doi:10.

1093/scan/nsq048

SCAN (2011) 6, 417^ 425

Priming for self-esteem influences the

monitoring of ones own performance
Sara L. Bengtsson,1 Raymond J. Dolan,1 and Richard E. Passingham1,2
1

Wellcome Trust Centre for Neuroimaging, University College London, 12 Queen Square, London, UK, and 2Department of Experimental
Psychology, University of Oxford, Oxford, UK
Social cues have subtle effects on a person, often without them being aware. One explanation for this influence involves implicit
priming of trait associations. To study this effect, we activated implicit associations in participants of being Clever or being
Stupid that were task relevant, and studied its behavioural impact on an independent cognitive task (the n-back task). Activating
a representation of Clever caused participants to slow their reaction times after errors on the working memory task, while the
reverse pattern was seen for associations to Stupid. Critically, these behavioural effects were absent in control conditions.
Using functional magnetic resonance imaging, we show that the neural basis of this effect involves the anterior paracingulate
cortex (area 32) where activity tracked the observed behavioural pattern, increasing its activity during error monitoring in the
Clever condition and decreasing in the Stupid condition. The data provide a quantitative demonstration of how implicit cues,
which specifically target a persons self-concept, influences the way we react to our own behaviour and point to the anterior
paracingulate cortex as a critical cortical locus for mediating these self-concept related behavioural regulations.
Keywords: self-regulation; implicit priming; paracingulate cortex; cognitive performance; social cognition; fMRI

INTRODUCTION
Leading theories on the self propose that cognitive concepts
of ones self are used to organize and guide self-relevant
information (Kelly, 1955; Markus, 1977) in a similar way
as we build internal representations of the environment on
the basis of our experiences of any mental process (Neisser,
1976). The formed concepts about ourselves are stored in
long-term memory and appear to be fairly stable over an
individuals lifespan (Markus and Kundra, 1986). The features of the self that guide behaviour, however, are those that
are currently active in the mind, i.e. the working
self-concept, which is minutely sensitive to environmental
cues and may change at any given time (Markus and
Wurf, 1987). This is exemplified in studies where individuals
are primed with trait associations. For example, if participants are required to solve a language task in which words
are synonyms for old, they walk more slowly than normal
after the language task, even when they are not explicitly
aware of the bias in the words (Bargh et al., 1996). It has
also been shown that priming for concepts associated with
being clever, such as Professor, influences intellectual performance as tested on a quiz (Dijksterhuis and van
Knippenberg, 1998). Priming can either lead to activation
of motivational goal concepts, or simple semantic constructs.
In tasks where participants are primed with neutral nouns,
Received 8 February 2010; Accepted 11 May 2010
Advance Access publication 15 June 2010
This work was supported by a personal grant for S.B. from the Swedish Research Council (VR), and a
Wellcome Trust program grant to RJD and REP. We are grateful to H. Bengtsson, D. Kumaran, and D. Hassabis
for valuable discussions.
Correspondence should be addressed to Dr Sara Bengtsson, Wellcome Trust Centre for Neuroimaging, UCL,
12 Queen Square, London WC1N 3BG, UK. E-mail: [email protected]

such as bottle, which results in a faster response time (RT)

to related words, such as drink, activation of a semantic
network has most likely triggered the priming effect. For
a prime to evoke motivational goals, certain effects need
to be observed. For example, if the priming results in behaviour that depends on the value of the prime and if the primes
influence the goal to complete the task in situations
when the task is uncompleted it is likely that a goal rather
than a semantic construct has been primed (Forster et al.,
2007).
The working self-concepts are associative networks
derived from the basic self-knowledge. Conway and
Pleydell-Pearce (2000) propose that the goals of the working
self form control processes that put constrains on cognition
and behaviour in a similar way to that of general working
memory processes (Baddeley, 1986). On this note, several
brain imaging studies have found activation in the frontal
lobe during retrieval and maintenance of highly self-relevant
knowledge. For example, retrieving autobiographical memories as compared to memories of films and news events
activate the anterior cingulate cortex and the middle frontal
gyrus (Summerfield et al., 2009), and in studies where participants reflect upon whether a trait applies to themselves or
not activation in the anterior cingulate cortex is typically
observed (Macrae et al., 2004; Ochsner et al., 2005;
Amodio and Frith, 2006). We have previously found that
this area is also sensitive to task instructions which make
participants specifically monitor their own performance
(Bengtsson et al., 2009). We manipulated participants
desire to do well on a working memory task by telling one
group that the task was a measure of their own ability, and

The Author(s) 2010. Published by Oxford University Press.

