ArticlePDF Available

Abstract and Figures

Human decisions can be shaped by predictions of emotions that ensue after choosing advantageously or disadvantageously. Indeed, anticipating regret is a powerful predictor of future choices. We measured brain activity using functional magnetic resonance imaging (fMRI) while subjects selected between two gambles wherein regret was induced by providing information about the outcome of the unchosen gamble. Increasing regret enhanced activity in the medial orbitofrontal region, the anterior cingulate cortex and the hippocampus. Notably, across the experiment, subjects became increasingly regret-aversive, a cumulative effect reflected in enhanced activity within medial orbitofrontal cortex and amygdala. This pattern of activity reoccurred just before making a choice, suggesting that the same neural circuitry mediates direct experience of regret and its anticipation. These results demonstrate that medial orbitofrontal cortex modulates the gain of adaptive emotions in a manner that may provide a substrate for the influence of high-level emotions on decision making.
Content may be subject to copyright.
Regret and its avoidance: a neuroimaging study of
choice behavior
Giorgio Coricelli
1
, Hugo D Critchley
2
, Mateus Joffily
1
, John P O’Doherty
2
, Angela Sirigu
1
& Raymond J Dolan
2
Human decisions can be shaped by predictions of emotions that ensue after choosing advantageously or disadvantageously.
Indeed, anticipating regret is a powerful predictor of future choices. We measured brain activity using functional magnetic
resonance imaging (fMRI) while subjects selected between two gambles wherein regret was induced by providing information
about the outcome of the unchosen gamble. Increasing regret enhanced activity in the medial orbitofrontal region, the anterior
cingulate cortex and the hippocampus. Notably, across the experiment, subjects became increasingly regret-aversive, a cumulative
effect reflected in enhanced activity within medial orbitofrontal cortex and amygdala. This pattern of activity reoccurred just
before making a choice, suggesting that the same neural circuitry mediates direct experience of regret and its anticipation. These
results demonstrate that medial orbitofrontal cortex modulates the gain of adaptive emotions in a manner that may provide a
substrate for the influence of high-level emotions on decision making.
Cost-benefit analyses in everyday decision making are often difficult
because our evidence about future outcomes is incomplete or, at
best, probabilistic. According to standard economic theory, rational
decision makers should optimize their choice strategies through
reliance on expected utility. However, it is known that human decisions
deviate from this ideal and are influenced by other, less rational,
considerations. For instance, the Dutch postal code lottery is popular,
although playing the lottery can be considered an irrational behavior.
Its success has been explained by the possibility that people anticipate
how bad they would feel if, not having bought a ticket, their postal
code is drawn.
A sense of responsibility in human decision making operates
through a process of counterfactual reasoning
1,2
that enables us to
relate the outcome of a previous decision with what we would have
obtained had we opted for a rejected alternative. Regrets are what we
experience when this comparison is to our disadvantage
3–6
, and it can
be defined as a cognitively enriched emotion embodying a feeling of
responsibility for negative outcomes of choices
7,8
. This contrasts with
disappointment, which is an emotion related to an unexpected negative
outcome, without an obligatory sense of personal responsibility
9,10
.As
regret is an unpleasant feeling that encapsulates a sense of responsi-
bility, we learn from past experience to minimize its likely reoccurrence
when considering a new choice decision
11
.
The interplay between decision making and emotional processing
can be assumed to involve a contribution from several brain structures,
including areas associated with executive and emotional processing.
Several lines of evidence indicate that orbitofrontal cortex (OFC)
assigns relative value to stimuli and updates the salience of primary
and secondary reinforcers
12–14
. Both simple and abstract complex
instrumental reinforcers such as monetary gain and loss evoke
emotions that serve to guide behavior, and the OFC is a
candidate substrate for the generation of such emotions
15–18
.We
would also suggest that the OFC modulates the gain of emotions
using a top-down process in which a paradigmatic cognitive process,
specifically counterfactual thinking, contributes to an emotional
response and ensuing choice behavior. Evidence that high-level emo-
tions such as regret depend on a specific neuroanatomical substrate
comes from a study showing that the normal expression of this
cognitively based emotion depends on the integrity of OFC
19
.Speci-
fically, patients with selective lesions to anterior medial OFC do not
experience regret and, unlike healthy controls, are unable to adjust their
behavior to avoid regret-inducing situations. In order to define pre-
cisely the conditions under which the OFC and related areas are
engaged by the experience and anticipation of regret and to determine
how the latter experience influences learning, we used functional
magnetic resonance imaging (fMRI) while subjects performed a
gambling task. We demonstrated that activity in the OFC mediates
the experience of adaptive emotions such as regret. Moreover, learning
to anticipate this emotion during choice reactivated the OFC in
conjunction with the amygdala.
On each trial, the subject viewed two gambles where different
probabilities involving financial gain or loss were represented by the
relative size of colored sectors of a circle (Fig. 1). There were two kinds
of trials, each with a different type of feedback information indicating
financial gain or loss for the subject. In the ‘partial feedback’ condition,
only the outcome of the selected gamble was shown, whereas in the
complete feedback’ condition, the outcome of both the selected and
unselected gambles were available to the subject. Complete feedback
Published online 7 August 2005; doi:10.1038/nn1514
1
Neuropsychology Group, Institut des Sciences Cognitives, Centre National de la Recherche Scientifique, 67 Boulevard Pinel 69675, Bron, France.
2
Wellcome Department
of Imaging Neuroscience, 12 Queen Square, London, WC1N 3BG, UK. Correspondence should be addressed to R.J.D. (rdolan@fil.ion.ucl.ac.uk) or A.S. (sirigu@isc.cnrs.fr).
NATURE NEUROSCIENCE VOLUM E 8
[
NUMBER 9
[
SEPTEMBER 2005 1255
ARTICLES
© 2005 Nature Publishing Group http://www.nature.com/natureneuroscience
trials enabled subjects to judge not only the financial consequence of
their decision but also the outcome they would have achieved had they
selected the unselected option.
As a sense of responsibility is critical to the experience of regret, we
compared subjects responses when they had a choice (the choose’
condition) with their responses when they had no choice, but rather
followed a computer-selected choice (the ‘follow’ condition), thus
removing any feeling of responsibility.
RESULTS
Ventral striatum response to wins and losses
Agency is reflected in cognitive and physiological engagement of
subjects. During task performance, subjects physiological responses
(heart rate) were significantly higher in ‘choose trials than in ‘follow’
trials (P ¼ 0.001). During fMRI, the processing of outcome was
modulated as a function of whether outcome (wins or losses) was
evaluated in the context of a ‘choose or ‘follow’ trial (that is, whether
subjects were agents). We found activation of anterior ventral striatum
during wins and a relative deactivation during losses solely in ‘choose
trials (Fig. 2), highlighting the dependency of reward-related signaling
in this region on instrumentality
20
. This pattern of activity in ventral
striatum for ‘choose’ trials is consistent with a ‘reward prediction error
response
21
insofar as in ‘follow’ trials (where choice was computer-
selected, meaning the subject had no agency) there is no need for
prediction
22
. In other words, this area processes mismatches between
predicted and actual outcome and is activated when an outcome is
better than expected and relatively deactivated in the alternative case. In
light of this agency effect, we restricted our subsequent analyses of
outcome-related activity to choose’ trials alone.
Disappointment and regret
The psychological and behavioral impact of outcome (wins and losses)
was influenced by the amount of feedback information provided to
subjects. Disappointment arises when, on a selected gamble, the
alternative outcome is more positive than an experienced outcome.
The magnitude of disappointment (that is, the discrepancy between the
unobtained outcome’ and actual outcome of the selected gamble
correlated with enhanced activity in middle temporal gyrus and dorsal
brainstem (including periaqueductal gray matter), a region implicated
in processing aversive signals such as pain
23
(Ta ble 1 a).
Regret represents an emotion based on counterfactual processing,
but it differs from disappointment in its abstract point of reference.
At outcome
2.0 s
2.0 s
4.0 s
4.0 s
1.0 s
At choice
Analysis of imaging data
CC
CF
'Choose'
'Choose'
'Follow'
'Follow'
Complete
feedback
feedback
PC
PF
Partial
1.0 s
4.0 s
4.0 s
2.0 s
2.0 s
12 + 1 (dummy)
trials per run
16 runs divided in
two sections: A & B
200 –200 –5050 200 –200 –5050 200 –200 –5050
200 –200 –5050 200 –200 –5050 200 –200 –5050
Trial 12Trial 11Trial 2Trial 1 Dummy
PC CC PF CF CC PF PC CF PF CC CF PC CF PF PC CC
Ventral striatum
0.8
0.6
0.4
0.2
0
–0.2
–0.4
–0.6
–0.8
–1.0
Win Loss Win Loss
Choose Follow
Relative % change in
BOLD signal
VStr VStr
Figure 1 Experimental design. On each trial, the
subject viewed two gambles where different
probabilities of financial gain or loss were
represented by the relative size of colored sectors
of a circle. The preferred gamble was indicated by
the subject by means of a left or right button
press. Once selected, the chosen gamble was
highlighted on the screen by a green square. A
rotating arrow then appeared in the center of the
gamble circle, stopping after 4 s. The outcome of
the selected gamble, indicated by the resting
position of the arrow, resulted in financial gain or
loss for the subject. Half of the trials were ‘choose’
trials; in half of those, only the outcome of the
selected gamble was given to the subject (‘partial
feedback choose’, PC). In the other half, the
outcome of both selected and unselected gambles
were available (‘complete feedback choose’, CC).
