Australian Journal of Psychology 2016; 68: 154–165
doi: 10.1111/ajpy.12130
Special Issue
Anxiously elaborating the social percept: Anxiety and age
differences in functional connectivity of the fusiform face area
in a peer evaluation paradigm
Joanne C. Beer,1 Ashley R. Smith,1 Johanna M. Jarcho,1 Gang Chen,2 Richard C. Reynolds,2 Daniel S. Pine,1
and Eric E. Nelson1
1
Section on Development and Affective Neuroscience and 2Scientific and Statistical Computing Core, National Institute
of Mental Health, Bethesda, MD, USA
Abstract
Objectives: Social anxiety disorder involves biased cognition and altered neural responses to social stimuli. This study further
assesses the precise ways in which neural activation associated with perceptual processing of faces differs in socially anxious
and non-anxious individuals. Method: Functional magnetic resonance imaging data were acquired as 90 anxious or healthy
juveniles and adults performed a peer feedback task. Psychophysiological functional connectivity analysis was performed on all
participants using a seed placed in the right fusiform face area (rFFA). Results: Social anxiety was associated with enhanced
rFFA coupling in several areas of ventral visual stream when viewing feedback from peers previously rejected by participants;
healthy participants demonstrated the opposite pattern. Moreover, anxious juveniles had greater positive rFFA coupling with
right inferior temporal gyrus during feedback from rejected peers; other groups showed the opposite pattern. Finally, anxious
juveniles had greater positive rFFA coupling with pregenual anterior cingulate cortex during negative peer feedback; other
groups displayed the opposite pattern. Conclusions: Anxious individuals, and juveniles in particular, may dedicate more neural
resources to stimuli associated with potentially negative, relative to positive, social outcomes; non-anxious individuals may do
the opposite.
Key words: development, face perception, psychophysiological interaction, social anxiety
What is already known about this topic?
1. Cognitive biases in attention, perception, interpretation, and memory underlie anxiety disorders.
2. Neural processing of stimuli involves multiple stages
of elaboration.
3. Anxiety expression changes over the course of
development.
Correspondence: Joanne C. Beer, Department of Biostatistics,
University of Pittsburgh, 311 Parran Hall, 130 De Soto Street,
Pittsburgh, PA 15261, USA.
Email: [email protected]
Received 13 January 2016. Accepted for publication 16
May 2016.
© 2016 The Australian Psychological Society
What this topic adds?
1. Anxiety is associated with aberrant weighting of perceptual processing of relevant stimuli in regions of
the core and extended face perception network.
2. Memory and brain responses may be a tractable and
malleable index of cognitive bias.
3. Marked developmental differences in anxious populations may reflect age differences in how biases
manifest.
Social anxiety disorder (SAD) typically emerges in adolescence and manifests as excessive worry and social avoidance
(Stein & Stein, 2008). Many cognitive biases have been
linked with SAD (Roth & Heimberg, 2001), and recently
functional neuroimaging has enabled assessment of the
neural mechanisms underlying these processes (Haller,
Cohen Kadosh, Scerif, & Lau, 2015). Functional neuroimaging studies in SAD suggest perturbed processing of social stimuli in a variety of regions related to processing at both
Anxiety and age differences in FFA connectivity
affective and cognitive levels (Brühl, Delsignore, Komossa, &
Weidt, 2014; Freitas-Ferrari et al., 2010; Goldin, Manber,
Hakimi, Canli, & Gross, 2009; Guyer et al., 2014; Jarcho
et al., 2015; Phan, Fitzgerald, Nathan, & Tancer, 2006;
Richey et al., 2014; Spielberg et al., 2015).
While brain imaging has long focused on functioning in
particular regions, recent approaches target functional networks, which may help illuminate how information is integrated across the brain (O’Reilly, Woolrich, Behrens,
Smith, & Johansen-Berg, 2012). Several such studies in
SAD have reported differential connectivity patterns
between the prefrontal cortex (PFC) and subcortical structures like the amygdala and striatum (Ding et al., 2011;
Hahn et al., 2011; Liao et al., 2010; Miskovic & Schmidt,
2012; Prater, Hosanagar, Klumpp, Angstadt, & Phan, 2013).
These differences are thought to reflect altered integration
of socio-emotional experiences with executive processes in
SAD (Jarcho et al., 2015; Nelson & Guyer, 2011).
In addition to dysfunction in affective and cognitive brain
network interactions, differential integration between perceptual networks and other brain systems may also underlie
SAD (Büchel & Friston, 1997; Pehrs et al., 2015). This study
assesses functional connectivity differences in the fusiform
face area (FFA), a relatively early node of the face perception network (Kanwisher, McDermott, & Chun, 1997; Weiner & Grill-Spector, 2012), as a function of age and anxiety.
Although the FFA is a region that demonstrates rapid and
specific activation in response to faces, it also receives
extensive ‘back projection’ from regions in limbic and prefrontal areas (Brühl et al., 2014). At a functional level, this
back projection likely amplifies neural signalling elicited by
a specific face and enhances processing of the facial percept.
This may manifest in a number of ways, including increased
attention and greater recall (Baldauf & Desimone, 2014;
Hasinski & Sederberg, 2016).
Prior studies find that anxiety relates to reduced negative
functional connectivity between face-selective regions of the
fusiform gyrus and posterior cingulate, precuneus, and sensorimotor regions; reduced positive coupling of the fusiform
with ventromedial PFC; and increased positive coupling with
the amygdala (Danti et al., 2010; Frick, Howner, Fischer, Kristiansson, & Furmark, 2013). Most of these studies examine
global patterns of connectivity rather than context-specific differences. Several studies also find differences in FFA connectivity patterns as a function of normative development,
generally noting an overall increase in size and more taskbased flexibility in seed-based FFA networks (Cohen Kadosh,
Johnson, Dick, Cohen Kadosh, & Blakemore, 2013; Golarai,
Liberman, Yoon, & Grill-Spector, 2010; Scherf & Scott, 2012).
We are aware of no study examining FFA connectivity in
anxiety and development concomitantly.
