NeuroImage 57 (2011) 686–694

Contents lists available at ScienceDirect

NeuroImage
j o u r n a l h o m e p a g e : w w w. e l s ev i e r. c o m / l o c a t e / y n i m g

Developmental influences on the neural bases of responses to social rejection:

Implications of social neuroscience for education
Catherine L. Sebastian a,b,⁎, Geoffrey C.Y. Tan c, Jonathan P. Roiser a, Essi Viding b,
Iroise Dumontheil a, Sarah-Jayne Blakemore a
a
Institute of Cognitive Neuroscience, University College London, 17 Queen Square, London, WC1N 3AR, UK
b
Clinical, Educational and Health Psychology, University College London, 26 Bedford Way, London, WC1H 0AP, UK
c
Wellcome Trust Centre for Neuroimaging, University College London, 12 Queen Square, London, WC1N 3BG, UK

a r t i c l e i n f o a b s t r a c t

Article history: Relational aggression such as social rejection is common within school peer groups. Converging evidence
Received 3 June 2010 suggests that adolescent females are particularly sensitive to social rejection. We used a novel fMRI adaptation
Revised 20 September 2010 of the Cyberball social rejection paradigm to investigate the neural response to social rejection in 19 mid-
Accepted 24 September 2010
adolescent (aged 14–16) and 16 adult female participants. Across all participants, social exclusion (relative to
Available online 12 October 2010
inclusion) elicited a response in bilateral medial prefrontal cortex (mPFC) extending into ventral and
Keywords:
subgenual anterior cingulate cortex and medial orbitofrontal cortex; and the left ventrolateral PFC (vlPFC);
Social rejection regions that have been associated in previous studies with social evaluation, negative affective processing,
Adolescence and affect regulation respectively. However, the exclusion-related response in right vlPFC, a region associated
fMRI in previous studies with the regulation of rejection-related distress, was attenuated in adolescents. Within
Prefrontal cortex mPFC, greater activation during exclusion vs. inclusion was associated with greater self-reported
Peer relationships susceptibility to peer influence in adolescents but not in adults. This suggests that the brain's response to
Education experimentally-induced social rejection relates to adolescent behaviour in real-world social interactions. We
Social neuroscience
speculate about the potential implications of these findings for educational settings. In particular, functional
development of affective circuitry during adolescence may influence social interaction within the school peer
group.
© 2010 Elsevier Inc. All rights reserved.

Introduction 14.3 years) reported being excluded or ignored by a group of peers

while at school. Being bullied (including relational aggression) is
In exploring the contributions that neuroscience can make to associated with decreased school achievement and psychological
educational practice, it is important to consider the pastoral role that wellbeing (Hawker and Boulton, 2000; Boulton et al., 2008); thus, a
schools must fill, particularly at secondary level. Adolescence is a time greater understanding of adolescent responses to phenomena such as
of opportunity for learning new skills and forging an adult identity. social rejection may contribute to greater understanding of the factors
However, it is also a time of vulnerability, as adolescents begin to face contributing to schooling success and failure (Blakemore, 2010).
adult challenges while still developing physically, socially and Recently, several influential fMRI studies have investigated the neural
cognitively (Steinberg, 2005). Understanding the brain basis of social bases of responses to social rejection. However, responses in adolescents
development and functioning during adolescence is crucial for the and adults have not been directly compared. Given evidence showing
fostering of social competence and psychological wellbeing inside and continuing structural and functional development in social brain regions
outside the classroom (Blakemore, 2010). during adolescence (Blakemore, 2008) the current study investigated
Managing social relationships is a significant challenge for adoles- whether adolescents differ from adults in the neural processing of social
cents. Relational or social forms of bullying are common within school rejection.
peer groups, particularly among adolescent girls. One recent study One of the difficulties in investigating social rejection with fMRI is
(Wang et al., 2009) found that 27.4% of adolescent girls (mean age the choice of an appropriate rejection paradigm. This must aim to
preserve some of the ecological validity of social rejection, while
maintaining sufficient experimental control. The paradigm most
⁎ Corresponding author. Clinical, Educational and Health Psychology, University commonly used with fMRI is the Cyberball game (Williams et al.,
College London, 26 Bedford Way, London, WC1H 0AP, UK. Fax: +44 20 7436 4276.
E-mail addresses: [email protected] (C.L. Sebastian), [email protected]
2000). Participants believe they are playing a game of ‘catch’ over an
(G.C.Y. Tan), [email protected] (J.P. Roiser), [email protected] (E. Viding), internet connection with two other players, whereas the actions of
[email protected] (I. Dumontheil), [email protected] (S.-J. Blakemore). these players are actually pre-programmed to include or exclude the

1053-8119/$ – see front matter © 2010 Elsevier Inc. All rights reserved.
doi:10.1016/j.neuroimage.2010.09.063
 C.L. Sebastian et al. / NeuroImage 57 (2011) 686–694 687

