A Functional Magnetic Resonance Imaging
Investigation of Uncertainty in Adolescents
with Anxiety Disorders
Amy L. Krain, Kristin Gotimer, Sara Hefton, Monique Ernst, F. Xavier Castellanos, Daniel S. Pine,
and Michael P. Milham
Background: Pediatric anxiety disorders, although highly prevalent, are understudied with little known about their pathophysiology.
Intolerance of uncertainty (IU) is a trait associated with worry, a key characteristic of these disorders. Neural responses to uncertainty in
healthy subjects involve the same frontal–limbic circuits that are hyper-responsive in pediatric anxiety. As such, the present study examines
the relationship between IU and neural responses to uncertainty in anxious adolescents.
Methods: Sixteen adolescents (ages 13–17) diagnosed with generalized anxiety disorder and/or social phobia (ANX) and 13 non-anxious
control subjects completed a decision-making task while functional magnetic resonance imaging scans were acquired.
Results: The ANX group endorsed greater task-related anxiety and less certainty than control subjects on a post-task questionnaire. Compared
with control subjects, the ANX group did not demonstrate hyper-responsivity of brain regions as hypothesized. Across groups, IU was positively
correlated with activity in several frontal and limbic regions. Further analyses identified subgroups within the ANX group: those with high IU
activated frontal/limbic regions, whereas those with low IU and less anxiety during the task deactivated the same regions in response to
Conclusions: Results substantiate the hypothesized link between IU and neural responses to uncertainty in some adolescents with anxiety
disorders. Our findings, if replicated, suggest that trait measures, such as IU, can significantly improve our understanding of the neurobio-
logical basis of pediatric anxiety disorders.
Key Words: Adolescence, anxiety, fMRI, uncertainty, decision-
studies focus exclusively on aspects of social processing. The
current brain imaging study takes an alternative approach to the
study of anxiety in this population.
Anxiety is defined, according to the Oxford English Dictionary,
as “uneasiness or trouble of mind about some uncertain event” (3).
As such, anxiety disorders are hypothesized to involve heightened
responses to uncertainty. Studies of intolerance of uncertainty (IU),
a trait characterized by negative beliefs about uncertainty, support
this characterization (see 4 for a review). In adolescents, IU strongly
correlates with worry, a central feature of anxiety disorders (5).
Furthermore, treatments aimed at reducing IU are beneficial in adult
(6,7) and adolescent generalized anxiety disorder (GAD) (8).
Although behavioral data link IU to impaired decision-making
(9,10), the neural mechanisms underlying this impairment have not
been examined. Studies of healthy adults implicate orbitofrontal
cortex (OFC), amygdala, and anterior cingulate cortex (ACC) in
processing uncertainty (11–13) . These same regions are hyper-
responsive to social threat stimuli (i.e., faces) in anxious adolescents
lthough anxiety disorders are highly prevalent among ado-
lescents (1,2), knowledge of their pathophysiological under-
pinnings remains limited. Furthermore, the few relevant
(14 –17). Despite this convergence, no study has probed relation-
ships among IU, anxiety, and function of these brain regions. Data
in adults suggest that individual differences in trait anxiety predict
differences in engagement of ACC by uncertain decision-making
(18). Additionally, prior work by our group demonstrated a signif-
icant relationship between IU and ACC activity in healthy adoles-
cents that was not seen in healthy adults, suggesting that IU might
be an especially relevant factor in adolescence (19).
The current study aims to extend these findings and clarify the
role of IU in anxiety disorders by investigating the relationship
between IU and uncertainty-related brain activity in anxious and
healthy adolescents. Using a simple decision-making task with
varied levels of certainty, we test predictions about the impact of
anxiety and IU on behavior and neural activity during uncertain
decision-making. We test the hypothesis that anxiety-disordered
youth endorse higher IU and consequently report greater anxiety
and less certainty when making uncertain decisions. We also
hypothesize that the amygdala–frontal hyper-reactivity shown in
previous studies of anxious adolescents is observed during the
neural processing of uncertainty and that this between-group dif-
ference in neural response is moderated by IU. More specifically, on
the basis of previous findings, we hypothesize that high IU is
associated with greater uncertainty-related neural activity and there-
fore predict that the greatest activation will be observed in the
anxious adolescents with the highest levels of IU.
