"of"consumer"motivation,"and"why"the"whole"concept"of"neuromarketing"
and"consumer"neuroscience"makes"sense."By"only"asking"people,"we"only"have"access"to"the"
liking"system."By"exploring"the"hidden"mental"value"operations"of"the"brain,"we"can"also"
understand"the"wanting"system.
 ORIGINAL RESEARCH ARTICLE
published: 03 June 2011
doi: 10.3389/fnins.2011.00077

Investigating the role of the ventromedial prefrontal cortex in

the assessment of brands
José Paulo Santos1,2*, Daniela Seixas 3,4,5, Sofia Brandão 6 and Luiz Moutinho 7
1
Superior Institute of Maia, Maia, Portugal
2
Socius – Research Centre in Economic and Organizational Sociology, Lisbon, Portugal
3
Faculty of Medicine, Department of Experimental Biology, University of Porto, Porto, Portugal
4
Instituto de Biologia Molecular e Celular, Porto, Portugal
5
Department of Imaging, Centro Hospitalar de Vila Nova de Gaia e Espinho, Vila Nova de Gaia, Portugal
6
Magnetic Resonance Imaging Unit, Department of Radiology, São João Hospital, Porto, Portugal
7
Department of Management, University of Glasgow, Glasgow, Scotland

Edited by:
Julia Trommershaeuser, New York
University, USA
Reviewed by:
Ming Hsu, University of California
Berkeley, USA
Jamie D. Roitman, University of
California San Francisco, USA
*Correspondence:
José Paulo Santos, Instituto Superior
da Maia, Avenida Carlos Oliveira
Campos, Castelo da Maia, 4475-690
Avioso (São Pedro), Maia, Portugal.
e-mail: jpsantos@ismai.pt
The ventromedial prefrontal cortex (vmPFC) is believed to be important in everyday preference
judgments, processing emotions during decision-making. However, there is still controversy in
the literature regarding the participation of the vmPFC. To further elucidate the contribution of
the vmPFC in brand preference, we designed a functional magnetic resonance imaging (fMRI)
study where 18 subjects assessed positive, indifferent, and fictitious brands. Also, both the
period during and after the decision process were analyzed, hoping to unravel temporally the
role of the vmPFC, using modeled and model-free fMRI analysis. Considering together the
period before and after decision-making, there was activation of the vmPFC when comparing
positive with indifferent or fictitious brands. However, when the decision-making period was
separated from the moment after the response, and especially for positive brands, the vmPFC
was more active after the choice than during the decision process itself, challenging some of
the existing literature. The results of the present study support the notion that the vmPFC may
be unimportant in the decision stage of brand preference, questioning theories that postulate
that the vmPFC is in the origin of such a choice. Further studies are needed to investigate in
detail why the vmPFC seems to be involved in brand preference only after the decision process.
Keywords: neuromarketing, brands, emotion, preference, multivariate analysis, FMRI

INTRODUCTION in the vmPFC persisted in their original choice, ignoring brand

In the last few years several articles were published involving a new
approach to the study of brands using neuroscientific techniques.
One of these first studies used photographs of soft drinks where
brands figured explicitly, inducing preference judgments (Paulus
and Frank, 2003). These authors hypothesized that a specific
area in the prefrontal cortex, the ventromedial prefrontal cortex
(vmPFC), was critical for everyday preference judgments. In fact,
they found important activations in this brain region when par-
ticipants selected preferred soft drinks in contrast with a visual
discrimination task of the same stimuli (liquids contained in bottles
or glasses). Also investigating brands, Deppe et al. (2005), largely
based on the work of Damásio, Bechara, and co-workers (Damásio,
1994; Bechara et al., 1997, 1999; Bechara and Damásio, 2005), pro-
posed a dichotomic theory in economic decision-making, “(…) one
chain involving emotional experience (…) and another one based
on reasoning strategies” (p. 180). Deppe et al. (2005) propose the
vmPFC to be central in the processing of emotions during decision-
making, whereas brain regions associated with working memory
could sustain reasoning.
Koenigs and Tranel (2008) recruited patients with a specific
damage of the vmPFC to select soft drinks in two conditions:
blinded or brand-cued. While healthy controls and patients with
damage in other brain areas changed their soda preference from
the blinded drinks to the brand-cued ones, patients with a lesion
information, showing that the vmPFC is also necessary in the inte-
gration of information in the decision-making process.
In addition the vmPFC was found to be important in signaling
risk probabilities (Tom et al., 2007; Rangel et al., 2008). Fellows and
Farah (2007) again working with patients with vmPFC impairment,
suggested that this brain region is necessary for all sorts of choice
tasks, either uncertain (including risky or ambiguous situations),
or certain.
However, there is still controversy in the literature regarding the
function of the vmPFC in decision-making in general, and in brand
preference in particular. For example, Schaefer and Rotte (2007) did
not report activations in this brain region when sport and luxury
car brands (rewarding stimuli) were compared with rational choices
of car brands. In another study, using fNIRS to compare luxury
and common handbags assessed individually, Lin et al. (2010) sug-
gest that the cognitive subprocesses that underlie the assessment of
branded handbags were only important after the choice was made.
To further elucidate the contribution of the vmPFC in brand
preference, we designed a functional magnetic resonance imaging
(fMRI) study where subjects assessed positive, indifferent and ficti-
tious brands, testing the participation of the vmPFC in the process-
ing of these different hedonic categories of brands. Moreover, both
the period during and after the decision process were analyzed,
hoping to unravel temporally the role of the vmPFC.

www.frontiersin.org June 2011 | Volume 5 | Article 77 | 1

Santos et al.
The designation vmPFC is ambiguous in the literature. The present
study relies on the probabilistic atlases Harvard-Oxford Cortical
Structural Atlas and Harvard-Oxford Subcortical Structural Atlas pro-
vided by the Harvard Centre for Morphometric Analysis1. We have
considered the vmPFC to include the ventral medial frontal pole,
frontal medial cortex, ventral paracingulate gyrus, ventral anterior
cingulate gyrus, and subcallosal cortex, limited dorsally by the plane
z = +10, and laterally by the planes x = ±20 (MNI152 coordinates).
MATERIALS AND METHODS
GENERAL STRUCTURE
To explore the research question, an event-related fMRI experiment
was designed. There were four different stimuli categories, plus the
interstimuli interval. Each category was composed by 35 slides (6 s
each). The interstimuli interval ranged from 4 until 9 s, in 0.5 s
steps. The experiment duration was 1200 s, plus 9 s added in the
end to ensure that all of the hemodynamic response was included.
The sequence was optimized with Optseq2 software (Athinoula A.
Martinos Center for Biomedical Imaging, USA)2.
Three of the four stimuli were brands’ logos grouped in the fol-
lowing categories: positive, indifferent, and fictitious brands. The
fourth stimulus was non-emotional words. During the interstimuli
interval participants fixated a cross.
BRAND SELECTION
In order to select the logos for the positive and indifferent brand
categories, participants completed an electronic survey in which
were shown 200 brand logos, that they had to rate using the pleasure
and arousal dimensions of the PAD – pleasure, arousal, dominance
scale (Russell and Mehrabian, 1977; Mehrabian and De Wetter, 1987;
Mehrabian, 1995), and the SAM – self assessment manikin, explained
in detail elsewhere (Morris, 1995; Bradley and Lang, 2007). Self-
reporting emotions is a complex task for most individuals, mainly
due to the difficulty in verbalizing such inner states (Chamberlain
and Broderick, 2007). SAM is a non-verbal pictorial assessment tech-
nique designed to represent each dimension of the PAD scale associ-
ated with a person’s affective reaction to a certain stimuli. Dominance
was not included in the brand assessment because with static pictures
this dimension correlates with pleasure (Bradley and Lang, 2007).
After this task, the responses were screened and categorized
according to the following criteria: positive brands if the score
was ≥7 in the pleasure dimension, and ≥5 in the arousal dimen-
sion; indifferent brands if the score was ≥4 and ≤6 in the pleasure
dimension, and ≤5 in the arousal dimension. With this procedure 35
positive and 35 indifferent brands were chosen for each participant,
and were randomized to enter the fMRI paradigm.
FICTITIOUS LOGOS
The fictitious brands were brands’ logos that did not exist in the
market. Each logo was designed by a marketer made to resemble a
real one, making it plausible for the consumer. The fictitious brands
did not represent a particular type of product. Instead, logos with
assorted shapes, colors, and fonts suggesting different products and
services were used (examples in Figure 1).
1
http://www.cma.mgh.harvard.edu
2
http://surfer.nmr.mgh.harvard.edu/optseq/
Frontiers in Neuroscience | Decision Neuroscience
The vmPFC in brands’ assessments
FIGURE 1 | Examples of some of the logos used as fictitious stimuli.
NON-EMOTIONAL WORDS
The fourth stimulus (a second baseline) was non-emotional words:
determiners, conjunctions, prepositions, or adverbs. Importantly,
nouns or verbs that could evoke emotions, objects, or actions were
not used. With this stimulus we hoped to avoid meditation during
the fMRI task (Gusnard and Raichle, 2001; Beckmann and Smith,
2005; De Luca et al., 2006), that could cloud possible self-reflexive
processes elicited by brands (Yoon et al., 2006).
STRUCTURING THE PARADIGM
The structure of the paradigm was the same for all participants. The
paradigm sequences were programmed with SuperLab 4.0 software
(version 4.0.6b; Cedrus Corporation, USA)3.
INSTRUCTIONS FOR THE SCANNING SESSION
Depending on the type of stimulus visualized, the participants were
instructed to either rate hedonically the brand (as positive, negative,
indifferent, or unknown), to read covertly non-emotional words, or
just to fixate a cross. Participants made their choices using a but-
ton box (model Lumina LU400-PAIR; Cedrus Corporation, USA)4.
HUMAN SUBJECTS
The participants were 18, 7 healthy male and 11 healthy female
volunteers, right handed, with neither history of neurological nor
psychiatric disturbances (mean age 28.2 ± 6.9 years, 19–41 years).
Informed consent was obtained in all cases. A safety form for
magnetic resonance imaging was filled by every participant. After
each session the participants were debriefed. This research project
was performed according to the Declaration of Helsinki and was
approved by the local Ethics Committee.
DATA ACQUISITION
Functional images with axial orientation were obtained using a
T2∗-weighted EPI sequence in a Siemens® Magnetom Trio high
field (3 T) MRI scanner (Siemens AG, Germany; TR = 3000 ms,
TE = 30 ms, 64 × 64 matrix, FOV = 192 mm, 3.0 mm axial slices).
The order of acquisition of the slices was interleaved, and they
covered the whole brain. The study consisted in one session where
407 volumes were acquired. The first four volumes were discarded
to ensure pulses stabilization.
http://www.superlab.com
3
http://www.cedrus.com
4
June 2011 | Volume 5 | Article 77 | 2
Santos et al.
A whole brain anatomical structural scan was acquired also for
each volunteer, using a T1-weighted MPRAGE protocol (256 × 256
matrix, FOV = 256 mm, 3.0 mm axial slices), for co-registration
purposes. Gradient field mapping was additionally acquired for
image quality control.
IMAGE ANALYSIS
Functional magnetic resonance imaging data processing was car-
ried out using FEAT (FMRI Expert Analysis Tool) version 5.98,
a model-based GLM (general linear model) analysis tool, and
also using probabilistic independent component analysis (PICA;
Beckmann and Smith, 2004) as implemented in MELODIC
(Multivariate Exploratory Linear Decomposition into Independent
Components) version 3.09, a model-free analysis tool, both part
of FSL – FMRIB’s Software Library5 (Smith et al., 2004; Woolrich
et al., 2009).
General linear model analysis – common procedures
In the FEAT analysis, the following pre-statistics processing was
applied: motion correction using MCFLIRT (Jenkinson et al.,
2002); slice-timing correction using Fourier-space time-series
phase-shifting; non-brain removal using BET (Smith, 2002);
spatial smoothing using a Gaussian kernel of FWHM 5 mm;
grand-mean intensity normalization of the entire 4D dataset
by a single multiplicative factor; high pass temporal filtering
(Gaussian-weighted least-squares straight line fitting, with
sigma = 30.0 s). Stimuli were convolved with a gamma func-
tion with canonical values (phase 0 s, SD 3 s, and mean lag 6 s).
To account for variations, temporal derivatives were added for
every explanatory variable (EV), in order to achieve a better
fit between the signal and the stimuli convolved hemodynamic
responses. Time-series statistical analysis was performed using
FILM with local autocorrelation correction (Woolrich et al.,
2001). Registration to high-resolution structural and/or stand-
ard space images was done using FLIRT (Jenkinson and Smith,
2001; Jenkinson et al., 2002).
5
http://www.fmrib.ox.ac.uk/fsl
FIGURE 2 | Splitting the duration of the stimulus for one subject. The figure represents the splitting of the first five stimuli of each category (positive, indifferent,
and fictitious logos). Lighter areas represent the period until the response (during decision), and darker areas represent the period after the response (passive
visualization of the stimulus).
www.frontiersin.org
The vmPFC in brands’ assessments
At the individual level two different strategies of analysis were
used for comparison. The first strategy was a traditional approach
where the hemodynamic response was investigated during the
complete time window of the stimulus (6 s). In the second approach
the stimulus duration was divided in two: the period before the
response (decision-making), and the period after the response
(passive period; see Figure 2).
General linear model analysis – conventional stimulus analysis
Before the scanning session, participants assessed a set of 200 brand
logos, from which the positive and indifferent stimuli were extracted.
Then, during the scanning, participants rated again the brands. In the
first model, in which the whole of the stimulus duration was consid-
ered, 13 EVs were included: the three types of stimulus (positive, indif-
ferent, and fictitious logos) times the four possible ratings (positive,
indifferent, negative, and unknown), and the non-emotional words.
Most of the assessments were consistent between the two study
sessions, before and during the scanning (see Brand Selection), but
some of the possible combinations received little or even no rat-
ings. Although all the possibilities were modeled with EVs aiming
to explain most of the variance, only those that were consistent
between sessions, i.e., positive brands that were rated as positive
during the scanning session (PosPos), indifferent brands that were
rated as indifferent (IndInd), or fictitious logos that were rated as
unknown inside the scanner (NoBUnk) were considered in the anal-
ysis. Hence, at the individual level analysis, stimuli, and baseline were
compared, resulting in the following contrasts: positive > indifferent,
positive > unrecognized logos, and indifferent > unrecognized logos.
General linear model analysis – stimulus detailed analysis
In the second model, 25 EVs were considered: the three types of
stimulus (positive, indifferent, and fictitious logos), times the four
possible ratings (positive, indifferent, negative, and unknown),
times the two epochs (before and after button pressing), and the
non-emotional words.
At the individual level and as before, stimuli and baseline
were subtracted, resulting in the following six contrasts (ar: after
response; br: before response): positive br > indifferent br, positive
June 2011 | Volume 5 | Article 77 | 3
Santos et al.
br > unrecognized logos br, indifferent br > unrecognized logos br,
positive ar > indifferent ar, positive ar > unrecognized logos ar, and
indifferent ar > unrecognized logos ar.
General linear model analysis – group analysis
For both models, group analysis was performed with FLAME
(FMRIB’s Local Analysis of Mixed Effects) stage 1 and stage 2 with
automatic outlier detection (Beckmann et al., 2003; Woolrich et al.,
2004; Woolrich, 2008). At this level, group means were calculated
from the individual level contrasts.
Z (Gaussianized T/F) statistic images were thresholded using
clusters determined by z > 2.3 and a (corrected) cluster significance
threshold of p = 1.00 (Worsley, 2001). Only clusters with more than
50 voxels survived the threshold.
Probabilistic independent component analysis
The following data pre-processing was applied: masking of non-
brain voxels, voxel-wise de-meaning of the data, and normaliza-
tion of the voxel-wise variance. Pre-processed data were whitened
and projected into a 164-dimensional subspace using probabilistic
Principal Component Analysis where the number of dimensions
was estimated using the Laplace approximation to the Bayesian
evidence of the model order (Minka, 2000; Beckmann and Smith,
2004). The whitened observations were decomposed into sets of
vectors, which describe signal variation across the temporal domain
(time-courses), the session/subject domain and across the spatial
domain (maps) by optimizing for non-Gaussian spatial source
distributions using a fixed-point iteration technique (Hyvärinen,
1999). Estimated component maps were divided by the SD of the
residual noise and thresholded by fitting a mixture model to the
histogram of intensity values (Beckmann and Smith, 2004).
The EVs basic shapes convolved with a gamma function and
including temporal derivatives were concatenated for all the par-
ticipants in the same order that time-courses were entered in
MELODIC, and the same contrasts used in FEAT were computed.
The parameter estimates of each spatial independent component
(164 total) were then calculated and tested using GLM for each case
(see Figure 3), and so the selection of significant spatial independ-
ent components was based on statistical criteria.
FIGURE 3 | Illustration of the application of a GLM analysis to each of the
164 independent components yielded by MELODIC. For each IC, 25
independent variables were modeled: the three types of stimulus (positive,
indifferent, and fictitious logos), times the four possible ratings (positive,
indifferent, negative, and unknown), times the two epochs (before and after
button pressing), and the non-emotional words.
Frontiers in Neuroscience | Decision Neuroscience
The vmPFC in brands’ assessments
RESULTS
CONSISTENCY IN THE ASSESSMENTS BETWEEN SESSIONS
Most of the ratings were coherent from one session to the other.
Results are summarized in Table 1. Five hundred fifty-four fictitious
brands’ logos out of 630 (87.9%) were rated as unknown, 590 posi-
tive brands out of 630 (93.7%) were rated as positive, and 427 indif-
ferent brands out of 630 (67.8%) were again rated as indifferent.
RESPONSE TIME
The graph in Figure 4 depicts the distribution of the subjects’
choices by response time. Response times were shorter with positive
ratings (1546 ms) than indifferent (2370 ms) or fictitious ratings
(2334 ms), suggesting a delayed decision process with the last two
ratings. These differences are significant between positive and indif-
ferent ratings (F426,589,0.01 = 1.702 – p-value < 0.000001), and posi-
tive and fictitious ratings (F553,589,0.01 = 1.708 – p-value < 0.000001),
but not significant between indifferent and fictitious ratings
(F553,426,0.01 = 1.004 – p-value = 0.969508).
GENERAL LINEAR MODEL ANALYSIS
In the conventional GLM analysis (the whole of the stimulus dura-
tion), the vmPFC was significantly and extensively activate for the
contrasts positive versus indifferent or fictitious logos (Figure 5).
Figure 6 represents the stimulus detailed analysis for the same
contrasts. For the period before the response (decision stage) the
vmPFC tendentiously deactivated. Conversely, after button press-
ing, i.e., after the decision was made and while subjects were pas-
sively visualizing the stimulus, the vmPFC was active.
Four local maxima from the cluster in the vmPFC in the contrast
positive versus indifferent in the conventional analysis were selected
for further analysis. The parameter estimates of these voxels are rep-
resented in Figure 7 both for the conventional analysis and for the
stimulus detailed analysis. For the conventional GLM analysis, all
the four local maxima significantly activated when positive brands
were involved. On the contrary, in the stimulus detailed analysis
there were deactivations, more prominent in the anterior subre-
gions (ventral paracingulate gyrus and ventral medial frontal pole).
After the response, however, the vmPFC was extensively activate.
PROBABILISTIC INDEPENDENT COMPONENT ANALYSIS
The 164 ICs yielded by PICA account for 86.95% of the variability.
To select the relevant ICs the criteria were: the z statistics of the
contrast between the parameter estimates of the positive brands
versus the parameter estimates of the indifferent brands, the
Table 1 | Assessments made during the scanning sessions separated
according to the type of stimuli.
Recorded ratings
Stimuli Total
Positive Indifferent Negative Unknown No
answer
Positive 590 29 3 6 2 630
Indifferent 82 427 74 44 3 630
Fictitious 33 36 2 554 5 630
Total 705 492 79 604 10 1890
June 2011 | Volume 5 | Article 77 | 4
Santos et al. The vmPFC in brands’ assessments

FIGURE 4 | Relative frequency of response times obtained during the scanning session grouped in 500 ms intervals.

FIGURE 5 | Statistical z maps (unthresholded in the upper row and thresholded in the lower row) for the contrasts positive versus indifferent brands and
positive versus fictitious logos in a conventional GLM analysis. In the unthresholded images the significant clusters are outlined in white (for z > 2.3), and the
vmPFC is outlined in green. Sagittal views for x = −04 (MNI152 coordinates).

fictitious logos, and the non-emotional words had to be superior
to 2.3 in all the three cases, or inferior to −2.3 in all three cases. This
procedure was implemented in the two situations, before and after
the response. In this way it was guaranteed that the ICs selected
would be significantly more active or more deactivated for positive
brands than in the remaining cases.
Neither before the response nor after the response there were
ICs with z value inferior to −2.3. On the contrary, two ICs (#17 and
#152) had all the considered z values superior to 2.3 in the situation
before the response, and four other (#24, #49, #96, and #135) had
z values superior to 2.3 in the situation after the response. Only IC
#24 included brain activations in the vmPFC. The z values for the

www.frontiersin.org June 2011 | Volume 5 | Article 77 | 5

Santos et al.
FIGURE 6 | Statistical z maps (unthresholded and thresholded) for the
contrasts between positive versus indifferent brands and positive versus
fictitious logos in the stimulus detailed analysis. The two top row maps
represent the decision stage (before the response), and the two bottom row
maps represent the period after the response (passive visualization). In the
unthresholded images the significant clusters are outlined in white (for
z > 2.3), the vmPFC is outlined in green. Sagittal views for x = −04 (MNI152
coordinates).
three cases are reported in Table 2. Three slices of IC #24 are repre-
sented in Figure 8. Besides the vmPFC activation this network also
includes active voxels in the precuneus, posterior cingulate gyrus,
right and left anterior divisions of the middle temporal gyrus, and
deactivation in the occipital fusiform gyrus.
Table 2 also reports the z values for all the ICs that encompass
at least one of the local maxima voxels considered in Figure 7 (acti-
vated or deactivated). Only IC #24 has this statistic consistently and
significantly positive (for the situation after response). However,
Frontiers in Neuroscience | Decision Neuroscience
The vmPFC in brands’ assessments
ICs #15 and #22 were significantly positive for the situation after the
response in the contrast with other logos (indifferent or fictitious).
In the situation before the response, IC #132 had significantly nega-
tive z values for the contrasts with logos.
DISCUSSION
Most of the neuroimaging studies involving brands use paradigms
consisting of choices between pairs of brands or products, i.e., both
stimuli are presented simultaneously and subjects have to choose
one or the other. However, the structure adopted in our study is
different, we believe closer to everyday life; each brand is presented
one at a time, meaning that subjects decide about the hedonic value
of a particular brand not by comparison. For example, when a
consumer chooses a product from a supermarket shelf, s/he does
not collect first all the available items and then choose. On the
contrary, there is a previous intention summarized in a concept
named consideration set, or evoked set (Roberts and Lattin, 1991;
Shocker et al., 1991; Petrof and Daghfous, 1996). The consumer
confronts each possibility in the shelf against the consideration
set until one brand/product is preferred. Thus, the process is not
a simple choice among several options, but instead an assessment
of the fit between one option and the inner expectations that were
previously constructed.
Damásio (1994) from his observations in neurologically
impaired patients, proposed that the prefrontal cortex is a crucial
structure in decision-making; the vmPFC in particular is thought
to be important in decisions of preference including preference for
certain brands (Paulus and Frank, 2003; Deppe et al., 2005; Knutson
et al., 2007; Koenigs and Tranel, 2008; Luu and Chau, 2009). The
results of our conventional GLM analysis, which included data
acquired both before and after decision of brand preference, cor-
roborate these findings: activation of the vmPFC was found when
comparing positive with indifferent or fictitious brands. However,
when we dissected the subjects’ responses and isolated the decision-
making period from the moment after the response, we found that,
especially for positive brands, the vmPFC was more active after the
choice than during the decision process itself, challenging some of
the existing literature. And this result was supported both by the
GLM time-split analysis and by the PICA analysis.
During the decision process itself, i.e., before the response, the
vmPFC was less active for positive brands than for indifferent or
fictitious logos. Conversely, the vmPFC was more active after the
brand choice was made. Considering the four local maxima in
the vmPFC (the subcallosal cortex, the frontal medial cortex, the
ventral paracingulate gyrus, and the ventral medial frontal pole),
although they were also involved in the conventional analysis when
it corresponded to all the period when the stimulus was present,
the same voxels of the vmPFC were deactive during the decision
period until the response, but active after the response. This pattern
was not found with indifferent brands (that subjects recognized
as having some meaning to them, but that were not preferred),
with fictitious logos (visualized for the first time and about which,
likewise, subjects could not have a preformed opinion), and also
with non-emotional words.
One of the ICs obtained with the multivariate model-free analy-
sis (PICA) was significantly more relevant in the choice of positive
brands than indifferent brands, fictitious logos, or non-emotional
June 2011 | Volume 5 | Article 77 | 6
Santos et al. The vmPFC in brands’ assessments

FIGURE 7 | Parameter estimates for the stimuli in four local maxima in the
vmPFC (subcallosal cortex: −6, 32, −10; frontal medial cortex: 2, 36, −14;
ventral paracingulate gyrus: −2, 48, −2; ventral medial frontal pole: −2, 58,
4). The bar graphs identified with the suffix 6 s are the conventional GLM-based
analysis of fMRI data. The bar graphs identified with the suffix br refer to the
participation of the voxel before the response (decision stage). The bar graphs
identified with the suffix ar refer to the participation after the decision instant but
before the stimulus offset. Pos: positive; Ind: indifferent; Fic: fictitious; NEW:
non-emotional words (baseline). MNI152 coordinates. Error bars correspond to
confidence intervals of 95%.

Table 2 | Statistic z for all the ICs that had at least one voxel activated or deactivated among those considered in Figure 7.

IC

15 22 24 41 50 89 104 110 132

Pos > Ind br −0.967 0.227 −1.876 −1.940 −1.588 0.476 −1.581 −0.239 −3.143
Pos > Fic br −2.560 0.329 1.441 2.296 0.471 0.573 −0.463 3.269 −2.497
Pos > NEW br −7.146 −1.021 0.275 2.961 1.417 0.388 0.413 −0.358 −1.753
Pos > Ind ar 4.805 3.136 2.562 −2.241 0.819 3.103 −0.173 1.348 −1.353
Pos > Fic ar 4.423 5.432 5.169 2.282 0.520 1.389 2.588 1.487 −0.448
Pos > NEW ar −4.001 0.693 2.892 −1.790 −1.562 −3.278 −4.711 3.839 −3.542

Pos, positive; Ind, indifferent; Fic, fictitious; NEW, non-emotional words (baseline); br refers to the participation of the voxel before the response (decision stage); ar
refers to the participation after the decision instant but before the stimulus offset.

words. IC 24 showed extensive activations in the vmPFC, among

FIGURE 8 | Two views of the network that constitutes the independent
component #24: sagittal (x = −04), and axial (z = −12). The vmPFC is
outlined in green. Radiological convention. MNI152 coordinates.
other brain structures (Figure 8). This network was significantly
more active with preferred brands than with indifferent brands,
fictitious logos, or non-emotional words only after the response,
which reinforces the fact that although important in decisions of
preference, the vmPFC is only so after the decision-making pro-
cess itself. The analysis of the participation of the vmPFC in brain
networks represented in other ICs corroborates this hypothesis,
because none of the ICs had consistent or significant statistics
to support the participation of the vmPFC in the period before
the response.
The results of the present study seem to contradict some
of the existing theories on the role of the vmPFC in the deci-
sion process. On the other hand, our data are supported by
Lin et al., (2010) work in which the brand stimuli were also
presented one at a time, suggesting as well a late participation
of the vmPFC in preference decision-making; or by Li et al.

www.frontiersin.org June 2011 | Volume 5 | Article 77 | 7

Santos et al. The vmPFC in brands’ assessments

(2010) study that used fMRI and the Iowa Gambling Task to
investigate the neural correlates of decision-making. They
have demonstrated a group of brain regions that included the
dorsolateral prefrontal cortex for working memory, and the
insula and posterior cingulate cortex for representations of
emotional states. However, the vmPFC was not part neither
of the memory nor the emotional networks, but instead was
coupling the two processes.
In summary, the results of the present study converge in
supporting the notion that the vmPFC may be unimportant in the
decision stage concerning brand preference, questioning theories
that postulate that the vmPFC is in the origin of brand choice. To
complement our findings, further studies that challenge as well
conventional research design and neuroimaging methodologies are
need to investigate in detail why the vmPFC seems to be involved
in brand preference only after the decision process.