participant. In adults, a positive relationship has been found between studies in adults have shown that social exclusion not only causes
self-reported distress during social exclusion and blood-oxygen-level- distress, but also increases the extent to which individuals conform to
dependent (BOLD) responses in brain regions associated with group norms, possibly because excluded individuals seek to increase
affective and pain processing, including the dorsal anterior cingulate their sense of ‘belonging’ within the group (Williams et al., 2000).
cortex (dACC), amygdala, periaqueductal grey and anterior insula Adolescence is a time when peer approval is of particular salience
(Eisenberger et al., 2003; Eisenberger et al., 2007a; Eisenberger et al., (Steinberg, 2008) and when ability to resist peer influence is still
2007b). A negative relationship between social distress and BOLD developing (Steinberg and Monahan, 2007). Therefore, we addition-
response has also been found in right ventrolateral prefrontal cortex ally investigated whether the neural response to rejection relates to
(vlPFC), a region associated with top-down regulatory control self-reported ability to resist peer influence, and whether this
(Eisenberger et al., 2003). relationship differs between adolescence and adulthood.
The key advantage of the Cyberball paradigm is its ecological The current study employed a modified version of the Cyberball
validity. However, it has been criticised in terms of experimental paradigm in order to explore neural responses to social rejection on a
control. For example, Somerville et al. (2006) argued that the task task with increased experimental control relative to previous studies.
confounds social distress with expectation violation, i.e. participants We predicted that social exclusion would activate a network of
expect to be thrown the ball, but during the exclusion condition they regions involved in the generation of negative affect (ACC, amygdala,
are not. To address this issue, Somerville et al. (2006) designed a task and insula), affect regulation (vlPFC), and social evaluation (mPFC).
in which social expectation and social distress served as separate We aimed to address conflicting findings across previous studies, such
factors within a social evaluation task. As predicted, dACC responses as the role of dACC in social rejection. We also compared adolescents
were elicited when participants' expectations differed from the and adults within the same study. Based on previous studies, we
feedback they received. In contrast, responses in ventral ACC specifically investigated the possibility of group differences in the
(vACC)/medial PFC (mPFC), regions associated with social and self- sgACC (Masten et al., 2009) and vlPFC (Sebastian et al., 2010b).
relevant evaluation, were associated with the difference between Finally, it was predicted that haemodynamic responses in rejection-
positive and negative feedback. This finding was replicated in a similar sensitive brain regions would correlate with real-world measures of
study by Somerville et al. (2010). The current study sought to address social behaviour, and with self-reported distress during Cyberball
these conflicting findings with a novel adaptation of the Cyberball (cf. Eisenberger et al., 2003, 2007a).
task, which improved experimental control while preserving the idea
of an online interaction. Methods
Using this framework, the current study aimed to explore brain
regions responding to social rejection across all participants, as well as Participants
those in which BOLD response may differ between adolescents and
adults. Behavioural studies have shown that adolescent females The study included 19 adolescent and 16 adult native-English
report lower mood and greater anxiety in response to social rejection, speaking female participants. Age and full IQ data (measured with the
relative to both younger children (O'Brien and Bierman, 1988) and two-subscale version of the WASI (1999)) are displayed in Table 1. IQ
adult females (Kloep, 1999; Sebastian et al., 2010a). However, the did not differ between the groups. Adolescents' mean Tanner stage
neurocognitive processes underlying these age differences remain based on a puberty questionnaire adapted from Carskadon and Acebo
unclear. Increased emotional reactivity measured behaviourally may (1993) was 3.49 (SD = .038; mid-late puberty). Participants had no
result from greater activity in regions associated with an affective history of psychiatric, neurological or neurodevelopmental disorder,
response, reduced activity in regulatory regions, or a combination of based on parent report (adolescents) or self-report (adults). Procedures
the two (Eisenberger, 2006). were approved by the local ethics committee. Only females were
One recent fMRI study from our laboratory (Sebastian et al., 2010b) included as previous studies have found sex differences in brain
found attenuated right vlPFC responses in adolescents relative to adults development trajectory during adolescence (Giedd et al., 1999).
during the processing of rejection-related words. As mentioned above, Averaging across males and females might therefore yield an inaccurate
previous studies have suggested that right vlPFC may regulate the picture and reduce statistical power. Females were chosen as they have
affective consequences of social rejection (Eisenberger et al., 2003; been reported to be more sensitive to social rejection than males (Crick
Masten et al., 2009). This finding was therefore consistent with models et al., 2002), and adolescent girls are more likely than boys to use or to be
of adolescent cognition such as the Social Information Processing subjected to social isolation as a bullying technique (Wang et al., 2009).
Network model (Nelson et al., 2005), which suggest that prefrontal
regulatory regions such as vlPFC regulate affective states less effectively Experimental task
in adolescents than in adults, resulting in a more acute subjective
experience of social rejection. There is also some indirect evidence that An fMRI-adapted version of the original Cyberball game was used. To
the response to social rejection in affective brain regions may be improve experimental control, the task was reprogrammed from the
increased in adolescents. One previous study to use the Cyberball original html into Matlab 6.5 using the Cogent toolbox (http://www.
paradigm with fMRI in adolescents (Masten et al., 2009) reported a
positive correlation between subgenual ACC (sgACC) responses and
exclusion-related distress in adolescents. A response in this region had Table 1
Participant variables for each Age group.
not been found in previous studies with adults (e.g. Eisenberger et al.,
2003), although adolescents and adults were not directly compared in Adolescents Adults
the same study. Therefore in the current study we tested whether Age Mean 15.44 28.70
adolescents engage this region to a greater extent than do adults in the Range 14.00–16.97 23.87–38.84
processing of social rejection, in addition to exploring possible group SD .81 3.91
WASI IQa Mean 114.00 120.56
differences in regulatory regions such as vlPFC.
Range 91–130 99–133
It is also useful to consider whether the brain's response to social SD 11.96 9.24
rejection relates to behaviour in the real world. Eisenberger et al. Tanner Stage Mean 3.49 –
(2007a) found that individuals with the highest ratings of perceived Range 2.63–4.00 –
social distress during real-world interactions also showed greater SD .38 –

responses in dACC and amygdala during Cyberball. Behavioural a

No significant Age group difference in IQ scores: t(33) = 1.79, p = .083.
688 C.L. Sebastian et al. / NeuroImage 57 (2011) 686–694