Methods and Materials
Eighteen adolescents with an anxiety disorder diagnosis (10
boys), ages 13–17 years (mean ? 15.4; SD ? 1.3), and 17 healthy
adolescents (9 boys), ages 13–17 years (mean ? 15.4; SD ? 1.3)
were scanned. Of these participants, 6 were excluded because of
scanner-related imaging artifacts (2 control subjects, 1 anxious) or
incomplete behavioral data (2 control subjects, 1 anxious). As a
From the New York University Child Study Center (ALK, KG, SH, FXC, MPM),
New York University School of Medicine, New York; Nathan Kline Insti-
tute for Psychiatric Research (FXC), Orangeburg, New York; and the
Section on Development and Affective Neuroscience (ME, DSP), Na-
tional Institute of Mental Health, Bethesda, Maryland.
Lexington, Avenue, 13th Floor, New York, N.Y. 10016; E-mail: amy.krain@
Received February 26, 2007; revised June 11, 2007; accepted June 12, 2007.
BIOL PSYCHIATRY 2008;63:563–568
© 2008 Society of Biological Psychiatry
result, complete data from 16 adolescents with an anxiety disorder
diagnosis (mean age ? 15.2 years; SD ? 1.3; 9 boys) and 13 healthy
control subjects (mean age ? 15.4 years; SD ? 1.4; 5 boys) were
analyzed for the current study. Parental consent and child assent
were obtained. This study was approved by the New York Univer-
sity (NYU) School of Medicine Institutional Review Board and NYU
Committee on Activities Involving Human Subjects.
Adolescents with either DSM-IV GAD or social phobia (ANX)
were recruited from the NYU Child Study Center and from the
community. Diagnoses were determined with the Anxiety Disorders
Interview Schedule for Children (ADIS-IV-C; 20), conducted by a
clinical psychologist and/or an advanced doctoral student. Desig-
nation of the disorder as primary indicated that it was the most
severe and would be a main focus of treatment. We included
adolescents with either one of these conditions, on the basis of the
significant comorbidity among these two disorders (21), coupled
with the convergent cognitive (22) and neuroimaging (15–17,23)
findings implicating a common neurobiological substrate.
Of the 16 adolescents with anxiety disorders who provided
complete data for analysis, 10 were diagnosed with comorbid
GAD and social phobia. Additional diagnoses in this group
included separation anxiety disorder (n ? 1), specific phobia
(n ? 2), obsessive-compulsive disorder (n ? 1), attention-
deficit/hyperactivity disorder (ADHD) (n ?3), oppositional de-
fiant disorder (n ? 2), and dysthymia (n ? 1). Five adolescents
were diagnosed with social phobia without GAD. Of these, one
adolescent also received diagnoses of specific phobia and dyst-
hymia. One adolescent was diagnosed with GAD with no
comorbid conditions. Mean IQ scores based on the two-subtest
version of the Wechsler Abbreviated Scale of Intelligence (WASI;
24) were in the average range (mean ? 105.6, SD ? 9.8).
Healthy control subjects were recruited from the community.
They were included in the current study if they scored within
normal limits (standardized score T ? 65) on the Child Behavior
Checklist (25) and if they failed to meet criteria for any current
psychiatric disorders on the basis of the ADIS-IV-C (20). Mean IQ
scores were in the average range (mean ? 100.9, SD ? 12.2) and
did not differ from the ANX group.
Across groups, adolescents were excluded if they had a
history of trauma, pervasive developmental disorder, psychosis,
current use of psychotropic medication, or IQ ? 80. To assess IU,
all participants completed the Intolerance of Uncertainty Scale
(IUS; 26). The Multidimensional Anxiety Scale for Children
(MASC; 27) and the Penn State Worry Questionnaire for Children
(PSWQ-C; 28) were completed as additional measures of anxiety.