REFERENCES
Bechara, A., and Damásio, A. R. (2005).
The somatic marker hypothesis: a
neural theory of economic decision.
Games Econ. Behav. 52, 336–372.
Bechara, A., Damásio, H., Damásio, A. R.,
and Lee, G. P. (1999). Different con-
tributions of the human amygdala
and ventromedial prefrontal cortex
to decision-making. J. Neurosci. 19,
5473–5481.
Bechara, A., Damásio, H., Tranel, D.,
and Damásio, A. R. (1997). Deciding
advantageously before knowing the
advantageous strategy. Science 275,
1293–1295.
Beckmann, C. F., Jenkinson, M., and
Smith, S. M. (2003). General multilevel
linear modeling for group analysis in
FMRI. Neuroimage 20, 1052–1063.
Beckmann, C. F., and Smith, S. M. (2004).
Probabilistic independent compo-
nent analysis for functional magnetic
resonance imaging. IEEE Trans. Med.
Imaging 23, 137–152.
Beckmann, C. F., and Smith, S. M. (2005).
Tensorial extensions of independent
component analysis for multisub-
ject FMRI analysis. Neuroimage 25,
294–311.
Bradley, M. M., and Lang, P. J. (2007).“The
international affective picture system
(IAPS) in the study of emotion and
attention,” in Handbook of Emotion
Elicitation and Assessment, eds J. A.
Coan and J. J. B. Allen (New York:
Oxford University Press), 29–46.
Chamberlain, L., and Broderick, A.
(2007). The application of physiologi-
cal observation methods to emotion
research. Qual. Mark. Res. Int. J. 10,
199–216.
Damásio, A. R. (1994). Descarte’s Error –
Emotion, Reason and the Human Brain.
New York: Penguin Putnam.
De Luca, M., Beckmann, C. F., De Stefano,
N., Matthews, P. M., and Smith, S. M.
(2006). FMRI resting state networks
define distinct modes of long-distance
interactions in the human brain.
Neuroimage 29, 1359–1367.
Deppe, M., Schwindt, W., Kugel, H.,
Plassmann, H., and Kenning, P.
(2005). Nonlinear responses within
the medial prefrontal cortex reveal
when specific implicit information
influences economic decision making.
J. Neuroimaging 15, 171–182.
Fellows, L. K., and Farah, M. J. (2007). The
role of ventromedial prefrontal cortex
in decision making: judgment under
uncertainty or judgment per se? Cereb.
Cortex 17, 2669–2674.
Gusnard, D. A., and Raichle, M. E. (2001).
Searching for a baseline: functional
imaging and the resting human brain.
Nat. Rev. Neurosci. 2, 685–694.
Hyvärinen, A. (1999). Fast and robust
fixed-point algorithms for independ-
ent component analysis. IEEE Trans.
Neural Netw. 10, 626–634.
Jenkinson, M., Bannister, P. R., Brady,
J. M., and Smith, S. M. (2002).
Improved optimization for the robust
and accurate linear registration and
motion correction of brain images.
Neuroimage 17, 825–841.
Jenkinson, M., and Smith, S. M. (2001). A
global optimisation method for robust
affine registration of brain images.
Med. Image Anal. 5, 143–156.
Knutson, B., Rick, S., Wimmer, G. E.,
Prelec, D., and Loewenstein, G. (2007).
Neural predictors of purchases.
Neuron 53, 147–156.
Koenigs, M., and Tranel, D. (2008).
Prefrontal cortex damage abolishes
brand-cued changes in cola prefer-
ence. Soc. Cogn. Affect. Neurosci. 3, 1–6.
Li, X., Lu, Z.-L., D’argembeau, A., Ng,
M., and Bechara, A. (2010). The iowa
gambling task in fMRI images. Hum.
Brain Mapp. 31, 410–423.
Lin, C.-H., Tuan, H.-P., and Chiu, Y.-C.
(2010). Medial frontal activity in
brand-loyal consumers: a behav-
ior and near-infrared ray study. J.
Neurosci. Psychol. Econ. 3, 59–73.
Luu, S., and Chau, T. (2009). Decoding
subjective preference from single-trial
near-infrared spectroscopy signals. J.
Neural Eng. 6, 016003.
Mehrabian, A. (1995). Framework for
a comprehensive description and
measurement of emotional states.
Genet. Soc. Gen. Psychol. Monogr. 121,
339–361.
Mehrabian, A., and De Wetter, R. (1987).
Experimental test of an emotion-
based approach to fitting brand
names to products. J. Appl. Psychol.
72, 125–130.
Minka, T. P. (2000). Automatic Choice of
Dimensionality for PCA. Technical
Report 514, Cambridge: MIT Media
Lab Vision and Modeling Group.
Morris, J. D. (1995). Observations: SAM: the
self-assessment manikin – an efficient
cross-cultural measurement of emo-
tional response. J. Advert. Res. 35, 63–68.
Paulus, M. P., and Frank, L. R. (2003).
Ventromedial prefrontal cortex acti-
vation is critical for preference judg-
ments. Neuroreport 14, 1311–1315.
Petrof, J. V., and Daghfous, N. (1996).
Evoked set: myth or reality? Bus. Horiz.
39, 72–77.
Rangel, A., Camerer, C. F., and Montague,
P. R. (2008). A framework for study-
ing the neurobiology of value-based
decision making. Nat. Rev. Neurosci.
9, 545–556.
Roberts, J., and Lattin, J. (1991).
Development and testing of a model
of consideration set composition. J.
Mark. Res. 28, 429–440.
Russell, J. A., and Mehrabian, A. (1977).
Evidence for a three-factor theory of
emotions. J. Res. Pers. 11, 273–294.
Schaefer, M., and Rotte, M. (2007). Favorite
brands as cultural objects modulate
reward circuit. Neuroreport 18, 141–145.
Shocker, A., Ben-Akiva, M., Boccara,
B., and Nedungadi, P. (1991).
Consideration set influences on con-
sumer decision-making and choice:
issues, models, and suggestions. Mark.
Lett. 2, 181–197.
Smith, S. M. (2002). Fast robust auto-
mated brain extraction. Hum. Brain
Mapp. 17, 143–155.
Smith, S. M., Jenkinson, M., Woolrich, M.
W., Beckmann, C. F., Behrens, T. E.,
Johansen-Berg, H., Bannister, P. R., De
Luca, M., Drobnjak, I., Flitney, D. E.,
Niazy, R. K., Saunders, J., Vickers, J.,
Zhang, Y., De Stefano, N., Brady, J. M.,
and Matthews, P. M. (2004). Advances
in functional and structural MR image
analysis and implementation as FSL.
Neuroimage 23(Suppl. 1), S208–S219.
Tom, S. M., Fox, C. R., Trepel, C., and
Poldrack, R. A. (2007). The neural
basis of loss aversion in decision-mak-
ing under risk. Science 315, 515–518.
Woolrich, M. W. (2008). Robust group
analysis using outlier inference.
Neuroimage 41, 286–301.
Woolrich, M. W., Behrens, T. E.,
Beckmann, C. F., Jenkinson, M., and
Smith, S. M. (2004). Multilevel linear
modelling for FMRI group analysis
using Bayesian inference. Neuroimage
21, 1732–1747.
Woolrich, M. W., Jbabdi, S., Patenaude,
B., Chappell, M., Makni, S., Behrens,
T. E., Beckmann, C. F., Jenkinson, M.,
and Smith, S. M. (2009). Bayesian
analysis of neuroimaging data in FSL.
Neuroimage 45, S173–S186.
Woolrich, M. W., Ripley, B. D., Brady, J.
M., and Smith, S. M. (2001). Temporal
autocorrelation in univariate linear
modeling of FMRI data. Neuroimage
14, 1370–1386.
Worsley, K. J. (2001). “Statistical analysis
of activation images,” in Functional
MRI: An Introduction to Methods,
eds P. Jezzard, P. M. Matthews, and
S. M. Smith (New York, NY: Oxford
University Press), 251–270.
Yoon, C., Gutchess, A. H., Feinberg, F.,
and Polk, T. A. (2006). A functional
magnetic resonance imaging study of
neural dissociations between brand
and person judgments. J. Consum.
Res. 33, 31–40.
Conflict of Interest Statement: The
authors declare that the research was con-
ducted in the absence of any commercial
or financial relationships that could be
construed as a potential conflict of interest.
Received: 15 December 2010; accepted: 21
May 2011; published online: 03 June 2011.
Citation: Santos JP, Seixas D, Brandão
S and Moutinho L (2011) Investigating
the role of the ventromedial prefron-
tal cortex in the assessment of brands.
Front. Neurosci. 5:77. doi: 10.3389/
fnins.2011.00077
This article was submitted to Frontiers
in Decision Neuroscience, a specialty of
Frontiers in Neuroscience.
Copyright © 2011 Santos, Seixas, Brandão
and Moutinho. This is an open-access
article subject to a non-exclusive license
between the authors and Frontiers Media
SA, which permits use, distribution and
reproduction in other forums, provided
the original authors and source are cred-
ited and other Frontiers conditions are
complied with.

Frontiers in Neuroscience | Decision Neuroscience June 2011 | Volume 5 | Article 77 | 8

 JCPS-00233; No. of pages: 7; 4C:

Available online at www.sciencedirect.com

Journal of
CONSUMER
PSYCHOLOGY
Journal of Consumer Psychology xx (2011) xxx – xxx

A neural predictor of cultural popularity

Gregory S. Berns ⁎, Sara E. Moore
Economics Department and Center for Neuropolicy, Emory University, Atlanta, GA 30322, USA

Received 7 February 2011; accepted 5 May 2011

Abstract

We use neuroimaging to predict cultural popularity — something that is popular in the broadest sense and appeals to a large number of
individuals. Neuroeconomic research suggests that activity in reward-related regions of the brain, notably the orbitofrontal cortex and ventral
striatum, is predictive of future purchasing decisions, but it is unknown whether the neural signals of a small group of individuals are predictive of
the purchasing decisions of the population at large. For neuroimaging to be useful as a measure of widespread popularity, these neural responses
would have to generalize to a much larger population that is not the direct subject of the brain imaging itself. Here, we test the possibility of using
functional magnetic resonance imaging (fMRI) to predict the relative popularity of a common good: music. We used fMRI to measure the brain
responses of a relatively small group of adolescents while listening to songs of largely unknown artists. As a measure of popularity, the sales of
these songs were totaled for the three years following scanning, and brain responses were then correlated with these “future” earnings. Although
subjective likability of the songs was not predictive of sales, activity within the ventral striatum was significantly correlated with the number of
units sold. These results suggest that the neural responses to goods are not only predictive of purchase decisions for those individuals actually
scanned, but such responses generalize to the population at large and may be used to predict cultural popularity.
© 2011 Published by Elsevier Inc. on behalf of Society for Consumer Psychology.

Keywords: fMRI; Neuroeconomics; Neuromarketing; Music; Prediction

Introduction popularity is a valuable skill that also can translate into

How can we predict popularity? Although superficially a
trivial question, the desire for popularity consumes a great
portion of the lives of many youths and adults. More than the
superficial teenager's quest for popularity, being popular is a
marker for social status. Consequently, popularity would seem
to confer a reproductive advantage in the evolution of the
human species, thus explaining its importance to humans. Such
importance extends to economic success as well because goods
and services that are popular command higher prices. Although
there are good economic and evolutionary rationales for
pursuing popularity, predicting who or what becomes popular
is a challenging problem. Even so, the ability to predict
⁎ Corresponding author at: 1602 Fishburne Dr., Emory University, Atlanta,
GA 30322, USA.
E-mail address: gberns@emory.edu (G.S. Berns).
1057-7408/$ - see front matter © 2011 Published by Elsevier Inc. on behalf of Society for Consumer Psychology.
doi:10.1016/j.jcps.2011.05.001
economic success.
In the domain of economic goods, traditional approaches to
forecasting popularity rely on standard marketing techniques.
These include focus groups, questionnaires, simulated choice
tests, and market tests. More recently, however, the widespread
use of neuroimaging has raised the possibility of using
functional magnetic resonance imaging (fMRI) in the marketing
process (Ariely & Berns, 2010). Neuroeconomic research
suggests that activity in reward-related regions of the brain,
notably the orbitofrontal cortex and ventral striatum is
predictive of future purchasing decisions of the individuals
who are scanned (Hare, O'Doherty, Camerer, Schultz, &
Rangel, 2008; Knutson, Rick, Wimmer, Prelec, & Loewenstein,
2007; Plassmann, O'Doherty, & Rangel, 2007; Plassmann,
O'Doherty, Shiv, & Rangel, 2008). For neuroimaging to be
useful in either a marketing or branding application, however,
these neural signals would need to generalize to a larger group
of individuals who themselves were not the direct object of

Please cite this article as: Berns, G.S., & Moore, S.E., A neural predictor of cultural popularity, Journal of Consumer Psychology (2011), doi:10.1016/
j.jcps.2011.05.001
2 G.S. Berns, S.E. Moore / Journal of Consumer Psychology xx (2011) xxx–xxx
brain scanning. Currently, it is unknown whether the neural
signals of a small group of individuals are predictive of the
purchasing decisions of the population at large.
Neuroimaging is often touted as a hot new tool for branding
(Lindstrom, 2008). Although branding and advertising have
been considered in a few neuroimaging papers (Kenning &
Plassmann, 2008; Lee, Broderick, & Chamberlain, 2007;
Yoon, Gutchess, Feinberg, & Polk, 2006), it is still unknown
whether neuroimaging can prospectively reveal whether a
particular ad or brand campaign will be effective. In a well-
known Coke–Pepsi study, participants who described them-
selves as Coke-drinkers showed significant activation in the
hippocampus and right DLPFC when they were cued about the
upcoming drink of Coke (McClure et al., 2004). Self-described
Pepsi-drinkers did not have this response. In the absence of
brand information, there was no significant difference in pref-
erence during a test-taste. This study suggested that any dif-
ferences in the neural response to the two brands must be
culturally derived. Although these results demonstrate that
branding does affect brain responses to nominally similar goods,
the question of whether brand effectiveness can be predicted in
advance remains an open question.
To demonstrate the efficacy of an fMRI study for branding,
three conditions must be met. First, the study participants—i.e.
the cohort of individuals who are actually scanned—should be
representative of the population that is the target of a brand
campaign. Second, to truly test whether the neural signals are
predictive of brand effectiveness, the scanning must be done
before the campaign is launched. Third, metrics of brand
effectiveness must be readily available for the target population.
For example, these might include sales data, web page views,
downloads, internet searches, etc. Finally, although not strictly a
condition, it is an open question as to what should actually be
scanned during fMRI. If the product can be consumed in the
scanner, then the product itself becomes the target. Alterna-
tively, an ad or branding campaign might be presented in the
scanner, in which case an abstract association between an ad
and a product becomes the scanned target.
One product that meets these requirements is music.
Everyone has musical preferences, and most people spend
money on this product. Thus, it is straightforward to find people
to scan who are representative of the music-consuming public,
which is almost everyone. Second, the rise of sites like
myspace.com has created a large repository of music which is
largely unadvertised and unbranded. Because much of this
music is provided directly by the artist, it can be used well in
advance of any ad campaign; moreover, the band is the brand.
Third, metrics of music success are simple and straightforward:
downloads, sales, and ticket receipts. Finally, music is ideally
suited to scanning because the act of listening to it is the same as
consuming it. Thus, imaging the neural response to music is a
direct measure of the consumption experience. Subsequent
success is then a combination of quality, branding, and
marketing.
In a previous study of adolescents, we measured the
interaction of social influence in the form of popularity ratings
with the consumption experience of music (Berns, Capra,
Please cite this article as: Berns, G.S., & Moore, S.E., A neural predictor of cultural popularity, Journal of Consumer Psychology (2011), doi:10.1016/
j.jcps.2011.05.001
Moore, & Noussair, 2010). Using fMRI, we found that although
an individual's musical preferences were strongly correlated
with activity in the caudate nucleus, the effect of social
information varied between participants. The tendency to
change one's evaluation of a song was positively correlated
with activation in the anterior insula and anterior cingulate, two
regions that are associated with physiological arousal and
negative affective states. While this earlier study examined the
effect of popularity information on individual preferences, here
we report a longitudinal analysis in which we examine the
relationship between brain responses and popularity of music
from the other direction: do neural responses to music in an
fMRI study predict subsequent commercial success of the song
and artist?
Material and methods
A total of 32 adolescent participants were studied. Five were
excluded from the fMRI analyses due to either excessive
movement or susceptibility artifacts. Although this was a
relatively high exclusion rate compared to adult studies, it was
comparable to previous fMRI studies in children and adoles-
cents, who tend to move more than adults (Galvan et al., 2006).
Prior to the experiment, they were screened for the presence of
medical and psychiatric diagnoses, and none were taking
medications. There were 14 female and 13 male participants
between the ages of 12 and 17.9 (mean 14.6). Fifteen were
Caucasian, eight were African-American, one was Hispanic,
and three were “Other.” The primary stimuli used were 15-s clips
from songs downloaded from MySpace.com. Songs were
downloaded between October 23 and November 8, 2006. In
order to minimize the possibility that participants would recog-
nize the artists, songs from unsigned musicians or relatively
unknown artists were used. A total of 20 songs were downloaded
in each of the following genres: Rock, Country, Alternative/
Emo/Indie, Hip-Hop/Rap, Jazz/Blues, and Metal (identified
by the MySpace category). At the time of download, the number
of times each song had been played was recorded, and this was
used to calculate the popularity of each song among MySpace
users. Each song was converted from MP3 to WAV format
and a 15-s clip was extracted that included either the hook or
chorus of the song. These 15-s clips were subsequently used in the
experiment.
At the beginning of each session, individuals' rankings of
musical genres were elicited. Participants were provided with
a list of the six musical genres, and were instructed to rank
the genres from 1 (“the type you like the best”) to 6 (“the type
you like the least”). Each participant's top three genres were
subsequently used in the experiment. Emory University's
Institutional Review Board approved all procedures. Individ-
uals then entered the scanner, and the total scan time was
approximately one hour. The scanning was performed on a
Siemens 3T Trio. Each subject received a T1-weighted structural
image (TR = 2600 ms, TE = 3.93 ms, flip angle = 8, 224 × 256
matrix, 176 sagittal slices, 1 mm cubic voxel size), a DTI scan
(TR = 6500 ms, TE = 90 ms, flip angle = 90, FOV = 220 mm,
128 × 128 matrix, 34 axial slices, 1.7 × 1.7 × 2.5 mm voxel size,
 G.S. Berns, S.E. Moore / Journal of Consumer Psychology xx (2011) xxx–xxx 3

6 sets of 12 directional b = 1000 and 1 b = 0 images), and 3
functional runs of BOLD-weighting (TR = 2000 ms, TE = 31 ms,
flip angle = 90, FOV = 192 mm, 64 × 64 matrix, 28 axial slices,
3 mm cubic voxel size). Each individual participated in 60 trials.
Each trial was divided into two stages; in each stage the subject
listened to the same 15-s song clip (Fig. 1a). During stage 1, no
popularity information was shown. After listening, subjects
were required to rate the song based on (a) how familiar it was and
(b) how much they liked it. Both ratings used a 1–5 star scaling
system. To prevent the subject from passively accepting a default
rating, each rating screen began with 0 stars, which could not be
accepted as a final selection. After the rating was entered, stage 2
of the trial took place. The clip was played again, after which
the subject provided another likability rating. Twenty songs in
each of the subject's top-three genres were presented in random
order throughout the experiment. In 2/3 of the trials, during the
second listen, the song's popularity was displayed in the 1–5 star
scaling system. The 40 trials in which the popularity display
appeared were sequenced randomly among the 60 trials. Only
brain activation data from the first listening period was used in
the subsequent analysis. As an incentive to accurately reveal their
song preferences, each subject received a CD with their top-rated
songs.
Nielsen SoundScan was used as the source of post factum
popularity information over the three years since the songs
were originally chosen for the study. The SoundScan database
was searched for information on the performers of each of the
120 songs in the study. Sales data were available for 87 songs,

Fig. 1. Design of popular music experiment and brain regions correlated with the average likability of each song. a) Timing of events for a typical trial. Each song was
played for 15 s. Each participant heard 60 songs from their favorite 3 genres. Following the song, participants rated the song for familiarity and likability. The trial was
then repeated with the average popularity shown on 2/3 of the trials and blocked for 1/3 of the trials. Only the initial listening period (red) was used for subsequent
analyses. The first listening period for each of the songs was modeled separately, for a total of sixty 15 second variable duration events for each participant. Second-
level models for each of the 120 songs were constructed as one-sample t-tests from the contrast images of the first listening period from the first-level model. A third-
level model was then built using the positive contrast images from the second-level model. This model also included a covariate of the first likability rating for each
song, averaged over the participants who heard that song. b) Brain regions positively correlated with the average likability of the song from the third-level model
(p b 0.005, cluster extent ≥ 56, yielding whole-brain FDR b 0.05) were limited to three areas: cuneus, orbitofrontal cortex, and ventral striatum. (For interpretation of
the references to color in this figure legend, the reader is referred to the web version of this article.)

Please cite this article as: Berns, G.S., & Moore, S.E., A neural predictor of cultural popularity, Journal of Consumer Psychology (2011), doi:10.1016/
j.jcps.2011.05.001
4 G.S. Berns, S.E. Moore / Journal of Consumer Psychology xx (2011) xxx–xxx
and this data was extracted from the SoundScan database during
May 2010. The obtained metric aggregated all recorded sales of
the song from release through May 2010, including singles,
albums, and compilations.
Preprocessing of the fMRI data was executed in SPM5
(Functional Imaging Laboratory, UCL, London). The preproces-
sing pipeline consisted of slice timing correction, motion
correction, spatial normalization, and smoothing (with an 8 mm
Gaussian kernel). A first-level GLM was constructed in SPM5
for each of the 27 participants. The first listening period for
each of the 60 songs was modeled separately, for a total of sixty
15-second variable duration events. All of the second listening
periods, also 15 s variable duration events, were collapsed into a
single condition for each run. All three variable duration rating
phases of the trial (familiarity, first likability, and second
likability) were also collapsed into one condition to model the
act of rating including the button presses. The motion parameters
were also included in the model as an effect of non-interest.
Second-level models for each of the 120 songs were constructed
as one-sample t-tests in SPM5 using contrast images of the first
listening period from the first-level model above. Since every
participant did not hear every song, the number of contrast images
in each of these second-level models ranged from 3 to 23. A
third-level model, also a one-sample t-test, was then built in
SPM5 using the positive contrast images from the second-level
model above. This model also included a covariate of the first
likability rating for each song, averaged over the participants
who heard that song. We used the likability covariate to identify
ROIs. Statistical thresholds were determined based on the
estimated smoothness of the 3rd-level contrasts. Using the
AlphaSim routine in AFNI, we estimated the combination of
height and extent thresholds that yielded a whole-brain
FDR b 0.05 (10,000 iterations). First, white matter and CSF
were masked out using the SPM probabilistic gray matter map.
With a gray matter probability N 0.6, this results in a mask that
retains most gray matter while effectively eliminating most
white matter and CSF, which would otherwise inflate the
required cluster size. Second, we used 3dFWHMx to estimate
the image smoothness from the square root of the masked SPM-
generated ResMS image and input into AlphaSim. Finally,
using a voxel level threshold of p b 0.005, the extent threshold
that yielded a cluster level alpha of 0.05 was determined to
be k ≥ 56 (Logan & Rowe, 2004; Zhang, Nichols, & Johnson,
2009). The 60% gray matter mask was applied to all contrasts
before using these thresholds. Only three brain regions showed
a significant correlation between activation and average
song likability: cuneus, orbitofrontal cortex (OFC) and ventral
striatum/nucleus accumbens (NACC) (Fig. 1b). The activations
for each song were then extracted from these regions of interest
(ROIs) for subsequent analyses with song sales data. This
approach ensures that the ROIs are defined independently from
the variable of interest (song sales).
Results
The vast majority of songs in our sample were not com-
mercially successful. The distribution of sales exhibited a long
Please cite this article as: Berns, G.S., & Moore, S.E., A neural predictor of cultural popularity, Journal of Consumer Psychology (2011), doi:10.1016/
j.jcps.2011.05.001
tail (Fig. 2a), with only three songs' albums meeting the
industry standard for “gold” (500,000 units). Given the large
number of songs released annually, this is not surprising. To
normalize the distribution, sales data were log-transformed for
subsequent analyses. First, we checked if either of the subjective
song ratings were predictive of future sales, but neither of the
two average ratings obtained for each song was correlated with
sales data (likability: R = 0.110, p = 0.313; familiarity: R = 0.106,
p = 0.330), nor was the average genre ranking of each song
(p= 0.102). This indicates that simple subjective reports collected
from study participants may not be good predictors of commercial
success.
Although subjective ratings of songs did not correlate with
future sales, the activation within the NACC did (Fig. 2b). Log
(sales + 1) was significantly correlated with the average activation
in NACC during the 15-sec listening period (R = 0.32, p = 0.004).
To see which part of the 15-sec period was responsible for this
correlation (e.g. initial or final reactions), we tested an alternative
model with the listening period divided into three 5-sec segments.
None of these three segments exhibited a greater correlation to
sales than the whole 15-sec period. This indicates that the
mechanism driving the correlation between NACC activity and
sales was integrated over the entire listening period.
To further understand the interrelationship between song
likability, NACC activity, and sales, we constructed a structural
equation model (SEM) (Fig. 3a). The SEM was based on known
anatomical connections between the OFC and NACC and their
relationships to subjective likability and purchase decisions
(Chib, Rangel, Shimojo, & O'Doherty, 2009; Knutson, et al.,
2007; Montague & Berns, 2002; O'Doherty, Kringelbach, Rolls,
Hornak, & Andrews, 2001; Plassmann et al., 2007; Rolls, 2000).
Consistent with this literature, the average song likability had
significant path coefficients to both the OFC and NACC, and
because of the direct connection between OFC and NACC, this
path was also significant. However, the final pathway linking the
brain to album sales was mediated only through the NACC. When
these relationships were visualized, it became clear that “hit”
songs did not result from a specific combination of NACC and
OFC activity, but that “non-hits” were associated with a com-
bination of both low OFC and low NACC activity (Fig. 3b). This
relationship was quantified through logistic regressions on
different hit/non-hit thresholds of sales (Fig. 3c). With thresholds
in the range of 15,000 to 35,000 units sold, the logistic model
achieved reasonable accuracy in correctly classifying hits and
non-hits. For example, with a hit-threshold of 15,000 units, the
logistic model correctly classified 80% of the non-hits; however,
this came at a cost of missing true hits (but still correctly classified
30% of the hits). To test the possibility that consistency of brain
responses might also be predictive of sales, we formulated a
model that included a term for the reciprocal of the variance of the
brain response across subjects that heard each song; however, this
term was not significant [F(1,83) = 0.28, p = 0.597].
Discussion
Our results demonstrate that not only are signals in reward-
related regions of the human brain predictive of individual
 G.S. Berns, S.E. Moore / Journal of Consumer Psychology xx (2011) xxx–xxx 5

Fig. 2. Distribution of number of albums sold and correlation with nucleus accumbens activation. a) Song sales were determined by sales data reported by Nielsen
SoundScan from the album release through May 2010. The distribution was positively skewed, indicating that most songs did not sell many units and with a long tail to
the right. Log-transformation of the sales data normalized this distribution for subsequent correlations (inset). b) The log of the number of units sold of a song was
significantly correlated with the average nucleus accumbens [MNI coordinates: 9, 6, − 9; p b 0.005, 59 voxels] activation during the listening of the song (R = 0.32,
p = 0.004). Exclusion of the far left outlier resulted in a decreased correlation but which was still significant (R = 0.27, p = 0.013).

purchase decisions, they are also modestly predictive of
population effects. While the nascent field of neuromarketing
has made claims to this effect, truly prospective data has been
lacking (Ariely & Berns, 2010). Surprisingly, our data suggest
some validity to these claims. Why might this be? If the specific
cohort of participants in an imaging study is representative of
a particular population, then it follows that the results should
generalize. When it comes to music, however, it may be difficult
to know on which dimensions of a population to match (e.g. age,
gender, SES, region). The Recording Industry Association of
America (RIAA) estimates that the age range of our cohort
accounts for approximately 20% of music sales (www.riaa.com/
keystatistics.php). To test whether the musical tastes of our cohort
were representative of the population, we compared our cohort's
pre-scan genre rankings to the 2009 Nielsen sales by category and
found a significant correlation (Kendall's τ = −0.733, p = 0.05;
assuming that our hip-hop category is equivalent to Nielsen's
R&B category), showing that our cohort had similar tastes as the
national population.
Our results also raise the question of why the brain activation
was predictive of future sales but the self-reported likability
ratings were not. One possibility is that the questions were not