vislab.ucl.ac.uk/cogent.php). Compared with previous studies using this with the constraint that the same block type was never presented more
paradigm with fMRI, the game was modified in the following ways. The than twice in a row. Block length was approximately 24 s. This balanced
use of a fixed condition order (inclusion followed by exclusion) was optimal signal-to-noise ratio with blocks long enough for participants to
replaced with short blocks of each condition in a pseudorandom order. be able to notice that they were being included or excluded. Exact block
This reduced the possibility of expectation violation, and order effects length varied slightly with each block due to a number of factors
including neural adaptation. Blocks were also shortened to increase designed to increase ecological validity. The game was self-paced, and
design efficiency, in line with a recent Cyberball fMRI study (Kumar the onset of the block was defined as the onset of the participant's first
et al., 2009). Additionally, motor responses were matched between throw. The offset of the block was defined as the end of the last video
conditions by requiring participants to ‘mark’ with a key press whenever preceding the onset of the fixation cross. Mean (SD) block length across
one of the other players caught the ball: this ensured that participants all participants and conditions was 24.35 (1.27) s. After each block, a
made motor responses during both inclusion and exclusion blocks. This fixation cross was displayed. Mean (SD) fixation time was 16.68 (1.11) s.
also removed any differences in inhibitory requirements between The timing and structure of each block is shown in Fig. 1.
conditions, as otherwise during exclusion blocks, participants may be
preparing movements that they never execute. It also ensured that Questionnaire measures
participants were paying attention to the game equally in both
conditions. Prior to scanning, participants completed the State/Trait Anxiety
As in previous Cyberball studies, participants were told prior to Inventory (Spielberger, 1983). Following scanning, participants
scanning that we were interested in ‘mental visualisation’ ability (to completed a measure of social distress comprising four need threat
avoid the topic of social interaction), and that they would play a game scales taken from Williams et al. (2000). The four needs comprise self-
of catch over the internet with two other players. As part of the esteem, belonging, meaningful existence and control, and combined
‘mental visualisation’ exercise, we instructed participants to imagine ratings have been used as a measure of social distress in previous
the experience was as real as possible. However, unlike previous studies (Eisenberger et al., 2003, 2007a,b; Masten et al., 2009). As
Cyberball studies, we did not explicitly state whether the other inclusion and exclusion conditions were pseudorandomised through-
players were real people or not because our more controlled out the scanning session, participants completed two sets of
experimental design rendered an elaborate cover story much less questionnaires following scanning, in which they were asked to
believable. This decision was taken on the basis of evidence showing think back to the times for which either ‘the other players were
that participants report feeling equally distressed by rejection when throwing the ball to you’ or ‘the other players were not throwing the
they know they are playing against a computer as when they believe ball to you’. After each social distress questionnaire, participants also
the other players are real (Zadro et al., 2004). This is in line with rated their current mood using a scale taken from Williams et al.
theories suggesting that the immediate distress caused by social (2000). The order in which participants completed these two
exclusion is an automatic and reflexive response (Williams, 2007). questionnaires was counterbalanced across participants.
Each block began with an instruction screen telling participants to Participants also completed the Resistance to Peer Influence
“Get ready to throw the ball”. Participants chose which of the other questionnaire (RPI; Steinberg and Monahan, 2007). This asks partici-
players to throw the ball to using a right index or middle finger key press pants to rate on a 4 point scale the extent to which 10 statements
response. Participants also marked each time one of the other players pertaining to peer influence apply to them, e.g. ‘Some people go along
received the ball with a key press response. Participants had to press this with their friends just to keep their friends happy.’ The questionnaire
‘mark’ key before the game could continue. To heighten the sense that has been validated as addressing susceptibility to peer influence as
the other players were deciding to whom to throw the ball, a random distinct from willingness to engage in anti-social behaviour. At the very
interval between .5 and 1 s was introduced after the participant had end of the session, participants completed a manipulation check
pressed the mark key before the ball was thrown again. In the inclusion measure from Williams et al. (2000) which asked participants whether
condition, the other players threw the ball to the participant with an 80% they noticed that they had been included and excluded in different
probability, i.e. the participant was over-included (a 50% probability conditions during the scan, and asked them to estimate percentage ball
would have yielded equal inclusion of the two other players). This figure possession for inclusion and exclusion conditions.
of 80% was chosen on the basis of task piloting. This showed that, in the
short blocks suitable for fMRI, participants needed to be over-included fMRI data acquisition
in inclusion blocks in order for them to register systematic differences
between the inclusion and exclusion conditions. A 1.5 T Siemens Sonata MRI scanner was used to acquire 12 min 3D
There was a total of 12 blocks (six each of inclusion and exclusion). T1-weighted structural images, and multislice T2*-weighted echo planar
The order of inclusion and exclusion blocks was pseudorandomised, volumes with BOLD contrast. The T2* EPI sequence was optimised to

Fig. 1. Timeline showing one block (of 12) of the Cyberball paradigm. The time of the participant's first throw defined the onset of each game. Following this throw, participants were
included or excluded by the other players with equal probability. Each game lasted approximately 24 s.
 C.L. Sebastian et al. / NeuroImage 57 (2011) 686–694 689