This is an Open Access article distributed under the terms of the Creative Commons Attribution Non-Commercial License (http://creativecommons.org/licenses/by-nc/2.5/uk/) which permits
unrestricted non-commercial use, distribution, and reproduction in any medium, provided the original work is properly cited.

418

SCAN (2011)

another group that the task was a piloting task so as to optimize certain task parameters and that they did not need to
care too much about their own performance. The task was
titrated to equate performance for all participants. The more
motivated group showed enhance activity in the anterior
paracingulate cortex (area 32) when they made errors on
the working memory task. We found subsequently that
there was increased activity in an overlapping cluster when
the participants were asked to explicitly rate their own
performance.
The aim of the present study was to investigate direct
evidence for the anterior paracingulate cortex playing a
role in monitoring ones own performance. We did so by
priming each participant with trait associations that were
relevant for their task performance. We studied the effect
of primes (Bargh and Chartrand, 2000) that evoked associations to Clever and to Stupid on performance on an n-2
back working memory task for letters (Cohen et al., 1997;
Figure 1A). The two categories of primes were chosen because of the familiar link between working memory capacity
and intelligence (Conway et al., 2003; Gray et al., 2003).
Clever-priming will generally evoke a network of selfconcept associations such as bright, competent and skilled,
whereas Stupid priming will trigger associations such as forgetful and inefficient. To make sure we evoked goal priming
it would be necessary to show that positive and negative trait
primes exert different effects on behaviour, and that they do
so only if the prime is task relevant. Therefore, to account for
effects of positive and negative affect, we included the primes
Happy and Sad as control conditions. In a pilot experiment, we had found that these primes had no influence on
n-back performance, most likely because they are irrelevant
for success on the task. There is, of course, considerable
evidence for the effects of emotional stimuli on attention
and perception (Dolan, 2002). However, the present study
involves priming with Happy and Sad when the task stimuli are themselves neutral (letters of the n-back task). Gray
et al. (2002) induced emotional states in participants by
having them view either happy or unhappy video clips
before performing an n-back task, but did not find an
effect. With a larger sample size of 66 participants there
was a significant emotion by working memory item interaction. This effect depended on whether words or faces were
presented as items in the n-back task, thus indicating that the
working memory items mattered in so far as they had an
emotional relevance.
Our design is a 2*2 factorial design where valence is varied
on one axis, and task relevance is varied on the other
(Figure 1B). Since a non-primed condition lack the control
of what associations the participants are making, we decided
to exclude such a condition. We measured the effect of
priming on RT and error rate on the n-back task. RT is a
measure that can often give a strong clue to the mental
processes going on. Investigating RT after an error response,
for example, provides a measure of executive cognitive

S. L. Bengtsson et al.

Fig. 1 (A) We used the 2-back task for letters as the task of interest. For each letter
that they saw, the participants made either a yes or a no response. If the current
letter also appeared two letters back a yes response should be made. The memory
task was preceded by the language task that served as the prime. (B) The conditions,
as defined by the trait primed, fitted in a 2*2 factorial design with valence varying
one axis, and task-relevance varying on the other.

control indicating how well participants adjust their behaviour after having made an error (Rabbitt, 1966). By using
functional magnetic resonance imaging (fMRI), we can
establish whether prime dependent activations occur in
brain areas similar, or distinct from brain areas involved
in task performance. It is unknown whether goal priming
acts by influencing a separate goal representation, as
opposed to simply influencing areas involved in task performance (Dijksterhuis and van Knippenberg, 1998;
Markus, 1999).
MATERIAL AND METHODS
Participants
Twenty-one neurologically healthy participants took part in
the study. Nine were males and 12 were females with a mean
age of 23.3  5.2 years. They were all native English speakers
while 19 of the participants were students. All gave written
informed consent with the study being approved by the joint
ethics committee of the Institute of Neurology and
University College London Hospital, London, UK.
Behavioural task
The participants performed a scrambled sentence task and a
working memory task with the scrambled sentence task