This study compares FFA connectivity between socially
anxious and non-anxious juveniles and adults using the
© 2016 The Australian Psychological Society
155
Chatroom Task, in which participants select or reject peers
for an online chat and receive peer feedback (see Jarcho
et al. (2015); Spielberg et al. (2015) for prior reports using
this dataset). Previous functional connectivity analyses of
Chatroom data have placed seeds in the amygdala and striatum and reveal differential connectivity with the PFC as a
function of both anxiety and development (Guyer et al.,
2008; Jarcho et al., 2015; Spielberg et al., 2015). These findings link SAD to perturbed processing of salient social information and further suggest that perturbations may manifest
differently in juveniles and adults.
By using an FFA seed, we mapped the ways in which
brain regions differentially modulate activation of face perception across anxiety and development. A variety of cognitive biases associated with SAD, including biases in
attention, perception, interpretation, or memory, could conceivably manifest in the context of the Chatroom Task. As
our seed was placed in a region associated with face perception, we expected to find evidence of perceptual biases, but
we also expected we might find differential connectivity
with neural systems that interact with the face perception
network, such as those underlying attention, interpretation,
and memory. As with previous findings, we predicted that
differential coupling between the FFA and other regions
would emerge as a function of both participants’ selection of
peers for a potential social interaction and type of feedback
received. We further predicted interactive effects between
age group and anxiety status due to the increased salience of
peers and the ongoing developmental processes during adolescence (Blakemore & Mills, 2014). As functional localisation studies have found activations specific to faces more
consistently and of greater magnitude in the right fusiform
(Haxby, Hoffman, & Gobbini, 2000; Kanwisher et al., 1997),
we restrict our attention to the right FFA (rFFA).
METHODS
Participants
Participants consisted of 90 clinically anxious and psychiatrically healthy juveniles and adults recruited from the local
community. Following initial screening, participants underwent a semi-structured clinical interview performed by a
trained clinician (masters or above). All anxious participants
met DSM-IV (Diagnostic and Statistical Manual of Mental
Disorders, 4th edition) criteria for anxiety, and although
only 82% of anxious participants met full DSM-IV criteria
for SAD, all expressed clinically significant levels of fear of
social situations in diagnostic interviews or scales (score >20
on Fear of Negative Evaluation (Watson & Friend, 1969) or
>40 on Liebowitz Social Anxiety Scale (Liebowitz, 1987)).
Exclusion criteria required that all participants be free of
selective serotonin reuptake inhibitor (SSRI) use for at least
156
J.C. Beer et al.
Table 1 Demographics
Juveniles
N
Female (%)
Age (years)
Age range
IQ
FNE
SCAREDa
Scanner 1 (%)
Non-anxious
Anxious
24
38
13.7 (2.4)
9.4–17.2
112.6 (11.7)
6.9 (5.0)
5.5 (5.4)
42
15
60
12.8 (3.4)
8.0–17.4
104.8 (12.5)
21.1 (8.3)
34.0 (6.9)
13
Adults
p value
.435
.035
<.001
<.0001
Non-anxious
Anxious
32
53
26.2 (5.2)
21.0–44.8
119.1 (11.4)
9.8 (4.9)
—
38
19
79
27.8 (7.8)
18.3–49.6
112.8 (12.6)
22.1 (8.1)
—
11
p value
.055
.633
.031
<.0001
—
.047
Notes. Values are mean (SD) unless otherwise specified. All p values were derived from Wilcoxon rank-sum test with continuity correction and
involved comparisons within age groups, with the exception of per cent female and per cent scanner 1, which used Fisher’s exact test and compared
proportions across all four groups. Missing data: IQ: 1 anxious adult; FNE: 1 non-anxious juvenile, 2 anxious juveniles, 4 anxious adults. IQ = intelligence quotient; FNE = Fear of Negative Evaluation; SCARED = Screen for Child Anxiety Related Disorders.
a
Average of juvenile and parent SCARED evaluations.
one month and must not have been treated with an SSRI
for their current episode. All participants were free of psychoactive medication at time of testing. Demographics of
the four groups are depicted in Table 1.
Chatroom Task
The Chatroom Task has been described in detail previously
(Guyer, McClure-Tone, Shiffrin, Pine, & Nelson, 2009;
Guyer et al., 2008, 2014; Jarcho et al., 2015; Lau et al.,
2011; Spielberg et al., 2015). The task requires that participants make two visits to the lab, separated by approximately
2 weeks. On the initial visit, participants were told that the
task is a multi-site study designed to assess how individuals
interact with previously unknown people on the internet.
Participants were then shown an array of 60 photographs of
other age-matched (within age categories 8–11, 12–14,
15–17, or 18+ years, equal numbers male and female)
peers, who, they were told, were other participants in the
study. Participants sorted these photos into two equal-sized
groups representing peers they would and would not want
to have an online chat with on a subsequent visit. A photograph was then taken of the participant, and they were told
that other participants in the study would be shown this rating and then rate their photograph in a similar manner.
After approximately 2 weeks (M = 11.87, SD = 7.42
days), participants returned for a functional magnetic resonance imaging (fMRI) scan. The Chatroom Task was
divided into two runs. In the first run (not considered here),
participants predicted how much the peer depicted in each
photograph liked them. In the second run, the peers’ ratings
were revealed (Fig. 1). Trials in the second run were divided
into two phases. Initially, each face was displayed for 3 s,
and participants were reminded how they rated this person
previously (e.g., ‘you were interested’ or ‘you were not
interested’). This anticipation phase was followed by a
Figure 1 The Chatroom Task consists of two
visits to the lab. On the initial visit (top), participants view 60 unknown age-matched peers
and sort them into two equal-sized groups
based on future desire to engage in an online
chat. The participants then have their own
photograph taken and are told other participants will rate them in a similar manner. On
the subsequent visit (middle), participants view
how others sorted them while undergoing
functional neuroimaging. Finally, immediately
after the scan, participants are asked to recall
the type of feedback they received from each
peer (bottom).