An equal number of trials were ‘follow’ trials, in
which the subject was informed that the computer
would randomly choose one of the two gambles. A
green square appeared behind one of the two
gambles, and the subject had to press a button on
the corresponding side. Follow trials were likewise
divided into complete feedback (‘complete
feedback follow’, CF) and partial feedback (‘partial
feedback follow’, PF) trials.
Figure 2 Activity at outcome is related to win and loss. Activity within the
striatum, encompassing regions of ventral striatum, discriminated between
financial gain and loss at trial outcome. This effect, however, was significant
only for ‘choose’ trials, in which the subject was responsible for the choice
(that is, when the subject rather than the computer selected between two
gambles). Group data (thresholded at P o 0.001, uncorrected) is plotted on
sagittal and coronal sections of a normalized canonical template brain.
Striatal activations (VStr) were centered on MNI coordinates (8, 18, 0),
(6, 18, 2) and (12, 24, 8). The bottom panel plots the average parameter
estimates (± s.e.m.) for relative difference in BOLD activity at outcome for
wins and losses in ‘choose’ and ‘follow’ trials.
1256 VOLUME 8
[
NUMBER 9
[
SEPTEMBER 2005 NATURE NEUROSCIENCE
ARTICLES
© 2005 Nature Publishing Group http://www.nature.com/natureneuroscience
Regret arises from a discrepancy between the actual outcome and an
outcome that would have pertained had an alternative choice been
taken. In our experiment, regret is represented as the difference between
the outcome of the unselected gamble and the outcome of the selected
gamble; hence, it occurs only on complete feedback trials. Across both
win and loss trials, the magnitude of the difference in unselected and
selected outcomes correlated with enhanced activity within anterior
cingulate, putamen, inferior parietal lobule (which is implicated in
processing number comparison
24
) and lateral OFC (Ta bl e 1 b).
Effect of the unobtained outcome
Activity in medial OFC clearly discriminated between trials resulting in
a win or a loss, but only in the complete feedback condition (Fig. 3). We
modeled the corresponding outcome of the nonselected gamble as a
parametric regressor for each outcome (win or loss); in partial feedback
trials, we used the unobtained outcome of the chosen gamble as a
parametric regressor. By adopting such a procedure, we could dis-
sociate comparative assessments underlying counterfactual thinking.
The counterfactual process in the complete feedback condition in
losing trials resulted in greater activation of bilateral regions of medial
OFC, and greater deactivation for winning trials. Notably, in this
analysis, losses in the complete feedback condition resulted in the
emotion of regret (with the exception of the case in which the subject
lost 50 and felt relief at forgoing an outcome of 200), whereas wins
resulted in relief (with the exception of the case in which the subject
won 50 and felt regret for a forgone outcome of +200; Fig. 3).
Activity in medial OFC, extending from subgenual cingulate, corre-
lated with the degree of regret (that is, the difference between the
outcome of the unchosen gamble and the obtained outcome) in the
choose’/complete feedback condition (referred to as the complete
choose’ condition). These data are plotted in Figure 4a, showing the
change in magnitude of the fitted response in OFC relative to the degree
of regret and relief experienced. In addition, anterior hippocampus
activity correlated with regret. From this analysis, we extracted data on
brain activity in response to the obtained outcomes of 50 and +50 as
a function of the outcome of the unselected gamble (200 and +200)
in the complete feedback condition (Fig. 4b). Notably, activity in the
OFC, dorsal anterior cingulate cortex (ACC) and anterior hippo-
campus discriminated between the two unobtained outcomes.
More specifically, these areas were activated when the actual outcome
was compared with a more advantageous forgone outcome (+200),
leading to regret, and are relatively deactivated when the same
actual outcome is compared with a less advantageous alternative
(200), leading to relief. Self-reported emotional ratings (in the
practice session before scanning) were consistent with this result in
that the ratings did not simply reflect wins or losses on the selected
gambles but rather were strongly influenced by the provision of
information regarding outcome of the alternative nonselected gamble
(Fig. 4c). These results suggest that there are three main regions that
contribute to the experience of regret: dorsal anterior cingulate cortex,
medial OFC and anterior hippocampus.
Choice behavior and brain activity
During the fMRI experiment, we predetermined the presentation of
individual gambles and the outcome of each gamble such that subjects
were exposed, in both ‘choose’ and ‘follow’ trials, to a range of positive
and negative outcomes and, for complete feedback trials, were exposed
to alternative outcomes that provided either positive and negative
values. However, for ‘choose’ trials, the selection of individual gambles
was determined by the subject, and the individual behavior in this
pattern of selection provided data on subject sensitivity to wins and
losses and counterfactual information.
It is indeed counterfactual thinking between the obtained and the
unobtained outcomes that modulates the experience of regret and
disappointment. When the subject obtains an outcome that is lower
than expected, he or she might feel disappointment. The greater the
difference between the expected and the obtained outcome, the more
intense is this negative feeling. Thus, the subject can avoid future
disappointment by choosing a gamble that minimizes a difference
between lowest and highest outcome, weighted by the probability of the
worst possible outcome.
We tested a model of choice (see Methods) using the data from the
partial-feedback/‘choose’ condition (the ‘partial choose condition)
that incorporated the effects of anticipating disappointment in addi-
tion to the maximization of the expected values. The result from a
regression analysis (Ta bl e 2a ) shows that subjects chose maximizing
Table 1 Activity at outcome
(a) Activity related to the comparison between the unobtained and the obtained
outcome of the selected gamble (‘partial choose’ condition)
Location Side Coordinates Z-score
Midbrain (periaqueductal gray region) 2, 34, 20 3.84
Precentral gyrus S2 L 60, 4, 8 3.74
Subcallosal gyrus L 4, 0, 8 3.61
Middle temporal gyrus L 42, 0, 20 3.54
(b) Activity related to the comparison between the outcome of the unselected gamble
and the obtained outcome of the selected gamble (‘complete choose’ condition)
Anterior cingulate cortex 10, 24, 34 4.68
Putamen L –14, 0, 6 3.92
Lateral OFC R 42, 42, –18 3.77
Inferior parietal lobule R 54, –50, 36 3.67
Win Loss Win Loss
PartialComplete
–0.6
–0.4
–0.2
0
0.2
0.4
0.6
Relative % change in
BOLD signal
Orbitofrontal cortex
OFC
Figure 3 The effect of the unobtained outcome: counterfactual processing of
value. Orbitofrontal cortex (OFC) activity for the comparison between the
outcome of the selected gamble and the alternative outcome in the ‘complete
choose’ versus ‘partial choose’ conditions (right). For the ‘complete choose’
condition we modeled the corresponding outcome of the nonselected gamble
as a parametric regressor for each actual outcome (win or loss), whereas in
partial feedback trials, we used the unobtained outcome of the chosen
gamble as a parametric regressor. The counterfactual process between losses
(or wins) and any forgone outcome in the complete condition resulted in
much greater activation (or deactivation) of bilateral regions of medial OFC.
Error bars indicate s.e.m. In the partial feedback condition, we observed only
a relative deactivation of the OFC when actual losses were compared with
unobtained outcome in the chosen gamble. Group data (thresholded at P o
0.001, uncorrected) is plotted on an axial section of a normalized canonical
template brain (left panel). Peak orbitofrontal activity occurred at MNI
coordinates (8, 32, 14).
NATURE NEUROSCIENCE VOLUM E 8
[
NUMBER 9
[
SEPTEMBER 2005 1257
ARTICLES
© 2005 Nature Publishing Group http://www.nature.com/natureneuroscience
expected values (P o 0.001) and minimizing future disappointment
(P ¼ 0.001). Notably, brain activity in motor and premotor cortex,
anterior cingulate cortex and superior parietal cortex related to a pre-
ceding choice of maximal expected value (Table 3a). Choice behavior
that deviates from the maximization of expected values, motivated by
avoidance of future disappointment, produced activity in the substan-
tia nigra, a dopaminergic midbrain area involved in anticipatory
reward (reward prediction, refs. 25,26; Supplementary Figure 1).
This activity is related to the activity at outcome observed when
subjects evaluated wins and loss in terms of prediction error (Fig. 2).