A self-representation revealed via priming

serving as the priming task (Bargh and Chartrand, 2000).
A scrambled sentence consisted of six words and participants
judged whether or not it could be made into a grammatically
coherent sentence by using five of the six words. Subjects
responded yes by pressing a button corresponding to their
right index finger or no by pressing a button corresponding
to their right middle finger. Each sentence was presented for
8 s during which time the participant had to respond. In each
session, 70% of the sentences had words that were synonyms
for one of the four category words and 30% of the sentences
were neutral. We ran one session for each condition: Clever,
Stupid, Happy and Sad. Examples of sentences for the different conditions are:
Clever: pupil apt Todd an is pencil, the brightest nothing
idea everything promoted.
Stupid: welcome not morons one are here, the room
obtuse had white green.
Happy: stop smiling could not she lips, most pink life
jolly this is.
Sad: quite gloomy were pictures the them, truth an is it
over unfortunate.
Neutral: coloured hair brown shopping her was CD
latest their brought playing he.
The sentences in each of the four conditions were matched
for first and third-person pronouns.
For the memory task, participants were tested on the
n-2 back task for letters. Sequences of letters were presented
on a screen, and for each letter the subject pressed a button.
If the letter appeared two letters back, the participant made a
yes response, otherwise they made a no response. The yes
response involved pressing a button corresponding to the
right index finger, while a no response was made by pressing the button corresponding to the right middle finger. A
block consisted of 19 letters, each letter being visible for
500 ms. Within a block, there were 30% hits. The
inter-stimulus interval (ISI) was set individually for each
participant prior to the data collection to achieve a performance level of 70% yielding an ISI that was either 500, 1000 or
1500 ms. This gave sufficient error trials to analyse these
separately. Note that the ISI was then kept constant between
conditions during the scanning.
To disguise a link between the two tasks, we told the participants that we would alternate between a language and
memory task. Our explanation was that the experimenters
had long experience of participants getting bored during
fMRI scanning, and this was a new way of trying to prevent
this from happening. After scanning, we asked the participants to fill in an adapted questionnaire (Bargh and
Chartrand, 2000) on what they thought was the purpose of
the study.
Prior to data collection, participants practised the task in a
separate test room, for one session with all sentences being
neutral in character. The participants were then given another practice session in the scanner, with the same material.
During this practise session, the ISI was set for each

SCAN (2011)

419

participant so that they reached a performance level on the

n-back task of 70% correct. Data from all four sessions
were collected during scanning with the order of the four
condition-specific sessions being pseudo-randomized between participants. In a session, eight sentences were presented followed by three blocks of the memory task, repeated
three times within a session.
Scanning parameters and data analysis
Imaging was performed using a 3 Tesla scanner (Siemens
Allegra: Erlangen, Germany). The functional images sensitive
to blood oxygen-level dependent (BOLD) contrasts were
acquired as gradient-echo, Epiplanar imaging T2*-weighted
images. Image volumes of the whole brain were built
up from contiguous axial slices (n 40), TR 2.6 s,
TE 10 ms, matrix size 64  64, slice thickness 2 mm with
a 1 mm gap.
Image processing and analysis used the SPM5 software
(www.fil.ion.ucl.ac.uk/spm/), Wellcome Trust Centre for
Neuroimaging, London, UK. The first five volumes were
discarded, and the remaining volumes realigned to the first
volume to correct for head movements. Subsequently, the
volumes were co-registered and normalized to the standard
space using the Montreal Neurological Institute reference
brain. The time series were smoothed spatially with an isotropic Gaussian filter of 8 mm full width at half-maximum.
Whole brain statistical parametric maps of t-statistics were
calculated for four condition specific effects within the general linear model. Covariates for transient activation in response to motor responses were modelled separately for
correct and incorrect responses separate for each condition.
All events were convolved with a canonical haemodynamic
response function. The data were high-pass filtered with a
frequency cut-off at 128 s.
To investigate whether priming acts by influencing
a self-representation or simply influences areas involved
in performance of the n-back task, we investigated if
prime dependent activations were found inside or outside
brain areas active in performing the n-back task. First, we
investigated the positive interaction between the valence
and task-specificity [(Clever Sad) vs (Stupid Happy)],
for error responses and for correct responses. Given results
of our previous study, where we also used the n-back
task and investigated the effect of overtly manipulating
participants view of the task, our a priori hypothesis
concerned activation of anterior cingulate cortex. In that
study, when participants were told that the task was an
indirect measure of intelligence, we found enhanced activation in the anterior paracingulate cortex (mean peakcoordinate: 10 46 36) on error trials, as compared to
participants who regarded the task as a pilot experiment
(Bengtsson et al., 2009). As a consequence, for the current data set, we hypothesized that implicit associations
to being clever and stupid would be reflected in anterior
paracingulate activity. This motivated a restricted region

420

SCAN (2011)