© 2016 The Australian Psychological Society
Anxiety and age differences in FFA connectivity
feedback phase, in which the words ‘they are interested’ or
‘they are not interested’ was overlaid on the face (2 s). After
the feedback was revealed, participants rated their expectation of the outcome (3–5 s). Equal numbers of positive and
negative feedback trials were presented, and the order was
randomised. A fixation jitter of 0–8 s was interspersed
between the anticipation and feedback phases to decouple
events in time. Finally, in order to assess potential memory
biases, immediately after the neuroimaging session, a surprise memory test was administered where participants
were asked to recall the responses of each peer.
Data processing and analysis
Data processing and analysis was conducted with AFNI software (Cox, 1996) and R version 3.2.2 (R Core Team, 2015).
Standard fMRI acquisition and preprocessing methods were
used (see Supporting Information).
157
An initial one-sample t-test (3dttest++ in AFNI) was performed on data grouped from all participants to identify the
area of maximal response to faces in the right fusiform
gyrus. The voxel in the right medial fusiform gyrus with
the highest t-value for task-baseline contrast (mean per cent
signal change = 0.54; t(89) = 16.20; uncorrected p = 6.0 ×
10−28; Talairach coordinates 34, −44, −19) was chosen as
the centre of the 6-mm radius sphere defining the rFFA.
The average task versus baseline contrast value across all
voxels in the rFFA sphere was extracted for each participant, and these values were subjected to a two-way
between-subjects sequential sum of squares analysis of variance (ANOVA). The results indicate no significant effect of
anxiety diagnosis (F(1, 86) = 0.56, p = .457), age group
(F(1, 86) = 0.77, p = .384), or their interaction
(F(1, 86) = 1.63, p = .205), and comparable levels of rFFA
activation across groups (F(3, 86) = 0.99, p = .403) (Fig. 2).
Seed identification
Task-dependent functional connectivity analyses
Individual-level general linear models (GLMs) were fit with
six task regressors corresponding to the face perception
events of the anticipation and feedback phases of the Chatroom Task; two anticipation regressors represented participant selection and rejection of peers on the initial visit, and
four feedback regressors represented the crossing of participant selection (‘selected peer feedback’ and ‘rejected peer
feedback’) and type of feedback received (‘positive peer
feedback’ and ‘negative peer feedback’). Events were modelled with BLOCK basis functions of the appropriate event
duration and convolved with the hemodynamic response
function included in AFNI. Models also included six motion
regressors and seven orthogonal polynomials to account for
head motion and low-frequency drift in the fMRI signal,
respectively.
We next performed psychophysiological interaction (PPI)
connectivity analyses (Friston et al., 1997; O’Reilly et al.,
2012) using the rFFA seed. Six PPI regressors corresponding
to the six event types were created for each participant; they
function essentially as interaction terms that cross the
expected BOLD response time intervals associated with each
event type with the individual participants’ rFFA seed region
time series. A group-level whole-brain PPI analysis (3dMVM
in AFNI; Chen, Adleman, Saad, Leibenluft, and Cox (2014))
modelled the four PPI feedback coefficient estimates in a
repeated measures ANOVA, including between-subject
factors age group, diagnosis group, their interaction; withinsubject factors participant selection, peer feedback, their
interaction; interactions between all between- and withinsubject factors; and fMRI scanner as a between-subject factor
Figure 2 The right fusiform face area (rFFA)
region of interest applied to all participants is
depicted by the red sphere overlaid on the
structural MRI images (left). The mean
(! SEM) task per cent signal change (relative
to baseline) across groups is depicted in the bar
graph (right).
© 2016 The Australian Psychological Society
158
J.C. Beer et al.
of no interest. The two anticipation PPI coefficients were not
of interest for the current analysis, so they were not included
in the group model to conserve statistical power. Because of
the low sensitivity of PPI (O’Reilly et al., 2012), voxel-wise
significance threshold was set to p < .005, with a minimum
cluster size of 50 voxels. This is similar to other connectivity
analytical approaches (Frick et al., 2013; Jarcho et al., 2015).
For statistically significant interactions involving both a
within-subject factor and diagnosis group, for each participant, we extracted the mean of each of the four PPI coefficient estimates over all voxels in the regions that survived
the significance threshold and calculated the appropriate
within-subject contrasts (selected peer feedback–rejected
peer feedback and positive peer feedback–negative peer
feedback). For interactions that only involved a withinsubject factor and diagnosis group, we performed Welch’s
t-test on the extracted data contrasts to confirm significant
differences across diagnosis. To control for potential confounding by group differences in baseline covariates, we
then fit linear regression models controlled by continuous
age, gender, IQ, and scanner. For interactions that additionally involved age group, we fit two-way between-subjects
ANOVA models to the extracted data contrasts, including
diagnosis group, age group, and interaction terms, and performed post hoc Bonferroni-corrected pair-wise comparisons (Welch’s t-test) of group means (six comparisons). We
then adjusted the ANOVA models by gender, IQ, and scanner. We also tested the association of the extracted data
contrasts with continuous age via linear regressions, unadjusted and adjusted for gender, IQ, and scanner. We examined the expectancy rating and memory data alone by
calculating analogous within-subject contrasts (e.g.,
memory per cent correct for positive peer feedback–
memory per cent correct for negative peer feedback) and
modelled these using two-way between-subjects ANOVA
with diagnosis, age group, and interaction terms. We then
assessed the association of the memory contrasts with corresponding extracted data contrasts using linear regression.
RESULTS
Several brain regions differed significantly in taskdependent rFFA functional connectivity across anxiety,
some of which also interacted with age group (Table 2).
Because the focus of this article is on anxiety and development, we highlight differences that varied as a function of
diagnosis or in which diagnosis and age group interacted.
Several other significant findings also emerged (see
Table S1, Supporting Information).