The HiLo Game has been used in previous research with healthy
adults and adolescents (19). Participants were instructed to watch a
computer screen where they were shown a numbered card (i.e.,
1–9) and a mystery card (i.e., “?”) (Figure 1). They were asked to
decide whether the value of the mystery card is lower or higher than
the numbered card; the two cards were never equal. Each card was
associated with a specific degree of uncertainty. For example, if the
displayed card was 1 or 9, one could predict with 100% certainty
that the mystery card would be higher or lower, respectively. In
contrast, the highest level of uncertainty was associated with the 5
card, where predictions of higher or lower were each correct 50% of
the time. Participants had 2 sec to respond with a button box, after
which a blank screen appeared for 2, 4, or 6 sec. This jittered interval
was included to better characterize hemodynamic responses. Then
the value of the mystery card and feedback regarding the accuracy
of the participant’s response were displayed (1.5 sec). Participants
won 100 points if correct and lost 50 points if incorrect. They were
instructed to try to win as many points as possible. Average trial
length was 8 sec. Twelve trials of each card (1–9) were randomly
presented resulting in a total of 108 HiLo trials equally distributed
across three blocks. Twenty-four fixation trials (4 sec) were ran-
domly inserted between HiLo trials to serve as a baseline condition.
After the scanning session, participants completed a question-
naire with two questions about each card: 1) “How anxious did
you feel when you saw this card?” (0 ? “Not at All,” 1 ? “A Little,”
2 ? “Somewhat,” 3 ? “Very”); and 2) “How certain were you that
you were correct?” (0 ? “Not Sure at All,” 1 ? “A Little Sure,” 2 ?
“Somewhat Sure,” 3 ? “Completely Sure”).
Functional Magnetic Resonance Imaging Data Acquisition
Scans were acquired on a 3.0T Siemens Allegra head-only
magnetic resonance imaging (MRI) scanner with a Nova array
coil. Functional scans of 35 contiguous 3-mm axial slices were
acquired with a T2*-sensitive gradient echo sequence (repetition
time ? 2000 msec; echo time ? 30; field of view ? 192; 64 ? 64
matrix; in-plane resolution 3 mm ? 3 mm). Stimuli were pro-
jected onto a screen approximately 57 cm from the subject with
an Eiki LC-XG100 projector. A high resolution T1-weighted
magnetization prepared rapid gradient echo (MPRAGE) image
was acquired for each participant as well as T2-weighted images
of functional acquisition slices.
Image processing and statistical analyses were carried out
with FEAT V5.4 (http://www.fmrib.ox.ac.uk/fsl). Images were
motion corrected with MCFLIRT (29). Across subjects, the mean
absolute displacement across the three runs of the task was .19
mm. The maximum absolute displacement for any one partici-
pant was .68 mm, which was still within the acceptable range
(?1.5 mm); therefore, no participants were excluded on the
basis of motion. Preprocessing also included slice timing correc-
tion, spatial smoothing (full-width-at-half-maximal [FWHM] ? 5
mm), mean-based intensity normalization of all volumes by the
same factor, highpass temporal filtering (Gaussian-weighted
least-squares straight line fitting, with sigma ? 30.0 sec), and
Gaussian lowpass temporal filtering (half-width-at-half-maximal
[HWHM] 2.8 sec). Registration to high resolution and standard
while viewing the screen with the numbered and mystery cards.
564 BIOL PSYCHIATRY 2008;63:563–568
A.L. Krain et al.
images was conducted with FLIRT (30). Finally, to ensure data
quality, we carried out independent components analysis
(ICA) with MELODIC to identify and exclude participants whose
data exhibited significant scanner-related artifacts (e.g., slice
dropout). On the basis of this, three participants were excluded.
Behavior. Behavioral analyses were carried out with SPSS
14.0 (SPSS, Chicago, Illinois). The nine trial types were com-
bined, for all dependent measures, to create five conditions on
the basis of the probability that one of the two possible responses
(higher or lower) would be correct (e.g., for card 7, choosing
“lower” would be correct 75% of the time and for card 3,
choosing “higher” would be correct 75% of the time): 100%
(cards 1 & 9), 87.5% (cards 2 & 8), 75% (cards 3 & 7), 62.5% (cards
4 & 6), and 50% (card 5). Mean reaction times, accuracy, and
variability of responses were calculated. Accuracy was deter-
mined by the number of correct responses for each condition
(i.e., participant indicated that the mystery card would be higher
when in fact it was higher). Response variability was represented
by the ratio of the percentage of responses made in a particular
direction (“higher” or “lower”) to 50%, which is the percentage of
responses made in one direction obtained by guessing. Repeated
measures analyses of variance (ANOVAs) were conducted to
compare these task performance measures across the five con-
ditions and between groups. The same analyses were performed
to examine responses to the post-scan questionnaire. The rela-
tionship between IU, as measured by mean scores on the IUS,
and task performance was examined with Pearson correlations.
