Failing to deactivate: Resting functional
abnormalities in autism
Daniel P. Kennedy*†‡, Elizabeth Redcay†§, and Eric Courchesne*¶
Departments of *Neurosciences, and§Psychology, University of California at San Diego, 9500 Gilman Drive, La Jolla, CA 92093; and¶Center for Autism
Research, Children’s Hospital Research Center, 8110 La Jolla Shores Drive, Suite 201, La Jolla, CA 92037
Edited by Marcus E. Raichle, Washington University School of Medicine, St. Louis, MO, and approved April 3, 2006 (received for review January 26, 2006)
Several regions of the brain (including medial prefrontal cortex,
rostral anterior cingulate, posterior cingulate, and precuneus) are
known to have high metabolic activity during rest, which is
suppressed during cognitively demanding tasks. With functional
magnetic resonance imaging (fMRI), this suppression of activity is
observed as ‘‘deactivations,’’ which are thought to be indicative of
an interruption of the mental activity that persists during rest.
Thus, measuring deactivation provides a means by which rest-
associated functional activity can be quantitatively examined.
Applying this approach to autism, we found that the autism group
failed to demonstrate this deactivation effect. Furthermore, there
was a strong correlation between a clinical measure of social
impairment and functional activity within the ventral medial pre-
frontal cortex. We speculate that the lack of deactivation in the
autism group is indicative of abnormal internally directed pro-
and emotional deficits of autism.
default mode ? functional MRI ? introspection ? medial prefrontal
cortex ? precuneus
sistently activate a medial cortical network involving several
brain regions, namely, the medial prefrontal cortex (MPFC) and
adjacent rostral anterior cingulate cortex (rACC), posterior
cingulate cortex (PCC), and precuneus (PrC) (1–4). Interest-
ingly, this network is active when normal subjects are passively
resting (5), leading many to speculate that these internally
directed thoughts dominate the resting state (6–9). Self-reports
from subjects while at rest further support this interpretation,
either recent or ancient, consisting of familiar faces, scenes,
dialogues, stories, and melodies’’ (8). Conversely, activity in this
midline ‘‘resting network’’ is reduced when subjects perform
externally directed, attention-demanding, goal-oriented tasks
(such as the Stroop task or math calculations), and the resulting
‘‘deactivation’’ of this network is thought to be an indicator of an
interruption of ongoing internally directed thought processes (5,
6, 9–11). In this context, the term deactivation simply refers to
activity that is greater during rest than during task performance
(i.e., the opposite of the more typically reported activations).
Thus, an objective method for testing the functioning of this
midline resting network is to measure whether there is deacti-
vation in these regions during externally directed tasks as
‘‘deactivation effect’’ have been used in studies of patients with
fragile X (12), a developmental disorder with some character-
istics that overlap with autism, and in patients with dementia of
the Alzheimer type (13) and Alzheimer’s disease (14).
network might be functioning abnormally in autism. First, the
functions it subserves [including emotional processing (15–17),
perception of social interactions (18), theory of mind (19–22),
experience of joint attention (23), and person familiarity (24,
25)] overlap remarkably well with the social and emotional
nternally directed processes, such as self-reflective thought
and most higher-order social and emotional processes, con-
deficits that characterize autism. Second, in anterior regions of
this network, researchers have documented volumetric, meta-
bolic, cellular, and developmental growth abnormalities in this
disorder (reviewed in ref. 26). Third, neuroimaging studies of
autism have often observed functional abnormalities in these
midline cortical regions during a variety of both socioemotional
(25, 27, 28) and nonsocioemotional tasks (29–31). These func-
tional abnormalities in nonsocioemotional tasks [e.g., a visually
cued motor task (30)] are particularly interesting because they
suggest that the results may be due to resting baseline differences
between groups, rather than task-related differences.
To objectively determine whether in autism this resting
network functions abnormally, we used functional magnetic
resonance imaging (fMRI) to measure this deactivation effect
in autistic and normal control subjects. To control for task
performance across patient and control groups, we used the
Stroop task because it is known that autistic patients are
unimpaired in this task relative to controls (32, 33). To
demonstrate that the magnitude of the deactivation effect in
these regions can be modulated in control but not autistic
subjects without changing the specific task demands, three
conditions of the counting Stroop task were used: one with
incongruent-number stimuli, one with emotional stimuli, and
one with neutral stimuli. Activity in each condition was
compared with activity during passive rest. To examine indi-
vidual subject and group effects, we used a whole-brain group
analysis followed-up with a region-of-activation approach in
which the pattern of activity for each autistic and control
individual could be characterized and quantified.