reduce dropout in the orbitofrontal cortex (Weiskopf et al., 2006), and significance threshold of p b .05 FWE was used. Interactions were
used the following acquisition parameters: 33 2 mm slices acquired in a interrogated using t-tests. Covariate analyses including participant age
descending trajectory with a 1 mm gap, TE=50 ms; TR= 90 ms; slice and self-report measures were also conducted in a similar manner.
tilt= −30° (TN C); flip angle=90°; field of view= 192 mm; matrix
size= 64 ×72. Functional data were acquired in a single scanning session Results
of approximately 9.5 min, in which 191 volumes were acquired.
Questionnaire data collected after scanning
fMRI data analysis
State and trait anxiety
Imaging data were analysed using SPM5 (www.fil.ion.ucl.ac.uk/spm). There were no group differences in trait anxiety: t(33) = − 1.40,
The first five functional image volumes from each run were discarded to p = .17, or state anxiety: t(33) = −.63, p = .54. Across all participants,
allow for T1 equilibrium effects, leaving 186 image volumes per trait anxiety M = 37.80, SD = 9.81; and state anxiety M = 33.31,
participant. Pre-processing included rigid-body transformation (realign- SD = 7.53.
ment) and normalisation into the standard space defined by the Montreal
Neurological Institute (MNI) template with a voxel size of 3 ×3 × 3mm, Resistance to peer influence
and smoothing with a Gaussian filter of 8 mm full width at half Adults reported significantly greater resistance to peer influence
maximum. (RPI) (M = 3.11, SD = .25) than did adolescents (M = 2.91, SD = .28;
A block analysis was conducted in order to compare neural activity t(33) = 2.15, p = .039).
associated with inclusion and exclusion conditions. The time series of
186 image volumes was deconstructed into 5 block types, each of Manipulation check
which was included as a separate regressor in the design matrix. One participant did not give an estimate of ball possession. For the
These consisted of inclusion, exclusion, baseline fixation, times during remaining 34 participants, a Condition (inclusion, exclusion) by Age
which the instruction screen was displayed, and times during which group (adult, adolescent) mixed model ANOVA showed a main effect of
the start of the game was displayed but before the participant had Condition: F(1,32) = 63.75, p b .001, η2p = .67, with a higher percentage
thrown the first ball (which defined block onset). These latter two estimated in inclusion (M= 43.34%, SD = 18.34) than in exclusion
block types were included as regressors of no interest. These 5 blocks (M= 12.09%, SD= 9.82%). There was no main effect of Age group,
regressors were modelled as boxcar functions convolved with a or Condition by Age group interaction (ps N .05). These estimates also
canonical haemodynamic response function. The six realignment did not differ from actual ball possession in inclusion (possession
parameters were modelled as effects of no interest, in order to account M = 42.01%, SD = 2.63%, t(34) = .43, p = .67) and exclusion (possession
for any variance due to head movement. For three participants M = 13.31%, SD= .68%, t(33) = −.73, p = .47) blocks.
(1 adult, 2 adolescents) extra regressors were included to model a Participants also rated how much they agreed with the statements
small number of corrupted images resulting from excessive motion. “I was ignored” and “I was excluded” during inclusion and exclusion
These images (amounting to 1–1.6% of those 3 participants' data) conditions. In a Condition by Age group ANOVA, there was a main
were removed and the adjacent images interpolated in order to effect of Condition: F(1,33) = 147.63, p b .001, η2p = .82, with greater
prevent distortion of the between-subjects mask. Data were high-pass feelings of exclusion in the exclusion condition (M = 3.94, SD = 1.04)
filtered at 128 s to remove low-frequency drifts. than in the inclusion condition (M = 1.40, SD = .71). There was no
At the first level, the main contrast of interest, exclusion N inclusion main effect of Age group. However, there was a Condition by Age
(as well as inclusion N exclusion), was conducted for each participant. group interaction: F(1,33) = 10.82, p = .002, η2p = .25. Both groups felt
Random effects analyses were performed by entering the resulting significantly more excluded in the exclusion condition than in
contrast images into a second-level analysis, where Age group served as inclusion (ps b .001), but adolescents reported feeling more excluded
a between-subjects variable. T-statistics were calculated at the whole during the exclusion condition (p = .011), and included during the
brain level for the main effects of Condition across all participants inclusion condition (p = .047) than did adults.
(exclusion N inclusion and inclusion N exclusion), and for the interaction
between Condition and Age group. Social distress
For the main effects of exclusion N inclusion and inclusion N exclu- Across all four needs, greater need threat was reported during
sion, regions reaching cluster-level significance at p b .05, FWE- exclusion than inclusion across all participants (ps b .001). Therefore,
corrected (following initial thresholding at p b .001, uncorrected), scores were averaged to create one overall ‘social distress’ variable,
are reported. Where peaks fell in predicted regions, post-hoc analyses with higher scores indicating greater distress (scale from 1 to 5). This
were conducted on contrast estimates from the peak voxel to allowed comparison with previous Cyberball fMRI studies, which have
investigate relationships with self-report measures. Peak voxel data also used this aggregate measure of distress. Mean distress across the
are a weighted average of the surrounding voxels due to smoothing. A whole sample was 1.86 (SD = .45) for inclusion, and 3.70 (SD = .87)
peak voxel approach was chosen because using peak voxels for for exclusion. In a Condition by Age group ANOVA, there was a main
correlation analyses avoids having to arbitrarily choose a shape or effect of Condition: F(1,33) = 125.05, pb.001, η2p = .79, with greater
statistical threshold to define a cluster of interest. distress after exclusion than inclusion. There was also a crossover
For the Condition by Age group interaction analysis, the SPM was interaction between Condition and Age group: F(1,33) = 13.45,
initially thresholded at p b .001, uncorrected. Small volume correction p = .001, η2p = .29, with lower distress during inclusion and higher
(SVC) was performed using spheres of 8 mm radius in two regions distress during exclusion in adolescents compared with adults
where the prediction of an Age group effect was strongest based on (ps b .001). There was no correlation between age and distress
previous studies: right vlPFC (adult N adolescent) and right sgACC following exclusion, or the difference score for distress following
(adolescentN adult). The coordinate for right vlPFC (MNI [45 30 −6]) exclusion relative to inclusion, within each Age group (psN.20).
was taken from Sebastian et al. (2010b). In this previous study, which
used a rejection-themed emotional Stroop task, this coordinate was Mood
associated with significantly greater activity in adults compared with Mood ratings were averaged to produce an overall mood score (from
adolescents in response to rejection words relative to acceptance words. 1 to 7), with higher scores indicating better mood. A Condition by Age
The coordinate for sgACC (MNI [8 22 −4]) was taken from the main group ANOVA showed a main effect of Condition: F(1,33)= 11.11,
effect peak (exclusion N inclusion) from Masten et al. (2009). A p= .002, η2p = .25, with lower mood reported after participants had
690 C.L. Sebastian et al. / NeuroImage 57 (2011) 686–694

thought about exclusion blocks (M= 5.44, SD =1.21) than inclusion contrast (inclusion N exclusion), significant clusters fell in the bilateral
blocks (M=5.89, SD =1.08). There was no main effect of Age group: occipital cortex, right superior temporal sulcus, and left dorsolateral PFC.
F(1,33) =.002, p= .97, or Condition by Age group interaction: F(1,33)=
.23, p =.64. Effects of age and self-report variables in predicted main effect peaks
Post-hoc analyses were conducted on contrast estimates in the two
Behavioural responses during scanning peak voxels from clusters in predicted regions: left mPFC ([−12 42 −9])
As the games were self-paced, the number of button presses and left vlPFC ([−48 33 −15]). Resistance to Peer Influence (RPI) score
(combining throws and mark responses) made in inclusion and exclusion and social distress were used as continuous self-report variables. There
conditions were compared in a mixed model Condition by Age group was an interaction between Age group and RPI score in the left mPFC
ANOVA. There was a main effect of Condition: F(1,33) = 162.88, peak: F(1,31) =5.72, p=.023. This was driven by a negative correlation
pb.001, η2p =.83, due to a higher mean number of throws per block in between BOLD response and RPI score in the adolescents: r= −.51,
the inclusion condition (M= 8.71, SD =.61) than in the exclusion p= .025, but not in the adults: r =.27, p= .31 (Fig. 3). No other
condition (M= 7.53, SD= .37). This was likely because the videos in relationships between self-report variables and BOLD signal reached
which the other characters passed the ball to each other were slightly significance in either peak.
longer than those involving the participant, as RTs were actually slightly
faster in the exclusion condition (M= 500 ms, SD = 160) than in Condition by Age group interaction
inclusion, (M= 555 ms, SD = 192; though this difference was not Small volume correction (SVC) performed in the right vlPFC revealed
significant (p=.094)). There was no main effect of Age group or a significant interaction in the predicted direction (adult N adolescent)
Condition by Age group interaction for either the mean number of throws (MNI peak coordinate [45 27 −3], k = 2, t = 3.46, z = 3.17, voxel-level
per game, or mean RTs to make/mark each throw. The choice of which FWE-corrected p = .048). Post-hoc t-tests showed greater activation
player to throw to on each trial also did not vary between Age groups during exclusion than inclusion in adults: t(15) = 3.00, p = .005,
(psN.40). Across inclusion and exclusion (in which the first throw was uncorrected; and greater activation during inclusion than exclusion in
made by the participant) 52.04% of throws were made to the player on adolescents: t(18)= −2.35, p = .015, uncorrected (Fig. 4). No interac-
the left, and 47.96% to the player on the right. tion was found between Condition and Age group in the sgACC.