In four regions, task-dependent differences emerged as a
function of anxiety. Three of these regions were in the
occipital lobe, and one was in the cerebellum (Table 2,
Fig. 3). Because the cerebellar regions have not been associated with either perceptual processing or anxiety, they will
not be considered further. The three occipital regions displayed similar patterns of task-based connectivity differences. For non-anxious participants, greater positive
coupling was found between the occipital regions and the
rFFA during feedback from previously selected, relative to
rejected, peers regardless of the type of feedback. Conversely, anxious participants had greater positive coupling
with rFFA during feedback from rejected, relative to
selected, peers. These group differences remained after
Table 2 Task-dependent functional connectivity of the right fusiform face area during feedback in the Chatroom Task
Talaraich
Region
Voxels
Diagnosis × participant selection
Left cuneus/lingual gyrus
290
Right cerebellum
105
Right declive (OFA)
81
Left declive (OFA)
55
Age group × diagnosis × participant selection
Left cerebellum
234
Right inferior temporal gyrus
55
Diagnosis × peer feedback
No suprathreshold activations
Age group × diagnosis × peer feedback
Left anterior cingulate
61
Diagnosis × participant selection × peer feedback
No suprathreshold activations
Age group × diagnosis × participant selection × peer feedback
No suprathreshold activations
x
y
z
F(1, 84)
p value a
−11
51
29
−29
−71
−44
−61
−64
−14
−36
−16
−16
14.12
27.06
15.88
19.24
<10−3
<10−5
10−4
<10−4
−4
41
−56
−4
−36
−36
16.15
24.86
10−4
<10−5
−4
36
6
15.38
<10−3
Notes. Voxel-wise p ≤ .005, cluster threshold of 50 voxels; coordinates are for maximum F-value. Significant differences across diagnosis group.
OFA = occipital face area.
a
Uncorrected.
© 2016 The Australian Psychological Society
Anxiety and age differences in FFA connectivity
159
Figure 3 A diagnosis by participant selection interaction emerged in the left cuneus/lingual gyrus (LLG; overlaid on structural images at
left) and left and right occipital face areas (LOFA, ROFA; overlaid on structural images at bottom right). Graphs at the top show mean
(! SEM) right fusiform face area (rFFA) connectivity of the regions during selected and rejected peer feedback. The bar graph displays the
mean (! SEM) contrast of selected–rejected connectivity.
controlling for continuous age, gender, IQ, and scanner differences across groups (all p < .005, Table S2).
There were two regions where an interaction between
anxiety and age group was found in rFFA coupling: one in
the right inferior temporal gyrus (rITG) and the other in the
cerebellum (Table 2, Fig. 4). In the inferior temporal lobe,
anxious juveniles exhibited greater positive coupling with
the rFFA during feedback from previously rejected peers,
whereas all other groups displayed greater positive coupling
with previously selected peers. Pair-wise comparisons of
extracted data indicated that the mean contrast (selected
peer feedback–rejected peer feedback) for anxious juveniles
was significantly lower than those for the other groups
(Bonferroni-adjusted p < .05), while the other groups did
not significantly differ from each other (Bonferroni-adjusted
p > .05). ANOVA analyses revealed a significant age group
by diagnosis interaction both before (F(1, 80) = 19.09,
p < .001) and after adjusting for gender, IQ, and scanner
(F(1, 77) = 20.09, p < .001; Table S3A). Unadjusted linear
regressions against continuous age resulted in a non© 2016 The Australian Psychological Society
significant negative slope in rITG–rFFA connectivity for the
non-anxious group (t(49) = −1.42, p = .163) and a significant positive slope for the anxious group (t(31) = 4.87,
p < .001). These results remained similar after controlling
for gender, IQ, and scanner and were robust to outlier sensitivity analyses for the anxious group (Table S3B, Fig. 4).
Finally, another three-way interaction was found as a
function of peer feedback (collapsed across participant selection at initial visit). A region in the left pregenual anterior
cingulate cortex (lACC) exhibited relatively greater rFFA
connectivity to negative than positive feedback in anxious
juveniles, but other groups displayed greater connectivity
during positive feedback (Table 2, Fig. 5). Pair-wise comparisons of extracted data indicated that the mean contrast
(positive peer feedback–negative peer feedback) for anxious
juveniles was significantly lower than for non-anxious juveniles (Bonferroni-adjusted p = .022) and anxious adults
(Bonferroni-adjusted p = .016), but not significantly different from non-anxious adults (Bonferroni-adjusted p =
.163). The other group means did not significantly differ
160
J.C. Beer et al.
=
=
=
=
Figure 4 A region of the right inferior temporal gyrus (rITG) exhibited a significant age group by diagnosis interaction for the participant
selection contrast (overlaid on structural images at left). Graphs at the top show mean (! SEM) right fusiform face area (rFFA) connectivity
during selected and rejected peer feedback (top left) and mean (! SEM) contrast of selected–rejected connectivity (top right). Scatter plots at
the bottom show linear regressions (with 95% confidence bands) of participant selection contrast for rITG–rFFA connectivity on continuous
age for non-anxious (bottom left) and anxious (bottom right) participants.
from each other (Bonferroni-adjusted p > .05). ANOVA
analyses revealed significant age group by diagnosis interaction both before (F(1, 86) = 16.61, p < .001) and after
adjusting for gender, IQ, and scanner (F(1, 83) = 15.36,
p < .001; Table S4A). Unadjusted linear regressions against
continuous age resulted in a non-significant negative slope
in lACC–rFFA connectivity for the non-anxious group (t
(54) = −1.40, p = .167) and a significant positive slope for
the anxious group (t(32) = 2.16, p = .038). Results were
similar after controlling for gender, IQ, and scanner
(Table S4B, Fig. 5).