The feeling of regret depends on the comparison between the
obtained and the unobtained outcome across the two gambles, a
comparison possible only in the complete feedback condition. The
greater the difference between the outcome of the unchosen gamble
and the obtained outcome, the more intense the feeling of regret. We
parameterized regret as the absolute value of the difference between the
lowest and the highest outcome across gambles. Thus, the subject can
avoid future regret by choosing a gamble that minimizes this difference
(see equation 3 in Methods). In our analysis of subjects’ behavior in the
complete choose condition, we considered anticipated disappointment,
anticipated regret and maximization of the expected values as the main
components of subjects choices. Results from a regression analysis
–50 +50
–50 +50–50 +50–50 +50
Obtained outcomeObtained outcomeObtained outcomeObtained outcome
1
3
5
7
9
Very bad
Very good
Emotional evaluation
–200
+200
–200
+200
Unobtained
outcome
Unobtained
outcome
–0.5
0.0
0.5
Relative % change in
BOLD signal
Relative % change in
BOLD signal
OFC ACC Hippocampus
11223344
Regret Relief
–0.4
–0.2
0.0
0.2
0.4
0.6
0.8
Data across subjects
OFC
ab c
Figure 4 Regret and relief. (a) Change in magnitude of fitted response in the medial OFC relative to the degree of regret or relief experienced in the CC
condition. Medial OFC activity (10, 30, 12) correlates with the level of regret experienced. Regret and relief are measured as the positive and negative
differences, respectively, between the outcome of the unselected gamble and the actual outcome. The relative change in BOLD signal between unobtained
and actual outcomes is plotted for each of the four levels of regret (numbered 1–4, from lowest to highest regret). The first value in each pair representsthe
unobtained outcome and the second value represents the actual outcome: 1: [50, 50]; 2: [200, 50], [50, 200]; 3: [200, 50], [50, 200]; and 4:
[200, 200]. The four levels of relief are the result of the following comparisons (numbered 1–4, as for regret): 1: [50, 50]; 2: [200, 50], [50, 200]; 3:
[200, 50], [50, 200]; and 4: [200, 200]. (b) Left plot: medial OFC activity; center: anterior cingulate cortex activity (coordinates 8, 32, 24); and right:
hippocampal activity (30, 10, 12) in processing the comparison between 50 and +50 obtained with a forgone outcome of 200 or + 200 in the CC
condition. Error bars, s.e.m. (c) Mean emotional evaluation (± s.e.m.) measured in the practice session before scanning for two obtained outcomes (50 or 50)
as a function of the outcome of the unchosen gamble (200 or 200) in the complete feedback condition.
Table 2 Regression analysis (panel logit procedure with individual
random effect)
(a) Subjects’ choice behavior as a function of anticipated disappointment (d)and
maximization of the expected value (e) in the ‘partial choose’ condition.
Variable name Coefficient Standard error ZP
Constant 0.22014 0.13002 1.69 0.09
d 0.00259 0.0008 3.23 0.001
e 0.01464 0.00106 13.79 0.000
Number of subjects ¼ 15;numberofobservations¼ 720. Log likelihood ¼331.4631,
Wald w
2
(2) ¼ 193.71, Prob 4w
2
¼ 0.000. The dependent variable ‘choice is equal to 1 if
subject chose gamble 1 and equal to 0 if subject chose gamble 2.
(b) Subjects’ choice behavior as a function of anticipated disappointment (d),
maximization of the expected value (e) and anticipated regret (r) in the ‘complete
choose’ condition.
Variable name Coefficient Standard error ZP
Constant 0.06958 0.09706 0.72 0.473
d 0.00211 0.00137 1.62 0.106
r 0.00463 0.00114 4.03 0.000
e 0.00726 0.00184 3.94 0.000
Number of subjects ¼ 15;numberofobservations¼ 720. Log likelihood ¼331.1425;
Wald w
2
(3) ¼ 205.82, Prob 4w
2
¼ 0.000. The dependent variable ‘choice’ is equal to 1
if subject chose gamble 1 and is equal to 0 if subject chose gamble 2.
Table 3 Activity at choice
(a) Activity preceding choice of maximal expected values in the ‘partial choose’
condition
Location Side Coordinates Z-score
Motor cingulate cortex L 12, 10, 46 4.85
Premotor cortex L 34, 20, 62 4.08
Motor/posterior cingulate cortex R 14, 28, 54 4.43
Genual anterior cingulate cortex R 18, 36, 8 4.04
Superior parietal cortex R 42, 52, 52 3.73
(b) Activity preceding choice of maximal expected values in the ‘complete choose’
condition
Lateral occipital cortex L 38, 76, 64.43
Lateral geniculate nucleus L 20, 31, 6 3.89
Somatomotor cortex L 28, 28, 52 3.98
Superior parietal cortex R 24, 56, 50 3.96
Parahippocampus L 32, 2, 26 3.29
Mid cingulate cortex 6, 4, 32 4.36
(c) Influence of cumulative experience of regret (CR) on choice-related activity
Somatomotor cortex R 68, 12, 20 5.19
Inferior parietal lobule L 38, 42, 40 4.8
Medial OFC L 10, 40, 24 4.24
Amygdala L 8, 4, 24 4.21
(d) Influence of immediately prior experience of regret (t –1)onchoice
Dorsolateral prefrontal cortex R 46, 28, 38 4.99
Lateral OFC R 42, 26, 16 4.72
Inferior parietal lobule R 54, 58, 48 4.62
1258 VOLUME 8
[
NUMBER 9
[
SEPTEMBER 2005 NATURE NEUROSCIENCE
ARTICLES
© 2005 Nature Publishing Group http://www.nature.com/natureneuroscience
(panel logit procedure with individual random effect, Ta ble 2b)show
that subjects chose maximizing expected values (P o 0.001) and
minimizing future regret (P o 0.001).
Anticipating disappointment (as defined in equation 2 in Methods)
would correspond to risk avoidance. The absence of this behavior (as
shown in Ta b l e 2 b for the variable d, P ¼ 0.11) in the complete choose
condition indicates a hierarchical relationship between risk and regret.
Indeed, the subjects chose minimizing regret independently of the risk
component in their choice responses.
How the experience of regret affects decision making
The experience of regret has a powerful influence on subsequent
behavioral choice, leading to a pattern of behavior that can be
characterized as regret-aversive. In our experiment, this was manifest
in two distinct ways. First, we observed a behavioral bias in subjects’
choices over the course of the gambling task away from choices that
mightengenderanemotionofregret(Table 2 b ). Second, regret aversion
was also evident in a bias away from choices that had previously led to
negative outcomes, on the basis of subjects’ own cumulative experience
of regret. The proportion of regret-avoiding choices increased over time
with the cumulative effect of the experience of regret (Fig. 5a).
On the basis of the above data, we determined how these behavioral
biases were expressed in patterns of neural activity at the time of
choice behavior. We found enhanced activity (during the epoch
between trial onset and subject response in the complete choose
condition) in the dorsal anterior cingulate cortex and in the substantia
nigra, when subjects chose minimizing regret over maximizing
expected values (Supplementary Figure 2). Both anterior cingulate
and midbrain activities are related to a reward anticipation process (in
terms of error prediction signal)
25–28
. Activity related to a preceding
choice of maximal expected values in the complete choose’ condition is
shown in Ta bl e 3 b.
We next assessed the effects of cumulative history using a reinforce-
ment-learning model based on past emotional experience (see Meth-
ods). For cumulative regret experience, we observed modulation of
choice-related activity in the medial OFC,
right somatomotor, inferior parietal lobule
and left amygdala (Fig. 5b and Ta ble 3c).
Notably, this expression of cumulative regret
at the time of decision making involved simi-
lar anatomical regions (medial OFC and ante-
rior medial temporal lobe) as that elicited by
regret at the time of outcome feedback. This
suggests that the experience-dependent influ-
ence of regret on decision making may be
supported by reactivation of processes med-
iating regret as a reactive emotion.
The more immediate experience of regret
in the preceding trial also influenced choice-
related activity, enhancing responses in right
dorsolateral prefrontal cortex, particularly
around the border between middle and infer-
ior frontal gyri, perhaps representing an influ-
ence of immediate regret on self-monitoring
at decision making. We also found that
enhanced activity during choice selection in
right inferior parietal lobule and right caudo-
lateral OFC correlated similarly with the mag-
nitude of regret experienced in the preceding
trial (Tabl e 3d and Fig. 5c).
DISCUSSION
Regret is a complex emotion based on a counterfactual process that
juxtaposes the outcome of choices we make with a better outcome for a
rejected alternative. We show that activity in response to this negative
emotion is distinct from activity seen for mere outcome evaluation. In
our brain imaging data, the influence of personal responsibility on the
processing of outcome was evident in contrasting outcome-related
activity for ‘choose trials (where the subject selected which gamble to
play’) with ‘follow trials (where the choice’ was computer-selected).