fMRI data Whole brain regression analyses

Results surviving cluster-level correction at p b .05 for the contrasts There were no significant correlations between BOLD signal and
exclusion N inclusion and inclusion N exclusion are presented in age within the adolescent group in predicted regions, potentially due
Table 2. to the small age range within this group. Controlling for significant
Age group differences in distress ratings and RPI scores, there were no
Main effect of condition significant relationships between BOLD signal and either of these self-
Significantly increased BOLD responses to exclusion relative to report measures in predicted regions.
inclusion falling in predicted regions were seen in the left vlPFC (BA 47),
and in a large bilateral cluster with its peak in mPFC (Fig. 2). As this Discussion
midline cluster was so extensive, Brodmann area masks were
systematically applied using WFU PickAtlas (Maldjian et al., 2003) in The present study investigated neural responses to social rejection in
order to determine which specific brain regions it comprised. Regions healthy adolescent and adult females using an adapted version of the
surviving cluster-level FWE correction were identified in each of Cyberball social exclusion paradigm. Self-report data collected outside
Brodmann areas 10, 11, 25 and 24/32 (medial PFC, medial OFC, the scanner replicated previous studies (Sebastian et al., 2010a), with
subgenual ACC, ventral ACC, respectively). Significant clusters in non- adolescents finding social exclusion more distressing than adults.
predicted regions were seen in the bilateral occipital cortex, left fusiform Neural responses to social exclusion (relative to inclusion) across all
gyrus, right anterior temporal cortex, and caudate. For the reverse participants were seen in a large cortical midline cluster and in left

Table 2
Main effects for exclusion N inclusion and inclusion N exclusion surviving FWE correction at the cluster level. BA = Brodmann area; L/R = Left/Right; Peak voxel = xyz voxel
coordinates in MNI space; k = cluster size in 3 × 3 × 3mm voxels; z = z-score; ⁎result in predicted region.

Brain region BA L/R Peak voxel k z Cluster correctedp-value

Main effect of Condition (exclusion N inclusion)

Occipital cortex (middle occipital gyrus) 18 R 24 − 87 − 9 225 5.43 b.001
17 R 15 − 90 − 9 5.23
17 R 9 − 93 − 3 4.36
Medial PFC (medial frontal gyrus) 10 L − 12 42 − 9 193 4.97 b.001⁎
(Ext. ventral/subgenual ACC) 24/32 R 3 24 − 9 4.62
(Ext. medial OFC) 11 L − 3 39 − 15 4.41
Occipital cortex (cuneus) 18 L − 24 − 99 − 6 79 4.58 b.001
Inferior temporal cortex (fusiform gyrus) 20 L − 42 − 15 − 24 83 4.40 b.001
Ventrolateral PFC (inferior frontal gyrus) 47 L − 48 33 − 15 32 4.39 .042⁎
47 L − 39 33 − 12 3.12
Caudate tail – L − 18 − 42 18 40 4.23 .016
Anterior temporal cortex (middle temporal gyrus) 21 R 48 9 − 33 71 4.15 .001

Main effect of Condition (inclusion N exclusion)

Occipital cortex (cuneus, ext. bilaterally) 30 – 0 − 69 3 328 5.71 b.001
– R 3 − 66 − 15 4.33
Superior temporal sulcus 39 R 48 − 51 15 35 4.30 .029
Dorsolateral PFC (middle frontal gyrus) 10 L − 36 51 18 35 3.92 .029
46 L − 42 45 12 3.80
10 L − 27 51 15 3.44
 C.L. Sebastian et al. / NeuroImage 57 (2011) 686–694 691

Fig. 2. Main effects for the contrast exclusion N inclusion. Elevated BOLD response to exclusion relative to inclusion was seen in the following predicted regions: a) a cluster in medial
PFC extending into subgenual/ventral ACC and medial OFC bilaterally; and b) left ventrolateral PFC. Results are shown at a threshold of p b .001, overlaid on an average structural scan
from all 35 participants.

vlPFC. In the mPFC peak, there was a negative relationship between Masten et al., 2009), in line with lesion evidence suggesting that right
BOLD response and RPI score in the adolescents, but not in the adult vlPFC is crucial for inhibitory control (Aron et al., 2004). However, recent
group. Comparing age groups directly revealed that adults activated work suggests that this region is recruited bilaterally during inhibitory
right vlPFC in response to exclusion (relative to inclusion), more than control (Hampshire et al., 2010), and a left vlPFC response in a very
did adolescents. This is in line with previous studies suggesting similar region to that reported here has been shown to correlate with
functional development between adolescence and adulthood in this emotion reappraisal success (Wager et al., 2008). One additional
regulatory region, and suggests that increased sensitivity to rejection in possibility is that the Condition by Age group crossover interaction
adolescence may result from reduced regulation of social distress. found in right vlPFC (discussed below), may have obscured a main effect
in this region. Previous studies have found activity in vlPFC to correlate
Effects of Cyberball task adaptations negatively with reported distress (Eisenberger et al., 2003; Masten et al.,
2009), suggesting that this region is involved in the regulation of
One aim of this study was to explore whether brain regions negative affect caused by exclusion. The inclusion of the ‘mark’ key press
involved in social and affective processing would respond to social response during exclusion blocks in the current study additionally
exclusion (relative to inclusion) in a version of the Cyberball paradigm reduced the possibility that vlPFC activity may reflect inhibition of a
with improved experimental control. The task was indeed capable of prepared motor response, and thus increases the specificity of the
eliciting robust effects, largely in predicted regions. finding to social rejection.
Across all participants, a response was seen in the left vlPFC (inferior The other significant cluster had its peak in left mPFC, but extended
frontal gyrus, orbital part). Effects of exclusion in this region in previous into vACC, sgACC and mOFC bilaterally. This result provides a partial
Cyberball studies have been right lateralised (Eisenberger et al., 2003, replication of previous studies. Responses in the current study were