S. L. Bengtsson et al.

of interest (ROI) analysis using the peak-coordinate from

this study (10 46 36). Thus, we tested for a significant
activation within a sphere with an 8 mm radius (P < 0.05
corrected).
To confirm that the two conditions of interest were driving the interaction, we carried out a whole brain analysis for
the simple effect (Clever vs Stupid). We then carried out a
ROI analysis for this comparison (P < 0.05 corrected), using
the same ROI as before. To investigate an interaction effect
of the primes on activation when participants make errors
on the n-back task, we applied an ROI analysis to
mid-anterior cingulate cortex. This area is implicated when
subjects make errors on cognitive tasks (Barch et al., 1997;
Bengtsson et al., 2009). The coordinate for this ROI was
again taken from the study of Bengtsson et al. (2009)
based upon a peak-coordinate (0 22 40) with a sphere of
8 mm radius. We also ensured we had not missed activation
in other areas by using an exploratory whole brain search,
with a threshold of P < 0.001 uncorrected, as is appropriate
when checking for false negatives.
We also investigated brain activation unaffected by priming by computing a whole brain global conjunction analysis
of the contrast (Clever and Stupid and Happy and Sad)
(P < 0.05 corrected) (Friston et al., 1999). To link behaviour
and brain imaging data, we tested for correlations between
participants reaction times (RTs) and BOLD on a
trial-to-trial basis for the four different conditions. Vectors
were created for each condition for errors and correct responses separately and a whole brain SPM analysis was then
implemented. Finally, to investigate if the primes Happy
and Sad had any effect on brain processes, we calculated
the main effect for valence [(Stupid Sad) vs
(Clever Happy)]. For this contrast, we did not have any
a priori hypothesis and used a contrast corrected for a whole
brain analysis.
All contrasts were created first for each participant separately, followed by a between subjects random-effects analysis
based on summary statistics from the images created from
each subject. On the second level, we used a one-sample
t-test for each contrast. The histograms in Figures 3 and 4
represent task-related BOLD signal changes as estimated
using the general linear model. These are included to illustrate statistical effects revealed from the SPM analyses, and
are not used to make any statistical inference. The parameter

estimates are given for peak voxels showing a difference on

the second level while the baseline represents null events,
which were not modelled. In Figure 3D, the time course of
activation is plotted for the four conditions at the time of
making errors on the n-back task. The time series of the
BOLD signals is taken at the coordinate 10 50 30, which
was the significant peak for the interaction. The signals
within each bin have been averaged across trials for all the
participants with a time bin corresponding to 2.6 s and representing the average value and standard error of the mean
for the group.
RESULTS
Prime dependent behavioural adjustment after error
When debriefed after the experiment, none of the participants reported any association between the tasks, a finding in
line with previous studies using scrambled sentence tasks
(Ferguson and Bargh, 2004). This is important because,
had the participants been aware of the link between the
prime and the n-back task, they might have adjusted their
behaviour to decrease the effect of the priming (Bargh and
Chartrand, 2000).
We assessed the behavioural effect of priming by analysing
the RT on the n-back task. We measured RTs for the subsequent correct response after an error, and compared these
with RTs for other correct responses. This provides a measure of executive cognitive control, indicating how well participants adjust their behaviour after having made an error.
Even though there was no explicit feedback, on tasks that
encouraged quick responses subjects can be aware of making
an error at the same time as they respond (Rabbitt, 1966).
Table 1 shows the data for all the RT measures.
We found that the prime only had an effect if it was task
relevant evident in a significant within-subject valence*task-relevance interaction (P < 0.05), as well as a
within-subject simple effect between task-relevant primes
(P < 0.005). These effects are illustrated in Figure 2. The
RTs per condition were; Clever 34.9 ms  15.8 ms; Stupid
0.94 ms  13.5 ms; Happy 19.0 ms  12.4 ms; Sad 21.4 ms 
12.4 ms. It can be seen that RT were longer when primed
Clever than Stupid. This was an effect driven by absolute
differences in RTs following errors (valence*task-relevance
interaction P < 0.005), rather than by absolute difference in
RTs for the correct trials (valence*task-relevance interaction

Table 1 A summary of the behavioural data. Reaction times are presented in ms  the standard error of the mean

RT_correct after error vs RT_corrects (ms)

RT_correct after error (ms)
RT_corrects (ms)
Total number of errors made
n.s. non-significant.

Clever

Stupid

Happy

Sad

Main
effect
Valance

Main effect
Task-specificity

Valence Taskspecificity
interaction

34.9  15.8
578.0  39.6
543.0  35.7
53.1  8.6

0.94  13.5
521.0  35.8
520.0  32.0
49.9  7.4

19.0  12.4
550.5  35.8
531.5  34.3
52.1  7.1

21.4  12.4
554.7  35.4
533.2  31.7
49.5  6.2

n.s.
P < 0.04
n.s.
n.s.

n.s.
n.s.
n.s.
n.s.

P < 0.05
P < 0.005
n.s.
n.s.

A self-representation revealed via priming

Fig. 2 As an estimate of cognitive control, we measured the RTs for the subsequent
correct response after an error, and compared this with the RTs for other correct
responses. We found a significant within-subject valence*task-relevance interaction
for this measure (P < 0.05), as well as a within-subject simple effect between
task-relevant primes (P < 0.004). There were no significant differences between
the four conditions when investigating correct responses after correct responses.