Please cite this article as: Berns, G.S., & Moore, S.E., A neural predictor of cultural popularity, Journal of Consumer Psychology (2011), doi:10.1016/
j.jcps.2011.05.001
6 G.S. Berns, S.E. Moore / Journal of Consumer Psychology xx (2011) xxx–xxx
Fig. 3. Structural equation model linking likability ratings to albums sold through
OFC and NACC, and logistic regression to categorize “hits.” a) Although the average
likability of each song was not directly correlated with the number of albums sold, the
relationship was moderated by activity in the OFC and NACC. All path coefficients
and main effects were significant (pb 0.05) except that linking OFC to albums sold.
b) The sales of each are represented by the area of each circle and plotted as a function
of both OFC and NACC activity. Although the “hits” (big circles) are scattered
throughout the activation space, there are a higher proportion of “non-hits” (small
circles) toward the lower left. c) To test the ability to correctly categorize “hits” and
“non-hits,” logistic regressions were performed. The threshold of hit vs. non-hit was
varied from 1000 to 50,000 units. Thresholds in the range 15,000 to 35,000 correctly
categorize a sizable fraction of hits while also correctly rejecting most non-hits.
Please cite this article as: Berns, G.S., & Moore, S.E., A neural predictor of cultural popularity, Journal of Consumer Psychology (2011), doi:10.1016/
j.jcps.2011.05.001
adequately specified to differentiate future success (Cronbach &
Meehl, 1955). A more detailed debriefing of the reactions to each
song might have yielded better predictive data. For example, a
choice-based conjoint model might have been superior to simple
rankings (Green & Srinivasan, 1990; Griffin & Hauser, 1993).
Another possibility is that above a certain quality threshold,
songs are too similar to prospectively differentiate them, but slight
differences in quality become magnified in superstar markets
(Rosen, 1981). Although our results do not invalidate any of these
approaches, they do suggest that brain imaging can augment
them. The SEM showed that even though likability was not
directly correlated with future sales, the OFC and NACC mod-
erated this relationship as essentially hidden variables within
the brain. Asking an individual how much they like something
requires several cognitive operations, including the initial
processing of the stimulus, referencing similar items with which
the individual has experience, projection of future utility, all of
which may be subject to framing effects of the experiment. In
contrast, brain responses in reward-related regions are likely to
reflect sub-conscious processes and may yield measurements that
are less subject to cognitive strategies. This would be especially
true during the consumption of music, which occurred during the
listening phase of our experiment. Thus, while the act of rating
something requires metacognition, the brain response during the
consumption of the good does not, and the latter may prove
superior to rating approaches.
Although our data raise the possibility of predicting future
popularity in the form of commercial sales, the actual per-
formance of such a model depended on the definition of a “hit.”
The scarcity of true hits (e.g. 500,000 units) in our sample,
underscores the difficulty in evaluating a hit-predictor and
confirms the previously noted shift toward superstars (Krueger,
2005). The logistic model performed well in identifying non-
hits, which may itself be valuable information, but given the
widely varying marketing approaches that are invested in bands
(Vogel, 2007), it is surprising that we found any predictive
power at all. The fact that we used a wide variety of songs in
different genres certainly made the prediction of hits more
difficult. A more focused presentation of songs, perhaps within
a single genre and pre-screened for minimal quality, would
increase the likelihood of hit-prediction. A more targeted group
of study participants that is representative of a particular music
consuming demographic might also increase predictive power.
However, predicting hit-songs may always be a particularly
difficult task. A recent study of internet searches found good
predictability for revenue of movies and video games but less
so for music (Goel, Hofman, Lahaie, Pennock, & Watts, 2010),
which makes our results even more surprising and refutes the
idea that hits are random (Bielby & Bielby, 1994).
Our results may also have implications for branding. In
commercial music, the band is the brand. We calculated com-
mercial success based on the number of units sold, but this
number included all sources of a particular song. As a result, the
sales numbers were dominated by album sales, which of course
contain many songs and may been heavily influenced by the
band/artist reputation (i.e. the band brand). It is hard to know
what marketing efforts might have been done to promote a
 G.S. Berns, S.E. Moore / Journal of Consumer Psychology xx (2011) xxx–xxx
band. However, it has been estimated that only 10% of new
releases end up making a profit for a record label (Vogel, 2007).
Consequently, marketing and branding efforts tend to be
minimal until a band shows signs of popularity. With more and
more artists having access to quality production equipment and
being able to release songs directly to the public, neuroimaging
tools may have real utility to help labels decide how to invest
limited marketing and branding resources.
Acknowledgments
This work was supported by grants from the National
Institute for Drug Abuse, the National Science Foundation, the
Air Force Office of Scientific Research, and the Office of Naval
Research.
References
Ariely, D., & Berns, G. S. (2010). Neuromarketing: The hope and hype of
neuroimaging in business. Nature Reviews Neuroscience, 11, 284–292.
Berns, G. S., Capra, C. M., Moore, S., & Noussair, C. (2010). Neural
mechanisms of the influence of popularity on adolescent ratings of music.
NeuroImage, 49, 2687–2696.
Bielby, W. T., & Bielby, D. D. (1994). “All hits are flukes”: Institutionalized
decision making and the rhetoric of network prime-time program
development. American Journal of Sociology, 99(5), 1287–1313.
Chib, V. S., Rangel, A., Shimojo, S., & O'Doherty, J. P. (2009). Evidence for a
common representation of decision values for dissimilar goods in human
ventromedial prefrontal cortex. The Journal of Neuroscience, 29(39),
12315–12320.
Cronbach, L. J., & Meehl, P. E. (1955). Construct validity in psychological tests.
Psychological Bulletin, 52, 281–302.
Galvan, A., Hare, T. A., Parra, C. E., Penn, J., Voss, H., Glover, G., et al. (2006).
Earlier development of the accumbens relative to orbitofrontal cortex might
underlie risk-taking behavior in adolescents. The Journal of Neuroscience,
26, 6885–6892.
Goel, S., Hofman, J. M., Lahaie, S., Pennock, D. M., & Watts, D. J. (2010).
Predicting consumer behavior with Web search. Proceedings of the
National Academy of Sciences of the United States of America, 107(41),
17486–17490.
Green, P. E., & Srinivasan, V. (1990). Conjoint analysis in marketing: New
developments with implications for research and practice. Journal of
Marketing, 54(4), 3–19.
Please cite this article as: Berns, G.S., & Moore, S.E., A neural predictor of cultural popularity, Journal of Consumer Psychology (2011), doi:10.1016/
j.jcps.2011.05.001
7
Griffin, A., & Hauser, J. R. (1993). The voice of the customer. Marketing
Science, 12(1), 1–27.
Hare, T. A., O'Doherty, J., Camerer, C. F., Schultz, W., & Rangel, A. (2008).
Dissociating the role of the orbitofrontal cortex and the striatum in the
computation of goal values and prediction errors. The Journal of Neuroscience,
28(22), 5623–5630.
Kenning, P. H., & Plassmann, H. (2008). How neuroscience can inform consumer
research. IEEE Transactions on Neural Systems and Rehabilitation Engineer-
ing, 16(6), 532–538.
Knutson, B., Rick, S., Wimmer, G. E., Prelec, D., & Loewenstein, G. (2007).
Neural predictors of purchases. Neuron, 53, 147–156.
Krueger, A. B. (2005). The economics of real superstars: The market for
rock concerts in the material world. Journal of Labor Economics, 23(1),
1–30.
Lee, N., Broderick, A. J., & Chamberlain, L. (2007). What is neuromarketing? A
discussion and agenda for future research. International Journal of
Psychophysiology, 63, 199–204.
Lindstrom, M. (2008). Buyology. Truth and lies about why we buy. New York:
Doubleday.
Logan, B. R., & Rowe, D. B. (2004). An evaluation of thresholding techniques
in fMRI analysis. NeuroImage, 22, 95–108.
McClure, S. M., Li, J., Tomlin, D., Cypert, K. S., Montague, L. M., &
Montague, P. R. (2004). Neural correlates of behavioral preference for
culturally familiar drinks. Neuron, 44, 379–387.
Montague, P. R., & Berns, G. S. (2002). Neural economics and the biological
substrates of valuation. Neuron, 36, 265–284.
O'Doherty, J., Kringelbach, M. L., Rolls, E. T., Hornak, J., & Andrews, C.
(2001). Abstract reward and punishment representations in the human
orbitofrontal cortex. Nature Neuroscience, 4(1), 95–102.
Plassmann, H., O'Doherty, J., & Rangel, A. (2007). Orbitofrontal cortex
encodes willingness to pay in everyday economic transactions. The Journal
of Neuroscience, 27(37), 9984–9988.
Plassmann, H., O'Doherty, J., Shiv, B., & Rangel, A. (2008). Marketing actions
can modulate neural representations of experienced pleasantness. Proceedings
of the National Academy of Sciences of the United States of America, 105(3),
1050–1054.
Rolls, E. T. (2000). The orbitofrontal cortex and reward. Cerebral Cortex, 10,
284–294.
Rosen, S. (1981). The economics of superstars. American Economic Review, 71
(5), 845–858.
Vogel, H. L. (2007). Entertainment industry economics. A guide for financial
analysis (7th ed.). Cambridge: Cambridge University Press.
Yoon, C., Gutchess, A. H., Feinberg, F., & Polk, T. A. (2006). A functional
magnetic resonance imaging study of neural dissociations between brand
and person judgments. Journal of Consumer Research, 33, 31–40.
Zhang, H., Nichols, T. E., & Johnson, T. D. (2009). Cluster mass inference via
random field theory. NeuroImage, 44, 51–61.
8024 • The Journal of Neuroscience, June 9, 2010 • 30(23):8024 – 8031

Behavioral/Systems/Cognitive

Neural Responses to Unattended Products Predict Later

Consumer Choices
Anita Tusche,1,2 Stefan Bode,1,2,3 and John-Dylan Haynes1,2
1Bernstein Center for Computational Neuroscience Berlin, Charité–Universitätsmedizin Berlin, 10115 Berlin, Germany, 2Max-Planck-Institute for Human
Cognitive and Brain Sciences, 04103 Leipzig, Germany, and 3Department of Neurology, Otto-von-Guericke University, 39120 Magdeburg, Germany

Imagine you are standing at a street with heavy traffic watching someone on the other side of the road. Do you think your brain is
implicitly registering your willingness to buy any of the cars passing by outside your focus of attention? To address this question, we
measured brain responses to consumer products (cars) in two experimental groups using functional magnetic resonance imaging.
Participants in the first group (high attention) were instructed to closely attend to the products and to rate their attractiveness. Partici-
pants in the second group (low attention) were distracted from products and their attention was directed elsewhere. After scanning,
participants were asked to state their willingness to buy each product. During the acquisition of neural data, participants were not aware
that consumer choices regarding these cars would subsequently be required. Multivariate decoding was then applied to assess the
choice-related predictive information encoded in the brain during product exposure in both conditions. Distributed activation patterns
in the insula and the medial prefrontal cortex were found to reliably encode subsequent choices in both the high and the low attention
group. Importantly, consumer choices could be predicted equally well in the low attention as in the high attention group. This suggests
that neural evaluation of products and associated choice-related processing does not necessarily depend on attentional processing of
available items. Overall, the present findings emphasize the potential of implicit, automatic processes in guiding even important and
complex decisions.

Introduction selective processing (Peelen et al., 2009). To date, no study has

Brain responses obtained during active evaluation of products
and explicit deliberation about purchases have been found to
predict consumer choices (Knutson et al., 2007). Interestingly,
some evidence suggests that automatic brain processes might
guide human judgments and choices, even in the absence of ex-
plicit deliberation and attention to the choice task. Brain re-
sponses were shown to engage automatically in assessing facial
attractiveness and preferences even when such judgments were
not part of the designated task (O’Doherty et al., 2003; Kim et al.,
2007). Likewise, brain activation was reported to reflect prefer-
ences when participants evaluate stimuli with respect to other,
non-preference-related aspects (Lebreton et al., 2009). However,
the precise role of stimulus-related attention in mediating such
automatic valuation processes remains unclear. On the one hand,
sensory responses to unattended stimuli have been shown be
strongly reduced (Rees et al., 1997; Martínez et al., 1999; Kastner
and Ungerleider, 2000). On the other hand, spatially unattended
stimuli have been reported to undergo substantial category-
Received Jan. 6, 2010; revised March 16, 2010; accepted April 8, 2010.
This work was funded by the Max Planck Society, the German Research Foundation, and the Bernstein Compu-
tational Neuroscience Program of the German Federal Ministry of Education and Research. We thank Christian
Kalberlah, Chun Siong Soon, and Carsten Bogler for their contributions to this project and Thorsten Kahnt for
comments on an earlier version of the manuscript.
Correspondence should be addressed to John-Dylan Haynes or Anita Tusche, Charité–Universitätsmedizin Berlin,
Bernstein Center for Computational Neuroscience, Haus 6, Philippstrasse 13, 10115 Berlin, Germany. E-mail:
haynes@bccn-berlin.de. or anita.tusche@bccn-berlin.de.
DOI:10.1523/JNEUROSCI.0064-10.2010
Copyright © 2010 the authors 0270-6474/10/308024-08$15.00/0
directly compared neural responses to attended versus unat-
tended products, and the impact of spatial attention on the pre-
diction of economic decisions. Here, we investigated whether
brain responses predict consumer choices even when products
are entirely task-irrelevant and presented outside the focus of
attention.
To examine the role of attention in the prediction of product
choices from brain activity, we performed an experiment with
two different groups of male participants. In each trial, partici-
pants were presented with an image of a car while their brain
responses were measured using functional magnetic resonance
imaging (fMRI). Participants in group 1 (high attention) were
instructed to actively evaluate and rate the attractiveness of each
particular car after its presentation (Fig. 1A). Hence, functional
brain responses were acquired while products were task-relevant
and in the focus of attention. In contrast, participants in group 2
(low attention), were engaged in a demanding visual fixation task
while task-irrelevant images of cars were passively presented out-
side the focus of attention in the background of the screen (Fig.
1B). After the scanning session, participants from both groups
were instructed to realistically picture themselves in a consumer
setting where they had to decide on a new car. For each of the
previously presented products, participants were then asked to
state whether they would like to purchase this car or not (Fig. 1C).
During scanning, participants from both groups were unaware
that they would later be asked about their potential purchases. Mul-
tivariate pattern classification (Haxby et al., 2001; Kriegeskorte et al.,
2006; Norman et al., 2006; Haynes et al., 2007) was then applied
Tusche et al. • Attention-Independent Prediction of Consumer Choices J. Neurosci., June 9, 2010 • 30(23):8024 – 8031 • 8025

Figure 1. Experimental paradigms. A, In each trial, participants in group 1 (high attention) were presented with images of single cars for 2.4 s followed by a randomized response-button mapping
screen (displayed for a variable duration of 7.2–12 s). Participants were instructed to closely attend to products and to actively evaluate the attractiveness of the particular car on a four-point scale
by pressing the corresponding response button. B, Participants in group 2 (low attention) were asked to perform a demanding fixation task. They responded to the opening of a centrally presented
fixation square (every 800 ms) with a corresponding left- or right-hand button press. The timing parameters applied in group 1 were retained. Every 7.6 –12 s, a car was passively presented on the
background of the screen for 2.4 s while the fixation task continued. C, After the scanning session, participants from both groups were instructed to realistically picture themselves in a consumer
setting where they had to decide upon a new car. Participants were then asked to state their willingness to buy each product. During scanning, participants of both groups were not informed that
such a consumer choice would subsequently be required.

to the brain responses obtained during product exposure to pre-
dict their stated product choices.
The present study examined the impact of attention on neural
predictors of economic decision making. More precisely, we in-
vestigated the predictability of consumer choices from brain re-
sponses for unattended products. A successful prediction would
indicate that product valuation and associated choice-related
processing can take place automatically— even with attention
diverted from products.
Materials and Methods
Participants and stimuli. Group 1 consisted of 17 participants (aged be-
tween 24 and 32 years) and group 2 of 15 participants (aged between 22
and 27 years). Participants were healthy male volunteers, had normal or
corrected-to-normal vision, were right-handed and stated that they were
interested in cars before the experiment. The sample was limited to males
interested in cars to ensure that participants were familiar with and pos-
sess stable mental representations about a wide range of items from the
product group. Both paradigms were approved by the local ethics com-
mittee. All participants were paid €12 to take part and all gave written
informed consent.
In both fMRI paradigms, monochrome images of 10 real life cars were
used as stimuli. The selection of cars was based on a behavioral pretest
with an independent sample of participants in a way that maximized
variability in ratings between participants. All pictures were obtained
from the Internet and were normalized with regard to size and contrast.
During both fMRI paradigms, images were centrally presented against a
white background using MATLAB 7.0 (The MathWorks) and the Cogent
toolbox (http://www.vislab.ucl.ac.uk/Cogent). Data from four partici-
pants (in group 1 and 2) as well as data for one car (group 1) and two cars
(group 2) was excluded because of missing variance in consumer choices.
Experimental paradigms. During each trial of both event-related fMRI
paradigms, participants were presented with a single image of a car for
2.4 s. Participants in group 1 (high attention) were instructed to judge the
subjective attractiveness of the specific car on a four-point-scale via a
button-press (1 ! “dislike it a lot,” 2 ! “dislike it a little,” 3 ! “like it a
little,” 4 ! “like it a lot”). The mapping of buttons to attractiveness values
was randomized on a trial-by-trial basis to avoid motor preparation
during product exposure. Thus, each presentation of a car was followed
by a randomized response-mapping screen, which was presented for
7.2–12 s. Responses were given using the index and middle fingers of
both hands operating separate button boxes. Participants in group 2 (low
attention) were instructed to attend to a black square presented centrally
on a white screen. Every 800 ms, the square opened to either the left or the
right side. Participants had to respond to each opening with a matching
left or right button press using the index fingers of both hands. After a
pseudo-randomized duration of 7.2–12 s, a single image of a car was
presented for 2.4 s in the background of the screen while the fixation
task continued. Thus, the visual presentation of cars in group 2 mir-
rored that of group 1 but differed in the direction of attention that was
diverted from products. Participants in group 2 were explicitly in-
structed to maximize performance on the fixation task throughout
the entire experiment.
For both paradigms, scanning was performed in a single measurement
session during which seven independent runs were acquired. The runs
were separated by breaks of "1 min during which no scanning data were
obtained. Within a run, each of the 10 cars (see stimuli section above) was
presented three times, resulting in a total number of 30 trials per run. For
each run, the presentation order of cars was pseudo-randomized such
that the same product was never shown in two consecutive trials. Pseudo-
randomized durations between presentations of cars varied between 7.2
to 12 s, meaning that each car was combined with different interstimulus
8026 • J. Neurosci., June 9, 2010 • 30(23):8024 – 8031
intervals equally often. This procedure ensured that the onset time of the
next stimulus was unpredictable and event-related brain responses were
clearly separable between trials.
Subsequent to the scanning sessions in both groups, participants were
given a questionnaire with the question “Would you buy this car?” and
the response options “No/Not sure/Yes.” Importantly, during the acqui-
sition of brain responses participants were unaware that it would be
necessary to make such a choice later on. In addition to consumer
choices, participants in group 2 had to explicitly judge the attractiveness
of each particular car on a four-point scale. Moreover, for both groups
familiarity ratings were obtained for all of the products presented. A
between-subject design was chosen to ensure that participants in the
“low attention” condition were completely unaware of the task-relevance
of the products presented.
Image acquisition. For both groups, functional imaging was performed
on a 3-Tesla Siemens TRIO scanner with a standard head coil. T2*-
weighted functional images were obtained using an echoplanar imaging
(EPI) sequence [repetition time (TR) ! 2.4 s, echo time ! 30 ms]. For
each run, 152 EPI volumes were collected (36 ascending axial slices per
volume, slice thickness 2 mm, in-plane resolution 3 mm " 3 mm, 1 mm
interslice gap, matrix size 64 " 64). The whole session consisted of seven
runs. Due to technical problems, only five runs were acquired for two
participants in group 2.
Data analysis. Data from both groups were analyzed in a similar man-
ner. In a first step, the acquired volumes were slice timed and realigned.
Preprocessed data were then analyzed using a general linear model
(GLM) (Friston et al., 1994) as implemented in SPM2 (http://www.fil.
ion.ucl.ac.uk/spm). For every run, parameter estimates for “purchase”
and “no purchase” choices were estimated, based on stated consumer
choices obtained after the scanning session. Additionally, biometric
pulse data acquired during the scanning were included as coregressors of
no interest to account for pulse artifacts. Four regressors were created by
using the integral of the pulse curve of four successive time bins (TR/4 !
0.6 s) during the product presentation phase.
In the second step, multivariate pattern classification using a support
vector machine (SVM) was applied to the parameter estimates of the
consumer choices (Cox and Savoy, 2003; Mitchell et al., 2004; Kamitani
and Tong, 2005; Haynes and Rees, 2006). To realize the classification a
standard radial basis function kernel as implemented in LIBSVM (http://
www.csie.ntu.edu.tw/#cjlin/libsvm) was used. In contrast to a classifier
restricted to linear effects, this approach allowed interactions between
features and nonlinear functions to drive the prediction of subsequent
choices. A comparative figure illustrating prediction accuracies obtained
with a linear classification compared with our nonlinear approach is
provided in supplemental Figure 1 (available at www.jneurosci.org as
supplemental material).
Multivariate pattern classification techniques take advantage of infor-
mation contained in multiple voxels distributed across space. They allow
investigating whether spatial patterns of brain activation contain stable
information about different experimental conditions (e.g., purchase vs
no purchase). To achieve best predictive accuracy, the classifier weights
the contributions of the different voxels optimally. In our case, some
voxels within a searchlight cluster were weighted positively and others
negatively by the classification algorithm. Moreover, the nonlinear clas-
sifier also takes their interactions into account.
To ensure an unbiased analysis of the neural activation patterns through-
out the whole brain, a “searchlight” approach was used (Kriegeskorte et al.,
2006; Haynes et al., 2007). Given that this approach does not depend on
a priori assumptions about informative brain regions or prior voxel se-
lection, the problem of circular analysis (or “double dipping”) can be
avoided (Kriegeskorte et al., 2009). For each participant, a sphere with a
radius of 4 voxels was created around every voxel vi of the measured
volume. For each sphere, we investigated whether the local pattern of
activation during product exposure predicted the willingness to buy that
was stated after scanning (purchase vs no purchase). It should be noted
that this analysis is predictive because it uses brain activity during expo-
sure to predict purchase ratings obtained after scanning.
For every run, parameter estimates from the GLM were extracted for
each of the N voxels in the sphere around voxel vi and transformed in an
Tusche et al. • Attention-Independent Prediction of Consumer Choices
N-dimensional pattern vector. For each run, two pattern vectors were
separately created for the purchase and no purchase conditions (see sup-
plemental Fig. 2A, available at www.jneurosci.org as supplemental ma-
terial). Initially, the pattern vectors of six of the seven runs were used for
the training (“training dataset”) of the nonlinear support vector machine
classifier with a fixed regularization parameter C ! 1. This provided the
basis of the subsequent classification of the pattern vectors of the remain-
ing seventh run (“test dataset”) as belonging to either the purchase or no
purchase condition (see supplemental Fig. 2B, available at www.
jneurosci.org as supplemental material). The procedure was indepen-
dently repeated seven times with a different run used as the test dataset to
achieve a sevenfold cross-validation. (A fivefold cross-validation was
realized for two participants in group 2 because complete data were
acquired only for 5 of the 7 runs). The amount of purchase-related in-
formation of the spatial activation pattern of each spherical cluster was
represented by the average decoding accuracy across all cross-validation
steps and was assigned to the central voxel vi of the cluster. Given that the
number of considered dimensions (equivalent to number of voxels in a
searchlight) exceeded the number of acquired data points, we have to
consider that noise was fitted to the training of the classifier. The use of
independent data for training and testing, however, controlled for the
impact of potential overfitting, and consequently for the overestimation
in the prediction accuracy (Vul et al., 2009). Moreover, to control for
“peaking” problems, decoding accuracies were calculated by estimating
the mean across seven cross-validation steps.
The described classification was successively performed for all clusters
created around every measured voxel, resulting in a three-dimensional
map of average classification accuracies for each participant. These accu-
racy maps were then spatially normalized to a standard brain [Montreal
Neurological Institute (MNI) EPI template as implemented in SPM2]
and resampled to an isotropic spatial resolution of 3 " 3 " 3 mm 3.
Finally, a standard second-level statistical analysis as implemented in
SPM2 was performed to identify brain regions that allowed classification
of consumer choices across participants in each group. For both groups,
analyses were based on the normalized three-dimensional accuracy maps
of each subject. Each single point of an individual accuracy map repre-
sented the average decoding accuracy of a searchlight surrounding this
position across all cross-validation steps. Thus, each value represented
the amount of choice-related information of the spatial activation pat-
tern of a surrounding spherical cluster. To assess the statistical signifi-
cance of these accuracy values across subjects, individual accuracy maps
of one group were submitted to a voxelwise one sample t test and con-
trasted against chance level. Since the classification was based on two
alternatives (purchase vs no purchase), chance level was 50%. Familywise
error (FWE) correction for multiple comparisons was implemented to
control for false positives. Only regions passing this stringent statistical
threshold ( p $ 0.05, FWE-corrected, whole brain) and showing signifi-
cant decoding accuracies above chance were considered relevant for in-
formation encoding (Haynes et al., 2007; Soon et al., 2008).
Supplemental data analysis. After each product presentation, partici-
pants in group 1 (high attention) had to rate the attractiveness of the
particular car on a four-point scale via a button-press. To ensure that
brain regions predictive for subsequent consumer choices do not mainly
reflect attractiveness, we conducted an additional multivariate decoding
analysis. Except for the GLM parameter estimates representing attrac-
tiveness ratings instead of consumer choices, the multivariate searchlight
decoding was identical to the one described for the main analysis. An-
other separate decoding analysis was performed on the button-presses to
confirm that regions predicting product choices did not simply encode
subsequent motor responses. Here, parameter estimates of the GLM
were created based on motor responses indicating subjective attractive-
ness judgments. Apart from that, the analysis was similar to the one
described above to predict consumer choices. In both additional decod-
ing analyses chance level was 25%, because the classification was always
based on four alternatives of either attractiveness or button-presses. We
considered only regions showing significant decoding accuracies above
chance as relevant for encoding of attractiveness and motor responses.
Furthermore, classic univariate analyses were conducted for data from
both groups. This enabled us to investigate whether there are single vox-
Tusche et al. • Attention-Independent Prediction of Consumer Choices
els whose mean activation is significantly more strongly activated in one
experimental condition of consumer choices compared with another
(purchase vs no purchase). The functional imaging data were prepro-
cessed using slice time correction, motion correction, were spatially nor-
malized to a standard stereotaxic space (MNI EPI template), resampled
to an isotropic spatial resolution of 3 ! 3 ! 3 mm 3 and smoothed with a
Gaussian kernel of 6 mm full-width half-maximum. Except for these
differences in the preprocessing, parameter estimates were created as
described in the first step of the multivariate decoding approach. Param-
eter estimates of purchase and no purchase choices were then contrasted
against each other on a single-subject level. Subsequently, a second-level
statistical analysis was performed to identify regions that were signifi-
cantly more strongly activated in one of the two conditions across
participants.
To test whether engagement of attention in one region of the visual
field (corresponding to the square of the fixation task in group 2) can
strongly decrease the processing of task-irrelevant background stimuli,
we compared the blood oxygenation level-dependent (BOLD) signal
change during product exposure of group 2 (low attention) and group 1
(high attention). In the first step, GLM parameter estimates for visual
responses during product exposure were created for each participant and
contrasted against baseline. Based on these contrasts, the individual
BOLD signal change was estimated for each participant. In the second
step, these estimates were used to calculate the average BOLD signal
change across participants by implementing a second-level analysis for
Groups 1 and 2 individually. Subsequently to contrasting the resulting
parameter estimates against baseline, we identified the peak activation
area of both hemispheres across these two contrast images (located in the
left [MNI "30, "81, "21] and right visual cortex [MNI 27, "84, "18]).
Assuming that the visual information was encoded in both hemispheres,
we pooled the signal change from both peak activation areas. This was
performed for group 1 and group 2 individually. Finally, a t test for
independent samples was applied to compare the average signal change
in group 1 and group 2.
Results
Behavioral results
Consumer choices
In group 1 (high attention), we obtained functional brain responses
while participants closely attended to and actively evaluated prod-
ucts. In group 2 (low attention), participants performed a dis-
traction task at fixation while task-irrelevant products were
presented outside the focus of attention on the back of the screen.
Subsequent to scanning, participants in both groups had to state
their willingness to purchase each particular car. The number of
trials assigned to either the purchase or no purchase condition
was found to be well balanced within both groups and compara-
ble across them. In group 1, the mean distribution of the declared
consumer choices for all products was 54% no purchase, 5%
“maybe,” and 41% purchase. In group 2, 46% of the products
were chosen to be no purchase trials, 4% maybe, and 50% purchase
trials across participants. For both groups, profiles of product selec-
tion were found to vary across participants (see supplemental Table
1, available at www.jneurosci.org as supplemental material). Addi-
tionally, the results of a t test for independent samples confirmed that
product-specific means (group 1: mean 1.87 # SD 0.83; group 2:
mean 1.86 # SD 0.93) of consumer choices were comparable across
both groups (t $ "0.43, p $ 0.67).
Familiarity
After scanning, participants from both groups were asked
whether they had been familiar with the presented products be-
fore the experiment. In group 1 (high attention), participants
reported the products were familiar in 85% of all cases and as
being unknown in 12%. In group 2 (low attention), participants
were acquainted with 87% of the products before the experiment
J. Neurosci., June 9, 2010 • 30(23):8024 – 8031 • 8027
while 10% of products were unknown. For the remaining 3% in
both groups the participants stated they were unsure whether
they were acquainted with the product before the experiment.
For both groups, these findings confirmed that most of the prod-
ucts presented were well known before the experiment.
Task performance and recognition rates
To examine whether attention was effectively removed from
products in the low attention condition as implemented in group
2, we analyzed the behavioral performance in the visual fixation
task during scanning. The high mean percentage of 89% correct
responses indicates that participants attended to the assigned de-
manding fixation task as instructed. Moreover, we compared the
recognition performance in a memory test for both groups (high
and low attention) conducted upon completion of the experi-
ment (independent sample of 41 healthy male volunteers: high
attention group: N $ 20, 26.5 # 2.8 years; low attention group:
N $ 21, 25.8 # 1.9 years, mean # SD). Subsequent to the com-
pletion of either the high attention or the low attention condi-
tion, participants were presented with single images of 20 cars in
randomized order. Ten of these cars were stimuli used in the
experiment while the remaining 10 images displayed previously
unseen cars. For each product, participants had to state whether
this car had been presented during the experiment or not. Size
and position of the products on the computer screen were kept
constant to facilitate recognition performance. Individual hit-
rates [HR(high attention) $ mean 0.97 # SD 0.07, HR(low at-
tention) $ mean 0.63 # SD 0.26] as well as false-alarm-rates
[FAR(high attention) $ mean 0.02 # SD 0.04, FAR(low atten-
tion) $ mean 0.22 # SD 0.18] were found to differ significantly
between high and low attention conditions (HR: t $ "5.41, p %
0.001; FAR: t $ 4.74, p % 0.001). Participants in the low attention
condition showed lower correct recognition of previously seen
cars than participants in the high attention condition. Moreover,
they were more likely to incorrectly rate a previously unseen car
as having been presented during the experiment. These findings
strongly suggest that attention was effectively removed from
products in the low attention condition of our experiment.
fMRI results
Multivariate decoding of subsequent consumer choices
In group 1 (high attention), spatial activation patterns in the
prefrontal cortex (PFC), namely in the left medial frontal gyrus
(82% decoding accuracy), the right dorsomedial PFC (75% de-
coding accuracy) and the bilateral ventromedial PFC (73% decod-
ing accuracy) predicted subsequent consumer choices. Moreover,
the left insula (73% decoding accuracy) and the right parahip-
pocampal gyrus (72% decoding accuracy) were found to contain
stable information about later product choices (see Table 1 for a
complete list of results).
In group 2 (low attention), activation patterns in the left me-
dial PFC (76% decoding accuracy) and the bilateral insula (right:
82% decoding accuracy, left: 72% decoding accuracy) predicted
subsequent consumer choices. Neural responses in the left infe-
rior parietal lobe (82% decoding accuracy) and the bilateral su-
perior temporal gyrus (left: 74% decoding accuracy, right: 70%
decoding accuracy) also encoded choices between cars (see Table
2 for a complete list of results). It should be noted that decoding
accuracies in brain regions predicting subsequent consumer
choices under high and low attention conditions were found to be
comparable (Fig. 2).
To provide further evidence for the statistical validity of the
results obtained by this approach, an additional decoding analy-
8028 • J. Neurosci., June 9, 2010 • 30(23):8024 – 8031 Tusche et al. • Attention-Independent Prediction of Consumer Choices

Table 1. Brain regions encoding the subsequent consumer choices among actively evaluated products in group 1 (high attention)
Accuracy MNI coordinates
Brain region Side BA M SE t value x y z
Frontal lobe
Medial frontal gyrus L 9 82 3.3 9.49 "21 27 30
mPFC (dorsal) R 10 75 2.6 9.53 9 57 21
mPFC (ventral) L 10 73 2.3 9.96 "9 63 12
Limbic lobe
Insula L 73 2.3 9.80 "36 "6 0
Parahippocampal gyrus R 30 72 1.9 10.98 30 "57 3
Occipital lobe
Inferior occipital lobe L 18 73 2.2 10.44 "15 "105 "6
L 18 70 2.0 9.96 "27 "93 "15
Cerebellum R 77 2.7 9.69 9 "66 "36
Results are reported on a statistical level of p ! 0.05, FWE-corrected; only peak activations of clusters are listed. L, Left hemisphere; R, right hemisphere; M, mean; BA, Brodmann area.