Fig. 3. In the medial PFC peak voxel for the contrast exclusion N inclusion ([− 12 42 − 9]), there was an inverse relationship between BOLD response and Resistance to Peer Influence
(RPI) score in the adolescents (in black) but not in the adults (in grey). The interaction between Age group and RPI score was significant.
692 C.L. Sebastian et al. / NeuroImage 57 (2011) 686–694

Fig. 4. Condition by Age group interaction in the right ventrolateral PFC ([45 27 − 3]) (shown inset, at a threshold of p b .005, overlaid on an average structural scan from all 35
participants). Adults activated this region more in response to exclusion than inclusion, while adolescents exhibited the reverse pattern. Bars indicate the pooled standard error of
the difference for exclusion–inclusion.

confined to the ventral portion of the PFC, and did not include the dorsal feedback relative to negative, in contrast to the present study. Since
ACC (dACC), a region activated during rejection in previous studies the cluster in the present study includes the peak reported by Somerville
(Eisenberger et al., 2003, 2007a,b). The lack of a dACC response may have et al. ([−6 49 −13]), task differences may explain the discrepancy. Both
been due to methodological differences between this and previous tasks involved social evaluation; however, there may be similarities
Cyberball studies. In particular, motor responses were matched between between the rejection condition in the present study, and the
conditions in the current study. In addition, the current study used acceptance condition in the studies by Somerville et al. For example,
multiple and randomised blocks, which aimed to reduce the possibility of Somerville et al. (2010) found activity in this region was enhanced to
expectation violation during the exclusion condition (Somerville et al., positive feedback specifically in individuals with low self-esteem. In
2006). these participants, positive feedback may have triggered a re-evaluation
In a study of adolescents aged 12–13, Masten et al. (2009) also of the self, while rejection feedback did not. In contrast, previous
failed to find a dACC response to rejection, but instead found a positive Cyberball studies have shown that it is only the exclusion condition (and
association between sgACC response and self-reported distress. In the not inclusion) that causes a shift in affective ratings and self-esteem
current study, sgACC responded to exclusion relative to inclusion relative to baseline measurement (Sebastian et al., 2010a). As such, it is
across the entire sample. This is in line with previous studies showing exclusion, rather than inclusion, that would be more likely to trigger
sgACC activation to negatively valenced stimuli in healthy adults self-evaluation processes mediated by mPFC.
(George et al., 1995; Haas et al., 2007), and with evidence showing a Finally, the medial cluster also extended into medial OFC (BA 11).
relationship between activity in this region and symptom severity in This region has not previously been reported in studies of social
major depression (Keedwell et al., 2009). The sgACC has autonomic rejection, possibly because this is a region commonly affected by
connections with the central nucleus of the amygdala and periaque- susceptibility artefact. The present study used an EPI sequence that
ductal grey (Neafsey et al., 1993), both of which are involved in the had been optimised to detect signal within the orbitofrontal cortex
perception of affect, including social pain (Eisenberger, 2006). It has and compensates for susceptibility artefact (Weiskopf et al., 2006).
been suggested that sgACC integrates autonomic information and is Medial OFC predicts reward and punishment outcomes in order to
preferentially activated by personally relevant emotional experience guide flexible behaviour (Schoenbaum et al., 2009), a role which may
as opposed to simple nociception (Vogt, 2005). The current data are in extend to encoding the value associated with a particular person in a
line with this account, suggesting a role for sgACC in the affective social encounter (Behrens et al., 2009). Over the course of an
experience of social pain. exclusion block, this value might need to be updated.
The response in mPFC (BA10) and vACC (BA24/32) is partially in line Although the new task aimed to preserve some of the ecological
with predictions. The ventral portion of mPFC commonly responds validity of the original, some compromises were inevitable. For
during self-evaluation (Kelley et al., 2002; Macrae et al., 2004), and self- example, the short, randomised blocks made the task less realistic,
evaluation processes are automatically triggered by social exclusion and so we decided not to explicitly deceive participants into believing
(Leary et al., 1995; Williams, 2007). This region may also regulate that the other players were real people. However, since Zadro et al.
negative affect via connections with rostral ACC and amygdala (2004) has shown that participants are equally distressed by rejection
(Buckholtz et al., 2008). However, Somerville et al. (2006, 2010) when they know they are playing Cyberball with a computer as with
found that activity in a region of the mPFC encompassing both regions ‘real’ people, it was decided that this was an acceptable compromise.
(termed vACC/mPFC) responded preferentially to positive social One further difference between this and previous Cyberball fMRI
 C.L. Sebastian et al. / NeuroImage 57 (2011) 686–694 693