Lastly, we analysed the expectancy rating and memory
data alone and the memory data in relation to functional
connectivity findings. A two-way sequential sum of squares
ANOVA performed on the expectancy rating peer feedback
contrast revealed a significant effect of anxiety diagnosis
(F(1, 86) = 4.43, p = .038) but no main effect for age group
(F(1, 86) = 0.60, p = .442) or age group by diagnosis interaction (F(1, 86) = 0.82, p = .367). We found a similar result
for the memory peer feedback contrast: ANOVA revealed a
significant effect of anxiety diagnosis (F(1, 85) = 4.80,
p = .031) but no main effect for age group (F(1, 85) = 1.23,
p = .270) or age group by diagnosis interaction
(F(1, 85) = 0.45, p = .506; one non-anxious juvenile with
missing memory data was excluded from the memory analyses). Thus, anxious participants were more likely to expect
negative than positive feedback and were more biased to
accurately remember negative peer feedback events than
were non-anxious participants. These measures were significantly positively correlated in a univariate regression
(R2 = .07, p = .010). Neither the expectancy rating nor the
memory data showed any significant group differences in
the participant selection contrasts. In an exploratory analysis of all significant regions discussed above, only
© 2016 The Australian Psychological Society
Anxiety and age differences in FFA connectivity
=
=
=
161
=
Figure 5 The left pregenual anterior cingulate cortex (lACC) exhibited a significant age group by diagnosis interaction for the peer feedback
contrast (overlaid on structural images at left). Graphs at the top show mean (! SEM) right fusiform face area (rFFA) connectivity during
positive and negative peer feedback (top left) and mean (! SEM) contrast of positive–negative connectivity (top right). Scatter plots at the
bottom show linear regressions (with 95% confidence bands) of peer feedback contrast for lACC–rFFA connectivity on continuous age for
non-anxious (bottom left) and anxious (bottom right) participants.
Figure 6 The box plot (left) displays distributions of memory bias for peer feedback recall in
the four participant groups. Positive values indicate greater memory for positive peer feedback,
and negative values indicate greater memory for
negative peer feedback. The scatter plot (right)
depicts the association between memory bias
(on the y-axis) and lACC–rFFA coupling (on the
x-axis) (with 95% confidence bands).
connectivity between the rFFA and lACC was significantly
correlated with the memory bias. Specifically, higher rFFA–
lACC connectivity when receiving positive versus negative
© 2016 The Australian Psychological Society
=
=
peer feedback was associated with greater memory bias for
positive versus negative peer feedback (Fig. 6). There were
no group differences in this association.
162
J.C. Beer et al.
DISCUSSION
We identified several group differences in functional rFFA
connectivity that related to participants’ selection or rejection of peers for a future online chat and the valence of
feedback associated with those peers’ faces. In particular,
anxious and non-anxious participants showed different patterns of rFFA connectivity when viewing faces of peers they
had previously selected versus rejected in areas of visual
cortex and cerebellum. Developmental differences in rFFA
connectivity patterns across anxiety groups emerged in the
right anterior temporal gyrus and left cerebellum as a function of participant selection, and in the left pregenual anterior cingulate cortex as a function of peer feedback. Finally,
we found evidence of an association across all participants
between rFFA–lACC co-activation and subsequent memory
bias. We will consider each of these findings in turn.
Among anxious participants, three regions in the occipital
lobe (left lingual gyrus/cuneus, bilateral regions of fusiform)
were more strongly coupled to the rFFA when viewing
faces of peers they had rejected than ones they had selected
on the initial visit, while non-anxious participants demonstrated the opposite pattern (Fig. 3). These regions have previously been characterised as part of the face processing
network (Haxby, Hoffman, & Gobbini, 2002; Weiner &
Grill-Spector, 2012), acting as components of the ventral
visual stream to support perception. Co-activation suggests
that socially anxious individuals dedicate more resources to
processing faces they had rejected than those they had
selected, consistent with other findings from this task concerning the salience of rejected peers (Guyer et al., 2008;
Spielberg et al., 2015), and may reflect greater neural processing of potentially negative than potentially positive outcomes among anxious individuals. This interpretation is
speculative, however, and awaits further study.
We observed a similar anxiety-based shift in co-activation
within the rITG (Fig. 4), although this difference interacted
with age. Post hoc comparisons confirmed that anxious
juveniles uniquely showed greater connectivity when
receiving feedback from peers they had previously rejected.
The rITG is also part of the ventral visual stream, although
this region is much further downstream from the occipital
regions noted above. As visual percepts advance along the
visual stream, they become increasingly integrated with
other information. In the inferior temporal cortex, social stimuli may be particularly sensitive to valence modulations
(Hadj-Bouziane et al., 2012; Morin, Hadj-Bouziane, Stokes,
Ungerleider, & Bell, 2014). Therefore, the finding that anxious juveniles co-activate this region more when viewing
previously rejected, relative to selected, faces may reflect
the same framing bias that was driving the activation in the
occipital lobe, but possibly more closely tied to affect than
attention. If so, this appears to be a feature of social anxiety
that is particularly marked in younger individuals and
diminishes across maturation, possibly reflecting the greater
affective potency of peer acceptance for adolescents than
adults.
A final age by diagnosis effect was found in a large cluster
in the pregenual ACC (Fig. 5). In contrast to other findings,
where connectivity differences related to participant selection or rejection of peers, these findings related to the
valence of peer feedback. Here, we found that in anxious
juveniles, left pregenual ACC–rFFA coupling was stronger
following negative than positive peer feedback, whereas this
coupling was reversed for the other three groups. Thus, as
with the findings with the rITG, the anxious juveniles
appear to have a unique pattern of rFFA–lACC coupling.
Several functions that have been attributed to the lACC
may relate to this task. These include mentalising, error
monitoring, self-appraisal, and affect regulation (Melcher,
Falkai, & Gruber, 2008; Rushworth, Mars, & Sallet, 2013;
Sturm et al., 2013; Tang, Posner, Rothbart, & Volkow,
2015). This region also has strong associations with internalising disorders (Murray, Wise, & Drevets, 2011). Although
it is difficult to say which psychological processes were
engaged in anxious juveniles relative to other groups, it is
clear that rFFA connectivity associated with faces in this
task was biased towards negative rather than positive social
information. These findings suggest that in contrast to all
other groups, anxious juveniles devote greater neuronal
resources to processing social information (e.g., regulating
affect, monitoring own errors, mentalising) following a negative occurrence. This pattern unexpectedly reversed in
anxious adults. This reversal may reflect a shift in the
potency of peer rejection as anxious individuals shift into
adulthood, a transition associated with psychopathology
progression, or may be related to other non-specific aspects
of this task. These results hint at compelling hypotheses
related to the interaction of anxiety with development,
which remain to be flushed out in future studies.