We had two main comparisons of interest: First, we predicted
that the greatest level of functional deactivation from a passive
resting state in control subjects would be seen in the most
cognitively demanding task condition, because this would draw
the most attention away from internal thought processes that
occur during rest. Based on previous research on the counting
Stroop task (17, 34), we chose the number condition to contrast
with rest, because counting the number of words while reading
incongruent numbers would create more interference (as re-
flected in longer reaction times and lower accuracy) than
counting the number of emotional or neutral words. We hypoth-
esized that in subjects with autism, although the number con-
dition will still be the most cognitively demanding condition, this
implicating differences in functional activity (and, therefore,
functional processes) during rest. Second, based on previous
Conflict of interest statement: No conflicts declared.
This paper was submitted directly (Track II) to the PNAS office.
Abbreviations: ASD, autism spectrum disorder; fMRI, functional magnetic resonance im-
aging; MPFC, medial prefrontal cortex; rACC, rostral anterior cingulate cortex; PCC, pos-
terior cingulate cortex; PrC, precuneus.
Data deposition: The fMRI data reported in this paper have been deposited with the fMRI
Data Center, www.fmridc.org (accession no. Z-2006-121FQ).
†D.P.K. and E.R. contributed equally to this work.
‡To whom correspondence should be addressed. E-mail: firstname.lastname@example.org.
© 2006 by The National Academy of Sciences of the USA
May 23, 2006 ?
vol. 103 ?
no. 21 ?
studies using negatively valenced words (17), we predicted that
control subjects would show relatively greater activity in anterior
midline regions (MPFC and rACC) in the emotional compared
with the neutral condition. However, we hypothesized that
autistic subjects would fail to show this activity related to implicit
emotional processing of negatively valenced words. Such a
pattern of results would suggest that there is abnormal function-
ing of the midline resting network in autism during rest and
emotion processing, which we speculate may reflect a more
general failure to engage in the types of internally directed
thoughts that normally recruit this network.
Behavior. As predicted, autism spectrum disorder (ASD) and
control participants showed similar behavioral responses while
performing the counting Stroop task (see Fig. 3 A and B, which
is published as supporting information on the PNAS web site).
Reaction time did not differ significantly between groups [ASD,
676.1 ms; control, 664.1 ms; F(1, 23) ? 0.097; P ? 0.759];
however, there was a main effect of condition [F(2, 22) ? 4.811;
P ? 0.018]. Contrasts corrected for multiple comparisons re-
vealed that this effect was due to significantly longer reaction
in the emotional (659.3 ms) and neutral (659.0 ms) conditions
[F(1, 23) ? 10.05, P ? 0.004]. There was no significant interac-
tion between condition and group for reaction time [F(2, 22) ?
0.730, P ? 0.493]. In the accuracy data, the ASD subjects did
have a significantly lower overall percent correct score than did
the control group [ASD, 95.8%; control, 98.8%; F(1, 22) ?
6.499; P ? 0.018]. However, mean accuracy for both groups was
?95% for all three conditions, indicating that, despite group
differences, both groups were performing the task with a high
level of accuracy. There was also a main effect of condition with
number showing a lower percent correct than the other two
conditions [neutral, 97.7%; emotional, 97.4%; number, 96.7%;
condition and group [F(2, 21) ? 0.679, P ? 0.518].
Immediately after image acquisition, subjects were given a
word recognition task for which they were asked to mark all of
the words they had seen during scanning. Repeated-measures
ANOVA revealed a main effect of word type (emotional vs.
neutral) [emotional, 86.0%; neutral, 73.8%; F(1, 19) ? 11.43;
P ? 0.003] and a trend toward an interaction between group and
word type [F(1, 19) ? 4.24, P ? 0.062] (Fig. 3C). Follow-up
within group t tests revealed a significant effect of word type in
control subjects [neutral, 72.2%; emotional, 91.1%; t(1, 8) ?
?4.02; P ? 0.004] but not in ASD subjects [neutral, 75.4%;
emotional, 80.8%; t(1, 11) ? ?1.15; P ? 0.274].
representing deactivations. The black outlines correspond to the area of deactivation derived from the number vs. rest condition in controls (A Left) mapped
onto the sagittal slices of the images shown. These outlines, which represent regions active during rest, highlight the presence or absence of activations or
deactivations in each image. Each sagittal slice location differs slightly because each image was chosen to best represent the midline activity for each group and
comparison. The comparisons shown are number vs. rest (A), neutral vs. rest (B), emotional vs. rest (C), and emotional vs. neutral (D).