studies, is that no relationships were found between BOLD responses In sum, this study demonstrated that social brain regions respond
and self-reported distress. The short blocks and randomised condition to rejection in a version of the Cyberball game optimised for
order meant that it was difficult to assess online subjective distress experimental control. This may be of use in future investigations
during inclusion and exclusion conditions. Instead, participants into the neural bases of responses to social rejection. Differences
completed a single questionnaire after the scanning session. The between adolescents and adults in regions including right vlPFC and
questionnaire may therefore have lacked specificity to capture a left mPFC suggest continuing maturation of regions subserving
particular affective response at a particular time. Overall, however, the emotional control and self-evaluative processing during social
new task was able to demonstrate that social, affective and regulatory rejection. Future studies could investigate the extent to which
brain regions respond to social exclusion, relative to inclusion, in a individual differences in the experience of rejection in everyday life
relatively minimal online social rejection paradigm which aimed to might mediate these effects. Future work could also build on the
optimise experimental control. current findings to increase applicability to the educational setting, for
example by exploring whether emotion regulation training might
mitigate some of the negative affective consequences of social
Developmental differences bullying. As we learn more about the extent to which social cognitive
abilities continue to develop during adolescence, it is possible that the
A further aim of this study was to investigate the neural bases of development of skills such as self-awareness and emotion regulation
behavioural evidence suggesting that adolescent females report will be factored into the school curriculum.
heightened negative affect relative to adults following social rejection
(Sebastian et al., 2010a). Social rejection is experienced by a Acknowledgments
significant minority of adolescents on a regular basis (Wang et al.,
2009). It is therefore important to understand how an experience like This study was supported by the BBSRC (CLS), A*STAR (GCYT), and
social rejection impacts on the developing brain as compared with the Royal Society (SJB). SJB is a Royal Society University Research
adults. Fellow.
An interaction between Condition and Age group was found in a
small cluster in right vlPFC, with a greater response in this region in
adults than in adolescents. This was driven by a greater response to References
exclusion than inclusion in the adults, and the reverse pattern in the Aron, A.R., Robbins, T.W., Poldrack, R.A., 2004. Inhibition and the right inferior frontal
adolescents. This finding is in agreement with previous studies cortex. Trends Cogn. Sci. 8, 170–177.
Behrens, T.E., Hunt, L.T., Rushworth, M.F., 2009. The computation of social behavior.
showing a reduced or reversed response in right vlPFC in adolescents
Science 324, 1160–1164.
compared with adults during affective and behavioural control tasks Blakemore, S.J., 2008. The social brain in adolescence. Nat. Rev. Neurosci. 9, 267–277.
(Sebastian et al., 2010b; Eshel et al., 2007). As discussed above, vlPFC Blakemore, S.J., 2010. The developing social brain: implications for education. Neuron
has been characterised in terms of its role in inhibitory aspects of 65, 744–747.
Boulton, M.J., Trueman, M., Murray, L., 2008. Associations between peer victimization,
cognitive control (Aron et al., 2004), including the regulation of fear of future victimization and disrupted concentration on class work among
negative affect (Eisenberger et al., 2003; Masten et al., 2009), junior school pupils. Br. J. Educ. Psychol. 78, 473–489.
potentially via connections with subcortical structures such as the Buckholtz, J.W., Callicott, J.H., Kolachana, B., Hariri, A.R., Goldberg, T.E., Genderson, M.,
Egan, M.F., Mattay, V.S., Weinberger, D.R., Meyer-Lindenberg, A., 2008. Genetic
amygdala (Wager et al., 2008; Ochsner and Gross, 2005). It may be variation in MAOA modulates ventromedial prefrontal circuitry mediating
that the atypical adolescent response seen in this region underlies individual differences in human personality. Mol. Psychiatry 13, 313–324.
increased levels of self-reported distress following rejection in this Carskadon, M.A., Acebo, C., 1993. A self-administered rating scale for pubertal
development. J. Adolesc. Health 14, 190–195.
group relative to adults. Understanding that adolescents may not have Crick, N.R., Grotpeter, J.K., Bigbee, M.A., 2002. Relationally and physically aggressive
the same regulatory resources to bring to bear on an experience of children's intent attributions and feelings of distress for relational and instrumental
social rejection may inform strategies for helping victims of social peer provocations. Child Dev. 73, 1134–1142.
Eisenberger, N.I., 2006. Identifying the neural correlates underlying social pain:
bullying to overcome their responses to it.
implications for developmental processes. Hum. Dev. 49, 273–293.
Correlational analyses revealed a further developmental effect. In Eisenberger, N.I., Lieberman, M.D., Williams, K.D., 2003. Does rejection hurt? An fMRI
the mPFC peak, an interaction was found between age and RPI score study of social exclusion. Science 302, 290–292.
Eisenberger, N.I., Gable, S.L., Lieberman, M.D., 2007a. Functional magnetic resonance
(accounting for the overall difference between Age groups). In
imaging responses relate to differences in real-world social experience. Emotion 7,
adolescents, but not in adults, mPFC activity was negatively correlated 745–754.
with RPI score, i.e. the less this region responded to social rejection, Eisenberger, N.I., Way, B.M., Taylor, S.E., Welch, W.T., Lieberman, M.D., 2007b.
the more adolescents reported being resistant to peer influence. The Understanding genetic risk for aggression: clues from the brain's response to
social exclusion. Biol. Psychiatry 61, 1100–1108.
specificity of the effect for the adolescent group suggests that the Eshel, N., Nelson, E.E., Blair, R.J., Pine, D.S., Ernst, M., 2007. Neural substrates of choice
relationship may be driven by a developmental process in this region. selection in adults and adolescents: development of the ventrolateral prefrontal
Behavioural data suggests that the ability to resist peer influence and anterior cingulate cortices. Neuropsychologia 45, 1270–1279.
George, M.S., Ketter, T.A., Parekh, P.I., Horwitz, B., Herscovitch, P., Post, R.M., 1995. Brain
develops in a linear fashion between the ages of 14 and 18 (a range activity during transient sadness and happiness in healthy women. Am. J.
which includes the adolescent group), and plateaus thereafter Psychiatry 152, 341–351.
(Steinberg and Monahan, 2007). The mPFC is known to mediate Giedd, J.N., Blumenthal, J., Jeffries, N.O., Castellanos, F.X., Liu, H., Zijdenbos, A., Paus, T.,
Evans, A.C., Rapoport, J.L., 1999. Brain development during childhood and
self-evaluative processes (Kelley et al., 2002; Macrae et al., 2004), and adolescence: a longitudinal MRI study. Nat. Neurosci. 2, 861–863.
adolescents may rely on others' appraisals of them during self- Haas, B.W., Omura, K., Constable, R.T., Canli, T., 2007. Emotional conflict and
evaluation more than do adults (Pfeifer et al., 2009). Although neuroticism: personality-dependent activation in the amygdala and subgenual
anterior cingulate. Behav. Neurosci. 121, 249–256.
speculative, the observed reduction in mPFC response with increasing
Hampshire, A., Chamberlain, S.R., Monti, M.M., Duncan, J., Owen, A.M., 2010. The role of
RPI score may reflect a similar reduction in the extent to which the right inferior frontal gyrus: inhibition and attentional control. NeuroImage. 50,
external cues (such as social rejection) trigger a self-evaluative 1313–1319.
Hawker, D.S., Boulton, M.J., 2000. Twenty years' research on peer victimization and
response in the adolescents. It was noted that RPI scores showed
psychosocial maladjustment: a meta-analytic review of cross-sectional studies.
relatively small variance in adults compared to adolescents. This may J. Child Psychol. Psychiatry 41, 441–455.
be one reason for the absence of neural association in adults. An Keedwell, P., Drapier, D., Surguladze, S., Giampietro, V., Brammer, M., Phillips, M., 2009.
alternative explanation is that the larger variance in adolescents is Neural markers of symptomatic improvement during antidepressant therapy in
severe depression: subgenual cingulate and visual cortical responses to sad, but not
mediated by developmental processes in regions such as the mPFC, happy, facial stimuli are correlated with changes in symptom score. J. Psycho-
with inter-individual variance on this trait decreasing in adulthood. pharmacol. 23, 775–788.
694 C.L. Sebastian et al. / NeuroImage 57 (2011) 686–694