We should note that two different analyses of these same
data or a subset of this data also found aberrant activity in
the anxious juvenile group in regions similar to or even
overlapping with the pregenual ACC. We have associated
these findings with differences in prediction error processing
(Jarcho et al., 2015) and anticipatory anxiety responses
(Spielberg et al., 2015) in anxious juveniles. While these
different models have generated slightly different interpretations of the role the ACC/medial PFC is playing in the
integration of social information, in this sample, anxious
juveniles are evidently engaging this region in a unique
manner.
Finally, we found a significant relationship between
rFFA–lACC connectivity and memory bias. Greater positive
rFFA–lACC coupling during positive social feedback was
associated with greater memory bias for positive feedback.
© 2016 The Australian Psychological Society
Anxiety and age differences in FFA connectivity
Although this relationship did not differ across groups,
when considered together with the finding that anxious
participants exhibited a bias towards expectation and recall
of negative social feedback, this suggests that functional
coupling between rFFA and pregenual ACC may relate to
negative information bias that is a common cognitive feature of SAD (Coles & Heimberg, 2002; Foa, Gilboa-Schechtman, Amir, & Freshman, 2000; Lundh & Öst, 1996;
Morrison & Heimberg, 2013; Roth & Heimberg, 2001).
One surprising outcome of this study is that we failed to
see group differences in functional connectivity between
rFFA and the amygdala or striatum. Hyperactivity within
the amygdala in response to fearful or threatening face
presentation is one of the more consistent findings in anxiety (Gentili et al., 2015), and as noted in the introduction,
we have observed alterations in both the amygdala and striatum in this task previously (Guyer et al., 2008, 2014;
Jarcho et al., 2015; Spielberg et al., 2015). Therefore, we
expected to see a differential pattern of connectivity with
the rFFA in this analysis. However, many of these effects
are stronger during the anticipatory period, which precedes
receipt of the response itself (Lau et al., 2011), and was not
considered in this study.
Limitations
There are several important limitations in the interpretations of this study. First, although we have found significant
age and diagnosis differences in functional coupling with
the rFFA, without further corroborating data, we can only
speculate on the clinical significance of brain differences.
Clinically meaningful correlates such as self-reports or behavioural measures are clearly needed. Second, as with all
developmental studies, cross-sectional measures are subject
to a variety of confounding effects and, as such, are not as
reliable as longitudinal measures to track maturational
changes. Third, there are clear limitations of the categorical
approach we adopted here for both age and diagnostic
measures—particularly as not all of our patients met full criteria for SAD. While a categorical approach offered straightforward analyses and interpretations in terms of four
distinct groups, use of continuous measures would enable a
more detailed analysis of the variability in the data. Finally,
the use of two scanners for data collection may have introduced artefacts between subjects. We used scanner as a covariate to help mitigate the potential impact of this on our
analyses, although this may have reduced our power to
detect differential functional coupling due to different scanner distributions across groups. In spite of these notable limitations, we feel that this study uncovered some important
neuronal differences as a function of both anxiety status
and age that warrant follow-up and replication in future
studies.
© 2016 The Australian Psychological Society
163
CONCLUSIONS
In sum, our results were largely consistent with our predictions in that co-activation with a face-responsive brain
region in socially anxious and healthy individuals varied in
a task-dependent manner. Results from healthy individuals
suggest that greater neuronal resources are dedicated to
perceptual processing of outcomes from potentially positive
social interactions (peers previously selected by participants), whereas anxious individuals dedicate more
resources to processing potentially negative interactions
(peers previously rejected by participants). In some regions,
this pattern was evident across development, whereas in
others, it was primarily evident in the juvenile group. We
attribute these brain differences to differential engagement
of attention and/or emotion in anxiety, and they are likely
a neuronal reflection of well-known cognitive biases in
social anxiety.
ACKNOWLEDGEMENT
This research (NCT# 00018057; Protocol ID 01-M-0192)
was supported by the NIMH intramural research
programme.
SUPPORTING INFORMATION
Additional Supporting Information may be found in the
online version of this article at the publisher’s web-site:
Table S1. Statistically significant regions showing functional
connectivity with the right fusiform face area during feedback in the Chatroom Task. Threshold of voxelwise p ≤ .005
and minimum cluster size of 50 voxels. Coordinates are for
maximum F-value.
Table S2. Anxiety diagnosis × participant selection interaction effect on right fusiform face area connectivity; analysis
of data extracted from significant regions.
Table S3. Anxiety diagnosis × age group × participant
selection interaction effect on right inferior temporal gyrus
(rITG)–right fusiform face area (rFFA) connectivity; analysis
of data extracted from significant region.
Table S4. Anxiety diagnosis × age group × peer feedback
interaction effect on left anterior cingulate cortex (lACC)–
right fusiform face area (rFFA) connectivity; analysis of data
extracted from significant region.
REFERENCES
Baldauf, D., & Desimone, R. (2014). Neural mechanisms of objectbased attention. Science, 344(6182), 424–427.
Blakemore, S.-J., & Mills, K. L. (2014). Is adolescence a sensitive
period for sociocultural processing? Annual Review of Psychology,
65, 187–207.
164
J.C. Beer et al.
Brühl, A. B., Delsignore, A., Komossa, K., & Weidt, S. (2014).
Neuroimaging in social anxiety disorder—A meta-analytic
review resulting in a new neurofunctional model. Neuroscience & Biobehavioral Reviews, 47, 260–280. doi:10.1016/j.
neubiorev.2014.08.003
Büchel, C., & Friston, K. (1997). Modulation of connectivity in visual pathways by attention: Cortical interactions evaluated with
structural equation modelling and fMRI. Cerebral Cortex, 7(8),
768–778.
Chen, G., Adleman, N. E., Saad, Z. S., Leibenluft, E., & Cox, R. W.
(2014). Applications of multivariate modeling to neuroimaging
group analysis: A comprehensive alternative to univariate general linear model. NeuroImage, 99, 571–588.
Cohen Kadosh, K., Johnson, M. H., Dick, F., Cohen Kadosh, R., &
Blakemore, S.-J. (2013). Effects of age, task performance, and
structural brain development on face processing. Cerebral Cortex,
23(7), 1630–1642.