Table 2. Brain regions encoding the subsequent consumer choices among passively presented, task-irrelevant products in group 2 (low attention)
Accuracy MNI coordinates
Brain region Side BA M SE t value x y z
Frontal lobe
Middle frontal gyrus L 8 79 2.8 10.45 "24 36 42
L 8 76 1.8 10.70 "21 36 51
L 8 72 1.9 11.35 "39 33 42
mPFC L 10 77 2.2 12.38 "15 57 24
L 10 76 2.0 12.83 "6 51 18
mOFC L 11 74 2.0 12.20 "12 42 "18
Inferior frontal gyrus L 45 73 2.1 10.90 "51 24 15
Limbic lobe
Insula R 82 2.6 5.19 39 "3 15
L 72 1.8 12.46 "42 15 "6
Parietal lobe
Inferior parietal lobe L 40 82 2.7 11.81 "57 "36 27
R 40 73 2.1 10.66 60 "39 30
Temporal lobe
Superior temporal gyrus L 38 74 1.8 13.31 "51 12 "18
R 42 75 1.8 13.40 54 "33 15
R 42 70 1.5 13.62 63 "36 18
Results are reported on a statistical level of p ! 0.05, FWE-corrected; only peak activations of clusters are listed. L, Left hemisphere; R, right hemisphere; M, mean; BA, Brodmann area; OFC, orbitofrontal cortex.

sis was conducted for data of both groups. When test datasets
were randomly assigned to either the purchase or no purchase
condition during the testing phase of the classifier, no statistical
significant prediction of consumer choices could be achieved
(FDR and FWE-corrected). This was true for data from both
groups independently. Moreover, decoding accuracies in brain
regions that were informative in the original analysis were at
chance level (50%) when the test data were randomly allocated to
the conditions (see supplemental Fig. 3, available at www.
jneurosci.org as supplemental material). This finding speaks
against potential methodological concerns such as possible biases
inherent in the testing procedure and insufficient corrections for
multiple comparisons. Finally, we investigated whether a combi-
nation of two informative classifiers [i.e., medial prefrontal cor-
tex (mPFC) and insula] could improve the overall prediction
accuracy. Compared with decoding results of single searchlights
in these areas, the weighted classification using decision values of
the first-level decoding enhanced the prediction by 7% and by 5%
respectively.
Neural activation in the ventral striatum has frequently been
implicated in financial decision-making, preference-related pro-
cessing of products as well as purchases (Erk et al., 2002; Kuhnen
and Knutson, 2005; Knutson and Bossaerts, 2007; Knutson et al.,
2007; Schaefer and Rotte, 2007a). To investigate whether spatial
activation patterns in these regions would be found to contain
stable information about product choices when smaller searchlights
(radius of 2 voxels) were used, an additional decoding analysis was
performed. At a more liberal statistical threshold of p ! 0.00001
(uncorrected), this analysis revealed predictive information in the
striatum (see supplemental Fig. 4, available at www.jneurosci.org as
supplemental material).
Univariate comparisons of subsequent consumer choices
In both groups, classic univariate comparisons did not reveal any
activation differences between products that participants were
willing to purchase and those they were not. This strongly sug-
gests that multivariate pattern classification is capable of extract-
ing information which conventional analyses fail to detect.
Multivariate decoding of attractiveness judgments
Participants in group 1 were instructed to judge the subjective
attractiveness of a particular car after each product presentation.
Spatial activation patterns in the right middle frontal gyrus (47%
decoding accuracy, [MNI 30, 12, 33]), medial frontal gyrus (43%
decoding accuracy, [MNI 15, 33, 45]) and left orbitofrontal cor-
tex (40% decoding accuracy, [MNI "27, 33, 18]) were found to
encode attractiveness judgments during product presentation.
Activity in the left (51% decoding accuracy, [MNI "15, 24, 30])
and right dorsal anterior cingulate cortex (49% decoding accu-
racy, [MNI 15, 24, 30]), left (41% decoding accuracy, [MNI "18,
Tusche et al. • Attention-Independent Prediction of Consumer Choices
Figure 2. Brain regions encoding subsequent consumer choices in both groups. Multivariate searchlight decoding (radius of 4
voxels) was applied to functional brain responses obtained during product exposure to predict subsequent consumer choices
(chance level 50%). Spatial activation patterns in the mPFC and the insula were found to encode these choices when participants
did not explicitly deliberate on purchases ( p " 0.05, FWE-corrected). Importantly, this applied to situations when participants
closely attended to and actively evaluated products (“high attention” group 1, blue) as well as when products were passively
presented outside the focus of attention (“low attention” group 2, red). The amount of predictive information in the brain
responses was found to be comparably high in both groups. The graph displays mean decoding accuracies and SEs across partici-
pants for both regions and both groups. For illustrative purposes, the contrasts are shown at p " 0.0001 (uncorrected) with a
cluster threshold of 10 voxels (L indicates left hemisphere).
!45, 12]) and right posterior cingulate cortex (42% decoding
accuracy, [MNI 9, !42, 12]) and left insula (40% decoding accu-
racy, [MNI !42, 9, !3]) were also found to be predictive of
evaluation scores ( p " 0.05, FWE-corrected, chance level 25%
for attractiveness ratings 1– 4). For a complete list of results, see
supplemental Table 3, available at www.jneurosci.org as supple-
mental material. Importantly, no direct overlap of regions pre-
dicting consumer choices and those predicting attractiveness
scores could be observed (see supplemental Fig. 5, available at
www.jneurosci.org as supplemental material). This finding
strongly supports the notion that the information encoding con-
sumer choices is not merely based on evaluative judgments of
attractiveness.
Multivariate decoding of button presses
To ensure that the neural activity during the product viewing
phase was not confounded by the preparation of the motor
response, we applied a randomized response-button mapping.
This mapping scheme was only presented after the removal of the
product to isolate the motor response from the processing of
product information. As expected, we found no brain region that
was informative about the subsequent motor response during the
product-viewing phase. This finding strongly suggests that the
prediction of subsequent consumer choices was not based on
motor preparation during product exposure. During the presen-
tation of the response-mapping scheme (subsequent to the prod-
uct presentation), the bilateral motor cortex ([MNI 48, !21, 57]
and [MNI !42, !21, 54]) predicted the current button-press
( p " 0.05, FWE-corrected, chance level 25% for buttons 1– 4).
Importantly, no neural structure predictive of subsequent con-
sumer choices contained information about the associated motor
responses (see supplemental Fig. 6, available at www.jneurosci.
org as supplemental material). This finding indicates that the
prediction of consumer choices was not related to neural pro-
cesses involved in the execution of motor responses.
Attention modulation in the visual cortex
Engagement of attention in one region of the visual field (corre-
sponding to the square of the fixation task) has been suggested to
J. Neurosci., June 9, 2010 • 30(23):8024 – 8031 • 8029
decrease the visual processing of task-
irrelevant background stimuli (cars). To
test this assumption, we compared the
BOLD signal change in the visual cortex
during product exposure in the high and
low attention conditions. Consistent with
our hypothesis, we found a significant dif-
ference in the BOLD signal change evoked
by the presentation of cars in the high at-
tention and low attention group (t test for
independent samples, t # 5.15, p "
0.001). More precisely, the results demon-
strated that the responses in the visual
cortex evoked by the task-irrelevant im-
ages of cars which were presented outside
the focus of attention were reduced com-
pared with when the images are actively
evaluated and attended to. This finding
indicates that the engagement of attention
in one region of the visual field (corre-
sponding to the square of the fixation
task) decreased the processing of task-
irrelevant products in the background.
Discussion
The present study investigated the impact
of product-related attention on neural predictors of consumer
choices. Activation patterns in the insula and the mPFC were
found to predict these choices under high and low attention pro-
cessing (Fig. 2). Thus, a close match of predictive brain regions
was revealed independent of spatial attention to products. This
demonstrates that processing of unattended stimuli can proceed
beyond object processing (Peelen et al., 2009) to a stage that even
allows the prediction of consumer choices. Importantly, the
amount of predictive information in these areas was comparably
as high when task-irrelevant products were presented outside the
focus of attention as when they were actively evaluated and at-
tended to. Congruent with the notion of unconscious environ-
mental triggers for automatic processes in consumer settings
(Chartrand, 2005), this indicates that a prediction of economic
choices from brain responses does not necessarily depend on
attentive processing of products.
Notably, these findings were achieved with participants who
were naive about the necessity of a subsequent choice. This is in
line with behavioral findings reporting that consumer-related
goals can automatically be activated, guiding subsequent con-
sumer behavior and choices even outside of conscious awareness
(Bargh, 2002; Chartrand, 2005; Dijksterhuis et al., 2005). More-
over, it is consistent with previous data showing that brain re-
sponses reflect preference choices when participants evaluate
stimuli with respect to other, non-preference-related aspects
(Kim et al., 2007; Lebreton et al., 2009). The current study goes
beyond these results by demonstrating that brain responses pre-
dict subsequent choices even in the absence of spatial attention to
choice options.
Neural activation in the insula and the mPFC has been shown
to predict consumer choices when participants closely attend to
products and explicitly deliberate about purchases (Knutson et
al., 2007). Given that the prediction of consumer choices in the
present study was achieved in the absence of explicit deliberation
and without a priori assumptions about informative regions, this
further supports the role of both areas in economic decision-
making. Spatial activation patterns in the insula and the mPFC
8030 • J. Neurosci., June 9, 2010 • 30(23):8024 – 8031
were also reported to predict behavioral choices in the following
trial of a reward-based decision-making task (Hampton and
O’Doherty, 2007). In line with this finding, both structures have
been suggested to underlie the neural representation of the ex-
pected reward value (Knutson et al., 2005; Preuschoff et al., 2006;
Hampton et al., 2007; Hare et al., 2008; Rolls et al., 2008), a
concept central to microeconomic and psychological models of
decision-making (Knutson and Bossaerts, 2007; Loewenstein et
al., 2008; Rangel et al., 2008). Cultural stimuli such as cars and
logos of car manufacturers can signal potential social dominance
or wealth and have been shown to modulate activity in the reward
circuit (Erk et al., 2002; Schaefer and Rotte, 2007a). It can be
assumed that products in the present experiment differed in their
expected reward value depending on whether they were subse-
quently chosen to be purchased or not (Sharot et al., 2009). We
suggest that distinguishable activation patterns in the insula and
the mPFC reflected these different reward values and— even
without full attention— contributed to subsequent hypothetical
consumer choices.
Activation in the mPFC—particularly the ventral part— has
also been reported to reflect product-related preference and at-
tractiveness judgments (Paulus et al., 2003; McClure et al., 2004;
Deppe et al., 2005; Plassmann et al., 2008; Lebreton et al., 2009;
Luu and Chau, 2009). The dorsal mPFC, on the other hand, has
been proposed to be involved in the processing of brand knowl-
edge (Schaefer et al., 2006; Schaefer and Rotte, 2007b), the impact
of which on consumer preferences and choices is well known
(McClure et al., 2004; Lee et al., 2006). Considering this evidence,
it is likely that the predictive information encoded in the mPFC
might have been influenced by subjective valuation and the brand
and price information.
It can be assumed that global product valuation was a major
source for subsequent choices. This is particularly likely, given
that no further product-related information was provided during
the experiment to avoid drawing participants’ attention to the
potential choice options. Therefore, it could be speculated that
the predictive information merely reflects global attractiveness or
desirability of products. However, many cognitive factors as well
as their interactions with automatic valuation processes might
contribute to a complex choice such as those for a new car. This is
particularly likely for participants who stated an interest in cars
and are assumed to possess relevant knowledge on the items.
Consistent with this notion product choices were not entirely
determined by attractiveness judgments. Thus, only partial cor-
relations between product-specific attractiveness ratings and
consumer choices were found for both groups in the present
experiment (see supplemental Table 2, available at www.
jneurosci.org as supplemental material). Additionally, no direct
overlap of brain regions predicting attractiveness judgments and
those being informative about consumer choices could be iden-
tified (see supplemental Table 3 and supplemental Fig. 5, avail-
able at www.jneurosci.org as supplemental material). Together,
these findings indicate that the prediction of product choices was
not mainly due to global evaluations of attractiveness but might
reflect automatic choice-related processing itself. This is in line
with previous results showing that brain activation reflecting
subsequent preference decisions were not merely responding to
attractiveness of stimuli (Lebreton et al., 2009). However, more
research is needed to specifically examine choice-relevant dimen-
sions with the current paradigms, possibly testing explicit models of
multiattribute decision-making (Dijksterhuis et al., 2006; Lassiter et
al., 2009). Given that consumer choices for each product were found
to vary from participant to participant, it is also unlikely that the
Tusche et al. • Attention-Independent Prediction of Consumer Choices
predictive information was based on physical properties of single
cars (see supplemental Table 1, available at www.jneurosci.org as
supplemental material). Results of an additional decoding anal-
ysis also demonstrate that the prediction of consumer choices was
not related to neural processes involved in the preparation of
motor responses (see supplemental Fig. 6, available at www.
jneurosci.org as supplemental material).
Another point that needs to be addressed concerns the ques-
tion of whether attention was effectively attenuated by engaging
in the distraction task in group 2. Evidence of successful distrac-
tion comes from differences in memory performance, obtained
for both groups after the experiment. Recognition rates of cars
were strongly decreased in participants who performed the dis-
traction task compared with participants who actively evaluated
products. Moreover, neural responses in the visual cortex evoked
by task-irrelevant, unattended images of cars were found to be
reduced compared with images that were actively evaluated and
fully attended to. This is consistent with previous research dem-
onstrating that the engagement of attention in one region of the
visual field strongly decreases the processing of irrelevant back-
ground stimuli (Rees et al., 1997). It can therefore be considered
that the attention to products in group 2 was strongly diminished
compared with group 1. As in most attention studies, there is a
possibility of weak but residual attention to unattended target
stimuli. However, our key finding is that a strong reduction in
attention does not affect the choice-predictive information.
Purchase-related information encoded in the brain was not re-
duced, but was comparably high in low and high attention con-
ditions. Furthermore, we found a close match of predictive brain
regions during high and low attentional processing of products.
Together, this indicates that choice-related processing does not
necessarily require close attention to products and can even occur
in cases where sensory signal processing is attenuated due to a
removal of attention. Although it remains possible that the choice
was not explicitly calculated during product presentation, the
high predictive accuracy reveals that choice-related processes
have already reached a high level of completion.
The present study implemented hypothetical choices, which
are commonly applied in marketing research. Nevertheless, it
needs to be explored whether the same activation patterns are
predictive for actual purchases— beyond the stated willingness to
buy. This would allow checking for potential biases such as the
tendency to overstate the willingness to pay or vote in case of
hypothetical choices (List and Gallet, 2001; Murphy et al., 2005).
Requiring expenses would likely involve losses and more relative
comparisons between available products, including changes in
the reference point for product valuation (FitzGerald et al., 2009;
Sharot et al., 2009). Real purchases might also engage processes of
perceived justification, anticipated regret, time pressure or self
esteem (Plassmann et al., 2007). Finally, it might be that making
actual purchases differs from our experimental setting in terms of
the strategies used for information acquisition and information
integration across multiple relevant dimensions. However, it
should be noted that informative brain regions as identified with
the present approach are strikingly consistent with previous find-
ings that implemented actual purchases (Knutson et al., 2007).
Further research might also address the generalizability of our find-
ings to other product categories, including different goods (e.g., cars
vs coffee) and different types of purchases (e.g., routine vs new), as
well as to people who are not interested in the product group.
In summary, we found a close match of brain regions predicting
consumer choices for both high and low attentional processing of
products. Importantly, the amount of predictive information was
Tusche et al. • Attention-Independent Prediction of Consumer Choices
found to remain persistently high when task-irrelevant products
were presented outside the focus of attention. Altogether, these find-
ings support the notion that even complex and important economic
choices can be prepared automatically, in the absence of explicit
deliberation and without attention to products.
References
Bargh JA (2002) Losing consciousness: automatic influences on consumer
judgment, behavior, and motivation. J Consum Res 29:280 –285.
Chartrand TL (2005) The role of conscious awareness in consumer behav-
ior. J Consum Psychol 15:203–210.
Cox DD, Savoy RL (2003) Functional magnetic resonance imaging (fMRI)
“brain reading”: detecting and classifying distributed patterns of fMRI
activity in human visual cortex. Neuroimage 19:261–270.
Deppe M, Schwindt W, Kugel H, Plassmann H, Kenning P (2005) Nonlin-
ear responses within the medial prefrontal cortex reveal when specific
implicit information influences economic decision making. J Neuroim-
aging 15:171–182.
Dijksterhuis A, Smith PK, van Baaren RB, Wigboldus DHJ (2005) The
unconscious consumer: effects of environment on consumer behav-
ior. J Consum Psychol 15:193–202.
Dijksterhuis A, Bos MW, Nordgren LF, van Baaren RB (2006) On making
the right choice: the deliberation-without-attention effect. Science 311:
1005–1007.
Erk S, Spitzer M, Wunderlich AP, Galley L, Walter H (2002) Cultural ob-
jects modulate reward circuitry. Neuroreport 13:2499 –2503.
FitzGerald THB, Seymour B, Dolan RJ (2009) The role of human orbito-
frontal cortex in value comparison for incommensurable objects. J Neu-
rosci 29:8388 – 8395.
Friston KJ, Holmes AP, Worsley KJ, Poline JP, Frith CD, Frackowiak RSJ
(1994) Statistical parametric maps in functional imaging: a general linear
approach. Hum Brain Mapp 2:189 –210.
Hampton AN, O’Doherty JP (2007) Decoding the neural substrates of
reward-related decision making with functional MRI. Proc Natl Acad Sci
U S A 104:1377–1382.
Hampton AN, Adolphs R, Tyszka MJ, O’Doherty JP (2007) Contributions
of the amygdala to reward expectancy and choice signals in human pre-
frontal cortex. Neuron 55:545–555.
Hare TA, O’Doherty J, Camerer CF, Schultz W, Rangel A (2008) Dissociat-
ing the role of the orbitofrontal cortex and the striatum in the computa-
tion of goal values and prediction errors. J Neurosci 28:5623–5630.
Haxby JV, Gobbini MI, Furey ML, Ishai A, Schouten JL, Pietrini P (2001)
Distributed and overlapping representations of faces and objects in ven-
tral temporal cortex. Science 293:2425–2430.
Haynes JD, Rees G (2006) Decoding mental states from brain activity in
humans. Nat Rev Neurosci 7:523–534.
Haynes JD, Sakai K, Rees G, Gilbert S, Frith C, Passingham RE (2007) Read-
ing hidden intentions in the human brain. Curr Biol 17:323–328.
Kamitani Y, Tong F (2005) Decoding the visual and subjective contents of
the human brain. Nat Neurosci 8:679 – 685.
Kastner S, Ungerleider LG (2000) Mechanisms of visual attention in the
human cortex. Annu Rev Neurosci 23:315–341.
Kim H, Adolphs R, O’Doherty JP, Shimojo S (2007) Temporal isolation of
neural processes underlying face preference decisions. Proc Natl Acad Sci
U S A 104:18253–18258.
Knutson B, Bossaerts P (2007) Neural antecedents of financial decisions.
J Neurosci 27:8174 – 8177.
Knutson B, Taylor J, Kaufman M, Peterson R, Glover G (2005) Distributed
neural representation of expected value. J Neurosci 25:4806 – 4812.
Knutson B, Rick S, Wimmer GE, Prelec D, Loewenstein G (2007) Neural
predictors of purchases. Neuron 53:147–156.
Kriegeskorte N, Goebel R, Bandettini P (2006) Information-based func-
tional brain mapping. Proc Natl Acad Sci U S A 103:3863–3868.
Kriegeskorte N, Simmons WK, Bellgowan PSF, Baker CI (2009) Circular
analysis in systems neuroscience: the dangers of double dipping. Nat Neu-
rosci 12:535–540.
Kuhnen CM, Knutson B (2005) The neural basis of financial risk taking.
Neuron 47:763–770.
Lassiter GD, Lindberg MJ, González-Vallejo C, Bellezza FS, Phillips ND
J. Neurosci., June 9, 2010 • 30(23):8024 – 8031 • 8031
(2009) The deliberation-without-attention effect: evidence for an arti-
factual interpretation. Psychol Sci 20:671– 675.
Lebreton M, Jorge S, Michel V, Thirion B, Pessiglione M (2009) An auto-
matic valuation system in the human brain: evidence from functional
neuroimaging. Neuron 64:431– 439.
Lee L, Frederick S, Ariely D (2006) Try it, you’ll like it: the influence of
expectation, consumption, and revelation on preferences for beer. Psy-
chol Sci 17:1054 –1058.
List JA, Gallet CA (2001) What experimental protocol influence disparities
between actual and hypothetical stated values. Environ Res Econ 20:241–
254.
Loewenstein G, Rick S, Cohen JD (2008) Neuroeconomics. Annu Rev Psy-
chol 59:647– 672.
Luu S, Chau T (2009) Decoding subjective preference from single-trial
near-infrared spectroscopy signals. J Neural Eng 6:016003.
Martínez A, Anllo-Vento L, Sereno MI, Frank LR, Buxton RB, Dubowitz DJ,
Wong EC, Hinrichs H, Heinze HJ, Hillyard SA (1999) Involvement of
striate and extrastriate visual cortical areas in spatial attention. Nat Neu-
rosci 2:364 –369.
McClure SM, Li J, Tomlin D, Cypert KS, Montague LM, Montague PR
(2004) Neural correlates of behavioral preference for culturally familiar
drinks. Neuron 44:379 –387.
Mitchell TM, Hutchinson R, Niculescu RS, Pereira F, Wang X, Just M, Newman
S (2004) Learning to decode cognitive states from brain images. Ma-
chine Learn 57:145–175.
Murphy J, Allen PG, Stevens TH, Weatherhead D (2005) A meta-analysis of
hypothetical bias in stated preference valuation. Environ Res Econ 30:
313–325.
Norman KA, Polyn SM, Detre GJ, Haxby JV (2006) Beyond mind-reading:
multi-voxel pattern analysis of fMRI data. Trends Cogn Sci 10:424 – 430.
O’Doherty J, Winston J, Critchley H, Perrett D, Burt DM, Dolan RJ (2003)
Beauty in a smile: the role of medial orbitofrontal cortex in facial attrac-
tiveness. Neuropsychologia 41:147–155.
Paulus MP, Rogalsky C, Simmons A, Feinstein JS, Stein MB (2003) In-
creased activation in the right insula during risk-taking decision making is
related to harm avoidance and neuroticism. Neuroimage 19:1439 –1448.
Peelen MV, Li FF, Kastner S (2009) Neural mechanisms of rapid natural
scene categorization in human visual cortex. Nature 460:94 –97.
Plassmann H, Ambler T, Brautigam S, Kenning P (2007) What can adver-
tisers learn from neuroscience? Int J Advertising 26:151–175.
Plassmann H, O’Doherty J, Shiv B, Rangel A (2008) Marketing actions can
modulate neural representations of experienced pleasantness. Proc Natl
Acad Sci U S A 105:1050 –1054.
Preuschoff K, Bossaerts P, Quartz SR (2006) Neural differentiation of
expected reward and risk in human subcortical structures. Neuron
51:381–390.
Rangel A, Camerer C, Montague PR (2008) A framework for studying
the neurobiology of value-based decision making. Nat Rev Neurosci
9:545–556.
Rees G, Frith CD, Lavie N (1997) Modulating irrelevant motion perception
by varying attentional load in an unrelated task. Science 278:1616 –1619.
Rolls ET, McCabe C, Redoute J (2008) Expected value, reward outcome,
and temporal difference error representations in a probabilistic decision
task. Cereb Cortex 18:652– 663.
Schaefer M, Rotte M (2007a) Favorite brands as cultural objects modulate
reward circuit. Neuroreport 18:141–145.
Schaefer M, Rotte M (2007b) Thinking on luxury or pragmatic brand prod-
ucts: brain responses to different categories of culturally based brands.
Brain Res 1165:98 –104.
Schaefer M, Berens H, Heinze HJ, Rotte M (2006) Neural correlates of cul-
turally familiar brands of car manufacturers. Neuroimage 31:861– 865.
Sharot T, De Martino B, Dolan RJ (2009) How choice reveals and shapes
expected hedonic outcome. J Neurosci 29:3760 –3765.
Soon CS, Brass M, Heinze HJ, Haynes JD (2008) Unconscious determinants
of free decisions in the human brain. Nat Neurosci 11:543–545.
Vul E, Harris C, Winkielman P, Pashler H (2009) Puzzlingly high correla-
tions in fMRI studies of emotion, personality, and social cognition. Per-
spect Psychol Sci 4:274 –290.
Neuron
Report
Subliminal Instrumental Conditioning
Demonstrated in the Human Brain
Mathias Pessiglione,1,2,* Predrag Petrovic,1 Jean Daunizeau,1 Stefano Palminteri,2 Raymond J. Dolan,1 and Chris D. Frith1
1Wellcome Trust Centre for NeuroImaging, Institute of Neurology, University College London, 12 Queen Square, London WC1N 3BG, UK
2Laboratoire INSERM U610, Centre de NeuroImagerie de Recherche (CENIR), Institut Fédératif de Recherche en Neurosciences, Hôpital
Pitié-Salpêtrière, Université Pierre et Marie Curie (Paris 6), 47 Boulevard de l’Hôpital 75013 Paris, France
*Correspondence: mathias.pessiglione@gmail.com
DOI 10.1016/j.neuron.2008.07.005
SUMMARY
How the brain uses success and failure to optimize
future decisions is a long-standing question in neuro-
science. One computational solution involves updat-
ing the values of context-action associations in
proportion to a reward prediction error. Previous
evidence suggests that such computations are ex-
pressed in the striatum and, as they are cognitively
impenetrable, represent an unconscious learning
mechanism. Here, we formally test this by studying
instrumental conditioning in a situation where we
masked contextual cues, such that they were not
consciously perceived. Behavioral data showed
that subjects nonetheless developed a significant
propensity to choose cues associated with monetary
rewards relative to punishments. Functional neuro-
imaging revealed that during conditioning cue values
and prediction errors, generated from a computa-
tional model, both correlated with activity in ventral
striatum. We conclude that, even without conscious
processing of contextual cues, our brain can learn
their reward value and use them to provide a bias
on decision making.
INTRODUCTION
Humans frequently invoke an argument that their intuition can re-
sult in a better decision than conscious reasoning. Such asser-
tions may rely on subconscious associative learning between
subliminal signals present in a given situation and choice out-
comes. For instance, clinicians may improve their therapeutic
decisions through learned associations between treatment
outcomes and subliminal signs presented by their patients.
Likewise, poker players can improve their gambles through
a learned association between monetary outcomes and sublim-
inal behavioral manifestations of their opponents (the so-called
‘‘gamblers’ tell’’).
The idea that such instrumental conditioning can occur sub-
consciously has been around for almost a century (Thorndike,
1911). This assumption originally rested on observations that re-
wards and punishments shape behavioral responses in species
allegedly lacking conscious awareness. However, subliminal
conditioning studies in humans have so far been restricted to
Pavlovian paradigms such as fear conditioning (Clark and
Squire, 1998; Knight et al., 2003; Morris et al., 1998; Olsson
and Phelps, 2004), where subliminal stimuli (like unseen faces)
are paired with unpleasant events (like white noise) to increase
automatic responses (like skin conductance). To our knowledge,
subliminal instrumental conditioning, where decision making
would be biased by unperceived cues predicting rewards or
punishments, has never been previously demonstrated.
Our subliminal conditioning task was adapted from a published
supraliminal task, wherein subjects selected between visual
cues so as to learn choices that maximized monetary outcomes
(Pessiglione et al., 2006). In our previous study, we modeled sub-
jects’ behavior by optimizing the free parameters of a standard
machine learning algorithm (termed Q learning), to get maximal
likelihoods for the observed decisions. When we regressed key
output variables of the optimized model against simultaneously
acquired functional neuroimaging data, we showed that predic-
tion errors were expressed in the striatum. Postexperimental
debriefing indicated that some subjects managed to understand
the statistical structure of the task, while others appeared to rely
on what they referred to as their intuition. These latter reports
suggest that subjects can improve their decisions without
consciously following the incremental steps of the Q-learning
procedure.
The motivating assumption of the current experiment was that
processes associated with striatal learning are not consciously
accessible but, nonetheless, influence choice decision making.
Indeed, if contextual cues reach awareness, other brain systems
are likely to play a role, as expressed in conscious reasoning or
strategic control, which allows one to develop explicit knowl-
edge of statistical contingencies. However, if the cues remain
unseen, learning would solely depend on a subconscious pro-
cessing that involves the striatum, with an algorithmic structure
similar to a Q learning, which does not embody explicit informa-
tion about statistical contingencies. Under these assumptions,
we predicted that, if in our task visual cues were masked, both
striatal activity and behavioral choices would still reflect Q-learn-
ing outputs.
RESULTS
A prerequisite for the present study was to demonstrate efficient
masking of the visual cues. These cues were novel abstract sym-
bols, which were scrambled and mixed to create mask images.
Neuron 59, 561–567, August 28, 2008 ª2008 Elsevier Inc. 561