Functional MRI. First level statistical analyses were conducted
for each participant with FILM (Oxford Centre for Functional
Magnetic Resonance Imaging of the Brain [FMRIB]’s Improved
Linear Model) with local autocorrelation correction (31). Event-
related responses were modeled with a double ? hemodynamic
response. Predictors for each of the five levels described earlier
were included in our models: 100% (cards 1 & 9), 87.5% (cards 2 &
8), 75% (cards 3 & 7), 62.5% (cards 4 & 6), and 50% (card 5). A
predictor for the fixation condition was also included (to provide low
level reference), as were temporal derivatives for each predictor.
The FMRIB’s Local Analysis of Mixed Effects (FLAME) was
used to carry out second and third level analyses (32). At the
second level, fixed-effects analyses were conducted for each
participant. Prior work examining brain-behavior relationships
with the HiLo task suggests that patterns of neural activity
associated with the pure uncertainty condition (i.e., 50% proba-
bility of a correct response) are relatively distinct compared with
the probabilistic uncertainty conditions (62.5%, 75%, 87.5%, and
100%), which seem to lie along a linear spectrum. The aim of the
previous study was to examine developmental differences in
probabilistic decision-making and therefore analyses focused pri-
marily on the linear contrast, excluding the 50% condition. The
current study examines anxiety-related differences between pure
uncertainty and probabilistic uncertainty, and therefore the contrast
of interest compares these two conditions. In the highest level
analysis, subject was treated as a “random” effect (so-called “ran-
dom effects” analysis), allowing inference at the population level.
To test for group differences, each group (ANX and healthy
control subjects) was modeled with an independent predictor,
allowing us to report within-group as well as between-group
findings. Then, to examine the specific effect of IU on neural
activity, demeaned IUS scores were added to the model as an
independent predictor. Z (Gaussianised t/F) statistic images were
thresholded with clusters determined by z ? 2.3 and a (cor-
rected) cluster significance threshold of p ? .05 (33). Additional
analyses were conducted to look for group differences in the
effect of IU on uncertainty-related neural activity. Owing to
significant differences in the variance of IUS scores across
groups, a categorical approach was used. To minimize the
variability in the ANX group, it was divided into two subgroups
with a median split of the IUS scores to maximize power. Group
analyses compared IU-related neural activity across the two ANX
subgroups and the healthy control group.
The ANX and control groups did not differ significantly on
age, IQ, SES, and gender. Furthermore, these variables were not
significantly related to behavioral variables of interest including
IUS scores, task-based measures (response time, accuracy, and
variability), and scores on the post-scan questionnaire.
Group Analyses. Independent sample t tests revealed signif-
icant group differences in clinical measures of anxiety and worry.
The IUS scores were significantly higher and more variable
(Levene statistic ? 7.44, p ? .01) in the ANX group than in
control subjects (Table 1). Similarly, the ANX group endorsed
greater worry on the PSWQ and greater anxiety on the MASC
(Total T) than the healthy control subjects. The IUS scores were
highly correlated with the PSWQ in both groups (anxious: r ?
.64, p ? .01; control subjects: r ? .67, p ? .01; overall: r ? .80,
p ? .01), consistent with previous studies.
Repeated measures ANOVAs were conducted to examine
within- and between-subjects differences in performance be-
tween the 50% condition and the other four task conditions.
Within subjects, response times (RTs) were longer during the
50% condition than during the more certain trials [F(1,27) ? 31.5,
p ? .001]. There were no group or interaction effects. Subjects
were less accurate [F(1,27) ? 189.0, p ? .001] and more variable
in their responses [F(1,27) ? 55.2, p ? .001] during the 50%
condition than during the other trial types, consistent with task
design. No group or interaction effects of either accuracy or
response variability were found. Longer RTs, decreased accu-
racy, and increased variability of responses suggest that partici-
pants are not simply “giving up” during the 50% condition but are
using alternative strategies when they cannot rely upon proba-
bility to guide their decisions.