In accordance with psychological theory
5,11
, we also nd a neuroana-
tomical dissociation of regret versus disappointment. Thus, outcome
evaluation is influenced by the level of responsibility in the process of
choice (agency) and by the available information regarding alternative
outcomes (complete or partial feedback). The level of regret, calculated
in terms of the magnitude of the difference between the forgone
outcome and the obtained outcome, was strongly correlated with
activity in the medial OFC.
In a number of studies (including tasks in which outcome is not
dependent on operant action) medial OFC activity reflects reward
attainment
17,18,29,30
. This has been interpreted as suggesting that
medial OFC may support positive emotions (and that lateral OFC
may support emotions with negative valence). Nevertheless, other
neuroimaging studies highlight a more complex role in reinforcement
representations that is also suggested by lesion data. Thus, enhance-
ment of medial OFC activity reflects devaluation in extinction
of conditioned aversive stimuli and inflation of aversive stimuli
12
.
Significant activity with monetary gain and loss has been reported in
both medial and lateral OFC
31
, whereas monetary gain in a probabil-
istic reversal task has been associated with activity in both medial and
lateral OFC. Similarly, lesions of medial OFC do not impair processing
of primary rewards but seem to interfere with relative reward dis-
crimination that includes conditions involving prospective and coun-
terfactual appraisal
19,32
. These findings point to a more complex
relationship in OFC than a simple medial-lateral specialization for
reward or punishment. Our data would suggest that cognitive context,
DLPFC
DLPFC
IPL
OFC
OFC
Amg
FinalMiddleInitial
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
Proportion
Anticipated regret
Blocks
ab
c
Figure 5 Activity at choice: learning from the experience of regret. (a) Proportion of choice (± s.e.m.)
related to anticipated regret in ‘complete choose’ trials. Anticipated regret increased over time as the
experiment proceeded. (b) Activity at choice reflecting cumulative regret. We found activity in the medial
left amygdala (Amg; coordinates 8to16, 4, 25) and medial OFC (10, 40, 24). Group data is
plotted on coronal and axial slices of a template brain in normal space at a threshold of P o 0.005,
uncorrected (for illustrative purposes in this figure only). (c) Activity reflecting prior regret at choice.
Individual subject analytic designs modeled the parametric modulation of activity during the epoch
between trial onset and subject response. Experience of regret in the preceding trial profoundly
influenced choice-related activity, enhancing responses in right dorsolateral prefrontal (DLPFC),
right lateral OFC, and inferior parietal lobule (IPL; Table 3c). Group activity is plotted at
P o 0.001, uncorrected.
NATURE NEUROSCIENCE VOLUM E 8
[
NUMBER 9
[
SEPTEMBER 2005 1259
ARTICLES
© 2005 Nature Publishing Group http://www.nature.com/natureneuroscience
exemplified by counterfactual thinking in relation to states of the world,
exerts a modulatory influence on OFC activation to reward and
punishment. On this basis, our neuroimaging data complement and
confirm results from lesion studies
19
that assign a critical role to the
medial OFC in experiencing and anticipating the emotion of regret.
To our knowledge, this is the first demonstration of involvement of
medial OFC and amygdala in choice behavior that reflects regret
avoidance. We suggest that, as a result of cumulative experience of
regret, inputs from these regions to decision-making processes provide
an updated representation of value that incorporates a weighting of the
relative emotional value of different options for choice
12,33
. In addition,
there is enhanced lateral OFC activity in the choice condition that
immediately follows the experience of regret, perhaps reflecting cogni-
tive processes associated with a need (across gamble comparisons) to
avoid future regrettable outcomes. Such processes are analogous to
those normally subsumed under procedures involving reversal learn-
ing, where subjects need to change behavioral strategies that are no
longer advantageous
34
. We also found enhanced activity in the parietal
cortex during choice selection, in accordance with findings in animal
studies related to action desirability
35
. Thus, the experience of regret
has a major impact on the process of choice that is expressed at two
levels, with a net result of biasing subjects to forgo choices that might
lead to future experience of this highly negative emotion.
Regret is also an emotion based on a declarative cognitive process
that requires an ability to assess the consequences of our actions within
conscious awareness. It is not surprising that there is involvement of
brain structures such as the hippocampus and dorsal anterior cingulate
regions that are critical to declarative memory (indicating what we
need to remember) and cognitive-induced arousal responses, respec-
tively
36,37
. Dorsal anterior cingulate activity is related to the experience
of regret at outcome and to regret-aversive decision making at choice.
In the context of decision making, regret involves an appraisal of
potential outcomes and is likely to evoke behavioral adjustments on
subsequent trials. Indeed, choice behavior in our complete task involves
either maximizing expected value in the face of potential regret, or
minimizing regret at a cost to expected value. Thus, the decision
process necessarily involves a degree of conflict. There is substantial
neuroimaging literature that highlights a role for dorsal anterior
cingulate and pre-supplementary motor area in response monitoring
and error detection
38–41
. Notably, dorsal anterior cingulate cortex
activity may also mediate an attentional focus on subjective emotional
states
42,43
and the cognitive and emotional processing engendered in
states of autonomic arousal
44,45
. Our findings would conform to a
unified model of error detection, response conflict and attendant
emotional arousal whereby anterior cingulate activation arises during
appraisal of conflicts (between maximizing expected value and mini-
mizing regret) in decision making that would include a signal for
potential behavioral adjustment.
However, the critical finding in our study concerns the role of OFC,
which we suggest integrates cognitive and emotional mechanisms after
a declarative process in which distinct counterfactual processes engen-
der a high-level emotion of regret. Our data suggest a mechanism by
which comparing a choice with its alternative outcome, along with the
associated feeling of responsibility, promotes behavioral flexibility and
exploratory strategies in dynamic environments so as to minimize the
likelihood of emotionally negative outcomes.
METHODS
Subjects. Fifteen healthy right-handed subjects were recruited to take part in a
study at the Wellcome Department of Imaging Neuroscience. These volunteers
gave fully informed consent for the project which was approved by the Joint
Ethics Committee of the National Hospitals and Institute of Neurology,
London. Each participant was screened to exclude medication and conditions
including psychological or physical illness or history of head injury. Mean age
of participants was 23.3 years.
Experimental design and task. Each participant underwent fMRI scanning
while performing a total of 192 trials of the experimental task illustrated in
Figure 1. During scanning, the subject viewed a projection of a computer
screen and made a two-choice button-press response with the right index or
middle finger. The task was adapted from refs. 5 and 19 and involved stimuli
resembling ‘wheels of fortune’. On each trial, the subject viewed two gambles
where different probabilities of financial gain or loss were represented by the
relative size of colored sectors of a circle.
Half the trials were designated choose trials in which the preferred gamble
was indicated by the subject by means of a left or right button press. Once
selected, the chosen gamble was highlighted on the screen by a green square,
and a rotating arrow then appeared in the center of the gamble circle, stopping
after 4 s. The outcome of the selected gamble, indicated by the resting position
of the arrow, resulted in financial gain or loss. There were two kinds of choose
trials: in 48 of the 96 ‘choose trials, the outcome (and the spinning arrow)
appeared for the selected gamble alone (‘partial feedback’). In the other 48
choose’ trials, the spinning arrow and outcome of both the selected and
unselected gambles were visible to the subject (‘complete feedback’). Complete
feedback trials enabled the subject to judge not only the financial consequence
of their decision, but also the outcome had they selected the other option.
An equal number of trials were designated ‘follow trials, which were also
divided into complete and partial feedback trials. In ‘follow’ trials, the subject
was informed that the computer would randomly choose one of the two
gambles. A green square appeared behind one of the two gambles, and the
subject was required to press the corresponding button. The rotating arrow
then appeared, as in the ‘choose’ condition, in the selected gamble circles (or in
both circles, for the complete feedback condition), stopping to indicate the
outcome. The ‘follow’ condition was a control condition, as it eliminated the
responsibility element of decision making but still produced outcomes repre-
senting financial gains and losses.
Partial and complete feedback and ‘choose’ and ‘follow’ conditions were
presented in pseudo-randomized blocks of 12 stimuli. Null events were also in-
cluded in each block to allow estimation of low-level baseline brain activity and
to desynchronize timings of event types. Behavioral responses were logged by
means of a desktop computer running Cogent software on a Matlab platform.
Incentive procedure. Each subject was told that the outcome of both choose’
and ‘follow’ trials would result in financial gain or loss. The subjects were
informed before starting the experiment that they would be paid a show-up fee
of 20 GBP and an additional amount that would depend on their performance
in terms of cumulative points earned during the experiment, in both ‘choose
and follow’ conditions: 0 GBP, 5 GBP and 10 GBP corresponded to low,
medium and high cumulative earnings, respectively. No quantitative references
about the ranges of the cumulative earnings were provided. The subjects were
informed about their earnings and paid in cash outside the fMRI scanner. No
information about their cumulative outcomes was provided during the
experiment. For ethical reasons, we were not able to pay different amounts
to subjects; thus, in practice we programmed the computer to always assign 5
extra GBP. Every subject ended up with final earnings of 25 GBP.