P > 0.13). The finding that participants took longer to make

a response following an error can be interpreted in terms of
becoming more careful in their task performance (Kerns
et al., 2004; Debener et al., 2005). There was no difference
in the total number of errors made in the four different
conditions (Table 1).

Representation specific to the primed trait

Behavioural data on their own do not discriminate between
the several possible psychological mechanisms behind this
effect. It is possible, for example, that priming for Clever
makes it more likely that participants notice errors, or because they activate a more specific high-level concept of
being an intelligent person. Therefore, by scanning the participants with functional MR while they performed the
working memory task, we could investigate whether a distinct representation to being clever and being stupid had
been activated. Thus, we tested for the interaction between
valence and task-specificity of brain activation data.
If priming simply led to enhanced detection of errors then
the activation cluster would be expected in the mid anterior
cingulate cortex (area 24). This region is classically activated
for errors on the n-back task, as well as when the difficulty
level on the n-back task increases (4 25 43) (Barch et al.,
1997). In fact, it is this activity that reflects the detection of
errors in a number of cognitive tasks, and where activity
increases as a function of task difficulty or monetary
reward (Bush et al., 2000; Yeung et al., 2004). If, on the
other hand, we detected a prime specific activation outside
of the mid-anterior cingulate cortex, in anterior paracingulate cortex as we had hypothesised, this would be consistent

SCAN (2011)

421

with a view that the prime exerted its effect by activating a

self-representation.
For a detailed analysis of the mid-anterior cingulate
cortex, we performed a search both without and within a
search volume around the coordinate 0 22 40 (Bengtsson
et al., 2009). No significant interaction was found in this
area. In fact, as revealed in a conjunction analysis, there
was common activation here for all four conditions
(Figure 3A). This is the precise pattern of activation we
would expect if the mid-anterior cingulate cortex is responsive to error detection, while being insensitive to primes. As
illustrated in Figure 3A, the common activation had a local
maxima in 6 32 30 (t 2.56, P < 0.05 corrected).
To investigate significant interaction outside this area, we
narrowed the search region to the anterior paracingulate
cortex (Bengtsson et al., 2009). We found a significant valence * task-relevance interaction (10 50 30, t 2.8, P < 0.02
corrected). The simple comparison between Clever and
Stupid also yielded a peak within the same region (6 44
44, t 3.3, P < 0.005, corrected). The histogram in Figure 3B
shows the peak lying within anterior paracingulate cortex
(area 32). The figure also illustrates that the pattern of activity resembled the pattern of post-error RTs for the different conditions (Figure 2). No other areas showed this
interaction (P < 0.001 uncorrected). As is evident from
Figure 3C, the effect of the priming was to influence paracingulate cortex activation in a zone anterior to the cluster
identified for error detection, and the clusters do not overlap
even at the low threshold of P < 0.05 uncorrected. Figure 3D
illustrates the time course of activation for the four conditions in 10 50 30.
To assess a link between the impact of the primes
and subsequent behaviour, we correlated the participants
reaction times with the BOLD signal on a trial-to-trial
basis, which was subsequently entered into a second-level
summary statistics. When priming for Clever, there was a
trend for a significant correlation between activation of the
anterior paracingulate cortex and increase in RTs on
the trials that followed errors (2 66 12, t 4.1, P < 0.001
uncorrected; 18 42 32, t 3.7, P < 0.001 uncorrected).
These peaks lie near the ROI in which we found significant
interaction, but they do not lie within it. That we found only
a statistical trend when correlating BOLD and RT could be
due to there being an indirect, non-linear link between the
factors.
Finally, we could distinguish the effect of a task relevant
prime from the effect of the valence of the primes. For errors,
the main effect for negative primes (Stupid Sad vs
Clever Happy) showed extensive activation in the insula
bilaterally (42 4 0, t 4.4, P < 0.001 corrected on cluster
level statistics with sub-peaks; 38 12 8, t 3.8; 44 8 6,
t 3.3) (Figure 4). We take this as consistent with the idea
that priming for Stupid enhances the negative response to
errors. There was no effect of the primes when the participants made correct responses (P < 0.001 uncorrected).

422

SCAN (2011)

S. L. Bengtsson et al.

Fig. 3 (A) The activation in the mid anterior cingulate cortex represents error detection on the n-back task. There was no significant difference in this area between the four
conditions [P < 0.05 corrected conjunction analysis (Clever and Stupid and Happy and Sad)]. (B) The imaging data revealed a significant positive interaction in the anterior
paracingulate cortex {P < 0.02 corrected [(Clever Sad) vs (Stupid Happy)]}. Note that the BOLD pattern resembles the pattern of post-error RT for the different conditions. (C)
Prime sensitive activation in the anterior paracingulate cortex (shown in green) lay anterior to that of medial prefrontal activation generally seen for error detection (shown in
red). The SPMs are thresholded at P < 0.05 uncorrected. (D) The time course of activation is plotted for the four conditions at the event of making errors on the n-back task. The
time series of the BOLD signals is taken at the coordinate 10 50 30. The plot represents the average value for the whole group, as well as the standard error of the mean.