Analyses of responses to the anxiety question on the post-
scan questionnaire revealed significant effects of task condition
[F(1,27) ? 16.7, p ? .001] and group [F(1,27) ? 39.8, p ? .001]
(Figure 2). Anxiety ratings were higher for the 50% condition
than for the others, and the ANX group endorsed significantly
greater anxiety than the control subjects. However, the interac-
tion of group and task condition was not significant [F(1,27) ?
1.9, p ? .18]. Conversely, certainty ratings were lower for the 50%
Table 1. Self-Report Measures of Anxiety, Worry, and IU
Mean (SD)t (df ? 27)
MASC Total T
IUS, Intolerance of Uncertainty Scale (range: 27–135); PSWQ-C, Penn
State Worry Questionnaire for Children (range: 0–42); MASC, Multidimen-
sional Anxiety Scale for Children (range: 25–90).
ap ? .001.
A.L. Krain et al.
BIOL PSYCHIATRY 2008;63:563–568 565
condition [F (1,27) ? 80.8, p ? .001] compared with the other
four conditions, and anxious adolescents endorsed less certainty
than control subjects [F (1,27) ? 4.3, p ? .05] with no interaction
effect [F (1,27) ? .01, p ? .91].
Relationship Between IU and Task Performance. Across
groups, IUS scores were significantly correlated with ratings on
the post-scan questionnaire. Greater IU was associated with
higher anxiety (r ? .69, p ? .01) and lower certainty (r ? ?.55,
p ? .01) across all task conditions. Correlations between IUS
scores and measures of task performance (RT, accuracy, variabil-
ity) were not significant.
Functional MRI Results
Group Analyses. Between-group analyses were conducted
on the primary contrast of interest comparing activation during
the pure uncertainty condition (50% condition) with the other
four task conditions. Contrary to prediction, no significant group
differences were found.
Relationship of IU to Neural Activity. Although we did not
find predicted group differences, we continued to explore the
relationship between IU and uncertainty-related neural activity as
planned, with the same contrast as discussed earlier (50% condition
vs. other task conditions). Examination of brain activity associated
with IUS scores independent of group revealed significant activation
in limbic and paralimbic regions (Figure 3). Higher IUS scores were
associated with greater activity in right and left amygdala, medial
frontal gyrus (Brodmann’s area [BA] 10), and ACC (Table 2).
Additionally, significant activation was found in posterior and
temporal regions, including precentral gyrus, posterior cingulate,
parietal cortex, and middle temporal gyrus.
Examination of IUS Subgroups. Further analyses examined
whether the relationship between IU and uncertainty-related
brain activity differed between anxiety-disordered youth and
healthy control subjects. We treated healthy subjects, all of who
showed low IUS scores, as a single group and used a median split
to divide the ANX group into high (mean ? 76.9, SD ? 14.1) and
low IU (mean ? 47.9, SD ? 7.3) subgroups—as noted in
Methods. The mean IUS score of the low IU subgroup was still
significantly higher than the control subjects [mean ? 38.0, SD ?
7.8; t (19) ? 2.89, p ? .01]. The two anxious groups did not differ
on measures of overall anxiety [MASC Total: t (14) ? ?.64, p ?
.54] and worry [PSWQ-C Total: t (14) ? ?1.85; p ? .09]. Behav-
iorally, compared with the low IU subgroup, the high IU
subgroup demonstrated longer RTs during the 50% condition
[t (2,14) ? ?2.11; p ? .05] and endorsed greater overall anxiety
during the task [t (2,14) ? ?2.12; p ? .05].
With the primary contrast of interest (50% condition vs. other
four task conditions), between-group analyses of the low and
high IU subgroups revealed significant differences. Compared
with the low IU subgroup, the high IU subgroup demonstrated
greater activity in orbitofrontal and limbic regions, including the left
amygdala and rostral and subgenual portions of ACC (Figure 4A).
Further comparisons of these regions of activation across all
three groups (high IU, low IU, healthy control subjects) yielded
significant differences [F (2,28) ? 22.03, p ? .001]. Post hoc
analyses revealed significantly less activation in the low IU
subgroup than in either the high IU subgroup (p ? .001) or
control subjects (p ? .001) and greater activation in the high IU
group than control subjects (p ? .05). Figure 4B shows that the
low IU subgroup showed significant deactivation and the high IU
subgroup showed activation, whereas the control subjects
showed little or no change in activity.