To assess visual awareness, we successively displayed two
masked cues on a computer screen and asked subjects whether
they perceived a difference or not. We reasoned that if subjects
are unable to correctly perceive any difference between the
masked cues, then they are also unable to build conscious rep-
resentations of cue-outcome associations. The procedure has
the advantage of not showing the cues unmasked, so that, by
the end of the experiment, subjects had no idea what the cues
look like.
The perceptual discrimination task was performed outside the
scanner at the beginning of the experiment, in order to adapt du-
ration of cue display to each individual, and in the scanner at the
end of the experiment, to check for any effect of learning or
change in visual conditions. For all subjects, cue duration was
set at either 33 or 50 ms and was kept fixed through the entire
experiment. In every individual, correct guessing on the final as-
sessment did not differ from chance (p > 0.05, chi-square test).
At group level, average percentage of correct responses for
the 20 subjects was 48% ± 3%, which again was not different
from chance (p > 0.5, two-tailed paired t test). Average d0 was
0.08 ± 0.20, showing that, even when correcting for response
bias, signal detection was not different from zero (p > 0.5, two-
tailed paired t test). Thus, subjects remained unable to discrim-
inate between the different masked cues, from the beginning
to the end of conditioning sessions.
We employed the same masking procedure in the subliminal
conditioning task, in which cues were paired with monetary out-
comes (Figure 1). From these outcomes (!£1, £0, +£1), subjects
had to learn when to take the risky response (either ‘‘Go’’ or
‘‘NoGo,’’ depending on subjects). Subjects were also told that,
for the risky response, the outcome would depend on the cue
hidden behind the masking image (see instructions in Supple-
562 Neuron 59, 561–567, August 28, 2008 ª2008 Elsevier Inc.
Neuron
fMRI Study of Subliminal Conditioning
Figure 1. Subliminal Conditioning Task
Successive screenshots displayed during a given
trial are shown from left to right, with durations in
milliseconds. After seeing a masked contextual
cue flashed on a computer screen, subjects
choose to press or not press a response button
and subsequently observe the outcome. In this
example, ‘‘Go’’ appears on the screen because
the subject has pressed the button, following
the cue associated with a rewarding outcome
(winning £1).
mental Data available online). As they would not see the cues,
we encouraged them to follow their intuition, taking the risky
response if they had a feeling they were in a winning trial and
choosing a safe response if they felt it was a losing trial. Note
that if subjects always made the same response, or if they
performed at chance, their final payoff would be zero.
As a dependent variable to assess for conditioning effects, we
used monetary payoff, which corresponds to the area below the
reward and above the punishment learning curves (Figure 2A).
Overall subjects won money in this task, on average £7.5 ± £1.8
(p < 0.001, one-tailed paired t test), indicating that the risky re-
sponse was more frequently chosen following reward predictive
relative to punishment predictive cues. Both reward and punish-
ment conditions also differed significantly from the neutral condi-
tion (p < 0.05, one-tailed paired t test). There was no significant
difference (p > 0.5, two-tailed paired t test) between the reward
and punishment condition: on average subjects won £24.3 ±
£1.9 and avoid losing £23.2 ± £2.1. Learning curves showed
that responses improved symmetrically for rewards and punish-
ments, ending with 63% ± 5% of correct responses on average.
Surprisingly, this plateau was reached at around the halfway
point of the learning session. The effects of subliminal condition-
ing were subsequently assessed with a preference judgment
task, in which cues were uncovered and rated by the subjects
from the most to least liked (Figure 2B). Ratings were significantly
higher for reward compared to punishment cues (p < 0.01, one-
tailed paired t test), consistent with subjects having learned the
affective values of subliminal cues, such that these values were
able to bias their preferences. When uncovering the cues, sub-
jects were also asked to signal any cue that they may have
seen during conditioning sessions; none was reported as previ-
ously seen.
Figure 2. Behavioral Data
(A) Learning curves. Colors indicate cues for which
button presses are rewarded (green), neutral
(blue), or punished (red). Diamonds represent,
across trials, percentages of subjects that pressed
the button. Left: continuous lines join the dia-
monds to illustrate actual choices made by
subjects. Right: continuous lines represent the
probabilities of button press estimated by an
optimized Q-learning model.
(B) Preferences. After the conditioning phase,
cues were unmasked and subjects rated them,
from the most (3) to the least liked (1). The graph
shows the average rating for reward (green), neu-
tral (blue), and punishment (red) cues. Bars are ±
intersubjects standard errors of the mean.
Neuron
fMRI Study of Subliminal Conditioning
Figure 3. Neuroimaging Data
Left: ventral striatal regions isolated by regression of monetary values against
BOLD responses to cue display. Right: visual cortical regions isolated by cor-
relation of cue-value regression coefficients with individual payoffs. Slices
were taken at global maxima of interest indicated by red pointers on the above
axial glass brains. Areas shown in gray/black on glass brains and in orange/
yellow on coronal slices showed significant effect. The [x y z] coordinates of
the maxima refer to the Montreal Neurological Institute space.
(A) Statistical parametric maps using conservative threshold (p < 0.05 after
familywise error correction for multiple comparisons).
(B) Statistical parametric maps using liberal threshold (p < 0.001 uncorrected).
(C) Regression coefficients of Q values (QV) and prediction errors (PE) against
BOLD responses to cue and outcome display, respectively. Bars are ± inter-
subjects standard errors of the mean.
To model instrumental conditioning, we implemented a stan-
dard Q-learning algorithm (Pessiglione et al., 2006), with inputs
from individual histories of cues, choices, and outcomes. On
every trial, the model estimates the likelihood of the risky re-
sponse from the value of the displayed cue. If the risky response
was actually taken, the model then updates the value of the dis-
played cue in proportion to the prediction error. The parameters
of the model were optimized such that likelihoods of risky
responses provided the best fit of subjects’ actual responses
across conditioning sessions (Figure 2A). The Q values and pre-
diction errors generated by this optimized algorithm were then
used as regressors for analysis of brain imaging data (see
Figure S1).
We recorded brain activity while subjects performed the sub-
liminal conditioning task, using functional magnetic resonance
imaging (fMRI). We first examined brain regions reflecting Q
value at the time of cue onset, increasing their response to re-
ward-predicting cues and decreasing their response to punish-
ment-predicting cues, across learning sessions. After correction
for multiple comparisons (family-wise error, p < 0.05), we noted
significant correlated activity in ventral striatum bilaterally (Fig-
ures 3A and 3B, left). The same region was also significantly ac-
tivated at the time of outcome in keeping with prediction errors
being expressed at this time point (Figure 3C, left). In a second
analysis, we computed regression coefficients for the different
conditions at the time of cue and outcome onsets, separately
for the first and second half of conditioning sessions. Contrasts
with the neutral condition were then averaged over all ventral
striatal voxels showing significant activation at the most conser-
vative threshold in the first analysis. This confirmed that from the
first to the second half of conditioning sessions, ventral striatal
responses increased for reward cues and decreased for punish-
ment cues (Figure 4A, left). At the time of outcome onset, the
same ventral striatal region reflected positive prediction errors
in the reward condition and negative prediction errors in the pun-
ishment condition. In keeping with the Q-learning model, both
positive and negative prediction errors decreased from the first
to the second half of conditioning sessions. Thus, across sublim-
inal conditioning, the ventral striatal response was consistent
with the expression of Q values (for unseen cues) and prediction
errors (based on visible outcomes).
We further examined variability in individual performance to
explain why some subjects won more money than others.
More precisely, we searched for brain regions where coefficients
of Q-value regressors correlated with individual payoffs. These
regions were confined to extrastriate visual cortex (Figure 3A,
right) at the most conservative threshold (familywise error, p <
0.05), spreading into the ventral occipitotemporal stream
(Figure 3B, right) with a more liberal threshold (uncorrected, p <
0.001). Contrast estimates confirmed that extrastriate voxels
progressively differentiated the reward and punishment cues
from the first to the second half of conditioning sessions
(Figure 4A, right). At the time of outcome onset, these extrastriate
regions responded positively for both rewards and punishments,
showing no evidence for encoding of prediction errors. Thus,
during the subliminal conditioning task, the extrastriate visual
cortex learned to discriminate between unseen cues according
to their reward value but did not express outcome-related
prediction errors (Figure 3C, right).
To further assess whether the ventral striatum and visual
cortex were able to discriminate between the subliminal cues,
Neuron 59, 561–567, August 28, 2008 ª2008 Elsevier Inc. 563

Figure 4. Model-free Analyses of Brain Activations
Ventral striatum (left) and visual cortex (right) correspond to voxels surviving
familywise error correction on statistical parametric maps.
(A) Regression coefficients. Histograms represent contrasts of the reward
(green) or the punishment (red) condition with the neutral condition, at the
times of cue and outcome display. For every contrast the two joint histograms
correspond to the first (empty) and second (filled) halves of the conditioning
sessions.
(B) Time courses. BOLD responses were averaged across trials over condi-
tioning sessions, for the reward (green) and punishment (red) conditions,
relative to the neutral condition. Bars are ± intersubjects standard errors of
the mean.
we extracted time courses of BOLD response. These time
courses were averaged over trials, sessions, and subjects,
separately for the reward and punishment conditions (Figure 4B).
We found that BOLD responses to reward and punishment cues
significantly differed after two acquisition volumes (3.9 s) in the
ventral striatum (one-tailed paired t test, p < 0.01) and after three
(5.85 s) in the visual cortex (one-tailed paired t test, p < 0.01).
Finally, we ascertained whether neuroimaging and behavioral
effects of subliminal conditioning were driven by subjects scor-
ing at the high end in perceptual discrimination performance.
We tested for correlations between correct guessing assessed
in the final perceptual discrimination test and coefficients of Q-
value regressors in both the ventral and extrastriate cortex.
None was significant; Pearson’s correlation coefficients were
respectively !0.25 and !0.18. We also tested correlation of cor-
rect guessing with monetary payoffs from conditioning sessions
and differential ratings in the preference judgment task. Again,
564 Neuron 59, 561–567, August 28, 2008 ª2008 Elsevier Inc.
Neuron
fMRI Study of Subliminal Conditioning
none was significant; Pearson’s correlation coefficients were
respectively 0.26 and 0.29.
DISCUSSION
We provide evidence that instrumental learning can occur with-
out conscious processing of contextual cues. This finding might
relate to previous evidence for implicit or procedural learning,
where behavioral responses can be adapted to the statistical
structure of stimuli that fails to be reported explicitly (Bayley
et al., 2005; Berns et al., 1997; Destrebecqz and Cleeremans,
2001; Seitz and Watanabe, 2003). Interestingly, implicit/proce-
dural learning has been suggested to involve the basal ganglia,
in contrast with explicit/declarative memory which would involve
the medial temporal lobe (Cohen et al., 1985; Milner et al., 1998;
Poldrack and Packard, 2003; Squire, 1992). In implicit learning
tasks, such as serial reaction time or probabilistic classification,
authors have claimed that subjects can achieve good acquisition
without explicit knowledge of the task structure. However,
methods for assessing awareness of statistical contingencies
have been criticized, principally on the issue that questions
were too demanding in terms of memory (Lagnado et al., 2006;
Lovibond and Shanks, 2002; Wilkinson and Shanks, 2004).
Thus, to formally test whether instrumental conditioning can
occur without awareness, we took a more stringent approach:
masking the cues, so that they remained unperceived.
We believe our methodology avoids most previous problems
related to assessing awareness, by demonstrating that subjects
were not able to discriminate between masked cues (without
the help of rewards and punishments), rather than retrospectively
assessing awareness of contingencies. Moreover, postcondi-
tioning recognition tests would not be sufficient in our case, since
subjects would not need to identify cues for associative learning
to be conscious. Indeed, they could learn associations between
risky response outcomes and tiny fragments of the visual dynamic
pattern formed by the mask/cue/mask flashing. However, post-
conditioning debriefing questions might be informative in explain-
ing why subjects could not discriminate between masked cues.
Thus, when we explicitly asked subjects to state what the cues
looked like, they reported in majority of cases that they had no
idea. When the subjects were presented with the cues, now
unmasked, they reported surprise at seeing the symbols while as-
serting that they had never seen them before. This suggests that
during conditioning, subjects had no a priori representational
knowledge to guide a visual search for cues hidden behind the
masks. We believe that absence of an a priori representation is
a crucial feature of our design, which, in addition to visual mask-
ing, prevented subjects from consciously seeing the cues.
Using this methodology, we show that pairing rewards and
punishments can guide behavioral responses and even condi-
tion preferences for abstract cues that subjects could not
consciously see. Note that if cues were visible, learning curves
would have been optimized in one trial or two; hence we are
not claiming that conscious awareness is unhelpful in supralim-
inal instrumental conditioning. However, in our subliminal condi-
tioning task, conscious strategies (such as win-stay/lose-switch)
might have been detrimental, which would explain why learning
curves were limited well below the optimum.
Neuron
fMRI Study of Subliminal Conditioning
We also identified brain circuitry associated with subliminal
instrumental conditioning. The ventral striatum responded to
subliminal cues and to visible outcomes in a manner that closely
approximates Q-learning algorithm, expressing reward ex-
pected values and prediction errors, just as was reported in
supraliminal instrumental conditioning studies (O’Doherty et al.,
2004; Pagnoni et al., 2002; Pessiglione et al., 2006; Yacubian
et al., 2006). Interestingly, there is no need for representing the
statistical structure of the task in Q learning, which is an incre-
mental procedure updating the expected values of chosen
actions according to the subject’s history of reward and punish-
ment outcomes. This accords well with views that the striatum is
a major player in implicit/procedural learning (Graybiel, 2005;
Hikosaka et al., 1999; Packard and Knowlton, 2002) and with ev-
idence that ventral striatum encodes reward-related information
(Delgado, 2007; Knutson and Cooper, 2005; Pecina et al., 2006).
For the sake of simplicity, we have described ventral striatum
activity as directly reflecting key outputs of Q-learning algorithm:
Q value at the time of cue onset and prediction error at the time of
outcome. There are nonetheless other variables in machine
learning literature that would also correlate with ventral striatum
activity and which could provide an alternative interpretational
framework for our study. In particular, it is important to note
that average Q values (over the reward, neutral, and punishment
conditions) remain around zero during our conditioning para-
digm. Hence, Q value is approximately equal to Q value minus
average Q value, which can be seen as equivalent to a cue pre-
diction error (actual Q value minus predicted Q value). Our data
are therefore equally compatible with the notion that the ventral
striatum encodes prediction errors at the time of both cue and
outcome onsets. However, because prediction errors represent
a function of Q values, the brain has to learn about Q values in
order to signal prediction errors. Thus, whether we consider
the ventral striatum as encoding a Q value or a prediction error
does not alter our central conclusion: namely, the human brain
can learn the reward value of subliminal cues, so as to later
influence behavioral choices.
It is of interest that extrastriate visual cortex also reflected the
reward value of subliminal cues, but not outcome-related predic-
tion errors. Modulation of visual cortex activity by monetary in-
centives has already been reported in neuroimaging studies of
supraliminal processes, such as visuomotor transformation, at-
tentional control, and working memory (Krawczyk et al., 2007;
Ramnani and Miall, 2003; Small et al., 2005). In our case, the
modulation suggests that conditioning involves an interaction
between the extrastriate cortex (which would discriminate cues
according to their visual properties) and the ventral striatum
(which would tag cues with affective values depending on
reward prediction errors). However, we acknowledge that we
do not as yet have a complete account of how the brain pro-
duces behavioral effects of subliminal conditioning. Notably,
we failed to identify the brain regions mapping affective values
onto motor commands, which would complete the circuit from
visual cues to behavioral responses. Further experiments will
be necessary to fill in these explanatory gaps.
More generally, our approach, combining perceptual masking
and computational modeling, can be extended over the field of
functional neuroimaging. Computational reinforcement learning
theory has proven useful to model both brain activity and behav-
ioral choices in human and nonhuman primates (Daw and Doya,
2006; McClure et al., 2004; O’Doherty et al., 2007). Brain activity
reflecting sophisticated computations are unlikely to be ac-
cessed by the conscious mind, which takes minutes to solve
even simple equations. This brain activity would therefore repre-
sent unconscious processes, which we formally demonstrated
here in the case of instrumental conditioning. Combining mask-
ing and modeling can, in principle, make more tractable the
identification of basic neuronal mechanisms shared within other
species, eliminating the use of reportable knowledge that might
be unique to humans. It might also help assess the integrity of
these same basic mechanisms in patients with neurological or
psychiatric conditions, avoiding confounds generated by
conscious strategic compensations.
EXPERIMENTAL PROCEDURES
Subjects
The study was approved by the National Hospital for Neurology and Neurosur-
gery and the Institute of Neurology joint Ethics Committee. Subjects were re-
cruited via Gumtree website and screened for exclusion criteria: left handed-
ness, age below 18 or above 39, regular taking of drug or medication,
history of psychiatric or neurological illnesses and contra-indications to MRI
scanning (pregnancy, claustrophobia, metallic implants). All subjects gave
informed consent prior to taking part. We scanned 20 subjects: 11 males
(mean age 26.8 ± 6.3 years) and 9 females (mean age 23.8 ± 3.3 years). Two
more subjects were scanned but discarded from the analysis, because they
eventually could describe parts of the visual cues, were above chance level
in the perception task, and won unusually large amounts of money in the con-
ditioning task. Subjects were told that they would play for real money, but at
the end of the experiment their winnings were rounded up to a fixed amount.
Behavioral Task and Analysis
Subjects first read the instructions (see Supplemental Data) about the different
tasks, which were later explained again step by step. Before scanning, sub-
jects were trained to perform the conditioning task and the perception task
on practice versions. In the scanner, they had to perform three sessions of
the conditioning task, each containing 120 trials and lasting 13 min, and one
session of the perception task, containing 120 trials and lasting about 7 min.
The abstract cues were letters taken from the Agathodaimon font. The 12
cues shown in the scanner were randomly assigned to the four task sessions,
each session hence employing 3 new cues. The same two masking patterns
(see Figure 1), one displayed before and the other after the cue, were used
in all task sessions. The sequence of display and the cue-outcome associa-
tions were also randomized for every subject.
The perceptual discrimination task was used to select the appropriate dura-
tion for cue display, which was then kept to either 33 or 50 ms for the entire
experiment. In this task, subjects were flashed two masked cues, 3 s apart,
displayed on the center of a computer screen, each following a fixation cross.
They had to report whether or not they perceived any difference between the
two visual stimulations. The response was given manually, by pressing one of
two buttons assigned to ‘‘same’’ and ‘‘different’’ choices. The perceptual dis-
crimination task was then employed as a control for awareness at the end of
conditioning sessions. We checked with a chi-square test that in all included
subjects performance was not significantly different from chance level (50%
of correct responses). We also calculated d0 measure, which is the difference
between normalized rates of hits (correct ‘‘different’’ response) and false
alarms (incorrect ‘‘different’’ responses). We ensured that this measure was
not significantly different from zero, at group level, using one-tailed paired t
test.
The instrumental conditioning task involved choosing between pressing or
not pressing a button, in response to masked cues. After showing the fixation
cross and the masked cue, the response interval was indicated on the
Neuron 59, 561–567, August 28, 2008 ª2008 Elsevier Inc. 565

computer screen by a question mark. The interval was fixed to 3 s and the re-
sponse was taken at the end: ‘‘Go’’ if the button was being pressed, ‘‘No’’ if the
button was released. The response was written on the screen as soon as the
delay had elapsed. Subjects were told that one response was safe (you do not
win or lose anything) while the other was risky (you can win £1, lose £1, or get
nothing). The risky response was assigned to Go for half of the subjects, and to
NoGo for the other half, such that motor aspects were counterbalanced
between reward and punishment conditions. Subjects were also told that
the outcome of the risky response would depend on the cue that was
displayed between the mask images. In fact, three cues were used, one was
rewarding (+£1), one was punishing (!£1), and the last was neutral (£0). Be-
cause subjects were not informed about the associations, they could only
learn them by observing the outcome, which was displayed at the end of the
trial. This was either a circled coin image (meaning +£1), a barred coin image
(meaning !£1), or a gray square (meaning £0).
Subjects were then debriefed about their visual perceptions and their re-
sponse strategies. They reported responding either at chance, following their
intuition, or following logical rules. None of them had the slightest idea of what
the cues looked like. For the preference judgment task, the cues were then
shown unmasked on a computer screen. The three cues used for a given ses-
sion were displayed side by side, the position being randomized. Subjects
were asked to rate them in order of preferences: 3 for the most liked, 2 for
the intermediate, and 1 for the least one.
To assess for instrumental conditioning, we used one-tailed paired t tests
comparing individual earnings with chance level (which is £0). Similarly, to as-
sess for preference conditioning, we used one-tailed paired t tests comparing
differential rating of winning and losing cues with chance level (which is 0).
Computational Model
We used a standard Q-learning algorithm (Sutton and Barto, 1998), which has
been shown previously to offer a good account of instrumental choice in both
humans and monkeys (Daw and Doya, 2006; McClure et al., 2004; O’Doherty
et al., 2007). For each cue, the model estimates the expected value of the risky
response, on the basis of individual sequences of choices and outcomes. This
value, termed a Q value, is essentially the amount of reward expected from
choosing the risky response given the contextual cue. These Q values were
set at 0.1 before learning, and after every risky response the value of the cue
was updated according to the Rescorla-Wagner rule: Q(t + 1) = Q(t) + a*d(t).
Following this rule, values are increased if the outcome is better than expected,
and decreased in the opposite case. The prediction error was d(t) = R(t) ! Q(t),
with R(t) defined as the reinforcement obtained from choosing the risky
response at trial t. In other words, the prediction error d(t) is the difference be-
tween the expected outcome, i.e., Q(t), and the actual outcome, i.e., R(t). The
reinforcement magnitude R was +1 and !1 for winning and losing £1, and 0 for
neutral outcomes. Given the Q value, the associated probability of choosing
the risky response was estimated by implementing the softmax rule: P(t) = 1/
(1 + exp(!Q(t)/b)). This rule ensures that likelihood will be superior to 0.5 for
positive values and inferior to 0.5 for negative values. The learning rate a con-
cerns the amplitude of value changes from one trial to the next. The tempera-
ture b concerns the randomness of decision making. These two free parame-
ters, a and b, were adjusted to maximize the probability (or likelihood) of the
actual choices under the model. With the constraint that the parameters
should be identical for reward and punishment cues we found: a = 0.1 and
b = 0.9. The model was then used to create statistical regressors correspond-
ing to the Q values and prediction errors, for analysis of brain images.
Images Acquisition and Analysis
T2*-weighted echo planar images (EPI) were acquired with blood oxygen-level
dependent (BOLD) contrast on a 3.0 Tesla magnetic resonance scanner. We
employed a tilted plane acquisition sequence designed to optimize functional
sensitivity in the orbitofrontal cortex and medial temporal lobes (Deichmann
et al., 2003). To cover the whole brain with a short TR (1.95 s), we used the fol-
lowing parameters: 30 slices, 2 mm slice thickness, 2 mm interslice gap. T1-
weighted structural images were also acquired, coregistered with the mean
EPI, normalized to a standard T1 template, and averaged across subjects to
allow group level anatomical localization. EPI images were analyzed in an
event-related manner, within a general linear model, using the statistical para-
566 Neuron 59, 561–567, August 28, 2008 ª2008 Elsevier Inc.
Neuron
fMRI Study of Subliminal Conditioning
metric mapping software SPM5 (Wellcome Trust center for NeuroImaging,
London, UK). The first 5 volumes of each session were discarded to allow
for T1 equilibration effects. Preprocessing consisted of spatial realignment,
normalization using the same transformation as structural images, and spatial
smoothing using a Gaussian kernel with a full-width at half-maximum of 6 mm.
We used two different statistical linear regression models for our analyses. In
both every trial was modeled as having two time points, corresponding to cue
and outcome onsets. In the first model, two separate regressors were created
for cues and outcomes, respectively modulated by the Q values and prediction
errors computed by our optimized algorithm. In the second model, 12 separate
regressors were created corresponding to the two time points (cues and
outcomes) times the two conditioning phases (first and second half of each
session) times the three conditions (reward, neutral, and punishment). In all
cases, the regressors of interest were convolved with a canonical hemody-
namic response function (HRF). To correct for motion artifact, subject-specific
realignment parameters were modeled as covariates of no interest. Linear
contrasts of regression coefficients were computed at the individual subject
level and then taken to a group level random-effects analysis. At group level,
we performed two statistical analyses: first a one-sample t test to find brain re-
gions where regression coefficients were significant across subjects, and sec-
ond a correlation with individual payoffs to find brain regions where regression
coefficients increased with higher conditioning effect. A threshold of p < 0.05
after familywise error (FWE) correction for multiple comparisons was applied to
avoid any a priori on brain localization. A more liberal threshold (p < 0.001,
uncorrected) was also used to observe the extension of significant activations.
To further illustrate activations, time courses were estimated by fitting a flexible
basis set of finite impulse responses (FIRs), separated from the next by one
scan (1.95 s). Both regression coefficients and time courses were then aver-
aged across subjects, pooling together the voxels that passed the conserva-
tive threshold in statistical parametric maps (SPMs).
SUPPLEMENTAL DATA
The Supplemental Data include one figure and supplemental text and can be
found with this article online at http://www.neuron.org/cgi/content/full/59/4/
561/DC1/.
ACKNOWLEDGMENTS
This work was funded by the Wellcome Trust research program grants to
C.D.F. and R.J.D. M.P. received a Young Researcher Prize from the Betten-
court-Schuller Foundation, P.P. was supported by the Swedish Research
Council, and J.D. is a beneficiary of the Marie Curie Intra-European research
fellowship program. We wish to thank Helen Bates and Maël Lebreton, who
independently replicated the behavioral findings outside the scanner.
Accepted: July 1, 2008
Published: August 27, 2008
REFERENCES
Bayley, P.J., Frascino, J.C., and Squire, L.R. (2005). Robust habit learning in
the absence of awareness and independent of the medial temporal lobe.
Nature 436, 550–553.
Berns, G.S., Cohen, J.D., and Mintun, M.A. (1997). Brain regions responsive to
novelty in the absence of awareness. Science 276, 1272–1275.
Clark, R.E., and Squire, L.R. (1998). Classical conditioning and brain systems:
the role of awareness. Science 280, 77–81.
Cohen, N.J., Eichenbaum, H., Deacedo, B.S., and Corkin, S. (1985). Different
memory systems underlying acquisition of procedural and declarative knowl-
edge. Ann. N Y Acad. Sci. 444, 54–71.
Daw, N.D., and Doya, K. (2006). The computational neurobiology of learning
and reward. Curr. Opin. Neurobiol. 16, 199–204.
Neuron
fMRI Study of Subliminal Conditioning
Deichmann, R., Gottfried, J.A., Hutton, C., and Turner, R. (2003). Optimized
EPI for fMRI studies of the orbitofrontal cortex. Neuroimage 19, 430–441.
Delgado, M.R. (2007). Reward-related responses in the human striatum. Ann.
N Y Acad. Sci. 1104, 70–88.
Destrebecqz, A., and Cleeremans, A. (2001). Can sequence learning be
implicit? New evidence with the process dissociation procedure. Psychon.
Bull. Rev. 8, 343–350.
Graybiel, A.M. (2005). The basal ganglia: learning new tricks and loving it. Curr.
Opin. Neurobiol. 15, 638–644.
Hikosaka, O., Nakahara, H., Rand, M.K., Sakai, K., Lu, X., Nakamura, K., Miya-
chi, S., and Doya, K. (1999). Parallel neural networks for learning sequential
procedures. Trends Neurosci. 22, 464–471.
Knight, D.C., Nguyen, H.T., and Bandettini, P.A. (2003). Expression of condi-
tional fear with and without awareness. Proc. Natl. Acad. Sci. USA 100,
15280–15283.
Knutson, B., and Cooper, J.C. (2005). Functional magnetic resonance imaging
of reward prediction. Curr. Opin. Neurol. 18, 411–417.
Krawczyk, D.C., Gazzaley, A., and D’Esposito, M. (2007). Reward modulation
of prefrontal and visual association cortex during an incentive working memory
task. Brain Res. 1141, 168–177.
Lagnado, D.A., Newell, B.R., Kahan, S., and Shanks, D.R. (2006). Insight and
strategy in multiple-cue learning. J. Exp. Psychol. Gen. 135, 162–183.
Lovibond, P.F., and Shanks, D.R. (2002). The role of awareness in Pavlovian
conditioning: empirical evidence and theoretical implications. J. Exp. Psychol.
Anim. Behav. Process. 28, 3–26.
McClure, S.M., York, M.K., and Montague, P.R. (2004). The neural substrates
of reward processing in humans: the modern role of FMRI. Neuroscientist 10,
260–268.
Milner, B., Squire, L.R., and Kandel, E.R. (1998). Cognitive neuroscience and
the study of memory. Neuron 20, 445–468.
Morris, J.S., Öhman, A., and Dolan, R.J. (1998). Conscious and unconscious
emotional learning in the human amygdala. Nature 393, 467–470.
O’Doherty, J., Dayan, P., Schultz, J., Deichmann, R., Friston, K., and Dolan,
R.J. (2004). Dissociable roles of ventral and dorsal striatum in instrumental
conditioning. Science 304, 452–454.
O’Doherty, J.P., Hampton, A., and Kim, H. (2007). Model-based fMRI and its
application to reward learning and decision making. Ann. N Y Acad. Sci.
1104, 35–53.
Olsson, A., and Phelps, E.A. (2004). Learned fear of ‘‘unseen’’ faces after
Pavlovian, observational, and instructed fear. Psychol. Sci. 15, 822–828.
Packard, M.G., and Knowlton, B.J. (2002). Learning and memory functions of
the Basal Ganglia. Annu. Rev. Neurosci. 25, 563–593.
Pagnoni, G., Zink, C.F., Montague, P.R., and Berns, G.S. (2002). Activity in hu-
man ventral striatum locked to errors of reward prediction. Nat. Neurosci. 5,
97–98.
Pecina, S., Smith, K.S., and Berridge, K.C. (2006). Hedonic hot spots in the
brain. Neuroscientist 12, 500–511.
Pessiglione, M., Seymour, B., Flandin, G., Dolan, R.J., and Frith, C.D. (2006).
Dopamine-dependent prediction errors underpin reward-seeking behaviour
in humans. Nature 442, 1042–1045.
Poldrack, R.A., and Packard, M.G. (2003). Competition among multiple
memory systems: converging evidence from animal and human brain studies.
Neuropsychologia 41, 245–251.
Ramnani, N., and Miall, R.C. (2003). Instructed delay activity in the human pre-
frontal cortex is modulated by monetary reward expectation. Cereb. Cortex
13, 318–327.
Seitz, A.R., and Watanabe, T. (2003). Psychophysics: Is subliminal learning
really passive? Nature 422, 36.
Small, D.M., Gitelman, D., Simmons, K., Bloise, S.M., Parrish, T., and Mesu-
lam, M.M. (2005). Monetary incentives enhance processing in brain regions
mediating top-down control of attention. Cereb. Cortex 15, 1855–1865.
Squire, L.R. (1992). Memory and the hippocampus: a synthesis from findings
with rats, monkeys, and humans. Psychol. Rev. 99, 195–231.
Sutton, R.S., and Barto, A.G. (1998). Reinforcement Learning (Cambridge, MA:
MIT Press).
Thorndike, E.L. (1911). Animal Intelligence: Experimental Studies (New York:
Macmillan).
Wilkinson, L., and Shanks, D.R. (2004). Intentional control and implicit
sequence learning. J. Exp. Psychol. Learn. Mem. Cogn. 30, 354–369.
Yacubian, J., Glascher, J., Schroeder, K., Sommer, T., Braus, D.F., and Bu-
chel, C. (2006). Dissociable systems for gain- and loss-related value predic-
tions and errors of prediction in the human brain. J. Neurosci. 26, 9530–9537.
Neuron 59, 561–567, August 28, 2008 ª2008 Elsevier Inc. 567
9984 • The Journal of Neuroscience, September 12, 2007 • 27(37):9984 –9988