Coles, M. E., & Heimberg, R. G. (2002). Memory biases in the anxiety disorders: Current status. Clinical Psychology Review, 22(4),
587–627.
Cox, R. W. (1996). AFNI: Software for analysis and visualization of
functional magnetic resonance neuroimages. Computers and Biomedical Research, 29(3), 162–173.
Danti, S., Ricciardi, E., Gentili, C., Gobbini, M. I., Pietrini, P., &
Guazzelli, M. (2010). Is social phobia a “Mis-Communication”
disorder? Brain functional connectivity during face perception
differs between patients with social phobia and healthy control
subjects. Frontiers in Systems Neuroscience, 4, 152. doi:10.3389/
fnsys.2010.00152
Ding, J., Chen, H., Qiu, C., Liao, W., Warwick, J. M., Duan, X., …
Gong, Q. (2011). Disrupted functional connectivity in social anxiety disorder: A resting-state fMRI study. Magnetic Resonance Imaging, 29(5), 701–711. doi:10.1016/j.mri.2011.02.013
Foa, E. B., Gilboa-Schechtman, E., Amir, N., & Freshman, M.
(2000). Memory bias in generalized social phobia: Remembering
negative emotional expressions. Journal of Anxiety Disorders,
14(5), 501–519.
Freitas-Ferrari, M. C., Hallak, J. E., Trzesniak, C., Filho, A. S.,
Machado-de-Sousa, J. P., Chagas, M. H., … Crippa, J. A. (2010).
Neuroimaging in social anxiety disorder: A systematic review of
the literature. Progress in Neuropsychopharmacology and Biological
Psychiatry, 34(4), 565–580. doi:10.1016/j.pnpbp.2010.02.028
Frick, A., Howner, K., Fischer, H., Kristiansson, M., & Furmark, T.
(2013). Altered fusiform connectivity during processing of fearful
faces in social anxiety disorder. Translational Psychiatry,
3(10), e312.
Friston, K. J., Buechel, C., Fink, G. R., Morris, J., Rolls, E., &
Dolan, R. J. (1997). Psychophysiological and modulatory interactions in neuroimaging. NeuroImage, 6(3), 218–229. doi:10.1006/
nimg.1997.0291
Gentili, C., Cristea, I. A., Angstadt, M., Klumpp, H., Tozzi, L.,
Phan, K. L., & Pietrini, P. (2015). Beyond emotions: A metaanalysis of neural response within face processing system in
social anxiety. Experimental Biology and Medicine, 241(3), 225–237.
Golarai, G., Liberman, A., Yoon, J. M., & Grill-Spector, K. (2010).
Differential development of the ventral visual cortex extends
through adolescence. Frontiers in Human Neuroscience, 3(80).
doi:10.3389/neuro.09.080.2009
Goldin, P. R., Manber, T., Hakimi, S., Canli, T., & Gross, J. J.
(2009). Neural bases of social anxiety disorder: emotional reactivity and cognitive regulation during social and physical threat.
Archives of General Psychiatry, 66(2), 170–180. doi:10.1001/
archgenpsychiatry.2008.525
Guyer, A. E., Benson, B., Choate, V. R., Bar-Haim, Y., PerezEdgar, K., Jarcho, J. M., … Nelson, E. E. (2014). Lasting associations between early-childhood temperament and late-adolescent
reward-circuitry response to peer feedback. Development and Psychopathology, 26(1), 229–243. doi:10.1017/S0954579413000941
Guyer, A. E., Lau, J. Y., McClure-Tone, E. B., Parrish, J.,
Shiffrin, N. D., Reynolds, R. C., … Nelson, E. E. (2008). Amygdala
and ventrolateral prefrontal cortex function during anticipated
peer evaluation in pediatric social anxiety. Archives of General Psychiatry, 65(11), 1303–1312. doi:10.1001/archpsyc.65.11.1303
Guyer, A. E., McClure-Tone, E. B., Shiffrin, N. D., Pine, D. S., &
Nelson, E. E. (2009). Probing the neural correlates of anticipated
peer evaluation in adolescence. Child Development, 80(4),
1000–1015.
Hadj-Bouziane, F., Liu, N., Bell, A. H., Gothard, K. M., Luh, W.-M.,
Tootell, R. B., … Ungerleider, L. G. (2012). Amygdala lesions disrupt modulation of functional MRI activity evoked by facial
expression in the monkey inferior temporal cortex. Proceedings of
the National Academy of Sciences, 109(52), E3640–E3648.
Hahn, A., Stein, P., Windischberger, C., Weissenbacher, A.,
Spindelegger, C., Moser, E., … Lanzenberger, R. (2011).
Reduced resting-state functional connectivity between amygdala
and orbitofrontal cortex in social anxiety disorder. NeuroImage,
56(3), 881–889. doi:10.1016/j.neuroimage.2011.02.064
Haller, S. P., Cohen Kadosh, K., Scerif, G., & Lau, J. Y. (2015).
Social anxiety disorder in adolescence: How developmental
cognitive neuroscience findings may shape understanding
and interventions for psychopathology. Developmental Cognitive
Neuroscience, 13, 11–20. doi:10.1016/j.dcn.2015.02.002
Hasinski, A. E., & Sederberg, P. B. (2016). Trial-level information
for individual faces in the fusiform face area depends on subsequent memory. NeuroImage, 124, 526–535.
Haxby, J. V., Hoffman, E. A., & Gobbini, M. I. (2000). The distributed human neural system for face perception. Trends in Cognitive
Sciences, 4(6), 223–233.
Haxby, J. V., Hoffman, E. A., & Gobbini, M. I. (2002). Human neural systems for face recognition and social communication. Biological Psychiatry, 51(1), 59–67.
Jarcho, J. M., Romer, A. L., Shechner, T., Galvan, A., Guyer, A. E.,
Leibenluft, E., … Nelson, E. E. (2015). Forgetting the best when
predicting the worst: Preliminary observations on neural circuit
function in adolescent social anxiety. Developmental Cognitive Neuroscience, 13, 21–31. doi:10.1016/j.dcn.2015.03.002
Kanwisher, N., McDermott, J., & Chun, M. M. (1997). The fusiform
face area: A module in human extrastriate cortex specialized for
face perception. The Journal of Neuroscience, 17(11), 4302–4311.