S p e cificity of Relationship Between IU and Neural Activity. To
examine whether IU- related neural activity was specific to this
measure or might also be related to other variables, such as age,
anxiety, and task performance, correlational analyses were con-
ducted with the significant clusters from the IU analysis. Across
groups, activity in these regions was not significantly correlated
with age, scores on the PSWQ-C or MASC, or task-based mea-
sures including accuracy, variability, and post-scan ratings of
anxiety and certainty. The RT during the 50% condition was
correlated with activation in bilateral amygdala (Right: r ? .45,
p ? .015; Left: r ? .43, p ? .02), posterior cingulate (r ? .42,
p ? .02), and precentral gyrus (r ? .40, p ? .03); however, these
correlations were no longer significant after Bonferroni correc-
tion (p ? .05/10).
This is the first study to link IU in participants with anxiety
disorders to the neural substrates believed to underlie these
disorders. High levels of IU characteristic of clinical anxiety were
associated with increased activation in frontal and limbic regions
in response to uncertainty, consistent with study hypotheses. An
Figure 3. Regions of significant Intolerance of Uncertainty Scale (IUS)- re-
lated neural activity for the primary contrast (50% condition vs. other four
conditions). Clusters survive Gaussian Random Field (GRF) correction at
p ? .05.
Figure 2. Mean post-scan questionnaire ratings (0–3) of anxiety and cer-
cents with anxiety disorders; HC, healthy control subjects.
566 BIOL PSYCHIATRY 2008;63:563–568
A.L. Krain et al.
unexpected finding was that adolescents with anxiety disorders
demonstrated different patterns of neural responsiveness de-
pending upon IU: those who endorsed low IU showed deacti-
vation, whereas those with high IU showed activation of frontal
and limbic regions. These findings suggest that IU impacts the
neurocognitive processing of uncertain decisions in adolescents
with anxiety disorders and, together with previous research
demonstrating a key role of IU in adult GAD, implicate this trait
in the development and maintenance of these disorders.
Behavioral findings provide support for the role of IU in
anxiety disorders and the impact of IU on affective responses to
uncertainty. The ANX group reported significantly greater IU
than the non-anxious control subjects, consistent with the adult
literature. Although the groups did not differ in their perfor-
mance on the HiLo Game, the ANX group endorsed greater
anxiety and less certainty on the post-scan questionnaire. This
suggests a relationship between IU and responses to uncertainty
that is further supported by the significant correlations between
IUS scores and ratings of anxiety and certainty during the task.
Greater IU, which is characteristic of anxiety-disordered popula-
tions (2), is associated with increased anxiety and decreased
certainty during uncertain decisions.
Previous neuroimaging and neuropsychological studies have
shown that GAD and social phobia in adult (34) and adolescent
(15–17,23) populations are characterized by vigilance toward social
threat cues (i.e., faces) coupled with hyper-reactivity of frontal–
amygdala limbic regions. On the basis of research demonstrating
that similar circuits are involved in processing uncertainty (11–13),
we hypothesized that anxious adolescents would demonstrate
greater uncertainty-related activity in these regions than healthy
control subjects and that this difference would be related to in-
creased IU. We found that, consistent with this prediction, IU was
significantly correlated with uncertainty-related neural activity in
limbic regions, including ACC, OFC, and bilateral amygdala. This
suggests that greater IU is associated with an elevated affective
response to uncertainty, which is further supported by the positive
correlation between IUS scores and ratings of anxiety during the
Despite this relationship between IU and neural activity, we did
not observe predicted group differences. Further analyses demon-
strated that anxious adolescents with low IU showed deactivation of
key regions during the 50% condition. We did not anticipate this
finding. However, we can speculate that patients with lower IU
suppress activity in these regions, allowing them to manage uncer-
tainty more effectively than the high IU subgroup. This suppression
might serve as an emotion regulation strategy, resulting in the lower
is consistent with recent findings suggesting that amygdala deacti-
vation might be an essential component of emotion regulation (35).
Given the preliminary nature of these results, further investigation
and replication is needed to draw firm conclusions about the
validity of these anxiety subgroups.