Parameter structure. Each individual gamble presented paired combinations of
200, 50, 50 and 200 points and represented one of three levels of outcome
probability (0.2, 0.5 and 0.8). Displayed and actual probabilities were identical.
The two gambles always differed in their expected values (that is, probability
times the outcome of each alternative choice) and in the value of their actual
outcomes. There were six possible outcome pairs (outcome obtained in the
chosen gamble and outcome of the unselected gamble, or vice versa): (i) 50
and 200, (ii) 50 and 200, (iii) 50 and 200, (iv) 50 and 200, (v) 50 and 50
and (vi) 200 and 200. The set of pairs of gambles was the same for each
condition. The order of presentation was pseudo-random and differed for each
condition. In the ‘follow’ condition, the predetermined pattern of choices did
not resemble any particular choice strategy. The set of trials and the computers
pattern of choice in ‘follow’ conditions can be found in Supplementary Table 1.
1260 VOLUME 8
[
NUMBER 9
[
SEPTEMBER 2005 NATURE NEUROSCIENCE
ARTICLES
© 2005 Nature Publishing Group http://www.nature.com/natureneuroscience
Emotional rating. Before scanning, the subject read a standard description of
the task and was familiarized with the computerized task events. The subject
then performed a practice session covering the four trial types (‘complete
choose’, ‘partial choose’, ‘complete follow’ and ‘partial follow’) and were asked
to rate how they felt about each outcome on a nine-point scale (where 9
signaled a very good outcome and 1 very bad).
Physiological monitoring. Blood volume pressure (BVP) was recorded with a
Nonin 8600 Pulse Oximeter (Nonin Medical) and sampled at 300 Hz at the
same time as task performance and fMRI data acquisition. BVP signal and
fMRI scanning pulses were coregistered by means of an analog-to-digital
converter (CED1901) and Spike 3.3 software (CED). Heart rate was estimated
from BVP inter-beat intervals and resampled at 2 Hz for later analysis. We
considered physiological responses in each trial of two time windows of 3 s
each. The first time window (anticipatory) started when the stimulus was
presented at the beginning of each trial and included the choice (or follow)
event and the period of waiting for the outcome; the second time window
(feedback) corresponded to the feedback presentation.
Analysis of behavioral data. We tested (by regression analysis and the panel
logit procedure with individual random effect; Ta ble 2 ) a model of choice that
incorporates the effects of anticipating disappointment and regret in addition
to maximization of expected values. x
1
and y
1
represent the highest and the
lowest outcome of gamble 1 (g
1
), and x
2
and y
2
represent the highest and the
lowest outcome of gamble 2 (g
2
). The probability of x
1
is p and the probability
of y
1
is 1 p; the probability of x
2
is q, and the probability of y
2
is 1 q.The
probability of choosing gamble 1 is
Prðg
1it
Þ¼1 Prðg
2it
Þ¼F½d
it
; r
it
; e
it
ð1Þ
where i ¼ individual and t ¼ time. The function F[y] denotes the function
e
y
/(1 + e
y
). The variables d and r, as described in equations 2 and 3 indicate the
process of minimizing future disappointment and future regret, respectively;
e indicates the result of maximizing expected values.
d ¼ðjy
2
x
2
1 qÞÞ ðjy
1
x
1
1 pÞÞ ð2Þ
r ¼jy
2
x
1
jjy
1
x
2
3Þ
e ¼ EVðg
1
ÞEVðg
2
Þ¼ðpx
1
+ð1 pÞy
1
Þðqx
2
+ð1 qÞy
2
Þð4Þ
Subjects would choose g
1
, minimizing disappointment (equation 2), if the
difference in absolute value between the lowest and the highest possible
outcome in g
2
, weighted by the probability of the lowest outcome, were larger
than the same weighted difference in g
1
. The process of anticipating regret is
described by the minimization of the difference between the lowest and the
highest outcome across gambles (equation 3). Finally, subjects would choose g
1
if its expected value (EV) is higher than that of g
2
(equation 4). A restricted
version of the model, where we considered only the effect of anticipated
disappointment (d) and maximization of expected values (e), was tested with
the data from the ‘partial choose’ condition (Table 2a).
Functional imaging data: acquisition, pre-processing and analysis. Subjects
were scanned at 3 T (Siemens Allegra) performing the experimental task over
two counterbalanced sessions. T2-weighted echoplanar images, optimized for
blood oxygenation level–dependent (BOLD) contrast, were acquired (TE:
30 ms, TR
vol
:2.86s,44slicesangledat301 in anterior-posterior axis). A
preparation pulse (duration 1 ms, amplitude 2 mT/m) was used in the slice
selection direction to compensate for through-plane susceptibility gradients for
enhancement of imaging of orbitofrontal and medial temporal lobe regions
46
.
The efficient imaging of these regions (cross-hairs) is illustrated in Supple-
mentary Figure 3, depicting smoothed normalized EPI images and corre-
sponding locations in a normalized structural template image.
Image pre-processing and subsequent analyses were done using statistical
parametric mapping (SPM2; http://www.fil.ion.ucl.uk/~spm/SPM2.html) on a
Matlab platform. Images were initially realigned and unwarped, correcting for
motion artifacts. Differences in the timing of image slices across each individual
volume were corrected, and each volume was transformed into standard
stereotaxic space and smoothed with a Gaussian filter (full-width half-max-
imum 8 mm). Voxel-wide differences in BOLD contrast within the smoothed
normalized images resulting from the different task conditions and trial types
were examined using SPM. Standard neuroimaging methods using the
general linear model were used with the first level (individual subject
analyses) providing contrasts for group effects analyzed at the second level.
All individual analyses modeled the period of stimulus presentation
up to choice selection as a mini-epoch. Activity at outcome was likewise
modeled as a mini-epoch, segregating complete and partial feedback trials
for choose’ and ‘follow’ conditions. In the fMRI data analysis, the outcomes
of each trial were modeled as 3-s epochs beginning at 1 s before the
outcome display stopped (that is, when the spinning arrow began to slow,
enabling prediction of its end point) and continuing for a further 2 s
while the trial outcome was displayed (and processed). Choice-related neural
activity at the time of choice was studied during the epoch between trial onset
and subject response. To model the hemodynamic lag of the BOLD response
relative to underlying evoked neural activity, regressors for task effects were
obtained by convolving these mini-epochs with a canonical hemodynamic
response function.
We did not jitter timings within trials to discriminate overlapping ‘raw’
evoked hemodynamic BOLD responses related to the choice and outcome
phases (although some temporal variability arose from the time it took the
subject to select between gambles on choose’ conditions). Thus, the temporal
proximity (and lack of jittering) between the decision-making and outcome
components of each trial resulted in overlapping of BOLD responses between
these different phases. However, our principal analyses focused on modulation
of outcome-related activity by parameters not embodied within the decision or
anticipation phases (regret, disappointment, win/loss). Furthermore, contrasts
(Students t-tests) within our regression analyses identify activity that is not
otherwise accountable for in the analytic design. Thus, the impact of shared
variance in activity between choice and parametric modulators of outcome is
minimized or excluded in results from first- and second-level analyses by virtue
of this orthogonalization with analyses modeling decision and outcome epochs
independently of their parametric modulation.
Three sets of analyses were performed. In the first, outcome trials were
partitioned according to whether the subject experienced a loss or a win and
additional regressors parametrically modeling the degree of disappointment
and, for complete-feedback trials, regret experienced at the time of outcome. In
two further analyses, activity for win and loss trials were not modeled
separately, and parametrical regressors of activity at choice modeled experience
of regret on the preceding trial and mean regret experienced over the course of
the task prior to the choice, respectively. For group analysis of outcome-related
activity, second-level analyses of contrast for regret and disappointment for
different trial types were computed as ANOVAs with sphericity correction for
repeated measures. Post-hoc exploration of individual data is also reported to
illustrate specific effects as a function of different trial types. For choice-related
activity, contrasts relating to prior experience of regret were modeled for
complete and partial trials in a second-level ANOVA, taking into account group
effects of prior regret experienced. Adjusted activity represents BOLD signal
changes proportionally adjusted for the analytic model. Although general
threshold significance was set at P o 0.05, corrected, we tabulate group effects
at P o 0.001, uncorrected, to highlight regions of interest. For illustrative
purposes, we show Figure 5b at P o 0.005, uncorrected.
Modeling cumulative regret. The regressor to test activity at choice that
reflected cumulative regret (Fig. 5b and Ta bl e 3c) represented the difference
between missed payoff and payoff realized from past choice in the complete
feedback condition (‘complete choose’ condition (CC)) over time, that is,
CR
t
¼ (A
unobtained, t – 1
A
obtained, t – 1
), where CR is cumulative regret, t is trial,
A
obtained
is the average payoff realized and A
unobtained
is the average payoff of the
unselected gambles.