Kelley, W.M., Macrae, C.N., Wyland, C.L., Caglar, S., Inati, S., Heatherton, T.F., 2002. Sebastian, C.L., Roiser, J.P., Tan, G.C., Viding, E., Wood, N.W., Blakemore, S.J., 2010a.
Finding the self? An event-related fMRI study. J. Cogn. Neurosci. 14, 785–794. Effects of age and MAOA genotype on the neural processing of social rejection.
Kloep, M., 1999. Love is all you need? Focusing on adolescents' life concerns from an Genes Brain Behav. 9, 628–637.
ecological point of view. J. Adolesc 22, 49–63. Sebastian, C., Viding, E., Williams, K.D., Blakemore, S.J., 2010b. Social brain development
Kumar, P., Waiter, G., Ahearn, T., Milders, M., Reid, I., Steele, J.D., 2009. Frontal and the affective consequences of ostracism in adolescence. Brain Cogn. 72, 134–145.
operculum temporal difference signals and social motor response learning. Hum. Somerville, L.H., Heatherton, T.F., Kelley, W.M., 2006. Anterior cingulate cortex responds
Brain Mapp. 30, 1421–1430. differentially to expectancy violation and social rejection. Nat. Neurosci. 9, 1007–1008.
Leary, M.R., Terdal, S.K., Tambor, E.S., Downs, D.L., 1995. Self-esteem as an interpersonal Somerville, L.H., Kelley, W.M., Heatherton, T.F., 2010. Self-esteem modulates medial
monitor — the sociometer hypothesis. J. Pers. Soc. Psychol. 68, 518–530. prefrontal cortical responses to evaluative social feedback. Cereb. Cortex advanced
Macrae, C.N., Moran, J.M., Heatherton, T.F., Banfield, J.F., Kelley, W.M., 2004. Medial online publication..
prefrontal activity predicts memory for self. Cereb. Cortex 14, 647–654. Spielberger, C., 1983. Manual for the State–Trait Anxiety Inventory. Consulting
Maldjian, J.A., Laurienti, P.J., Kraft, R.A., Burdette, J.H., 2003. An automated method for Psychologists Press, Palo Alto, CA.
neuroanatomic and cytoarchitectonic atlas-based interrogation of fMRI data sets. Steinberg, L., 2005. Cognitive and affective development in adolescence. Trends Cogn
NeuroImage 19, 1233–1239. Sci. 9, 69–74.
Masten, C.L., Eisenberger, N.I., Borofsky, L.A., Pfeifer, J.H., McNealy, K., Mazziotta, J.C., Steinberg, L., 2008. A social neuroscience perspective on adolescent risk-taking. Dev.
Dapretto, M., 2009. Neural correlates of social exclusion during adolescence: Rev. 28, 78–106.
understanding the distress of peer rejection. Soc. Cogn Affect. Neurosci. 4, 143–157. Steinberg, L., Monahan, K.C., 2007. Age differences in resistance to peer influence. Dev.
Neafsey, E.J., Terreberry, R.R., Hurley, K.M., Ruit, K.G., Frysztak, R.J., 1993. Anterior cingulate Psychol. 43, 1531–1543.
cortex in rodents: connections, visceral control functions, and implications for Vogt, B.A., 2005. Pain and emotion interactions in subregions of the cingulate gyrus. Nat.
emotion, In: Vogt, B.A., Gabriel, M. (Eds.), Neurobiology of Cingulate Cortex and Limbic Rev. Neurosci. 6, 533–544.
Thalamus: A Comprehensive Handbook, 2nd ed. Birkhäuser, Boston. Wager, T.D., Davidson, M.L., Hughes, B.L., Lindquist, M.A., Ochsner, K.N., 2008. Prefrontal–
Nelson, E.E., Leibenluft, E., McClure, E.B., Pine, D.S., 2005. The social re-orientation of subcortical pathways mediating successful emotion regulation. Neuron 59, 1037–1050.
adolescence: a neuroscience perspective on the process and its relation to Wang, J., Iannotti, R.J., Nansel, T.R., 2009. School bullying among adolescents in the
psychopathology. Psychol. Med. 35, 163–174. United States: physical, verbal, relational, and cyber. J. Adolesc. Health 45, 368–375.
O'Brien, S.F., Bierman, K.L., 1988. Conceptions and perceived influence of peer groups — Wechsler Abbreviated Scale of Intelligence, 1999. Psychological Corp, San Antonio, TX.
interviews with preadolescents and adolescents. Child Dev. 59, 1360–1365. Weiskopf, N., Hutton, C., Josephs, O., Deichmann, R., 2006. Optimal EPI parameters for
Ochsner, K.N., Gross, J.J., 2005. The cognitive control of emotion. Trends Cogn. Sci. 9, reduction of susceptibility-induced BOLD sensitivity losses: a whole-brain analysis
242–249. at 3 T and 1.5 T. NeuroImage. 33, 493–504.
Pfeifer, J.H., Masten, C.L., Borofsky, L.A., Dapretto, M., Fuligni, A.J., Lieberman, M.D., 2009. Williams, K.D., 2007. Ostracism. Annu. Rev. Psychol. 58, 425–452.
Neural correlates of direct and reflected self-appraisals in adolescents and adults: Williams, K.D., Cheung, C.K., Choi, W., 2000. Cyberostracism: effects of being ignored
when social perspective-taking informs self-perception. Child Dev. 80, 1016–1038. over the Internet. J. Pers. Soc. Psychol. 79, 748–762.
Schoenbaum, G., Roesch, M.R., Stalnaker, T.A., Takahashi, Y.K., 2009. A new perspective Zadro, L., Williams, K.D., Richardson, R., 2004. How low can you go? Ostracism by a
on the role of the orbitofrontal cortex in adaptive behaviour. Nat. Rev. Neurosci. 10, computer is sufficient to lower self-reported levels of belonging, control, self-
885–892. esteem, and meaningful existence. J. Exp. Soc. Psychol. 40, 560–567.