There are several limitations to the current study. First, the ANX
group demonstrated significant comorbidity. However, this is con-
sistent with most studies of this population, where comorbidity is
the rule. For example, in recent functional MRI studies of adoles-
cents with GAD, 40%–60% of the samples had comorbid social
phobia (15,23). Clinical samples show similarly elevated rates of
comorbidity (36). The current sample was also notable for comorbid
ADHD (n ? 3). Comparison of comorbid ADHD and non-ADHD
anxious subjects on behavioral and neural measures showed no
differences. As such, comorbid ADHD is not a likely contributor to
the present findings. Second, we were unable to fully examine the
interaction between diagnostic status (anxiety disorder vs. control)
and IU because of the limited range of IUS scores in the healthy
control subjects. Inclusion of a group of healthy control subjects
with high IUS scores would be needed to conduct such analyses;
however, finding such a group is difficult, given the close linkages
Figure 4. (A) Orbitofrontal and limbic regions that show significantly
greater activation in the high intolerance of uncertainty (IU) subgroup of
Random Field (GRF) correction at p ? .05. (B) Group differences in uncer-
tainty-related neural activity in orbitofrontal and limbic regions [F(2,28) ?
22.03, p ? .001]: High IU ? Low IU (p ? .001), High IU ? healthy control
subjects (HC) (p ? .05), HC ? Low IU (p ? .001).
Table 2. Thresholded Clusters of Activity Significantly Associated with IUS Scores Across Groups
RegionL/R BANo. Voxels
Anterior cingulate cortex/
middle frontal gyrus
Anterior cingulate cortex
Middle frontal gyrus
Inferior frontal gyrus/amygdala
Middle temporal gyrus
R 32/101888 1446
p ? .05. IUS, Intolerance of Uncertainty Scale; COI, center of intensity in Talaraich coordinates; BA, Brodmann’s area.
A.L. Krain et al.
BIOL PSYCHIATRY 2008;63:563–568 567
between high IU and anxiety disorders (4,37). Finally, recent work Download full-text
suggests possible confounds associated with the use of a fixation
baseline (38). Task-related deactivations, such as those observed in
the low IU subgroup, can result from increased activation to
fixation, particularly in limbic regions. This alternative explanation
was examined, and the groups were not found to differ in their
activity related to the fixation baseline.
In summary, the current findings suggest that, within pediatric
anxiety disorders, subgroups of patients exist with similar diag-
nostic and symptom presentations, who demonstrate signifi-
cantly different pathophysiological and affective responses to
uncertainty. By using a trait measure such as IU, we were able to
explain significant variability in uncertainty-related activations of
limbic and paralimbic regions in adolescents with anxiety disor-
ders, highlighting the potential merits of examining such traits in
future studies rather than relying solely on diagnostic categories.
This research was supported by the National Alliance for
Research in Schizophrenia and Depression and the National
Institute of Mental Health (K23MH074821).
We would like to thank Mauricio Delgado, Ph.D., for his
assistance designing the HiLo Game and Rachel G. Klein, Ph.D., for
her helpful comments and feedback. Finally, we would like to thank
the study participants and their families who dedicated their time to
The authors have no financial relationships to disclose.
1. Achenbach TM, Howell CT, McConaughy SH, Stanger C (1995): Six-year
2. Costello EJ, Mustillo S, Erkanli A, Keeler G, Angold A (2003): Prevalence
and development of psychiatric disorders in childhood and adoles-
3. “Anxiety.” The Oxford English Dictionary, 2nd ed. 1989. OED Online.
Oxford University Press. 26 Feb, 2007. http://dictionary.oed.com/cgi/
4. Dugas MJ, Buhr K, Ladouceur R (2004): The role of intolerance of uncer-
tainty in etiology and maintenance. In: Heimberg RG, Turk CL, Mennin
tice. New York: Guilford Press, 143–163.
5. Laugesen N, Dugas MJ, Bukowski WM (2003): Understanding adoles-
cent worry: The application of a cognitive model. J Abnorm Child Psy-
6. Dugas MJ, Ladouceur R, Leger E, Freeston MH, Langlois F, Provencher
MD, et al. (2003): Group cognitive-behavioral therapy for generalized
anxiety disorder: Treatment outcome and long-term follow-up. J Con-
7. Ladouceur R, Dugas MJ, Freeston MH, Leger E, Gagnon F, Thibodeau N
iety disorder: Evaluation in a controlled clinical trial. J Consult Clin Psy-
8. Leger E, Ladouceur R, Dugas MJ, Freeston MH (2003): Cognitive-behav-
ioral treatment of generalized anxiety disorder among adolescents: A
erance of uncertainty in worry. Experimental findings. Behav Modif 21:
10. Tallis F, Eysenck M, Mathews A (1991): Elevated evidence requirements
responding to degrees of uncertainty in human decision-making. Sci-
12. Kim H, Somerville LH, Johnstone T, Alexander AL, Whalen PJ (2003):
Inverse amygdala and medial prefrontal cortex responses to surprised
faces. Neuroreport 14:2317–2322.