DISCUSSION
The interaction between the residual prime effect and
making errors on the working memory task caused participants to slow down their RT immediately after an error

when they had been primed Clever, whereas their RT

decreased after an error when they had been primed
Stupid. The primes Happy and Sad revealed no such
influence. This suggests that most likely motivational goal

A self-representation revealed via priming

SCAN (2011)

423

Fig. 4 The insula was active in response to errors for the two negative primes {P < 0.01 corrected [(Stupid Sad) vs (Clever Happy)]}.

priming has taken place. Priming specific brain activation

was seen in an area of the brain outside those involved in
performing the task, namely the anterior paracingulate
cortex (area 32). When participants had been primed
Clever, there was increased activation in the anterior paracingulate cortex on error trials. By contrast, when they had
been primed Stupid there was decreased activity on error
trials. This result is consistent with a role of the anterior
paracingulate cortex in monitoring task performance in
line with our working self-concept. In the present study,
the number of errors for the two primes was the same, perhaps reflecting the fact that each n-back sequence was relatively short (19 letters per block). We have in later priming
experiments found that priming for Stupid leads to more
errors being made compared to when participants are
primed Clever (submitted data).
The prefrontal cortex is traditionally viewed as dealing
with conscious processes (Jack and Shallice, 2001; Dehaene
et al., 2003). Dehaene et al. (2003) tested whether an area
that processes conflicts in cognitive tasks, the mid-anterior
cingulate cortex, would be sensitive to priming, but they
found no task-specific changes as a function of prime. In
their task, participants compared a target number to a
fixed standard, and either another number or a random consonant string served as prime. However, Lau and Passingham
(2007) observed that the lateral prefrontal cortex responded
to subliminal information in a task where participants set up
a complex cognitive task. In their study, the primes were
symbols made to distract participants to prepare for an
irrelevant task and the prime sensitive activation was found
within task areas. Finding priming sensitive activation in
the prefrontal cortex is in line with the behavioural results
showing that priming can influence higher order concepts
(Dijksterhuis and van Knippenberg, 1998).
The activation of self-related information in the medial
prefrontal cortex is consistent with a general hypothesis
that the medial wall supports processes related to

introspection. There is enhanced activation in the caudal

cingulate motor area when one attends to ones own movements and there is activation in the pre-supplementary
motor area when one attends to ones own intention to
move (Lau et al., 2006). Similarly, there is enhanced activation in the anterior cingulate cortex when one attends to
ones own internal states (Critchley et al., 2004). Amodio
and Frith (2006) suggest in their review that there is an
anterior-posterior gradient within the cingulate cortex,
with anterior paracingulate cortex being involved in
meta-cognition, i.e. thinking about ones thoughts.
Supporting evidence for our interpretation comes from studies reporting activation in a similar location when participants explicitly reflect upon themselves; [(19 40 32) Macrae
et al., 2004; (12 48 36) Ochsner et al., 2005; (10 46 24)
Bengtsson et al., 2009] or retrieve autobiographical memories [(6 33 24) Summerfield et al., 2009]. It is true that
other studies using self-referential tasks have found activation in the ventral part of the medial pre-frontal cortex
(Mitchell et al., 2005, Jenkins et al., 2008). Whereas the ventral part is interconnected with areas which are regulating
emotions, the dorsal part has fewer such connections (Ongur
et al., 2003). Activations in the dorsal region often reflect
cognitive and memory processes, such as reasoning about
oneself (Ochsner et al., 2005). In the present study, the participants were primed Clever and Stupid in relation to a
cognitive task which most likely evoked cognitive rather than
emotional self-associations, hence the dorsal activation.
The anterior medial area which we find priming sensitive
has a different connectivity profile than the area we find
sensitive in detecting errors on the n-back task (the
mid-anterior cingulated cortex). This has been identified
in a connectivity study based on probabilistic tractography
of diffusion tension imaging data, where nine distinct areas
of the cingulate cortex were identified as having different
connectivity patterns (Beckmann et al., 2009). The latter
area has connections to premotor and dorsal prefrontal