Note: Supplementary information is available on the Nature Neuroscience website.
ACKNOWLEDGMENTS
This work was supported by grants from the Human Frontier Science Program
(RGP 56/2005), the Action Concerte
´
e Incitative, Systemes Complexes from the
Centre National de la Recherche Scientifique to A.S. and G.C., the Coordenac¸a
˜
o
NATURE NEUROSCIENCE VOLUM E 8
[
NUMBER 9
[
SEPTEMBER 2005 1261
ARTICLES
© 2005 Nature Publishing Group http://www.nature.com/natureneuroscience
de Aperfeic¸oamento de Pessoal de
´
vel Superior to M.J., a Wellcome Trust
Programme Grant to R.J.D. and a Wellcome Senior Fellowship in Clinical Science
to H.D.C.
COMPETING INTERESTS STATEMENT
The authors declare that they have no competing financial interests.
Received 29 April; accepted 11 July 2005
Published online at http://www.nature.com/natureneuroscience/
1. Roese, N.J. & Olson, J.M. What Might Have Been: The Social Psychology of Counter-
factual Thinking (Erlbaum, Mahwah, New Jersey, USA, 1995).
2. Byrne, R.M.J. Mental models and counterfactual thinking. Trends Cogn. Sci. 6,426
445 (2002).
3. Kahneman, D. & Tversky, A. The psychology of preferences. Sci. Am. 246, 136–142
(1982).
4. Kahneman, D. & Miller, D. Norm theory: comparing reality to its alternatives. Psychol.
Rev. 93, 136–153 (1986).
5. Mellers, B., Schwartz, A. & Ritov, I. Emotion-based choice. J. Exp. Psychol. Gen. 128,
332–345 (1999).
6. Zeelenberg, M. & van Dijk, E. On the comparative nature of regret. in The Psychology
of Counterfactual Thinking (eds. Mandel, D., Hilton D. & Catelani, P.) 147–161
(Routledge, London, 2005).
7. Bell, D.E. Regret in decision-making under uncertainty. Oper. Res. 30, 961–981
(1982).
8. Loomes, G. & Sugden, R. Regret theory: an alternative theory of rational choice under
uncertainty. Econ. J. 92, 805–824 (1982).
9. Bell, D.E. Disappointment in decision making under uncertainty. Oper. Res. 33,127
(1985).
10. Loomes, G. & Sugden, R. Disappointment and dynamic inconsistency in choice under
uncertainty. Rev. Econ. Stud. 53, 271–282 (1986).
11. Zeelenberg, M. et al. Consequences of regret aversion: effects of expected feedback on
risky decision making. Organ. Behav. Hum. Decis. Process. 65, 148–158 (1996).
12. Gottfried, J.A. & Dolan, R.J. Human orbitofrontal cortex mediates extinction learning
while accessing conditioned representations of value. Nat. Neurosci. 7, 1144–1152
(2004).
13. Rolls, E.T. The orbitofrontal cortex and reward. Cereb. Cortex 10, 284–294 (2000).
14. Tremblay, L. & Schultz, W. Relative reward preference in primate orbitofrontal cortex.
Nature 398, 704–708 (1999).
15. Elliott, R. et al. Dissociable neural responses in human reward systems. J. Neurosci. 20,
6159–6165 (2000).
16. Kringelbach, M. & Rolls, E. The functional neuroanatomy of the human orbitofrontal
cortex: evidence from neuroimaging and neuropsychology. Prog. Neurobiol. 72,341
372 (2004).
17. Breiter, H.C., Ahron, I., Kahneman, D., Dale, A. & Shizgal, P. Functional imaging of
neural responses to expectancy and experience of monetary gains and losses. Neuron
30, 619–639 (2001).
18. O’Doherty, J. et al. Abstract reward and punishment representations in the human
orbitofrontal cortex. Nat. Neurosci. 4, 95–102 (2001).
19. Camille, N. et al. The involvement of the orbitofrontal cortex in the experience of regret.
Science 304, 1167–1170 (2004).
20. Corlett, P.R. et al. Prediction error during retrospective revaluation of causal associations
in humans: fMRI evidence in favor of an associative model of learning. Neuron 44,
877–888 (2004).
21. Schultz, W., Dayan, P. & Montague, P.R. A neural substrate of prediction and reward.
Science 275, 1593–1599 (1997).
22. Miceli, M. & Castelfranchi, C. The mind and the future: The (negative) power of
expectations. Theory Psychol. 12, 335–366 (2002).
23. Peyron, R., Laurent, B. & Garcia-Larrea, L. Functional imaging of brain responses to
pain. A review and meta-analysis. Neurophysiol. Clin. 30, 263–288 (2000).
24. Gobel, S.M., Johansen-Berg, H., Behrens, T. & Rushworth, M.F.S. Response-selection-
related parietal activation during number comparison. J. Cogn. Neurosci. 16, 1536–
1551 (2004).
25. Tobler, P.N., Fiorillo, C.D. & Schultz, W. Adaptive coding of reward value by dopamine
neurons. Science 307, 1642–1645 (2005).
26. Dreher, J.C., Kohn, P. & Berman, K.F. Neural coding of distinct statistical properties of
reward information in humans. Cereb. Cortex (in the press).
27. Holroyd, C.B. et al. Dorsal anterior cingulated cortex shows f
MRIresponse to internal and
external error signal. Nat. Neurosci. 7, 497–498 (2004).
28. Shidara, M. & Richmond, B.J. Anterior cingulated: single neuronal signal related to
degree of reward expectancy. Science 296, 1709–1711 (2002).
29. Berns, G.S., McClure, S.M., Pagnoni, G. & Montague, P.R. Predictability modulates
human brain response to reward. J. Neurosci. 21, 2793–2798 (2001).
30. Gottfried, J.A., O’Doherty, J. & Dolan, R.J. Encoding predictive reward value in human
amygdala and orbitofrontal cortex. Science 301, 1104–1107 (2003).
31. Elliott, R., Newman, J.L., Longe, O.A. & Deakin, J.F. Differential response patterns in the
striatum and orbitofrontal cortex to financial reward in humans: a parametric functional
magnetic resonance imaging study. J. Neurosci. 23, 303–307 (2003).
32. Bechara, A., Tranel, D. & Damasio, H. Characterization of the decision-making deficit of
patients with ventromedial prefrontal cortex lesions. Brain 123, 2189–2202 (2000).
33. Schoenbaum, G., Chiba, A.A. & Gallagher, M. Orbitofrontal cortex and basolateral
amygdala encode experience outcomes during learning. Nat. Neurosci. 1, 155–159
(1998).
34. Rolls, E.T., Hornak, J., Wade, D. & McGrath, J. Emotion-related learning in patients with
social and emotional changes associated with frontal lobe damage. J. Neurol. Neurosurg.
Psychiatry 57, 1518–1524 (1994).
35. Dorris, M.C. & Glimcher, P.W. Activity in posterior parietal cortex is correlated with the
relative subjective desirability of action. Neuron 44, 365–378 (2004).
36. Bush, G., Phan, L. & Posner, M.I. Cognitive and emotional influences in anterior
cingulate cortex. Trends Cogn. Sci. 4, 215–222 (2000).
37. Critchley, H.D. The human cortex responds to an interoceptive challenge. Proc. Natl.
Acad. Sci. USA 101, 6333–6334 (2004).
38. Carter, C.S., Botvinick, M.M. & Cohen, J.D. The contribution of the anterior cingulate
cortex to executive processes in cognition. Rev. Neurosci. 10, 49–57 (1999).
39. Kiehl, K.A., Liddle, P.F. & Hopfinger, J.B. Error processing and the rostral anterior
cingulate: an event-related fMRI study. Psychophysiology 37, 216–223 (2000).
40. Garavan, H., Ross, T.J., Murphy, K., Roche, R.A. & Stein, E.A. Dissociable executive
functions in the dynamic control of behavior: inhibition, error detection, and correction.
Neuroimage 17, 1820–1829 (2002).
41. Kerns, J.G. et al. Anterior cingulate conflict monitoring and adjustments in control.
Science 303, 1023–1026 (2004).
42. Lane, R.D. et al. Neural correlates of levels of emotional awareness. Evidence of an
interaction between emotion and attention in the anterior cingulate cortex. J. Cogn.
Neurosci. 10, 525–535 (1998).
43. Phan, K.L., Liberzon, I., Welsh, R.C., Britton, J.C. & Taylor, S.F. Habituation of rostral
anterior cingulate cortex to repeated emotionally salient pictures. Neuropsychopharma-
cology 28, 1344–1350 (2003).
44. Critchley, H.D. et al. Activity in the human brain predicting differential heart rate
responses to emotional facial expressions. Neuroimage 24, 751–762 (2005).