Behavioral/Systems/Cognitive

Orbitofrontal Cortex Encodes Willingness to Pay in Everyday

Economic Transactions
Hilke Plassmann, John O’Doherty, and Antonio Rangel
California Institute of Technology, Division of Humanities and Social Sciences, MC 228-77, Pasadena, California 91125

An essential component of every economic transaction is a willingness-to-pay (WTP) computation in which buyers calculate the maxi-
mum amount of financial resources that they are willing to give up in exchange for the object being sold. Despite its pervasiveness, little
is known about how the brain makes this computation. We investigated the neural basis of the WTP computation by scanning hungry
subjects’ brains using functional magnetic resonance imaging while they placed real bids for the right to eat different foods. We found that
activity in the medial orbitofrontal cortex and in the dorsolateral prefrontal cortex encodes subjects’ WTP for the items. Our results
support the hypothesis that the medial orbitofrontal cortex encodes the value of goals in decision making.
Key words: decision making; reward; neuroeconomics; orbitofrontal cortex; fMRI; buying

Introduction
An essential component of every marketplace transaction is a
willingness-to-pay (WTP) computation in which buyers calcu-
late the maximum amount of resources that they are willing to
give up in exchange for the object being sold. The WTP compu-
tation is used to evaluate whether a proposed trade is beneficial
(e.g., when the WTP exceeds the price at which the item is being
offered) or to decide how much to bid for an item (e.g., when
competing with other individuals at an auction). To make good
trades, individuals must be able to assign a WTP to an item that is
commensurate to the benefits that it will generate. Otherwise they
would end up purchasing items for a price that exceeds their
worth to them. Despite its pervasiveness and importance for eco-
nomic well being, little is known about how the brain performs
the WTP computation in everyday transactions, or about how its
ability to do so is affected by diseases such as addiction or obses-
sive compulsive disorders. This makes understanding how and
where the brain makes these computations one of the most im-
portant open questions in the nascent field of neuroeconomics
(Glimcher and Rustichini, 2004; Camerer et al., 2005).
Based on the results of several previous studies, we hypothe-
sized that activity in the medial orbitofrontal cortex (mOFC)
encodes WTP. Monkey electrophysiology studies of binary
choice have found that activity in the OFC encodes the value of
the available actions (Wallis and Miller, 2003; Padoa-Schioppa
and Assad, 2006). Using functional magnetic resonance imaging
(fMRI), Paulus and Frank (2003) found greater medial OFC ac-
tivation during hypothetical choices than during a visual discrim-
Received May 9, 2007; revised July 2, 2007; accepted July 24, 2007.
This work was supported by the Moore Foundation, National Science Foundation Grant SES-0134618, and Ger-
man Academic Exchange Service Grant DAAD D/05/47698. We thank Vivian Valentin, Jan Glaescher, Alan Hampton,
and Axel Linder for their help with this work.
Correspondence should be addressed to Antonio Rangel, California Institute of Technology, Division of Human-
ities and Social Sciences, MC 228-77, Pasadena, CA 91125. E-mail: rangel@hss.caltech.edu.
DOI:10.1523/JNEUROSCI.2131-07.2007
Copyright © 2007 Society for Neuroscience 0270-6474/07/279984-05$15.00/0
ination task (see also Arana et al., 2003). Erk et al. (2002) found
that mOFC activity during a hypothetical liking rating task in-
creased with the reported attractiveness of the stimuli. Finally, a
series of stimulus– outcome learning studies (in which no deci-
sions were made) have shown that the mOFC maintains a repre-
sentation of the expected reward associated with particular cues
(Rolls, 1996; Schoenbaum et al., 1998; Tremblay and Schultz,
1999; Roesch and Olson, 2004). Although all of these studies
suggest that the medial OFC plays a critical role in the evaluation
of choices, none of them have established that activity in the
medial OFC correlates with the economic computation of WTP.
We investigated the neural basis of the WTP computation by
scanning hungry subjects’ brains using fMRI while they placed
bids for the right to eat different foods in a Becker–DeGroot–
Marshak auction (Becker et al., 1964). The results described be-
low confirmed our hypothesis: we found that activity in the right
medial OFC encodes subjects’ WTP for items.
Materials and Methods
Subjects. Nineteen normal-weight subjects participated in the experi-
ment (16 males, mean age, 25.45; age range, 18 – 46). One additional
subject participated in the experiment but was excluded from the analysis
because she did not understand the instructions. All subjects were right-
handed, healthy, had normal or corrected-to-normal vision, had no his-
tory of psychiatric diagnoses, neurological or metabolic illnesses, and
were not taking medications that interfere with the performance of fMRI.
All subjects had no history of eating disorders and were screened for
liking and at least occasionally eating the types of foods that we used.
Subjects were told that the goal of the experiment was to study food
preferences and gave written consent before participating. Caltech’s in-
stitutional The review board of the California Institute of Technology
(Pasadena, CA) approved the study.
Stimuli. Subjects bid on 50 different sweet and salty junk foods (e.g.,
chips and candy bars). We selected the foods based on pilot data to satisfy
several characteristics. First, we wanted items to be highly familiar and to
be sold in local convenience stores, to remove uncertainty considerations
from the WTP computation as much as possible. Familiarity data col-
lected at the end of the experiment shows that we were successful [famil-
Plassmann et al. • fMRI of Willingness to Pay
Figure 1. Experimental design. A, Timeline of the experiment. B, Time course for free bid and forced bid trials. Free and forced the true WTP is $2.3, our measure is $2. Simi-
bid trials were identical except that in forced bid trials visual presentation of the food items was paired with the forced bid amount. larly, subjects with a WTP larger than $3 enter a
In addition, the forced bid amount was repeated during the bidding cue. Food items, trial type, and forced bid amounts were fully bid of $3. However, the bids are a monotonic
randomized within subjects.
iarity scores: mean, 3.97; SD, 1.34; scale, 1 (not familiar) to 5 (very famil-
iar)]. Second, we wanted items to be positive for the subjects (in the sense
that their WTP for them is greater or equal than zero). The foods were
presented to the subjects using high-resolution color pictures (72 dpi).
The stimulus presentation and response recording was controlled by
E-prime (Psychology Software Tools, Pittsburgh, PA). The visual stimuli
were presented using video goggles.
Task. Figure 1 describes the time structure of the experiment. Subjects
were instructed not to eat for 4 h before the experiment, which increased
the value that they placed on the foods. They were also instructed that
they would have to remain in the lab for 30 min at the conclusion of the
experiment, and that the only thing that they will be able to eat is what-
ever food they purchased from us during the task. In addition to a $35
participation fee, each subject received three $1 bills in “spending
money” to purchase food from us. Whatever money they did not spend
was theirs to keep.
Subjects placed bids for the right to eat a snack at the end of the
experiment in 100 different bidding trials. In each trial they were allowed
to bid $0, $1, $2, or $3 for each food item. At the end of the experiment,
one of those trials was randomly selected, by drawing a ball from an urn,
and only the outcome of that trial was implemented. As a result, subjects
did not have to worry about spreading their $3 dollar budget over the
different items and they could treat each trial as if it were the only deci-
sion that counted. Objects were sold using a Becker–DeGroot–Marschak
auction. The rules of the auction are as follows. Let b denote the bid made
by the subject for a particular item. After the bid is made a random
number n is drawn from a known distribution (in our case, $0, $1, $2,
and $3 were chosen with equal probability). If b ! n, the subject got the
item and paid a price equal to n. In contrast, if b ! n, the subject did not
get the object but also did not have to pay anything.
We used this auction institution as our model of market transactions
in the laboratory because it has three very useful properties. First, it is
characterized by a simple set of rules. Second, the optimal strategy for a
buyer is to bid exactly her WTP for the item being sold (Becker et al.,
1964). The intuition for why this is the case is as follows. There is no
incentive to bid less than the WTP because the price paid is determined
by the random number n and, thus, the bids do not affect the price paid.
J. Neurosci., September 12, 2007 • 27(37):9984 –9988 • 9985
There is also no incentive to increase the bid
above the WTP because this may lead to a situ-
ation in which the subject gets the item but ends
up paying a price larger than his WTP (e.g.,
consider the case WTP " $1, b " $3, and n "
$2). The fact that bidding the WTP is the opti-
mal strategy was explained and emphasized ex-
tensively during the instruction and training
period. We performed an extensive amount of
pilot work to find a set of instructions that led to
100% reported compliance with the best strat-
egy. The instructions, included in supplemental
material (available at www.jneurosci.org), em-
phasized that the subject’s best strategy is to
look at the item, ask themselves how much is
worth, and simply bid that amount. Third, be-
cause individuals always bid their exact WTP,
we got a measure of the WTP computed by the
brain for every bidder and item at the time of
decision making, which we could then compare
with the blood oxygenation level-dependent
(BOLD) measure of neural activity.
To keep the task simple, subjects were only
allowed to bid discrete amounts for the items
($0, $1, $2, or $3). A consequence of this is that
the bids are only approximations of the true
WTP computed by subjects. For example, when
function of the true WTP and highly correlated
with it.
We used two different kinds of trials: free-bid
trials and forced-bid trials. Each of the 50 items was shown twice, once in
a bid trial and once in a forced trial. These trials were fully randomized
within and across subjects. Both types of trials had an equal probability of
being selected to be the trial that counted. The timing for each type of trial
is shown in Figure 1 B. The only difference between the two types of trials
is that whereas subjects were free to select the amount of their bid in the
free trials, they were told how much to bid in the forced trials. The forced
bids were drawn uniformly and independently from $0, $1, $2, or $3 on
each trial. The set of rules described above applied to both trials. Note
that subjects needed to make a willingness-to-pay computation in free
trials to decide how much to bid, but they did not need to do so in forced
trials.
After receiving the instructions, subjects were trained on using the
response boxes with their right hand and on the bidding procedure. To
avoid activation artifacts caused by the assignment of buttons to bid
amounts, the assignment was counterbalanced across subjects.
The existence of two types of bidding trials is a novel and essential
component of the experimental design. A difficulty in searching for the
neural basis of the WTP computation is that, when the brain is exposed to
a picture of a food item, it might calculate other variables that are corre-
lated with WTP. For example, the brain may simulate the anticipated
taste of the food, or it may asses its caloric content. If this issue is not
properly addressed, one could erroneously attribute WTP computations
to areas that are calculating different albeit correlated variables. The pres-
ence of free and forced trials provides a solution to the problem. The only
difference between both types of trials is that the subject needs to perform
a WTP computation in the free trials, because she needs to decide how
much to bid, but not in the forced trials, because she is told what her bid
should be. Every other computation, such as the anticipated taste of the
food, should be performed equally in both types of trials. As a result, we
can conclude that a brain area encodes the WTP computation whenever
its activity increases with the WTP in the free trials, but not in the forced
trials.
fMRI data acquisition. The functional imaging was conducted using a
Siemens (Erlangen, Germany) 3.0 Tesla Trio MRI scanner to acquire
gradient echo T2*-weighted echoplanar (EPI) images with BOLD con-
9986 • J. Neurosci., September 12, 2007 • 27(37):9984 –9988 Plassmann et al. • fMRI of Willingness to Pay

trast. To optimize functional sensitivity in the OFC, we used a tilted

acquisition in an oblique orientation of 30° to the anterior commissure–
posterior commissure line (Deichmann et al., 2003). In addition, we used
an eight-channel phased array coil which yields a 40% signal increase in
signal in the medial OFC over a standard head coil. Each volume com-
prised 32 axial slices. A total of 1100 volumes (two sessions, !18 min
each) were collected during the experiment in an interleaved-ascending
manner. The imaging parameters were as follows: echo time, 30 ms; field
of view, 192 mm; in-plane resolution and slice thickness, 3 mm; repeti-
tion time, 2 s. Whole-brain high-resolution T1-weighted structural scans
(1 " 1 " 1 mm) were acquired from the 19 subjects and coregistered with
their mean EPI images and averaged together to permit anatomical lo-
calization of the functional activations at the group level. Image analysis
was performed using SPM5 (Wellcome Department of Imaging Neuro-
science, Institute of Neurology, London, UK). Temporal normalization
was applied to the scans with a time of acquisition of 1.9375 referenced to
the last volume. To correct for subject motion, the images were realigned
to the last volume, spatially normalized to a standard T2* template with
a resampled voxel size of 3 mm, and spatially smoothed using a Gaussian
kernel with a full width at half maximum of 8 mm. Intensity normaliza-
tion and high-pass temporal filtering (using a filter width of 128 s) were
Figure 2. Behavioral results. A, Distribution of bids in free and forced trials. B, Reaction times
also applied to the data.
for free and forced trials as a function of bid. Error bars denote SEs.
fMRI data analysis. The data analysis proceeded in three steps. First, we
estimated a general linear model with AR(1) and the following regressors
that capture the main events in our experiment: free bid and picture cluster size of 10. Anatomical localizations were then performed by over-
presentation (R1), free bid and response (R2), forced bid and picture laying the t maps on a normalized structural image averaged across sub-
presentation (R3), forced bid and response (R4), missed bid trial and jects, and with reference to an anatomical atlas (Duvernoy, 1999)
picture presentation (R5), and missed bid trial and response (R6). The
regressors that capture the presentation of the food pictures were mod- Results
eled using 4 s box-car functions. The regressors for the bid responses were Behavioral
modeled using stick functions. Figure 2 A shows the distribution of bids during free- and forced-
To take advantage of the parametric nature of our design, the general bid trials. The average free bid was $1.4 (SD, 0.27) and over 75%
linear model also included the following parametric modulators: free bid of the free bids were greater than zero. The bid amounts for the
and picture presentation modulated by bid (M1), free bid and picture
forced bid trials were randomly drawn from a uniform distribu-
presentation modulated by surplus# (M2), free bid and response mod-
ulated by bid (M3), free bid and response modulated by surplus# (M4), tion on $0, $1, $2, and $3. Although there is substantial variability
forced bid and picture presentation modulated by bid (M5), forced bid on value that subjects place on particular items, the average WTP
and item presentation modulated by surplus# (M6), forced bid and item was significantly greater than zero ( p & 0.001), which suggests
presentation modulated by surplus$ (M7), forced bid and response that most items were rewarding for most subjects.
modulated by bid (M8), forced bid and response modulated by surplus# Figure 2 B shows the distribution of bidding reaction times for
(M9), and forced bid and response modulated by surplus$ (M10). The free and forced trials. The reaction times were entered in a two-
parametric modulators are defined as follows. “Bid” equals the amount way repeated-measures ANOVA with two factors: bid amount
bid for the item sold in that trial during the corresponding free trial and, ($0, $1, $2, or $3) and trial type (free- or forced-bid trial). The
thus, is a measure of the subject’s WTP for the item being shown. “Sur- analysis revealed no significant main effects or interactions.
plus” is a variable that measures the expected “profit” from the trial given
the bid that was placed (and conditional on the trial being selected to be
the one that counts). For example, suppose that a subject’s true value is $2
Neuroimaging results
and that he bids $2. Then the surplus equals 0.25 " $2 # 0.25 " $1 # Identifying the neural correlates of WTP in free trials
0.50 " $0% $0.75, where the first term measures the probability that the We performed a whole-brain analysis to identify areas that cor-
random number in the auction is 0 times the profit made in that case, the related with WTP in the free trials at the time of evaluation (i.e.,
second number measure the probability that the random number is 1 when the food picture is displayed). This contrast is interesting
times the profit made in that case, and so on. Surplus# % max[0,surplus] because it identifies areas that might encode for WTP. Our hy-
and Surplus$ % min[surplus,0]. (Note that for the free trials, the surplus pothesis was that activity in the OFC would be positively corre-
variable is always non-negative, whereas for the forced trials it can be lated with WTP. The hypothesis was supported by the data: ac-
positive or negative.) We orthogonalized the modulators for each of the tivity in the medial OFC (x % 6, y % 30, z % $17; p & 0.001,
main regressors (M1 and M2, M3 and M4, M5 to M7, and M8 to M10).
uncorrected) was correlated with WTP. Other areas identified by
Each of the regressors was convolved with a canonical hemodynamic
response function. We also included a constant term and six motion
this contrast were the dorsal anterior cingulate cortex (x % $4,
parameters as regressors of no interest. Note that the inclusion of the y % 34, z % $20; p & 0.001, uncorrected), and the dorsolateral
surplus modulator is important to avoid confounding areas that code for prefrontal cortex (DLPFC; x % 44, y % 44, z %12; p & 0.001,
WTP with areas that code for economic surplus. uncorrected).
Second, we calculated the following first-level single-subject contrasts:
(1) free-bid trials when exposed to item modulated by bid (regressor Identifying the neural correlates of value during the
M1), (2) forced-bid trials when exposed to item modulated by bid (re- forced-bid trials
gressor M5), and (3) free- minus forced-bid trials when exposed to an Our experimental design is based on the idea that the brain com-
item modulated by bid (regressors M1 minus M5). putes a WTP during free-bid trials, but not during forced-bid
Third, we calculated second-level group contrasts using a one-sample trials. To test this hypothesis, we performed a whole-brain anal-
t test. The figures shown below are constructed using these second-level ysis to identify areas correlated with WTP in the forced-bid trials
contrasts at a threshold of p & 0.001, uncorrected, and a minimum at the time of evaluation (i.e., when the food picture is displayed).
Plassmann et al. • fMRI of Willingness to Pay
Figure 3. Neural correlates of WTP. A, B, Activity in the medial OFC and the DLPFC was positively correlated with WTP at the Olson, 2004; Padoa-Schioppa and Assad,
time of evaluation in the free trials more than in the forced trials. Activation maps shown at a threshold of p ! 0.001 uncorrected 2006). In addition to providing cross-
and 10 voxel clusters. C, Averaged time courses for the medial OFC voxels during free trials as a function of WTP (error bars denote species and cross-modality validation, this
SEs). D, Averaged time courses for the medial OFC voxels during forced bid trials as a function of WTP. E, Averaged time courses for experiment shows that the OFC plays a
the medial OFC voxels during forced bid trials as a function of the forced bid. A comparison of the time courses shows that the central role in the encoding of decision
medial OFC encodes WTP in free trials, but not in forced trials, and that it does not encode the forced bid amounts.
No areas showed the desired correlation at a level of p ! 0.001
(uncorrected) and an extent threshold of 10 voxels.
Another possibility is that during the forced trials the brain
encodes either the size of the forced bid, or the disagreement
between the forced bid and the WTP. We tested for both possi-
bilities using minor variants of the general linear model described
above and found no regions of correlation at a level of p ! 0.001
(uncorrected) and an extent threshold of 10 voxels (for details,
see the supplemental material, available at www.jneurosci.org).
Test of the main hypothesis
As discussed above, a limitation of the previous contrast is that it
identifies areas with activity that is correlated with WTP, but also
areas that encode for variables that are correlated with it, such as
anticipatory taste. To address this potential confound, we looked
for areas that (1) showed increasing activation with WTP in the
free trials and (2) were significantly more activated in the free
trials than in the forced trials. As predicted, we found that the
right mOFC satisfied these conditions (x " 4, y " 30, z " #18;
p ! 0.001, uncorrected). Unexpectedly, we also found that right
DLPFC satisfied them (x " 44, y " 44, z " 18; p ! 0.001, uncor-
rected). Figure 3, A and B, describes the results of this contrast.
We also extracted trial averaged time-course data from peak
voxels in the mOFC for each subject, which were then averaged
across subjects (Fig. 3C–E). The time courses show that activity in
this area during free trials showed an increase in activation that
was correlated with the subjects’ bids. The time courses also show
that activity during forced trials did not discriminate the subjects’
WTP for the items (as measured by their bids for that item during
free trials) or the magnitude of the forced bids. We thus con-
cluded that activity in the right medial OFC and dorsolateral
prefrontal cortex encode for WTP in everyday economic
decisions.
J. Neurosci., September 12, 2007 • 27(37):9984 –9988 • 9987
Discussion
In this study, we provide evidence that
mOFC encodes subjects’ WTP during sim-
ple economic transactions. Critically, we
used a parametric experimental design
that allowed us to identify areas that en-
code for WTP, as opposed to areas that are
active during economic choice, but that do
not correlate with WTP (Blair et al., 2006;
Arana et al., 2003; Paulus and Frank,
2003).
Our findings are consistent with data
from human lesion studies showing that
lesions to the ventromedial prefrontal cor-
tex impair the ability to make consistent
pairwise choices (Fellows and Farah,
2007). Our findings are also consistent
with monkey electrophysiology studies of
simple choice behavior. For example, sev-
eral studies have found that OFC neurons
encode for the decision or incentive value
of the stimuli at the time of decision mak-
ing (Wallis and Miller, 2003; Roesch and
values in decision-making tasks that are
significantly more abstract and complex
that those that can be studied in monkey
experiments. As a result, we speculate that the mOFC might en-
code for the decision value of choices in a wide class of economic
settings. However, because this study and much of the previous
literature has focused on the valuation of primary appetitive re-
wards, such as desirable foods, additional work is needed to in-
vestigate whether the mOFC area also encodes for the value of
nonprimary rewards, such as a book or a DVD, and of negative or
undesirable items, such as electric shocks of different
magnitudes.
An open question in behavioral neuroscience is which parts of
the OFC play a role in learning the encoding of stimulus-outcome
associations (Tremblay and Schultz, 1999; Rolls, 2000) and which
parts are involved in guiding decisions by encoding the value of
alternative goals (Arana et al., 2003; Schoenbaum and Roesch,
2005; Feierstein et al., 2006; Padoa-Schioppa and Assad, 2006)?
Our results suggest that the mOFC plays a critical role in goal
directed behavior by encoding economic value.
We also found that activity in the DLPFC was correlated with
the subject’s WTP. This is consistent with previous monkey neu-
rophysiology studies that have found neurons in the DLPFC that
encode aspects of the decision or the incentive value of stimuli
(Watanabe, 1996; Wallis and Miller, 2003). This raises an impor-
tant open question in the neuroeconomics literature: what are the
relative contributions of the DLPFC and the mOFC to the valu-
ation of stimuli during economic decision making? The neuro-
anatomy of these two regions suggests a potential explanation.
The OFC receives inputs from multiple sensory areas, which are
likely to be used in valuation, whereas the DLPFC does not (Price,
2006). In contrast, the DLPFC is heavily connected with motor
output areas, whereas the OFC is not connected with these areas
directly (Petrides and Pandya, 1999). Finally, the DLPFC and the
9988 • J. Neurosci., September 12, 2007 • 27(37):9984 –9988
OFC are interconnected (Petrides and Pandya, 1999). This pat-
tern of connectivity suggests that economic values might be first
computed in the mOFC and then passed to the DLPFC to influ-
ence motor commands. This pathway is not unique as the OFC
might also be able to affect actions through its connections to the
striatum, which, in turn, is also heavily connected to the motor
system (Yeterian and Pandya, 1991). This conjecture is consistent
with our data, but our experimental design does not allow us to
reach this directly.
It is interesting to compare our findings with those of Knutson
et al. (2007), who study which brain areas are involved in making
economic purchase decisions for unfamiliar items sold at exog-
enously given prices. Their task proceeds in three steps: (1) sub-
jects are shown a picture of the item that is for sale, (2) the sale
price is added, and (3) subjects decide whether to purchase the
item or not. Knutson et al. (2007) obtained a measure of the value
that the subjects placed on the items, albeit not a WTP, and
looked for neural correlates of this value at the time of picture
presentation (step 1). They found a correlation with nucleus ac-
cumbens (NAcc) activity, but not with OFC activity. In contrast,
in our study activity in the medial OFC, but not in the NAcc,
encoded the subject’s WTP for items. Note that the difference in
results is not attributable to an inability to image the NAcc be-
cause we get strong striatal activation in other contrasts of interest
(supplemental Tables 3, 4, available at www.jneurosci.org as sup-
plemental material). Instead, it is likely that there are subtle but
important differences between the computations that the brain
makes in the two tasks, and that the mOFC and NAcc play a
differential role in such computations. For example, in the Knut-
son et al. (2007) task, the value of purchasing an item is the WTP
minus the price and the information needed to compute this “net
value” is revealed over time. In contrast, in our experiment all of
the information needed to compute the WTP is revealed at the
beginning of the trials. As a result, anticipatory reward signals,
which are known to be correlated with NAcc activity (Knutson et
al., 2001), might be computed in the Knutson et al. (2007) task,
but not in the current study. Additional experiments are needed
to systematically explore the differences between the computa-
tions made in the two experiments.
Part of the research agenda in neuroeconomics is to under-
stand how the brain evaluates potential goals and outcomes at the
time of decision making, and how other cognitive, emotional,
and visceral processes affect the computation of economic value.
A first step in this research agenda is to understand what are the
brain structures responsible for the computation of value in sim-
ple everyday choices. Our results suggest that the medial OFC is a
place where a variety of variables computed in other brain regions
are integrated into a single representation of value. If this hypoth-
esis is correct, other brain processes may be able to influence
decision making by modulating activity in the medial OFC.
References
Arana FS, Parkinson JA, Hinton E, Holland AJ, Owen AM, Roberts AC
(2003) Dissociable contributions of the human amygdala and orbito-
Plassmann et al. • fMRI of Willingness to Pay
frontal cortex to incentive motivation and goal selection. J Neurosci
23:9632–9638.
Becker G, DeGroot M, Marschak J (1964) Measuring utility by a single-
response sequential method. Behav Sci 9:226 –232.
Blair K, Marsh AA, Morton J, Vythilingam M, Jones M, Mondillo K, Pine DC,
Drevets WC, Blair JR (2006) Choosing the lesser of two evils, the better
of two goods: specifying the roles of ventromedial prefrontal cortex and
dorsal anterior cingulate in object choice. J Neurosci 26:11379 –11386.
Camerer C, Loewenstein G, Prelec D (2005) Neuroeconomics: how neuro-
science can inform economics. J of Economic Literature 43:9 – 64.
Deichmann R, Gottfried JA, Hutton C, Turner R (2003) Optimized EPI for
fMRI studies of the orbitofrontal cortex. NeuroImage 19:430 – 441.
Duvernoy HM (1999) The human brain: surface, three-dimensional sec-
tional anatomy with MRI, and blood supply. Berlin: Springer.
Erk S, Spitzer M, Wunderlich A, Galley L, Walter H (2002) Cultural objects
modulate reward circuitry. NeuroReport 13:2499 –2503.
Feierstein CE, Quirk MC, Uchida N, Sosulski DL, Mainen ZF (2006) Rep-
resentation of spatial goals in rat orbitofrontal cortex. Neuron
51:495–507.
Fellows LK, Farah MJ (2007) The role of ventromedial prefrontal cortex in
decision making: judgment under uncertainty or judgment per se? Cere-
bral Cortex, in press.
Glimcher PW, Rustichini A (2004) Neuroeconomics: the concilience of
brain and decision. Science 306:4447– 4452.
Knutson B, Adams CM, Fong GW, Hommer D (2001) Anticipation of in-
creasing monetary reward selectively recruits nucleus accumbens. J Neu-
rosci 21:RC159.
Knutson B, Rick S, Wimmer GE, Prelec D, Loewenstein G (2007) Neural
predictors of purchases. Neuron 53:147–156.
Padoa-Schioppa C, Assad JA (2006) Neurons in the orbitofrontal cortex
encode economic value. Nature 441:223–226.
Paulus MP, Frank LR (2003) Ventromedial prefrontal cortex activation is
critical for preference judgments. NeuroReport 14:1311–1315.
Petrides M, Pandya DN (1999) Dorsolateral prefrontal cortex: comparative
cytoarchitectonic analysis in the human and the macaque brain and cor-
ticocortical connection patterns. Eur J Neurosci 11:1011–1036.
Price JL (2006) Connections of orbital cortex. In: the orbitofrontal cortex
(Zald DH, Raunch SL, eds), pp 39 –55. Oxford: Oxford UP.
Roesch MR, Olson CR (2004) Neuronal activity related to reward value and
motivation in primate frontal cortex. Science 304:307–310.
Rolls ET (1996) The orbitofrontal cortex. Philos Trans R Soc Lond B Biol Sci
351:1433–1443, discussion 1443–1434.
Rolls ET (2000) The orbitofrontal cortex and reward. Cereb Cortex
10:284 –294.
Schoenbaum G, Roesch M (2005) Orbitofrontal cortex, associative learn-
ing, and expectancies. Neuron 47:633– 636.
Schoenbaum G, Chiba AA, Gallagher M (1998) Orbitofrontal cortex and
basolateral amygdala encode expected outcomes during learning. Nat
Neurosci 1:155–159.
Tremblay L, Schultz W (1999) Relative reward preference in primate or-
bitofrontal cortex. Nature 398:704 –708.
Wallis JD, Miller EK (2003) Neuronal activity in primate dorsolateral and
orbital prefrontal cortex during performance of a reward preference task.
Eur J Neurosci 18:2069 –2081.
Watanabe M (1996) Reward expectancy in primate prefrontal neurons. Na-
ture 382:629 – 632.
Yeterian EH, Pandya DN (1991) Prefrontostriatal connections in relation to
cortical architectonic organization in rhesus monkeys. J Comp Neurol
312:43– 67.
 The Journal of Neuroscience, September 29, 2010 • 30(39):13095–13104 • 13095

Behavioral/Systems/Cognitive

The Architecture of Reward Value Coding in the Human

Orbitofrontal Cortex
Guillaume Sescousse,1,2 Jérôme Redouté,1,2,3 and Jean-Claude Dreher1,2
1Center for Cognitive Neuroscience, Reward and Decision-Making Group, Centre National de la Recherche Scientifique, Unité Mixte de Recherche 5229,
69675 Bron, France, 2Université Lyon 1, 69003 Lyon, France, and 3CERMEP–Imagerie du Vivant, 69003 Lyon, France

To ensure their survival, animals exhibit a number of reward-directed behaviors, such as foraging for food or searching for mates. This
suggests that a core set of brain regions may be shared by many species to process different types of rewards. Conversely, many new brain
areas have emerged over the course of evolution, suggesting potential specialization of specific brain regions in the processing of more
recent rewards such as money. Here, using functional magnetic resonance imaging in humans, we identified the common and distinct
brain systems processing the value of erotic stimuli and monetary gains. First, we provide evidence that a set of neural structures,
including the ventral striatum, anterior insula, anterior cingulate cortex, and midbrain, encodes the subjective value of rewards regard-
less of their type, consistent with a general hedonic representation. More importantly, our results reveal reward-specific representations
in the orbitofrontal cortex (OFC): whereas the anterior lateral OFC, a phylogenetically recent structure, processes monetary gains, the
posterior lateral OFC, phylogenetically and ontogenetically older, processes more basic erotic stimuli. This dissociation between OFC
representations of primary and secondary rewards parallels current views on lateral prefrontal cortex organization in cognitive control,
suggesting an increasing trend in complexity along a postero-anterior axis according to more abstract representations. Together, our
results support a modular view of reward value coding in the brain and propose that a unifying principle of postero-anterior organization
can be applied to the OFC.