424

SCAN (2011)

cortex. The former area has connections to hippocampus,

striatum and orbitofrontal cortex. Activation in the anterior
paracingulate cortex is seen in motor tasks that involves cognitive manipulations (Pickard and Strick, 2001; Beckmann
et al., 2009), in memory studies (Beckmann et al., 2009) and
in studies involving social interactions (Amodio and Frith,
2006). We note that activation has been reported in paracingulate cortex when people reflect on the mental states of
others (Jenkins et al., 2008) but this finding would not explain the motivational changes in participants own performance that we observe on error trials specifically. Taking on
the perspective of Conway and Pleydell-Pearce (2000), these
anatomical and functional data suggest that the anterior
paracingulate cortex function as the cognitive control process, formed by the working self-concept, which puts constrains on behaviour. It appears likely that this area would
control behaviour on other cognitive tasks as well, but it
remains to be investigated.
We find that the anterior paracingulate cortex processes
trait associations in a dynamic and specific manner. The
results of the present study themselves suggest that the increase in activation in this area when primed Clever relates
to heightened monitoring of errors, and that this leads to
increased caution on the next trial. Interestingly, we found
that while activation in the anterior paracingulate cortex
goes down on error trials when participants have been
primed Stupid, activation in the anterior insula increases.
As a result, faster RT after errors is observed. Noteworthy,
priming for associations to Sad enhance responses in the
insula as well (Figure 4). The insula is commonly activated
by aversive stimuli, as seen with pain (Seymour et al., 2004)
or following monetary loss (ODoherty et al., 2003). Thus,
our results demonstrate two distinct effects of priming on
errors where enhancement of activation of the anterior paracingulate cortex reflects monitoring of errors so as to influence future responses, and activation in anterior insula
reflecting a likely negative emotional response to errors.
Our data point to a mechanism by which incidental
events, in this case primes, influence our behaviour via effects on a self-representation. The finding provides support
for the view that an individuals working self-concept, and
consequent behaviour, is guided by distinct and organized
cortical structures and representations, in the same way as
other concepts (Kelly, 1955; Markus, 1977, 1999).
REFERENCES
Amodio, D.M., Frith, C.D. (2006). Meeting of minds: the medial
frontal cortex and social cognition. Nature Reviews Neuroscience, 7,
26877.
Baddeley, A.D. (1986). Working memory. Oxford, England: Clarendon Press.
Barch, D.M., Braver, T.S., Nystrom, L., Noll, D.C., Cohen, J.D. (1997).
Activation of prefronal cortex by the representation and maintenance
of context information. Schizophrenia Research, 24, 16366.
Bargh, J.A., Chartrand, T.L. (2000). The mind in the middle: A practical
guide to priming and automaticity research. In: Reis, H.T., Judd, C.M.,
editors. Handbook of Research Methods in Social and Personality
Psychology. Cambridge: Cambridge University Press, pp. 253.