45. Critchley, H.D. et al. Human cingulate cortex and autonomic control: converging
neuroimaging and clinical evidence. Brain 126, 2139–2152 (2003).
46. Deichmann, R., Gottfried, J.A., Hutton, C. & Turner, R. Optimized EPI for fMRI studies of
the orbitofrontal cortex. Neuroimage 19, 430–441 (2003).
1262 VOLUME 8
[
NUMBER 9
[
SEPTEMBER 2005 NATURE NEUROSCIENCE
ARTICLES
© 2005 Nature Publishing Group http://www.nature.com/natureneuroscience
... You choose one of two alternative queues (queue I) and then find out that the other queue (queue II) is moving much faster than you expected. In this case, the advantage you would compute for your choice of queue is negative, appropriately signaling a missed opportunity (Bennett, Davidson, et al., 2021;Coricelli et al., 2005;Loomes & Sugden, 1982), but you would experience no prediction error with respect to this queue since it is moving as expected. Moreover, you will in fact experience a positive prediction error with respect to the state 'about to choose a queue', regardless of whether it is your queue or the alternative queue that is faster than expected. ...
... This suggestion can explain not only the general involvement of emotion in how we evaluate and adjust our actions, but also more subtle empirical observations. First, it inherently explains why we feel bad when an action we could have taken, but didn't, is found to have better outcomes than expected (Coricelli et al., 2005;Loomes & Sugden, 1982). In addition, it explains why we typically have stronger emotional responses to outcomes of actions that we don't normally take (Kahneman & Miller, 1986;Kutscher & Feldman, 2019). ...
Article
Full-text available
Emotions ubiquitously impact action, learning, and perception, yet their essence and role remain widely debated. Computational accounts of emotion aspire to answer these questions with greater conceptual precision informed by normative principles and neurobiological data. We examine recent progress in this regard and find that emotions may implement three classes of computations, which serve to evaluate states, actions, and uncertain prospects. For each of these, we use the formalism of reinforcement learning to offer a new formulation that better accounts for existing evidence. We then consider how these distinct computations may map onto distinct emotions and moods. Integrating extensive research on the causes and consequences of different emotions suggests a parsimonious one-to-one mapping, according to which emotions are integral to how we evaluate outcomes (pleasure & pain), learn to predict them (happiness & sadness), use them to inform our (frustration & content) and others’ (anger & gratitude) actions, and plan in order to realize (desire & hope) or avoid (fear & anxiety) uncertain outcomes.
... Our version of the two-box task therefore provides a novel means to further illuminate the consequences and correlates of children's regret. For example, our task could be used to investigate whether an experience of regret over a poor choice can encourage better future choices in children, as it does in adults [6,[30][31][32]. Previous studies employing the two-box task have found that children who appear to feel regret tend to make better subsequent decisions [19,21] and are more likely to delay gratification [33], though these studies did not include a non-counterfactual control condition as we did in the present study, and so it remains possible that frustration or other non-counterfactual emotions were driving the effects. ...
Article
Adults often reason about what might have happened had they chosen an alternative course of action in the past, which can elicit the counterfactual emotion of regret. It is unclear whether young children's emotions are similarly impacted by counterfactual thinking about past possibilities. In this study, 4- to 9-year-old children ( N = 160) opened one of two boxes, which concealed small and large prizes, respectively. Some children had the means to open either box, whereas other children only had the means to open one box. After seeing that the prize they did not obtain was larger than the one they did obtain, children were significantly more likely to report a negative change in emotion when the non-obtained prize had been a straightforward counterfactual possibility than when it had not. This shows that even young children experience counterfactual emotions following choices, which may ultimately drive them to make better choices in the future. This article is part of the theme issue ‘Thinking about possibilities: mechanisms, ontogeny, functions and phylogeny’.
Chapter
This up-to-date, superbly illustrated book is a practical guide to the effective use of neuroimaging in the patient with sleep disorders. There are detailed reviews of new neuroimaging techniques – including CT, MRI, advanced MR techniques, SPECT and PET – as well as image analysis methods, their roles and pitfalls. Neuroimaging of normal sleep and wake states is covered plus the role of neuroimaging in conjunction with tests of memory and how sleep influences memory consolidation. Each chapter carefully presents and analyzes the key findings in patients with sleep disorders indicating the clinical and imaging features of the various sleep disorders from clinical presentation to neuroimaging, aiding in establishing an accurate diagnosis. Written by neuroimaging experts from around the world, Neuroimaging of Sleep and Sleep Disorders is an invaluable resource for both researchers and clinicians including sleep specialists, neurologists, radiologists, psychiatrists, psychologists.
Chapter
This up-to-date, superbly illustrated book is a practical guide to the effective use of neuroimaging in the patient with sleep disorders. There are detailed reviews of new neuroimaging techniques – including CT, MRI, advanced MR techniques, SPECT and PET – as well as image analysis methods, their roles and pitfalls. Neuroimaging of normal sleep and wake states is covered plus the role of neuroimaging in conjunction with tests of memory and how sleep influences memory consolidation. Each chapter carefully presents and analyzes the key findings in patients with sleep disorders indicating the clinical and imaging features of the various sleep disorders from clinical presentation to neuroimaging, aiding in establishing an accurate diagnosis. Written by neuroimaging experts from around the world, Neuroimaging of Sleep and Sleep Disorders is an invaluable resource for both researchers and clinicians including sleep specialists, neurologists, radiologists, psychiatrists, psychologists.
Chapter
This up-to-date, superbly illustrated book is a practical guide to the effective use of neuroimaging in the patient with sleep disorders. There are detailed reviews of new neuroimaging techniques – including CT, MRI, advanced MR techniques, SPECT and PET – as well as image analysis methods, their roles and pitfalls. Neuroimaging of normal sleep and wake states is covered plus the role of neuroimaging in conjunction with tests of memory and how sleep influences memory consolidation. Each chapter carefully presents and analyzes the key findings in patients with sleep disorders indicating the clinical and imaging features of the various sleep disorders from clinical presentation to neuroimaging, aiding in establishing an accurate diagnosis. Written by neuroimaging experts from around the world, Neuroimaging of Sleep and Sleep Disorders is an invaluable resource for both researchers and clinicians including sleep specialists, neurologists, radiologists, psychiatrists, psychologists.
Chapter
This up-to-date, superbly illustrated book is a practical guide to the effective use of neuroimaging in the patient with sleep disorders. There are detailed reviews of new neuroimaging techniques – including CT, MRI, advanced MR techniques, SPECT and PET – as well as image analysis methods, their roles and pitfalls. Neuroimaging of normal sleep and wake states is covered plus the role of neuroimaging in conjunction with tests of memory and how sleep influences memory consolidation. Each chapter carefully presents and analyzes the key findings in patients with sleep disorders indicating the clinical and imaging features of the various sleep disorders from clinical presentation to neuroimaging, aiding in establishing an accurate diagnosis. Written by neuroimaging experts from around the world, Neuroimaging of Sleep and Sleep Disorders is an invaluable resource for both researchers and clinicians including sleep specialists, neurologists, radiologists, psychiatrists, psychologists.
Chapter
This up-to-date, superbly illustrated book is a practical guide to the effective use of neuroimaging in the patient with sleep disorders. There are detailed reviews of new neuroimaging techniques – including CT, MRI, advanced MR techniques, SPECT and PET – as well as image analysis methods, their roles and pitfalls. Neuroimaging of normal sleep and wake states is covered plus the role of neuroimaging in conjunction with tests of memory and how sleep influences memory consolidation. Each chapter carefully presents and analyzes the key findings in patients with sleep disorders indicating the clinical and imaging features of the various sleep disorders from clinical presentation to neuroimaging, aiding in establishing an accurate diagnosis. Written by neuroimaging experts from around the world, Neuroimaging of Sleep and Sleep Disorders is an invaluable resource for both researchers and clinicians including sleep specialists, neurologists, radiologists, psychiatrists, psychologists.
Chapter
This up-to-date, superbly illustrated book is a practical guide to the effective use of neuroimaging in the patient with sleep disorders. There are detailed reviews of new neuroimaging techniques – including CT, MRI, advanced MR techniques, SPECT and PET – as well as image analysis methods, their roles and pitfalls. Neuroimaging of normal sleep and wake states is covered plus the role of neuroimaging in conjunction with tests of memory and how sleep influences memory consolidation. Each chapter carefully presents and analyzes the key findings in patients with sleep disorders indicating the clinical and imaging features of the various sleep disorders from clinical presentation to neuroimaging, aiding in establishing an accurate diagnosis. Written by neuroimaging experts from around the world, Neuroimaging of Sleep and Sleep Disorders is an invaluable resource for both researchers and clinicians including sleep specialists, neurologists, radiologists, psychiatrists, psychologists.