Introduction
A basic concern about the functional organization of the prefron-
tal cortex is to delineate the functional divisions of the orbito-
frontal cortex (OFC). A number of lesion, electrophysiological,
and neuroimaging studies indicate a general role for the OFC in
encoding the value assigned to different goods in both human
(O’Doherty et al., 2003a; Plassmann et al., 2007; Chib et al., 2009;
FitzGerald et al., 2009) and nonhuman primates (Tremblay and
Schultz, 1999; Padoa-Schioppa and Assad, 2008). An important
remaining issue, which is key to our understanding of the func-
tional organization of the OFC, is to determine whether distinct
parts of the OFC encode rewards of different nature. The anterior
and posterior parts of the OFC are considered to belong to two
distinct cytoarchitectonic trends. The anterior part of the OFC,
especially well developed in humans and characterized by a gran-
ular cell layer, is thought to be phylogenetically and ontogeneti-
cally more recent than the posterior and medial parts, which
consist of agranular and dysgranular cortices (Ongür and Price,
2000; Wise, 2008). Although never tested empirically, one funda-
mental hypothesis, based on this increasing trend in complexity
Received July 7, 2010; revised July 19, 2010; accepted Aug. 9, 2010.
This work was funded by a Marie Curie International Reintegration Grant (J.-C.D.) and a Fyssen Foundation Grant
(J.-C.D.). G.S. was funded by a scholarship from the French Ministry of Research and the Medical Research Founda-
tion. We thank the staff of CERMEP–Imagerie du Vivant for helpful assistance with data collection.
Correspondence should be addressed to either Guillaume Sescousse or Jean-Claude Dreher, Centre de Neuro-
science Cognitive, CNRS UMR 5229, 67 Boulevard Pinel, 69675 Bron Cedex, France, E-mail: gsescousse@isc.cnrs.fr or
dreher@isc.cnrs.fr.
DOI:10.1523/JNEUROSCI.3501-10.2010
Copyright © 2010 the authors 0270-6474/10/3013095-09$15.00/0
along a postero-anterior axis, is that the anterior part of the OFC
would process secondary rewards, whereas the posterior part
would process primary rewards (Kringelbach and Rolls, 2004).
Here, we directly tested this hypothesis with functional mag-
netic resonance imaging (fMRI) by comparing the brain re-
sponses to two experienced rewards: money and erotic pictures.
These two rewards present significant evolutionary differences
likely to be reflected at the cerebral level: whereas money is a
secondary reward that appeared recently in human history and
whose abstract value needs to be learned by association with pri-
mary reinforcers, erotic stimuli can be considered as primary
rewards because they have an innate value and satisfy biological
needs. We therefore hypothesized that monetary gains would
recruit anterior OFC regions and erotic pictures would engage
more posterior OFC regions. Despite their critical sociobiological
importance, erotic stimuli have never been studied as reinforcers
but rather as arousing stimuli in passive viewing paradigms fo-
cusing on sexual function (Redouté et al., 2000; Ponseti et al.,
2006). However, erotic stimuli are clearly rewarding (Hamann et
al., 2004), probably because sexual attractiveness, which may
have evolved to enhance reproductive fitness, is an important cue
for mate choice (Thornhill and Gangestad, 1999; Rhodes, 2006).
In addition to specialized OFC regions processing different
types of rewards, we hypothesized the existence of common brain
structures supporting general hedonic representations indepen-
dent of reward type. To evaluate and compare the relative value of
different rewards on a unique scale, it has been proposed that the
brain may use a “common neural currency,” likely to be imple-
mented in integrative reward regions such as the ventral striatum
13096 • J. Neurosci., September 29, 2010 • 30(39):13095–13104
and ventromedial prefrontal cortex (vmPFC) (Montague and
Berns, 2002; Sugrue et al., 2005; Rangel et al., 2008; Dreher,
2009). Consistent with this claim, a number of studies have
shown that goal/decision values, reflecting the anticipated re-
warding properties of a stimulus at the time of choice, were en-
coded in these regions regardless of reward type (McClure et al.,
2007; Hare et al., 2008; Chib et al., 2009; Peters and Buchel, 2010).
However, very few studies have investigated whether shared ce-
rebral substrates are engaged when actually experiencing differ-
ent types of rewards, i.e., during the computation of outcome/
hedonic values (Izuma et al., 2008; Smith et al., 2010). Such a
common representation of outcome value may be particularly
useful to support the trading ability demonstrated by primates.
For example, humans are willing to sacrifice money to view at-
tractive faces (Hayden et al., 2007), and similarly male monkeys
exchange meat for sex or sacrifice fluid for the opportunity to
view female perinea (Deaner et al., 2005; Gomes and Boesch,
2009).
Building on these considerations, we designed a reward para-
digm aiming to (1) determine whether the OFC is functionally
divided depending on reward type and (2) identify common
brain structures processing the experienced value of both mone-
tary and erotic rewards. To further characterize the role of these
brain regions in reward processing, we also manipulated reward
intensity and probability and collected hedonic ratings after re-
ward outcomes inside the scanner. Because erotic pictures are not
quantifiable like money, it is unclear whether changes in the in-
tensity of erotic pictures would affect the same brain regions as
those responding to changes in monetary amounts. Likewise, it is
unknown whether the concept of prediction error, reflecting the
discrepancy between expected and actual rewards and primarily
used with quantifiable rewards (such as money and juice), can be
extended to erotic stimuli.
Materials and Methods
Participants
Eighteen right-handed volunteers (mean ! SD age, 24 ! 3.3 years) with
no history of neurological or psychiatric disorders participated in this
study. All of them were heterosexual males, because men are generally
more responsive to visual sexual stimuli than women (Hamann et al.,
2004) and to avoid the potential influence of the menstrual cycle known
to have an effect on reward processing in women (Caldú and Dreher,
2007; Dreher et al., 2007). All subjects gave written informed consent to
be part of the experiment, which was approved by the local ethics com-
mittee (Centre Léon Bérard, Lyon, France).
Motivation, which was a crucial element of our study, was closely
controlled. First, sexual arousability was assessed at the time of screening
through specific questionnaires, namely the Brief Sexual Function ques-
tionnaire (Reynolds et al., 1988) and the Sexual Arousability Inventory
(Hoon and Chambless, 1998). Of an initial pool of 22 subjects, two of
them were excluded because they scored too low on the Sexual Arous-
ability Inventory (mean score for all subjects, 91.1 ! 14.6; scores of
excluded subjects, 54 and 64). To further ensure that all participants
would be in a similar state of motivation to see erotic stimuli, we asked
them to avoid any sexual contact during a period of 24 h before the
scanning session. Second, we sought to enhance the motivation for
money by telling the subjects that the financial compensation for their
participation would be calculated based on their winnings during the
task. We also excluded two subjects presenting symptoms of depression
as assessed by the 13-item version of the Beck Depression Inventory
(Beck and Beck, 1972) (mean score for all subjects, 1.9 ! 2.7; scores of
excluded subjects, 6 and 10).
Task
Our protocol was inspired from the typical design of incentive delay tasks
(Knutson et al., 2005; Abler et al., 2006) but included several modifica-
Sescousse et al. • Reward Value Coding in the Human OFC
tions related to our questions. Experimental trials were divided into two
phases: reward anticipation and outcome. During reward anticipation, a
cue was presented, followed by a delay period and a discrimination task
(Fig. 1). The cue carried three types of information regarding the upcom-
ing reward: the red portion of a pie chart in the background indicated its
probability (25, 50, or 75%), and the pictogram in the foreground indi-
cated its type (monetary or erotic) and intensity (high or low, depending
on the size of the pictogram). This led to a total of 12 different cues plus
a control condition associated with no chance of winning (supplemental
Fig. 1, available at www.jneurosci.org as supplemental material). After a
variable delay period (question mark representing a pseudorandom draw
depending on probability), subjects were asked to perform a discrimina-
tion task, in which they had to respond correctly to a target within a
maximum time of 1 s. The shape of the target was drawn at random on
each trial and could be either a triangle (left button press) or a square
(right button press). Success on this discrimination task (indicated by a
magnified target) allowed the subjects to view the outcome of the pseu-
dorandom draw, whereas erroneous or slow response (indicated by no
change in target size) led to no reward. In rewarded trials, the reward was
either an erotic image (with high or low erotic content) or the picture of
a safe mentioning the amount of money won (high or low amount). After
each reward outcome, subjects were asked to provide a hedonic rating by
moving a cursor along a 1-to-9 continuous scale (1 for very little pleased;
9 for very highly pleased). In non-rewarded and control trials, the sub-
jects were presented with “scrambled” pictures. A fixation cross was fi-
nally used as an intertrial interval of variable length.
Stimuli
Two categories (high and low intensity) of erotic pictures and monetary
gains were used. Nudity being the main criteria driving the reward value
of erotic stimuli, we separated them into a “low intensity” group display-
ing women in underwear or bathing suits and a “high intensity” group
displaying naked women in an inviting posture. Each erotic picture was
presented only once during the course of the task to avoid habituation. A
similar element of surprise was introduced for the monetary rewards by
randomly varying the amounts at stake: the low amounts were €1, €2, or
€3 and the high amounts were €10, €11, or €12. The pictures displayed in
non-rewarded and control trials were scrambled versions of the pictures
used in rewarded trials and hence contained the same information in
terms of chromaticity and luminance.
fMRI data acquisition
Imaging was conducted on a 1.5 T Siemens Sonata scanner, using an
eight-channel head coil. The scanning session was divided into four runs.
Each of them included four repetitions of each cue, with the exception of
the control condition, repeated nine times. This yielded a total of 228
trials. Within each run, the order of the different conditions was pseudo-
randomized and optimized to improve signal deconvolution. The order
of the runs was counterbalanced between subjects. Before scanning, all
subjects were given oral instructions and familiarized with the cognitive
task in a short training session.
Each of the four functional runs consisted of 296 volumes. Twenty-six
interleaved slices parallel to the anterior commissure–posterior commis-
sure line were acquired per volume (field of view, 220 mm; matrix, 64 "
64; voxel size, 3.4 " 3.4 " 4 mm; gap, 0.4 mm), using a gradient-echo
echoplanar imaging (EPI) T2*-weighted sequence (repetition time, 2500
ms; echo time, 60 ms; flip angle, 90°). To improve the local field homo-
geneity and hence minimize susceptibility artifacts in the orbitofrontal
area, a manual shimming was performed within a rectangular region
including the OFC and the basal ganglia. A high-resolution T1-weighted
structural scan was subsequently acquired in each subject.
fMRI analysis
Preprocessing. Preprocessing of fMRI data was conducted using SPM2.
The first four functional volumes of each run were removed, and the
remaining images were corrected for slice-timing artifacts and spa-
tially realigned to the first image of each time series. We then searched
for residual artifacts in the time series with the tsdiffana utility (http://
imaging.mrc-cbu.cam.ac.uk/imaging/DataDiagnostics) and modeled
them with dummy regressors in our general linear model (two subjects
Sescousse et al. • Reward Value Coding in the Human OFC J. Neurosci., September 29, 2010 • 30(39):13095–13104 • 13097

Figure 1. Paradigm and behavior. A, Sequence of events during a typical trial. Subjects first saw a cue informing them about the type, probability, and intensity of an upcoming reward
(supplemental Fig. 1, available at www.jneurosci.org as supplemental material). Three cases are represented here: a 75% chance of receiving a high amount of money (top), a 25% chance of seeing
a low erotic content picture (middle), and a sure chance of getting nothing (control trials; bottom). After a short delay and a target discrimination task, subjects saw the outcome, which was
contingent on both the announced probability and their performance on the discrimination task. Reward outcomes consisted either in a monetary amount displayed on a safe (top) or an erotic
picture (middle) and were followed by the rating of their subjective value on a continuous scale. Non-rewarded and control trials displayed a scrambled picture at outcome (bottom). B, Behavioral
results on the discrimination task: mean reaction times according to reward intensity (left) and probability (middle) and mean hit rates according to reward intensity (right). C, Mean subjective
ratings according to reward intensity, on a 1-to-9 scale. Error bars indicate SEM. *p # 0.05; **p # 0.01; ***p # 0.001 by Tukey’s HSD tests.

had three artifacts and one subject had one artifact in their time series).
The functional images were then normalized to the Montreal Neurolog-
ical Institute (MNI) stereotaxic space using the EPI template of SPM2
and spatially smoothed with a 10 mm full-width at half-maximum iso-
tropic Gaussian kernel. Anatomical scans were normalized to the MNI
space using the icbm152 template brain and averaged across subjects.
Identification of common and specific brain regions. The event-related
statistical analysis was performed according to the general linear model
as implemented in SPM2. Anticipation-related responses were modeled
as boxcar functions time locked to the onset time of the cue with a
duration of 2.5 s. The 2 rewards (monetary/erotic) ! 2 intensities (high/
low) were modeled as four separate conditions. For each of them, a
first-order parametric regressor modeled reward probability. The con-
trol condition was modeled in a separate regressor. Outcome-related
responses were modeled as events time locked to the appearance of the
reward (or scrambled picture). Four main conditions were defined:
“monetary reward” (MR), “erotic reward” (ER), “no-monetary reward”
(NoMR), and “no-erotic reward” (NoER). Two orthogonalized covari-
ates linearly modeling the expected probability and the ratings were
added (in this order) to the MR and ER regressors. A last regressor
modeled the appearance of a scrambled picture in the control condi-
tion (C). All regressors were subsequently convolved with the canon-
ical hemodynamic response function and entered in a first-level
analysis. A high-pass filter with a cutoff of 128 s was applied to the
time series to remove low-frequency noise and baseline drifts. The
resulting images of parameter estimates were then passed in a second-
level group analysis in which between-subject variability was treated
as a random effect.
Although the anticipatory period was explicitly modeled in our anal-
ysis, we only report results concerning the outcome phase because our
focus was on the coding of experienced reward value. Following the
forward inference approach (Henson, 2006) (see Results), brain regions
responding specifically to monetary (or erotic) rewards resulted from the
contrast MR " ER (or ER " MR), masked inclusively with MR " C (or
ER " C) and exclusively with ER " C (or MR " C). The main contrasts
MR " ER and ER " MR and the masks were thresholded independently
using a whole-brain correction for multiple comparisons [p # 0.05 fami-
lywise error (FWE) and p # 0.01 false discovery rate (FDR), respectively],
thereby ensuring the absence of any “selection bias” in the analysis
(Kriegeskorte et al., 2009). Brain regions activated by both monetary and
erotic rewards were identified in two steps. We first performed a con-
junction analysis of the contrasts MR " C and ER " C based on the
minimum statistic (Nichols et al., 2005) ( p # 0.05 FWE whole-brain
corrected). Because this conjunction may be sensitive to other types of
computations (such as attention or image processing), we then masked it
inclusively with the regions responding parametrically with both mone-
tary and erotic hedonic ratings (each mask thresholded at p # 0.05 FDR
whole-brain corrected).
Anatomical localization of functional clusters was performed based on
a probabilistic atlas (Hammers et al., 2003).
Prediction error model. Positive prediction errors were defined at each
trial t by PE(t) $ V(t) % B(t), where V(t) is the outcome value and B(t) is
the expected value (Yacubian et al., 2006). Whereas monetary amounts
could have been used to assess V(t) for monetary rewards, erotic rewards
could not be similarly quantified. Hence, to use an equivalent measure
for both rewards, we used the hedonic rating to assess V(t) on each trial.
13098 • J. Neurosci., September 29, 2010 • 30(39):13095–13104 Sescousse et al. • Reward Value Coding in the Human OFC

Figure 2. Functional postero-anterior dissociation in the orbitofrontal cortex depending on reward type. Brain regions responding specifically to monetary reward outcomes are displayed in
blue– green, and those responding specifically to erotic reward outcomes are displayed in red–yellow. Plots of mean percent signal change, which are not independent of the whole-brain analysis,
are shown only to illustrate the double dissociation between monetary/erotic rewards and anterior (Ant.)/posterior (Post.) OFC. Activations are overlaid on an average anatomical scan of all subjects
( p " 0.05 FWE whole-brain corrected). Error bars indicate SEM.

B(t) was defined as the product of reward probability P(t) by expected

intensity E(t). P(t) was simply the probability given explicitly in the cue
(25, 50, or 75%). E(t), presented as either “high” or “low” to the partic-
ipants, was transformed into a numerical value by using the past ratings:
for instance, E(t) for a high monetary reward was estimated as the average
of all the ratings given to high monetary rewards since the beginning of
the task up to trial t.
Prediction error values were entered into two parametric regressors
separately modeling monetary and erotic reward prediction errors
(PEMR and PEER). Note, however, that prediction error was relatively
correlated with outcome value, i.e., with the hedonic ratings (mean r !
0.72 for monetary rewards and mean r ! 0.75 for erotic rewards). This
correlation is inherent to the nature of these signals and is a classical
shortcoming of fMRI studies on reward processing (Hare et al., 2008). As
a consequence, if prediction errors and hedonic ratings were entered in
the same general linear model, they would both end up with a rather low Figure 3. Specific response of amygdala to erotic rewards. Activations are overlaid on an
explanatory power, because only the orthogonal component of each re- average anatomical scan of all subjects ( p " 0.05 FWE whole-brain corrected). Left and right
gressor would be allowed to compete for variance (Hunt, 2008) (such a plots of mean percent signal change, which are not independent of the whole-brain analysis,
model was estimated and produced poor results in expected brain re- are shown only to illustrate the specificity of amygdalar response. Error bars indicate SEM.
gions such as the ventral striatum and the OFC). For this reason, we built
a separate general linear model, in which reward outcomes were modu-
lated only by prediction errors (instead of probability and ratings). We condition (! value) divided by the mean activity of that ROI and multi-
should emphasize, however, that this procedure makes it difficult to plied by 100.
distinguish the contribution of prediction error and outcome value com-
putations at the brain level. Results
The resulting T-maps showing positive correlations between the Behavior
blood oxygenation level-dependent (BOLD) signal and monetary or Hit rates and reaction times (RTs), obtained at the time of the
erotic prediction errors were subsequently entered in a conjunction discrimination task, as well as hedonic ratings obtained at the
analysis (Nichols et al., 2005) thresholded at p " 0.001 uncorrected time of outcome, were analyzed in separate three-way ANOVAs
for multiple comparisons at the voxel level. including reward type, probability, and intensity as within-
Region of interest analyses. Region of interest (ROI) analyses were con- subject factors. The analysis on hit rates and RT was performed
ducted with the extension of SPM MarsBaR (http://marsbar.sourceforge.
on 17 subjects only, because data were accidentally lost for one
net/) within ROIs defined functionally from the whole-brain analyses.
Each ROI was created by taking the intersection of the functional cluster subject. The mean hit rate across subjects on the discrimination
of interest and a 10-mm-radius sphere centered on the highest peak voxel task was 96%.
of the cluster (to isolate distinct brain areas pertaining to the same clus- There was no significant main effect of reward type on hit
ter). In keeping with the approach of MarsBaR, percent signal change for rates ( p ! 0.38) and RT ( p ! 0.20), suggesting that monetary
a given condition in a given ROI was calculated as the effect size of that gains and erotic pictures had comparable incentive values.
Sescousse et al. • Reward Value Coding in the Human OFC
Figure 4. Response pattern of the reward-specific brain regions as a function of reward intensity. Percent signal change is plotted in the circled ROIs for monetary and erotic rewards according
to the following conditions: high intensity, low intensity, and no reward. In each region, brain activity increases with reward intensity only for the reward for which it is specific. Error bars indicate
SEM. The signal is averaged across the right and left hemispheres in each brain region (similar patterns of activity were observed in each hemisphere). Ant., Anterior; Post., posterior.
Subjects were faster (F(1,16) ! 34.2, p " 0.001) and more ac-
curate (F(1,16) ! 7.7, p " 0.05) for high intensity incentives and
were also faster for more likely rewards (F(2,32) ! 5.3, p " 0.05)
(Fig. 1 B). These results reflect increased motivation for higher
reward intensity and more certain rewards. They also confirm
that subjects were engaged in the task and effectively encoded the
cue information. Importantly, these effects were similar for mon-
etary and erotic rewards, as shown by the absence of significant
interaction between intensity and reward type (hit rate, p ! 0.67;
RT, p ! 0.20) (supplemental Fig. 2, available at www.jneurosci.
org as supplemental material) and between probability and re-
ward type (RT, p ! 0.11).
No significant effect of reward type was observed on the he-
donic ratings ( p ! 0.40), suggesting that monetary and erotic
rewards had similar subjective values. Conversely, we found a
robust main effect of intensity on the ratings (F(1,17) ! 150.8, p "
0.001), which remained significant for each type of reward taken
separately [Tukey’s honestly significant difference (HSD) tests:
monetary rewards, T(17) ! 20.4, p " 0.001; erotic rewards,
T(17) ! 4.4, p " 0.001] (Fig. 1C). This shows that, for both re-
wards, the two intensity categories chosen a priori (high vs low)
were effectively perceived by the subjects. The ratings also showed
an interaction between reward type and intensity (F(1,17) ! 111.5,
p " 0.001), simply because the subjects used a smaller portion of the
scale to rate erotic pictures. Finally, the ratings were not influenced
by reward probability (F(2,34) ! 1.4, p ! 0.26), confirming that they
reflected a purely hedonic evaluation.
Neuroimaging data
Reward-specific brain regions
Brain regions specific for each type of reward were identified
based on the forward inference approach proposed by Henson
(2006) to demonstrate qualitative differences in brain imaging
data. Specifically, money-specific regions were defined as those
stemming from the comparison MR # ER, masked inclusively
with MR # C and exclusively with ER # C, and conversely erotic-
specific regions were defined as those responding in the compar-
ison ER # MR, masked inclusively with ER # C and exclusively
with MR # C. This procedure ensures that the resulting brain areas
J. Neurosci., September 29, 2010 • 30(39):13095–13104 • 13099
meet two criteria: (1) they are more activated by one reward com-
pared with the other (“dissociation” criterion); (2) they respond to
either monetary or erotic rewards, but not to both, compared with a
common control condition (“association” criterion).
As hypothesized, monetary rewards specifically recruited the
anterior lateral OFC (MNI [x y z] [$30, 51, 0], T ! 5.92; [30, 54,
$3], T ! 6.80), spanning the anterior orbital gyrus, the lateral
orbital gyrus, and the ventral part of the middle frontal gyrus (Fig.
2). In contrast, erotic rewards elicited activity specifically in the
posterior part of the lateral OFC ([$30, 33, $15], T ! 7.56; [30,
33, $15], T ! 7.54), straddling the posterior and lateral orbital
gyri (Fig. 2). These results demonstrate a double dissociation
between monetary/erotic rewards and the anterior/posterior
OFC, which is further illustrated in the bar graphs of Figure 2,
representing the MR # C and ER # C differences in percent
signal change extracted from these regions. Among erotic-
specific areas, a large cluster was also present in the medial OFC
([$6, 45, $15], T ! 8.90), encompassing the medial orbital gy-
rus, the straight gyrus, and the most ventral part of the superior
frontal gyrus (Fig. 2). Subcortically, the only structure specifically
activated by erotic pictures was the bilateral amygdala ([$21, $6,
$27], T ! 6.94; [24, 0, $27], T ! 5.45) (Fig. 3). Other money-
specific and erotic-specific foci are reported in supplemental Ta-
bles 1 and 2 (available at www.jneurosci.org as supplemental
material), respectively.
This segregated representation of reward types in the OFC was
not merely attributable to visual or hedonic differences between
monetary and erotic outcomes, because an identical dissociation
emerged when we repeated the same analysis using no-reward
outcomes instead of reward outcomes. That is, when comparing
the NoMR and NoER conditions, which only differ with respect
to the type of reward being expected (while comparing visually
identical scrambled pictures), the anterior OFC specifically re-
sponded to no-monetary outcomes ([$30, 51, 0], T ! 6.16; [33,
51, 0], T ! 6.20), whereas the posterior OFC specifically re-
sponded to no-erotic outcomes ([$21, 33, $12], T ! 6.39) (sup-
plemental Fig. 3, supplemental Tables 1, 2, available at www.
jneurosci.org as supplemental material). This result excludes a mere
13100 • J. Neurosci., September 29, 2010 • 30(39):13095–13104 Sescousse et al. • Reward Value Coding in the Human OFC