S. L. Bengtsson et al.
Bargh, J.A., Chen, M., Burrows, L. (1996). Automaticity of social behavior:
Direct effects of trait construct and stereotype activation on action.
Journal Of Personality And Social Psychology, 71, 23044.
Beckmann, M., Johansen-Berg, H., Rushworth, M.F.S. (2009).
Connectivity-based parcellation of human cingulated cortex and
its relation to functional specialization. Journal of Neuroscience, 29,
117590.
Bengtsson, S.L., Lau, H.C., Passingham, R.E. (2009). Motivation to do well
enhances responses to errors and self-monitoring. Cerebral Cortex, 19,
797804.
Bush, G., Luu, P., Posner, M.I. (2000). Cognitive and emotional influences
in anterior cingulate cortex. TICS, 4, 21522.
Cohen, J.D., Perlstein, W.M., Braver, T.S., et al. (1997). Temporal
dynamics of brain activation during a working memory task. Nature,
386, 60408.
Conway, M.A., Pleydell-Pearce, C.W. (2000). The construction of autobiographical memories in the self-memory system. Psychological Review, 107,
26188.
Conway, A.R.D., Kane, M.J., Engle, R.W. (2003). Working memory capacity
and its relation to general intelligence. TICS, 7, 54752.
Critchley, H.D., Wiens, S., Rotshtein, P., Ohman, A., Dolan, R. (2004).
Neural systems supporting interoceptive awareness. Nature
Neuroscience, 7, 18995.
Debener, S., Ullsperger, M., Siegel, M., et al. (2005). Trial-by-trial coupling
of concurrent electroencephalogram and functional magnetic resonance
imaging identifies the dynamics of performance monitoring. Journal of
Neuroscience, 25, 1173037.
Dehaene, S., Artiges, E., Naccache, L., et al. (2003). Conscious and subliminal conflicts in normal subjects and patients with schizophrenia: the role
of the anterior cingulate. Proceedings of the National Academy of Sciences
USA, 100, 1372227.
Dijksterhuis, A., van Knippenberg, A. (1998). The relation between
perception and behavior, or how to win a game of trivial pursuit.
Journal Of Personality And Social Psychology, 74, 86577.
Dolan, R.J. (2002). Emotion, cognition, and behaviour. Science, 298,
119194.
Ferguson, M.J., Bargh, J.A. (2004). How social perception can automatically
influence behavior. TICS, 8, 339.
Forster, J., Liberman, N., Friedman, R.S. (2007). Seven principles of goal
activation: a systematic approach to distinguishing goal priming from
priming of non-goal constructs. Personality and Social Psychology
Review, 11, 21133.
Friston, K.J., Holmes, A.P., Price, C.J., Buchel, C., Worsley, K.J. (1999).
Multisubject fMRI studies and conjunction analyses. Neuroimage, 10,
38596.
Gray, J.R., Braver, T.S., Raichle, M.E. (2002). Integration of emotion and
cognition in the lateral prefrontal cortex. Proceedings of the National
Academy of Sciences USA, 99, 411520.
Gray, J.R., Chabris, C.F., Braver, T.S. (2003). Neural mechanisms of genral
fluid intelligence. Nature Neuroscience, 6, 31622.
Jack, A.I., Shallice, T. (2001). Introspective physicalism as an approach to
the science of consciousness. Cognition, 79, 16196.
Jenkins, A.C., Macrae, C.N., Mitchell, J.P. (2008). Repetition
Suppression of Ventromedial Prefrontal Activity During Judgments of
Self and Others. Proceedings of the National Academy of Sciences USA,
105, 450712.
Kelly, G. (1955). The psychology of personal constructs. New York: Norton.
Kerns, J.G., Cohen, J.D., MacDonals, A.W., Cho, R.Y., Stenger, V.A.,
Carter, C.S. (2004). Anterior cingulate conflict monitoring and adjustments in control. Science, 303, 102326.
Lau, H., Passingham, R.E. (2007). Unconscious activation of the cognitive
control system in the human prefrontal cortex. Journal of Neuroscience,
27, 580511.
Lau, H., Rogers, R.D., Passingham, R.E. (2006). On measuring the
perceived onsets of spontaneous actions. Journal of Neuroscience, 26,
726571.

A self-representation revealed via priming

Macrae, C.N., Moran, J.M., Heatherton, T.F., Banfield, J.F., Kelley, W.M.
(2004). Medial prefrontal activity predicts memory for self.
Cerebral Cortex, 14, 64754.
Markus, H. (1977). Self-Schemata and Processing Information About Self.
Journal Of Personality And Social Psychology, 35, 6378.
Markus, H., Kunda, Z. (1986). Stability and malleability of the self-concept.
Journal of Personality and Social Psychology, 51, 85866.
Markus, H., Wurf, E. (1987). The dynamic self-concept. A social psychological perspective. Annual Review of Psychology, 38, 299337.
Markus, H. (1999). Culture and the self: Implications for cognition, emotion, and motivation. In: Baumeister, R.F., editor. The self in social
psychology: essential readings. New York, US: Psychology Press, p. 123.
Mitchell, J.P., Banaji, M.R., Macrae, C.N. (2005). The link between social
cognition and self-referential thought in the medial prefrontal cortex.
Journal of Cognitive Neuroscience, 17, 130615.
Neisser, U. (1976). Cognition and reality: Principles and implications of
cognitive psychology. San Fransisco: Freeman.
Ochsner, K.N., Beer, J.S., Robertson, E.R., et al. (2005). The neural correlates of direct and reflected self-knowledge. Neuroimage, 28, 797814.

SCAN (2011)

425

Ongur, D., Ferry, A.T., Price, J.L. (2003). Architectonic subdivision of the
human orbital and medial prefrontal cortex. Journal of Comparative
Neurology, 460, 42549.
ODoherty, J.P., Dayan, P., Friston, K., Critchley, H., Dolan, R.J. (2003).
Temporal difference models and reward-related learning in the human
brain. Neuron, 38, 32937.
Pickard, N., Strick, P.L. (2001). Imaging the premotor areas. Current
Opinion in Neurobiology, 11, 66372.
Rabbitt, P.M.A. (1966). Error correction time without external error signals.
Nature, 212, 4389.
Seymour, B., ODoherty, J.P., Dayan, P., et al. (2004). Temporal difference
models describe higher-order learning in humans. Nature, 429, 6647.
Summerfield, J.J., Hassabis, D., Maguire, E.A. (2009). Cortical midline
involvement in autobiographical memories. Neuroimage, 44, 1188200.
Yeung, N., Botvinick, M.M., Cohen, J.D. (2004). The neural basis of error
detection: Conflict monitoring and the error-related negativity.
Psychological Review, 111, 93159.