Chapter
This up-to-date, superbly illustrated book is a practical guide to the effective use of neuroimaging in the patient with sleep disorders. There are detailed reviews of new neuroimaging techniques – including CT, MRI, advanced MR techniques, SPECT and PET – as well as image analysis methods, their roles and pitfalls. Neuroimaging of normal sleep and wake states is covered plus the role of neuroimaging in conjunction with tests of memory and how sleep influences memory consolidation. Each chapter carefully presents and analyzes the key findings in patients with sleep disorders indicating the clinical and imaging features of the various sleep disorders from clinical presentation to neuroimaging, aiding in establishing an accurate diagnosis. Written by neuroimaging experts from around the world, Neuroimaging of Sleep and Sleep Disorders is an invaluable resource for both researchers and clinicians including sleep specialists, neurologists, radiologists, psychiatrists, psychologists.
Chapter
This up-to-date, superbly illustrated book is a practical guide to the effective use of neuroimaging in the patient with sleep disorders. There are detailed reviews of new neuroimaging techniques – including CT, MRI, advanced MR techniques, SPECT and PET – as well as image analysis methods, their roles and pitfalls. Neuroimaging of normal sleep and wake states is covered plus the role of neuroimaging in conjunction with tests of memory and how sleep influences memory consolidation. Each chapter carefully presents and analyzes the key findings in patients with sleep disorders indicating the clinical and imaging features of the various sleep disorders from clinical presentation to neuroimaging, aiding in establishing an accurate diagnosis. Written by neuroimaging experts from around the world, Neuroimaging of Sleep and Sleep Disorders is an invaluable resource for both researchers and clinicians including sleep specialists, neurologists, radiologists, psychiatrists, psychologists.
Article
Full-text available
We try to identify the configurations of beliefs and goals typical of the various kinds of representation of the future: forecasts, hopes and fears, and a particular kind of anticipatory representations that we call `hope-casts' and `fear-casts', which are supposed to imply not only forecasts and either hopes or fears, but also a normative component according to which the expected event `ought' to happen. We address the psychological consequences of hope-casts, either before or after the expected event comes true or false, and point to the sense of injustice and loss produced by violated hope-casts. We also address individual differences in dealing with violated hope-casts, and the possible role played by just-world beliefs, optimism and defensive pessimism. Finally, we compare our model with related approaches, and stress the negative potential of hope-casts in terms of the negative attitudes associated with them, which hamper people's ability to cope with disappointment.
Article
Anterior cingulate cortex (ACC) is a part of the brain's limbic system. Classically, this region has been related to affect, on the basis of lesion studies in humans and in animals. In the late 1980s, neuroimaging research indicated that ACC was active in many studies of cognition. The findings from EEG studies of a focal area of negativity in scalp electrodes following an error response led to the idea that ACC might be the brain's error detection and correction device. In this article, these various findings are reviewed in relation to the idea that ACC is a part of a circuit involved in a form of attention that serves to regulate both cognitive and emotional processing. Neuroimaging studies showing that separate areas of ACC are involved in cognition and emotion are discussed and related to results showing that the error negativity is influenced by affect and motivation. In addition, the development of the emotional and cognitive roles of ACC are discussed, and how the success of this regulation in controlling responses might be correlated with cingulate size. Finally, some theories are considered about how the different subdivisions of ACC might interact with other cortical structures as a part of the circuits involved in the regulation of mental and emotional activity.
Article
Reciprocal connections between the orbitofrontal cortex and the basolateral nucleus of the amygdala may provide a critical circuit for the learning that underlies goal-directed behavior. We examined neural activity in rat orbitofrontal cortex and basolateral amygdala during instrumental learning in an olfactory discrimination task. Neurons in both regions fired selectively during the anticipation of rewarding or aversive outcomes. This selective activity emerged early in training, before the rats had learned reliably to avoid the aversive outcome. The results support the concept that the basolateral amygdala and orbitofrontal cortex cooperate to encode information that may be used to guide goal-directed behavior.
Article
Associative learning theory assumes that prediction error is a driving force in learning. A competing view, probabilistic contrast (PC) theory, is that learning and prediction error are unrelated. We tested a learning phenomenon that has proved troublesome for associative theory - retrospective revaluation - to evaluate these two models. We previously showed that activation in right lateral prefrontal cortex (PFC) provides a reliable signature for the presence of prediction error. Thus, if the associative view is correct, retrospective revaluation should be accompanied by right lateral PFC activation. PC theory would be supported by the absence of this activation. Right PFC and ventral striatal activation occurred during retrospective revaluation, supporting the associative account. Activations appeared to reflect the degree of revaluation, predicting later brain responses to revalued cues. Our results support a modified associative account of retrospective revaluation and demonstrate the potential of functional neuroimaging as a tool for evaluating competing learning models.
Article
The human orbitofrontal cortex is an important brain region for the processing of rewards and punishments, which is a prerequisite for the complex and flexible emotional and social behaviour which contributes to the evolutionary success of humans. Yet much remains to be discovered about the functions of this key brain region, and new evidence from functional neuroimaging and clinical neuropsychology is affording new insights into the different functions of the human orbitofrontal cortex. We review the neuroanatomical and neuropsychological literature on the human orbitofrontal cortex, and propose two distinct trends of neural activity based on a meta-analysis of neuroimaging studies. One is a mediolateral distinction, whereby medial orbitofrontal cortex activity is related to monitoring the reward value of many different reinforcers, whereas lateral orbitofrontal cortex activity is related to the evaluation of punishers which may lead to a change in ongoing behaviour. The second is a posterior–anterior distinction with more complex or abstract reinforcers (such as monetary gain and loss) represented more anteriorly in the orbitofrontal cortex than simpler reinforcers such as taste or pain. Finally, we propose new neuroimaging methods for obtaining further evidence on the localisation of function in the human orbitofrontal cortex.
Article
Presents a theory of norms and normality and applies the theory to phenomena of emotional responses, social judgment, and conversations about causes. Norms are assumed to be constructed ad hoc by recruiting specific representations. Category norms are derived by recruiting exemplars. Specific objects or events generate their own norms by retrieval of similar experiences stored in memory or by construction of counterfactual alternatives. The normality of a stimulus is evaluated by comparing it with the norms that it evokes after the fact, rather than to precomputed expectations. Norm theory is applied in analyses of the enhanced emotional response to events that have abnormal causes, of the generation of predictions and inferences from observations of behavior, and of the role of norms in causal questions and answers. (3 p ref) (PsycINFO Database Record (c) 2012 APA, all rights reserved)
Article
This study explores the implications of dissapointment, a psychological reaction caused by comparing the actual outcome of a lottery to one's prior expectations, for decision making under uncertainty. Explicit recognition that decision makers may be paying a premium to avoid potential disappointment provides an interpretation for some known behavioral paradoxes, and suggests that decision makers may be sensitive to the manner in which a lottery is resolved. The concept of disappointment is integrated into utility theory in a prescriptive model.
Article
EVIDENCE EXISTS THAT PEOPLE DO NOT ALWAYS MAKE DECISIONS INVOLVING UNCERTAIN MONETARY REWARDS AS IF THEY WERE MAXIMIZING EXPECTED UTILITY OF FINAL ASSETS. EXPLANATIONS FOR THIS BEHAVIOR POSTULATE THAT THE COGNITIVE DEMANDS OF CONSISTENCY TO SUCH A THEORY ARE TOO GREAT. HOWEVER, SITUATIONS EXIST IN WHICH MORE THAN MENTAL SHORTCUTS ARE INVOLVED AND THESE ANOMALIES RAISE EQUATIONS ABOUT EXPECTED UTILITY THEORY AS A GUIDE TO BEHAVIOR. THIS STUDY EXPLORES THE POSSIBILITY THAT EXPECTED UTILITY THEORY APPEARS TO FAIL BECAUSE THE SINGLE OUTCOME DESCRIPTOR - MONEY - IS NOT SUFFICIENT. AFTER MAKING A DECISION UNDER UNCERTAINTY, A PERSON MAY DISCOVER, ON LEARNING THE RELEVANT OUTCOMES, THAT ANOTHER ALTERNATIVE WOULD HAVEBEEN PREFERABLE. THIS KNOWLEDGE MAY IMPART A SENSE OF LOSS,OR REGRET. THE DECISION MAKER WHO IS PREPARED TO TRADEOFF FINANCIAL RETURN IN ORDER TO AVOID REGRET WILL EXHIBIT SOME OF THE BEHAVIORAL PARADOXES OF DECISION THEORY.
Article
The capacity to predict future events permits a creature to detect, model, and manipulate the causal structure of its interactions with its environment. Behavioral experiments suggest that learning is driven by changes in the expectations about future salient events such as rewards and punishments. Physiological work has recently complemented these studies by identifying dopaminergic neurons in the primate whose fluctuating output apparently signals changes or errors in the predictions of future salient and rewarding events. Taken together, these findings can be understood through quantitative theories of adaptive optimizing control.