perceptual account of the functional disso-

ciation observed in the OFC and supports
the idea that monetary and erotic rewards
are encoded in distinct OFC regions.
Moreover, ROI analyses on reward
intensity coding brought additional evi-
dence in support of a segregated represen-
tation of monetary and erotic outcomes.
For both rewards, we extracted the per-
cent signal change for the high reward,
low reward, and no-reward conditions
(Fig. 4). The results show that, whereas
activity in the anterior OFC increased with
monetary reward intensity, such a mono- Figure 5. Brain regions reflecting hedonic ratings regardless of reward type. Activations show the brain areas in which activity
tonic variation was not present for increas- positively correlates with both monetary and erotic ratings (intersection of T-maps thresholded at p # 0.01 FDR whole-brain
ing levels of erotic reward intensity. corrected shown in yellow or thresholded at p # 0.05 FDR whole-brain corrected shown in red). The regions displayed in red were
Conversely, in the posterior OFC, medial used as an inclusive mask in the analysis of Figure 6. Activations are overlaid on an average anatomical scan of all subjects.
OFC, and amygdala, activity was found to
increase monotonically with erotic reward
intensity but not with monetary reward in-
tensity. Together, these findings indicate
that reward-specific brain regions only re-
flected intensity for the reward type they
specifically encoded.
Finally, to further confirm the specificity
of these regions with respect to the subjec-
tive experience of monetary and erotic out-
comes, we performed a whole-brain
analysis comparing the coding of their he-
donic value. For each reinforcer, reward and
no-reward outcomes were pooled together,
while the hedonic value was modeled in a
parametric regressor with the correspond-
ing continuous rating (from 1 to 9) or a 0,
respectively. As expected, the contrast of
these parametric regressors between the two
rewards revealed the same brain regions as
the previous categorical analysis. In particu-
lar, BOLD activity in the right anterior lat-
eral OFC was found to scale best with
monetary hedonic value, whereas activity in
the posterior lateral OFC, medial OFC, and
amygdala was found to scale best with erotic
hedonic value (supplemental Fig. 4, avail- Figure 6. Common reward brain regions. A, T-map showing the brain regions encoding experienced reward value for both
able at www.jneurosci.org as supplemental monetary and erotic reward outcomes. Activations are overlaid on an average anatomical scan of all subjects ( p # 0.05 FWE
material). whole-brain corrected). B, Percent signal change is plotted in the circled ROIs for monetary and erotic rewards according to the
following conditions: high intensity, low intensity, and no reward. Note that these plots are not independent of the whole-brain
Common reward brain regions analysis and are only shown as an illustration for easier visual comparison with Figure 4. Error bars indicate SEM. The signal is
To identify the brain regions commonly averaged across the right and left hemispheres in each brain region (similar patterns of activity were observed in each hemisphere).
activated by monetary and erotic out- Ant., Anterior.
comes, we first compared each reward
with the control condition and performed 39], T " 8.42), and the anterior insula ([!27, 21, !6], T " 7.48;
a conjunction of these two comparisons (supplemental Fig. 5, [33, 24, 3], T " 8.14) (Fig. 6 A). Other foci are reported in sup-
available at www.jneurosci.org as supplemental material). Be- plemental Table 3 (available at www.jneurosci.org as supplemen-
cause this analysis remains sensitive to non-reward-related com- tal material).
putations such as attentional or image processing effects, it was The relative coding of reward intensity in these regions was
masked by the brain regions responding parametrically with the further illustrated using the same ROI approach as with reward-
subjective value of both monetary and erotic rewards (i.e., hedo- specific regions (Fig. 6 B). Note, however, that this representation
nic ratings) (Fig. 5). This procedure revealed significant bilateral is not independent of the previous whole-brain analysis, because
activations in a set of brain regions classically involved in reward reward intensity is highly correlated with the hedonic ratings that
processing: the ventral striatum ([!12, 9, !9], T " 6.38; [9, 6, have served in the identification of the “common network.” Con-
!9], T " 6.20), the midbrain ([!3, !24, !24], T " 7.30), the sequently, the resulting bar graphs are purely illustrative of the
anterior cingulate cortex (ACC) [!6, 27, 39], T " 8.15; [9, 18, conclusion drawn from the T-map, which is that activity in the
Sescousse et al. • Reward Value Coding in the Human OFC
Figure 7. Brain regions reflecting prediction errors regardless of reward type. Activations result from a conjunction analysis that it is organized hierarchically, whereby
showing the brain regions in which activity positively correlates with both monetary and erotic prediction errors (p # 0.001 cognitive control involving temporally
uncorrected). Activations are overlaid on an average anatomical scan of all subjects.
ventral striatum, midbrain, ACC, and anterior insula reflects ex-
perienced reward value regardless of reward type.
Coding of reward prediction errors
Reward-related brain regions are thought to mediate prediction
error signals, coding the difference between expected outcomes
and those effectively delivered (McClure et al., 2003; O’Doherty
et al., 2003b; Schultz, 2006). These prediction error signals occur
not only in conditioning procedures and can also be computed in
non-learning situations, such as in the present study (Dreher et
al., 2006; Yacubian et al., 2006). To determine whether such sig-
nals are supported by similar brain networks for monetary and
erotic rewards, the fMRI data were fitted with parametric regres-
sors modeling positive prediction errors related to reward out-
comes. Our results revealed that monetary and erotic reward
prediction errors were processed in a similar set of brain regions,
essentially overlapping with the previously identified common
network. Specifically, a conjunction analysis showed that activity
in the ventral striatum ([!9, 9, !6], T " 3.98; [6, 9, !9] T "
4.57), anterior insula ([!33, 18, !18], T " 4.80; [39, 21, !15],
T " 3.37), and rostral ACC ([!3, 42, 15], T " 5.57) correlated
positively with prediction errors regardless of reward type (Fig. 7)
(supplemental Table 4, available at www.jneurosci.org as supple-
mental material). Moreover, the comparison of monetary and
erotic reward prediction errors revealed almost no activity in
reward-related areas (at p # 0.001 uncorrected, only a tiny cluster
appeared in the left pallidum for the contrast of erotic minus
monetary prediction errors) and especially not in the brain re-
gions labeled as “reward specific” (even at a liberal threshold of
p # 0.05).
Discussion
As predicted, the OFC was found to be functionally organized
along a postero-anterior axis with respect to reward type, with the
anterior part responding exclusively to money and the posterior
part responding exclusively to erotic stimuli. Additional erotic-
specific activations were also found in the bilateral amygdala and
medial OFC. Importantly, brain activity in these reward-specific
regions only scaled with the hedonic value of the reward they
specifically encoded. In parallel, our results support the idea of a core
reward system processing experienced rewards regardless of their
nature. In this common network, including the ventral striatum,
ACC, anterior insula, and midbrain, functional activity correlated
with hedonic value and prediction error for both monetary gains
and erotic pictures. Together, our results reveal the existence of both
reward-specific and nonspecific brain networks, challenging the
view of a unique reward system for all reinforcers.
J. Neurosci., September 29, 2010 • 30(39):13095–13104 • 13101
Postero-anterior dissociation in the
orbitofrontal cortex
The segregated responses to monetary
and erotic outcomes along a postero-
anterior axis in the OFC suggest a func-
tional division of experienced reward
value representation according to an ab-
stractness gradient. Paralleling this result
in the domain of cognitive control, recent
theories on the functional divisions of the
human lateral prefrontal cortex proposed
proximate and concrete action represen-
tations is supported by posterior lateral
prefrontal regions, and cognitive control
involving temporally extended and ab-
stract representations is supported by more anterior lateral pre-
frontal regions such as the frontopolar cortex (Koechlin and
Summerfield, 2007; Badre, 2008; Dreher et al., 2008a). Our study
shows that a similar unifying principle of caudo-rostral hierar-
chical organization can be applied to the OFC. Notably, patients
with lesions in the anterior OFC have been reported to be specif-
ically impaired in making decisions entailing abstract, i.e., dis-
tant, consequences, and not in making decisions leading to
concrete, i.e., immediate, consequences, further supporting a
postero-anterior trend in the representation of abstractness in the
OFC (Bechara and Damasio, 2005).
Although an anatomical gradient in the postero-anterior
axis of the OFC had been suggested based on cytoarchitectonic
data (Ongür and Price, 2000; Wise, 2008), its functional rele-
vance for reward processing had never been tested empirically.
Our data bring strong empirical support to this hypothesis.
Although a generalization to all primary and secondary re-
wards cannot be ascertained in a single fMRI study, it is con-
sistent with the pattern of activation observed in the recent
literature for other rewards (Kringelbach and Rolls, 2004). In
particular, the medial and posterior lateral OFC, responding
to erotic pictures in the current and previous studies (Ponseti
et al., 2006), were shown to respond to other primary rewards,
such as attractive faces (O’Doherty et al., 2003a), pleasant
odors (Gottfried et al., 2006), and pleasant taste (Small et al.,
2001). Conversely, other studies manipulating monetary (Re-
uter et al., 2005; Vollm et al., 2007) or social (Izuma et al.,
2008) rewards have reported a similar anterior OFC region as
the one we found.
One intrinsic property of erotic pictures is that they are the
reward, whereas monetary rewards delivered in the scanner
are a representation of what the participant will receive at the
end of the experiment. Thus, it could be argued that the OFC
dissociation relates to the immediate rewarding effect of erotic
pictures compared with the delayed rewarding effect of mon-
etary gains. This is unlikely to be the case, because (1) our
pattern of OFC activation was not observed for immediate
versus delayed rewards in intertemporal choice studies (Kable
and Glimcher, 2007; Prevost et al., 2010) and (2) the same
OFC functional dissociation emerged when monetary and
erotic rewards were expected but not effectively delivered.
This finding also suggests that the segregated representation of
reward types in the OFC was not merely attributable to visual
or intrinsic differences between monetary and erotic out-
comes, such as saliency or arousal.
13102 • J. Neurosci., September 29, 2010 • 30(39):13095–13104
Electrophysiological recordings in monkeys indicate that
OFC neurons encode the economic value assigned to different
rewarding juices when choosing between them (Padoa-
Schioppa and Assad, 2008). Some OFC neurons were also
found to encode taste responses reflecting the identity of a
chosen juice. However, no neuronal recording study has yet
investigated whether primary and secondary rewards (e.g.,
juice vs social dominance) are coded in distinct OFC subre-
gions. Our finding of a clear OFC dissociation in humans
supports a hierarchical organization along a continuum from
the posterior to the anterior part of the OFC. Whether and
how these distinct representations of value in the OFC are
preserved across species remains an important open question.
Our results also suggest that learning through secondary rein-
forcement may depend more on anterior OFC regions,
whereas primary reinforcement may depend more on the pos-
terior OFC. This hypothesis, which remains to be tested, may
shed light on studies across a range of animal species indicat-
ing effects of OFC lesions on behavior maintained or acquired
through secondary reinforcement (Murray et al., 2007).
In addition, our findings clarify how value signals in the
OFC are integrated with those from other brain structures.
Together with the posterior and medial OFC, we found that
the amygdala responded exclusively to erotic pictures. Previ-
ous neuroimaging studies also reported that erotic pictures
evoke amygdala response (Redouté et al., 2000; Karama et al.,
2002; Hamann et al., 2004), whereas money, as a secondary
reinforcer, often failed to do so in studies using experimental
designs similar to the present paradigm (Knutson et al., 2001).
This is consistent with the underlying anatomy, showing that
the amygdala is more connected with the posterior and medial
OFC than with the anterior OFC (Carmichael and Price,
1995). Moreover, the parametric modulation of the amygdala
with erotic hedonic value is in accordance with its general role
in emotional arousal for both appetitive and aversive stimuli.
However, amygdala response to erotic stimuli is not solely
determined by arousal, as suggested by a previous study re-
porting higher amygdala response in men than in women
viewing sexual stimuli, despite similar arousal ratings in both
groups (Hamann et al., 2004).
Brain regions common to monetary and erotic rewards
An important strength of our experimental design, compared
with previous reward studies, is that it made it possible to directly
test whether monetary and erotic reward outcomes are truly en-
coded within the same brain regions. The enhanced response to
increasing hedonic value observed in the common network re-
gardless of reward type suggests that this network processes ex-
perienced reward value in a general manner. This is consistent
with its reported implication (in separate studies) in the hedonic
processing of rewards ranging from primary reinforcers, such as
sexual stimuli (Redouté et al., 2000), attractive faces (Bray and
O’Doherty, 2007; Smith et al., 2010), and pleasant taste (Small et
al., 2001), to secondary rewards, such as money or social approval
(Izuma et al., 2008). Our common network activity is also com-
patible with an interpretation in terms of general arousal or sa-
liency (Zink et al., 2004).
To efficiently compare the goal values of different rewards
during decision making, it has been proposed that the brain may
convert them into a common neural currency (Montague and
Berns, 2002). A wealth of electrophysiological and fMRI studies
has since confirmed this hypothesis, emphasizing in particular
the role of the ventral striatum and vmPFC (Kable and Glimcher,
Sescousse et al. • Reward Value Coding in the Human OFC
2007; Knutson et al., 2007; Chib et al., 2009). The present results
suggest that a similar computation might be performed in the
same brain network at the time of reward consumption. Indeed,
along with the ventral striatum, activity in the vmPFC also re-
flected the hedonic experience of the participants regardless of
reward type (Fig. 5). Such a mechanism encoding heterogeneous
outcome values on a common scale might be helpful for efficient
comparison of these values during subsequent value-based deci-
sion making.
Finally, we found prediction-error-related activity in the
ventral striatum, anterior insula, and ACC for both monetary
and erotic stimuli. This finding demonstrates that prediction
errors are computed in a network independent of reward type
and generalizes this concept to the domain of erotic stimuli,
paralleling previous work performed with monetary gains
(Yacubian et al., 2006), pleasant taste (O’Doherty et al.,
2003b), and attractive faces (Bray and O’Doherty, 2007).
Thus, prediction errors may be a primitive neural signal com-
puted in midbrain dopaminergic neurons regardless of reward
type and primarily delivered to the common reward network
that may be responsible for making predictions. However, it is
important to emphasize that, because of the inherent correla-
tion existing between prediction error and hedonic value, it is
difficult to disentangle these computations at the brain level
(Behrens et al., 2008; Hare et al., 2008). Consequently, al-
though the response pattern observed in our common net-
work is consistent with both interpretations, we cannot
ascribe with certainty one role or the other to these regions.
Concerning the ventral striatum, however, a recent study aim-
ing to dissociate reward value from prediction error found
that activity in this region best correlated with the latter (Hare
et al., 2008). Although this study focused on the computation
of goal values in the context of decision making, and therefore
cannot be directly compared with the present one, it brings
evidence favoring the prediction error hypothesis in the ven-
tral striatum. Note that midbrain activity was observed in our
common network, but not in the prediction error analysis,
whereas single-neuron recordings classically report prediction
error signals in midbrain dopaminergic structures (Schultz,
2006). Although further work is needed to gain a better un-
derstanding of the relationship between dopaminergic neuron
firing and the BOLD signal observed in reward paradigms,
recent findings combining fMRI with FDOPA positron emis-
sion tomography measures of midbrain dopamine synthesis
(Dreher et al., 2008b) and analyses of variations in genes in-
volved in dopamine transmission established a link between
higher prefronto-striatal BOLD signal and dopamine synaptic
availability during reward processing (Dreher et al., 2009).
Conclusion
Our results provide plausible functional mechanisms explaining
the existence of two separate reward networks in the brain. The
nature of their interactions remains to be determined, but one
possibility is that outcome value signals computed in the reward-
specific OFC regions would be sent to the common network for
additional integration and comparison processes. From an evo-
lutionary perspective, the distinct cytoarchitectonic properties of
the anterior and posterior parts of the OFC suggest that the ability
to process primary rewards may occur phylogenetically and on-
togenetically earlier than the ability to process secondary rewards,
which represent more evolved adaptive behavior. Our findings
also have important clinical implications for a range of neuropsy-
chopathological disorders characterized by major deficits in mo-
Sescousse et al. • Reward Value Coding in the Human OFC
tivation and behavioral control, such as pathological gambling or
hypersexuality. The dissociable representation of various rewards
along a postero-anterior axis in the OFC may shed light on this
important question.
References
Abler B, Walter H, Erk S, Kammerer H, Spitzer M (2006) Prediction error as
a linear function of reward probability is coded in human nucleus accum-
bens. Neuroimage 31:790 –795.
Badre D (2008) Cognitive control, hierarchy, and the rostro-caudal organi-
zation of the frontal lobes. Trends Cogn Sci 12:193–200.
Bechara A, Damasio A (2005) The somatic marker hypothesis: A neural
theory of economic decision. Games Econ Behav 52:336 –372.
Beck AT, Beck RW (1972) Screening depressed patients in family practice. A
rapid technic. Postgrad Med 52:81– 85.
Behrens TE, Hunt LT, Woolrich MW, Rushworth MF (2008) Associative
learning of social value. Nature 456:245–249.
Bray S, O’Doherty J (2007) Neural coding of reward-prediction error sig-
nals during classical conditioning with attractive faces. J Neurophysiol
97:3036 –3045.
Caldú X, Dreher JC (2007) Hormonal and genetic influences on processing
reward and social information. Ann NY Acad Sci 1118:43–73.
Carmichael ST, Price JL (1995) Limbic connections of the orbital and
medial prefrontal cortex in macaque monkeys. J Comp Neurol
363:615– 641.
Chib VS, Rangel A, Shimojo S, O’Doherty JP (2009) Evidence for a common
representation of decision values for dissimilar goods in human ventro-
medial prefrontal cortex. J Neurosci 29:12315–12320.
Deaner RO, Khera AV, Platt ML (2005) Monkeys pay per view: adaptive
valuation of social images by rhesus macaques. Curr Biol 15:543–548.
Dreher JC (2009) Decomposing brain signals involved in value-based deci-
sion making. In: Handbook of reward and decision making (Dreher JC,
Tremblay L, eds), pp 137–164. New York: Academic/Elsevier.
Dreher JC, Kohn P, Berman KF (2006) Neural coding of distinct statistical
properties of reward information in humans. Cereb Cortex 16:561–573.
Dreher JC, Schmidt PJ, Kohn P, Furman D, Rubinow D, Berman KF (2007)
Menstrual cycle phase modulates reward-related neural function in
women. Proc Natl Acad Sci U S A 104:2465–2470.
Dreher JC, Koechlin E, Tierney M, Grafman J (2008a) Damage to the
fronto-polar cortex is associated with impaired multitasking. PLoS
One 3:e3227.
Dreher JC, Meyer-Lindenberg A, Kohn P, Berman KF (2008b) Age-related
changes in midbrain dopaminergic regulation of the human reward sys-
tem. Proc Natl Acad Sci U S A 105:15106 –15111.
Dreher JC, Kohn P, Kolachana B, Weinberger DR, Berner KF (2009) Varia-
tion in dopamine genes influences responsivity of the human reward
system. Proc Natl Acad Sci U S A 106:617– 622.
FitzGerald TH, Seymour B, Dolan RJ (2009) The role of human orbitofron-
tal cortex in value comparison for incommensurable objects. J Neurosci
29:8388 – 8395.
Gomes CM, Boesch C (2009) Wild chimpanzees exchange meat for sex on a
long-term basis. PLoS One 4:e5116.
Gottfried JA, Small DM, Zald DH (2006) The chemical senses. The orbito-
frontal cortex, pp 125–171. Oxford: Oxford UP.
Hamann S, Herman RA, Nolan CL, Wallen K (2004) Men and women differ
in amygdala response to visual sexual stimuli. Nat Neurosci 7:411– 416.
Hammers A, Allom R, Koepp MJ, Free SL, Myers R, Lemieux L, Mitchell TN,
Brooks DJ, Duncan JS (2003) Three-dimensional maximum probability
atlas of the human brain, with particular reference to the temporal lobe.
Hum Brain Mapp 19:224 –247.
Hare TA, O’Doherty J, Camerer CF, Schultz W, Rangel A (2008) Dissociat-
ing the role of the orbitofrontal cortex and the striatum in the computa-
tion of goal values and prediction errors. J Neurosci 28:5623–5630.
Hayden BY, Parikh PC, Deaner RO, Platt ML (2007) Economic principles
motivating social attention in humans. Proc Biol Sci 274:1751–1756.
Henson R (2006) Forward inference using functional neuroimaging: disso-
ciations versus associations. Trends Cogn Sci 10:64 – 69.
Hoon E, Chambless D (1998) Sexual arousability inventory and sexual
arousability inventory— expanded. In: Handbook of Sexuality-
Related Measures, pp 71–74. Washington, DC: American Psychiatric
Association.
J. Neurosci., September 29, 2010 • 30(39):13095–13104 • 13103
Hunt LT (2008) Distinctive roles for the ventral striatum and ventral pre-
frontal cortex during decision-making. J Neurosci 28:8658 – 8659.
Izuma K, Saito DN, Sadato N (2008) Processing of social and monetary
rewards in the human striatum. Neuron 58:284 –294.
Kable JW, Glimcher PW (2007) The neural correlates of subjective value
during intertemporal choice. Nat Neurosci 10:1625–1633.
Karama S, Lecours AR, Leroux JM, Bourgouin P, Beaudoin G, Joubert S,
Beauregard M (2002) Areas of brain activation in males and females
during viewing of erotic film excerpts. Hum Brain Mapp 16:1–13.
Knutson B, Fong GW, Adams CM, Varner JL, Hommer D (2001) Dissocia-
tion of reward anticipation and outcome with event-related fMRI. Neu-
roreport 12:3683–3687.
Knutson B, Taylor J, Kaufman M, Peterson R, Glover G (2005) Distributed
neural representation of expected value. J Neurosci 25:4806 – 4812.
Knutson B, Rick S, Wimmer GE, Prelec D, Loewenstein G (2007) Neural
predictors of purchases. Neuron 53:147–156.
Koechlin E, Summerfield C (2007) An information theoretical approach to
prefrontal executive function. Trends Cogn Sci 11:229 –235.
Kriegeskorte N, Simmons WK, Bellgowan PS, Baker CI (2009) Circular
analysis in systems neuroscience: the dangers of double dipping. Nat Neu-
rosci 12:535–540.
Kringelbach ML, Rolls ET (2004) The functional neuroanatomy of the hu-
man orbitofrontal cortex: evidence from neuroimaging and neuropsy-
chology. Prog Neurobiol 72:341–372.
McClure SM, Berns GS, Montague PR (2003) Temporal prediction errors in
a passive learning task activate human striatum. Neuron 38:339 –346.
McClure SM, Ericson KM, Laibson DI, Loewenstein G, Cohen JD (2007)
Time discounting for primary rewards. J Neurosci 27:5796 –5804.
Montague PR, Berns GS (2002) Neural economics and the biological sub-
strates of valuation. Neuron 36:265–284.
Murray EA, O’Doherty JP, Schoenbaum G (2007) What we know and do
not know about the functions of the orbitofrontal cortex after 20 years of
cross-species studies. J Neurosci 27:8166 – 8169.
Nichols T, Brett M, Andersson J, Wager T, Poline JB (2005) Valid conjunc-
tion inference with the minimum statistic. Neuroimage 25:653– 660.
O’Doherty J, Winston J, Critchley H, Perrett D, Burt DM, Dolan RJ (2003a)
Beauty in a smile: the role of medial orbitofrontal cortex in facial attrac-
tiveness. Neuropsychologia 41:147–155.
O’Doherty JP, Dayan P, Friston K, Critchley H, Dolan RJ (2003b) Temporal
difference models and reward-related learning in the human brain. Neu-
ron 38:329 –337.
Ongür D, Price JL (2000) The organization of networks within the orbital
and medial prefrontal cortex of rats, monkeys and humans. Cereb Cortex
10:206 –219.
Padoa-Schioppa C, Assad JA (2008) The representation of economic value
in the orbitofrontal cortex is invariant for changes of menu. Nat Neurosci
11:95–102.
Peters J, Büchel C (2010) Neural representations of subjective reward value.
Behav Brain Res 213:135–141.
Plassmann H, O’Doherty J, Rangel A (2007) Orbitofrontal cortex encodes
willingness to pay in everyday economic transactions. J Neurosci
27:9984 –9988.
Ponseti J, Bosinski HA, Wolff S, Peller M, Jansen O, Mehdorn HM, Büchel C,
Siebner HR (2006) A functional endophenotype for sexual orientation
in humans. Neuroimage 33:825– 833.
Prevost C, Pessiglione M, Météreau E, Clery-Melin ML, Dreher JC (2010)
Distinct valuation subsystems in the human brain for effort and delay. J
Neurosci, in press.
Rangel A, Camerer C, Montague PR (2008) A framework for studying the neu-
robiology of value-based decision making. Nat Rev Neurosci 9:545–556.
Redouté J, Stoléru S, Grégoire MC, Costes N, Cinotti L, Lavenne F, Le Bars D,
Forest MG, Pujol JF (2000) Brain processing of visual sexual stimuli in
human males. Hum Brain Mapp 11:162–177.
Reuter J, Raedler T, Rose M, Hand I, Gläscher J, Büchel C (2005) Patholog-
ical gambling is linked to reduced activation of the mesolimbic reward
system. Nat Neurosci 8:147–148.
Reynolds CF 3rd, Frank E, Thase ME, Houck PR, Jennings JR, Howell JR,
Lilienfeld SO, Kupfer DJ (1988) Assessment of sexual function in de-
pressed, impotent, and healthy men: factor analysis of a brief sexual func-
tion questionnaire for men. Psychiatry Res 24:231–250.
Rhodes G (2006) The evolutionary psychology of facial beauty. Annu Rev
Psychol 57:199 –226.
13104 • J. Neurosci., September 29, 2010 • 30(39):13095–13104
Schultz W (2006) Behavioral theories and the neurophysiology of reward.
Annu Rev Psychol 57:87–115.
Small DM, Zatorre RJ, Dagher A, Evans AC, Jones-Gotman M (2001)
Changes in brain activity related to eating chocolate: from pleasure to
aversion. Brain 124:1720 –1733.
Smith DV, Hayden BY, Truong TK, Song AW, Platt ML, Huettel SA (2010)
Distinct value signals in anterior and posterior ventromedial prefrontal
cortex. J Neurosci 30:2490 –2495.
Sugrue LP, Corrado GS, Newsome WT (2005) Choosing the greater of two
goods: neural currencies for valuation and decision making. Nat Rev
Neurosci 6:363–375.
Thornhill R, Gangestad SW (1999) Facial attractiveness. Trends Cogn Sci
3:452– 460.
Sescousse et al. • Reward Value Coding in the Human OFC
Tremblay L, Schultz W (1999) Relative reward preference in primate or-
bitofrontal cortex. Nature 398:704 –708.
Völlm B, Richardson P, McKie S, Elliott R, Dolan M, Deakin B (2007) Neuronal
correlates of reward and loss in Cluster B personality disorders: a functional
magnetic resonance imaging study. Psychiatry Res 156:151–167.
Wise SP (2008) Forward frontal fields: phylogeny and fundamental func-
tion. Trends Neurosci 31:599 – 608.
Yacubian J, Gläscher J, Schroeder K, Sommer T, Braus DF, Büchel C (2006)
Dissociable systems for gain- and loss-related value predictions and errors
of prediction in the human brain. J Neurosci 26:9530 –9537.
Zink CF, Pagnoni G, Martin-Skurski ME, Chappelow JC, Berns GS (2004)
Human striatal responses to monetary reward depend on saliency. Neu-
ron 42:509 –517.
 !
!
ABOUT)THIS)COMPENDIUM)
!
!
The$original$purpose$of$this$compendium$has$been$for$the$use$in$my$own$lectures$in$consumer$
neuroscience$and$neuromarketing$at$the$Copenhagen$Business$School.$However,$I$also$
recognise$that$this$volume$can$also$be$a$potentially$valuable$resource$for$both$newcomers$as$
well$as$experienced$people$within$this$discipline.$Neuromarketing$is$today$very$much$a$
conglomerate$of$divergent$solutions;$hyped$up$talks;$and$a$mixture$of$true$science$and$pop$
science$gone$terribly$wrong.$This$collection$of$papers$represent$my$own$take$on$what$the$
basics$should$entail$
!
This$book$is$also$intended$as$a$supplement$to$my$book$“Introduction$to$Neuromarketing$&$
Consumer$Neuroscience”,$which$you$can$read$more$about$here:$http://neuronsinc.com/
publications/introductionPtoPneuromarketingPconsumerPneuroscience/$(also$see$next$page).$
!
The$selection$of$texts$are$not$intended$to$be$an$exhaustive$listing$of$all$relevant$articles.$I$have$
worked$from$two$basic$premises:$1)$that$the$article$is$available$freely$on$the$web;$and$2)$that$
the$article$represents$some$of$the$leading$thoughts$(and$scholars)$in$this$field.$$
!
If$you$have$suggestions$or$comments,$please$send$me$an$email$at$tzramsoy@gmail.com$$
!
!
DISCLAIMER)
!
All$materials$in$this$compendium$–$texts$and$images$–$have$either$been$collected$from$freely$
available$resources,$or$written$by$myself.$I$claim$no$ownership$or$rights$over$these$materials.$
All$materials$in$this$compendium$–$except$my$own$freely$distributed$materials$–$can$be$
collected$and$compiled$by$any$individual.$Please$note$that$there$may$be$restrictions$on$
materials$in$this$compendium$for$sharing,$distributing$or$selling.$$
!
If$you$find$that$this$compendium$contains$materials$that$are$not$permitted$for$sharing,$please$
send$me$an$email$at$tzramsoy@gmail.com$and$I$will$adjust$accordingly.$
!
!
!
Happy$reading!$
!
!
All$the$best,$
$
!
!
!
!
!
!
!
v"2.0"–"August"2014

 WHO)MADE)THIS?)
!
$
Thomas"Zoëga"Ramsøy,"b."1973"in"Oslo,"Norway"
!
Thomas$is$considered$one$of$the$leading$experts$on$
neuromarketing$and$consumer$neuroscience,$and$he$is$an$
innovator$by$heart.$With$a$background$in$economics$and$
neuropsychology,$he$holds$a$PhD$in$neurobiology$from$the$
University$of$Copenhagen.$$
!
Thomas$has$published$extensively$on$the$application$of$
neuroimaging$and$neurophysiology$to$consumer$behaviour$and$
decision$making.$He$is$the$Director$of$the$Center$for$Decision$
Neuroscience,$where$his$research$team$uses$an$eclectic$mix$of$
technologies$and$the$sciences$of$economics,$$psychology$and$
neuroscience.$Beyond$this,$Thomas$is$the$CEO$of$Neurons$Inc,$
where$he$consults$companies$around$the$globe$on$the$use$of$
science$and$technology$in$business.$
!
!
!
More$information$about$Thomas$can$be$found$on$the$following$resources:$
!
Professional)pages)
DNRG$CBS$–$http://cbs.dk/DNRG$$
DNRG$HH$–$http://drcmr.dk/research/DecisionNeuroscience$$
Neurons$Inc$–$http://neuronsinc.com$$
!
Social)media)
Twitter$–$https://twitter.com/NeuronsInc$$
Neurons$Inc$–$http://NeuronsInc.com$$
BrainEthics$–$http://brainethics.org$
!
Societies)
Neuromarketing$Science$&$Business$Association$–$http://www.neuromarketingPassociation.com$$
Society$for$Mind$Brain$Sciences$–$http://mbscience.org/$
!
Publications)
ResearchGate$–$https://www.researchgate.net/profile/Thomas_Z_Ramsoy/$$
!
!
!
!