A functional MRI study of human amygdala responses to facial expres-
sions of fear versus anger. Emotion 1:70–83.
14. McClure EB, Monk CS, Nelson EE, Parrish JM, Adler A, Blair RJR, et al.
(2006): Abnormal attention modulation of fear circuit function in pedi-
atric generalized anxiety disorder. Arch Gen Psychiatry 64:97–106.
response to angry faces in adolescents with generalized anxiety disor-
(2001): Amygdala response to fearful faces in anxious and depressed
17. Killgore WD, Yurgelun-Todd DA (2005): Social anxiety predicts amyg-
dala activation in adolescents viewing fearful faces. Neuroreport 16:
activation in high trait anxious subjects is related to altered error pro-
cessing during decision making. Biol Psychiatry 55:1179–1187.
examination of developmental differences in the neural correlates of un-
20. Silverman WK, Albano AM (1996): The Anxiety Disorders Interview
Schedule for Children for DSM-IV, Child and Parent Versions. San Anto-
nio, Texas: Psychological Corporation.
21. Kendall PC, Brady EU, Verduin TL (2001): Comorbidity in childhood
anxiety disorders and treatment outcome. J Am Acad Child Adolesc Psy-
22. Vasey MW, MacLeod C (2001): Information-processing factors in child-
hood anxiety: A review and developmental perspective. In: Vasey MW,
don: Oxford University Press, 253–277.
23. McClure EB, Monk CS, Nelson EE, Parrish JM, Adler A, Blair RJR, et al.
(2007): Abnormal attention modulation of fear circuit function in pedi-
atric generalized anxiety disorder. Arch Gen Psychiatry 64:97–106.
Antonio, Texas: Psychological Corporation.
25. Achenbach TM (1991): Manual for the Child Behavior Checklist 4-18 and
1991 Profile. Burlington: University of Vermont.
26. Freeston MH, Rheaume J, Letarte H, Dugas MJ (1994): Why do people
27. March JS (1998): Multidimensional Anxiety Scale for Children. North
Tonawanda, New York: Multi-Health Systems.
28. Chorpita BF, Tracey SA, Brown TA, Collica TJ, Barlow DH (1997): Assess-
ment of worry in children and adolescents: An adaptation of the Penn
State Worry Questionnaire. Behav Res Ther 35:569–581.
29. Jenkinson M, Bannister P, Brady M, Smith S (2002): Improved optimiza-
tion for the robust and accurate linear registration and motion correc-
tion of brain images. Neuroimage 17:825–841.
30. Jenkinson M, Smith S (2001): A global optimisation method for robust
affine registration of brain images. Mel Image Anal 5:143–156.
31. Woolrich MW, Ripley BD, Brady M, Smith SM (2001): Temporal autocor-
relation in univariate linear modeling of FMRI data. Neuroimage 14:
32. Beckmann CF, Jenkinson M, Smith SM (2003): General multilevel linear
modeling for group analysis in FMRI. Neuroimage 20:1052–1063.
33. Worsley KJ, Evans AC, Marrett S, Neelin P (1992): A three-dimensional
statistical analysis for CBF activation studies in human brain. J Cereb
34. Phan KL, Fitzgerald DA, Nathan PJ, Tancer ME (2006): Association be-
tween amygdala hyperactivity to harsh faces and severity of social
anxiety in generalized social phobia. Biol Psychiatry 59:424–429.
within childhood anxiety disorders. J Clin Child Adolesc Psychol 32:
37. Dugas MJ, Gagnon F, Ladouceur R, Freeston MH (1998): Generalized
anxiety disorder: A preliminary test of a conceptual model. Behav Res
transporter genotype (5-HTTLPR): Effects of neutral and undefined condi-
568 BIOL PSYCHIATRY 2008;63:563–568
A.L. Krain et al.