Lau, J. Y., Guyer, A., Tone, E. B., Jenness, J., Parrish, J. M.,
Pine, D. S., & Nelson, E. E. (2011). Neural responses to peer
rejection in anxious adolescents: Contributions from the
amygdala-hippocampal complex. International Journal of Behavioral Development, 36(1), 36–44.
Liao, W., Chen, H., Feng, Y., Mantini, D., Gentili, C., Pan, Z., …
Zhang, W. (2010). Selective aberrant functional connectivity of
resting state networks in social anxiety disorder. NeuroImage,
52(4), 1549–1558. doi:10.1016/j.neuroimage.2010.05.010
Liebowitz, M. R. (1987). Social phobia. Modern Problems of Pharmacopsychiatry, 22, 141–173.
Lundh, L.-G., & Öst, L.-G. (1996). Recognition bias for critical faces
in social phobics. Behaviour Research and Therapy, 34(10),
787–794.
Melcher, T., Falkai, P., & Gruber, O. (2008). Functional brain
abnormalities in psychiatric disorders: Neural mechanisms to
detect and resolve cognitive conflict and interference. Brain
Research Reviews, 59(1), 96–124.
Miskovic, V., & Schmidt, L. A. (2012). Social fearfulness in the
human brain. Neuroscience & Biobehavioral Reviews, 36(1), 459–478.
Morin, E. L., Hadj-Bouziane, F., Stokes, M., Ungerleider, L. G., &
Bell, A. H. (2014). Hierarchical encoding of social cues in primate
inferior temporal cortex. Cerebral Cortex, 25(9), 3036–3045.
© 2016 The Australian Psychological Society
Anxiety and age differences in FFA connectivity
Morrison, A. S., & Heimberg, R. G. (2013). Social anxiety and social
anxiety disorder. Annual Review of Clinical Psychology, 9, 249–274.
doi:10.1146/annurev-clinpsy-050212-185631
Murray, E. A., Wise, S. P., & Drevets, W. C. (2011). Localization of
dysfunction in major depressive disorder: prefrontal cortex and
amygdala. Biological Psychiatry, 69(12), e43–e54.
Nelson, E. E., & Guyer, A. E. (2011). The development of the ventral prefrontal cortex and social flexibility. Developmental Cognitive
Neuroscience, 1(3), 233–245.
O’Reilly, J. X., Woolrich, M. W., Behrens, T. E., Smith, S. M., &
Johansen-Berg, H. (2012). Tools of the trade: psychophysiological interactions and functional connectivity. Social Cognitive and
Affective Neuroscience, 7(5), 604–609. doi:10.1093/scan/nss055
Pehrs, C., Zaki, J., Schlochtermeier, L. H., Jacobs, A. M.,
Kuchinke, L., & Koelsch, S. (2015). The temporal pole top-down
modulates the ventral visual stream during social cognition. Cerebral Cortex, bhv226. doi:10.1093/cercor/bhv226
Phan, K. L., Fitzgerald, D. A., Nathan, P. J., & Tancer, M. E. (2006).
Association between amygdala hyperactivity to harsh faces and
severity of social anxiety in generalized social phobia. Biological
Psychiatry, 59(5), 424–429. doi:10.1016/j.biopsych.2005.08.012
Prater, K. E., Hosanagar, A., Klumpp, H., Angstadt, M., &
Phan, K. L. (2013). Aberrant amygdala–frontal cortex connectivity during perception of fearful faces and at rest in generalized
social anxiety disorder. Depression and Anxiety, 30(3), 234–241.
R Core Team (2015). R: A language and environment for statistical computing. Vienna, Austria: R Foundation for Statistical Computing.
Richey, J. A., Rittenberg, A., Hughes, L., Damiano, C. R.,
Sabatino, A., Miller, S., … Dichter, G. S. (2014). Common and
distinct neural features of social and non-social reward
© 2016 The Australian Psychological Society
165
processing in autism and social anxiety disorder. Social Cognitive
and Affective Neuroscience, 9(3), 367–377. doi:10.1093/scan/nss146
Roth, D. A., & Heimberg, R. G. (2001). Cognitive-behavioral models
of social anxiety disorder. The Psychiatric Clinics of North America,
24(4), 753–771.
Rushworth, M. F., Mars, R. B., & Sallet, J. (2013). Are there specialized circuits for social cognition and are they unique to humans?
Current Opinion in Neurobiology, 23(3), 436–442.
Scherf, K. S., & Scott, L. S. (2012). Connecting developmental trajectories: Biases in face processing from infancy to adulthood.
Developmental Psychobiology, 54(6), 643–663.
Spielberg, J. M., Jarcho, J. M., Dahl, R. E., Pine, D. S., Ernst, M., &
Nelson, E. E. (2015). Anticipation of peer evaluation in anxious
adolescents: Divergence in neural activation and maturation.
Social Cognitive and Affective Neuroscience, 10(8), 1084–1091.
doi:10.1093/scan/nsu165
Stein, M. B., & Stein, D. J. (2008). Social anxiety disorder. Lancet,
371(9618), 1115–1125. doi:10.1016/S0140-6736(08)60488-2
Sturm, V. E., Sollberger, M., Seeley, W. W., Rankin, K. P.,
Ascher, E. A., Rosen, H. J., … Levenson, R. W. (2013). Role of right
pregenual anterior cingulate cortex in self-conscious emotional
reactivity. Social Cognitive and Affective Neuroscience, 8(4), 468–474.
Tang, Y.-Y., Posner, M. I., Rothbart, M. K., & Volkow, N. D. (2015).
Circuitry of self-control and its role in reducing addiction. Trends
in Cognitive Sciences, 19(8), 439–444.
Watson, D., & Friend, R. (1969). Measurement of social-evaluative
anxiety. Journal of Consulting and Clinical Psychology, 33(4), 448.
Weiner, K. S., & Grill-Spector, K. (2012). The improbable simplicity
of the fusiform face area. Trends in Cognitive Sciences, 16(5),
251–254.