Significant functional activity derived from group whole-brain analyses (P ? 0.01, cluster corrected). t values are displayed, with negative t values
www.pnas.org?cgi?doi?10.1073?pnas.0600674103Kennedy et al.
Functional Imaging. Group analysis. For control subjects, whole-
brain analysis of the number vs. rest contrast revealed large
cluster-corrected) (Fig. 1A; see also Table 1, which is published
as supporting information on the PNAS web site). However, this
deactivation was entirely absent in ASD subjects. Furthermore,
a direct group comparison between control and ASD subjects
revealed a significant difference between groups in MPFC?
rACC and PrC (Fig. 1A; see also Table 2, which is published as
supporting information on the PNAS web site). The right
superior temporal sulcus and bilateral angular gyrus also deac-
tivated in control subjects but not in ASD subjects, although
these regions were not significantly different in the direct group
comparison (see Tables 1 and 2).
Similar but weaker effects were seen in the neutral vs. rest and
emotional vs. rest comparisons. In the neutral vs. rest contrast
(Fig. 1B), cluster volume in control subjects was reduced com-
pared with that seen in the number vs. rest comparison, likely
reflecting differences in difficulty between the conditions (10).
Again, the autism group showed no significant deactivations in
1C), significant deactivations were seen in the PCC and PrC in
controls but not in ASD subjects. Furthermore, a significant
cluster of positive activity was observed in the dorsal MPFC in
both groups, although this activity was of a greater extent in
control subjects. In both of these comparisons, there were no
significant differences in these midline regions in the direct
To determine functional activity specific to emotional pro-
1D). As predicted, significantly increased activity was observed
in the ventral MPFC?rACC, extending into the medial portion
of the orbital frontal cortex, for control subjects, but this
activation was absent in ASD subjects (Table 1). Direct group
comparison revealed significantly greater activity in the control
subjects in this medial orbital region of the resting network (Fig.
1D and Table 2).
Finally, to ensure that there were no between-group differ-
ences in functional activity relating to the interference aspect of
the task [namely, in the dorsal anterior cingulate cortex (34)], we
examined group differences between the number and neutral
conditions. Even at a more relaxed threshold (P ? 0.05, uncor-
rected), dorsal anterior cingulate cortex activity was still not
significantly different between groups (see Fig. 4, which is
published as supporting information on the PNAS web site),
although differences were seen in MPFC and PrC.
Exploratory correlational analyses.Wefoundaclusterofvoxelsinthe
ventral MPFC (P ? 0.005, uncorrected) whose activity during
the number vs. rest contrast significantly correlated with the
social subscale score on the Autism Diagnostic Interview–
Revised in ASD individuals [r(11) ? 0.939, P ? 0.0001)] (Fig. 2).
This region was partially overlapping with the region found to be
significantly different between the ASD and control groups in
the number vs. rest contrast (see Fig. 2 Inset). ASD subjects with
greater deactivation had lower social impairment scores,
whereas those that showed less deactivation (or, in certain
individuals, activation) had greater social impairment scores.
Furthermore, this correlation was still significant when the two
extreme points (i.e., subjects with the lowest and highest scores
on the social subscale) were removed from the analysis [r(9) ?
0.669, P ? 0.034].
Additional Analyses. Because we noted individual variability in
deactivation and activation patterns in the ASD subjects, we
chose to examine the data at the individual subject level. These
results are published in Supporting Text and Fig. 5, which are
published as supporting information on the PNAS web site.
Our results demonstrate that ASD subjects fail to show the
PCC?PrC) in the number vs. rest comparison (Fig. 1A). These
results cannot be accounted for merely by differences in task
performance because (i) no differences were seen in reaction
time between groups; (ii) slightly reduced accuracy (and, thus,
increased difficulty) in the ASD group should lead to greater
deactivation (10), although we observed the exact opposite
result; (iii) there were no significant correlations between task
performance and functional activity in the anterior region-of-
activation mask; and (iv) there were no neurofunctional differ-
ences in regions involved in Stroop task performance (namely,
the dorsal ACC) (see Figs. 1A, 4, and 5B). ASD subjects also
of emotional stimuli. Although control subjects showed a rec-
an effect of word type on recognition ability, consistent with
previous behavioral reports in autism (35). Furthermore, we
demonstrated that autistic subjects failed to show the normal
pattern of functional activation in the medial orbital frontal
region of the resting network when processing emotional words
compared with neutral words (Fig. 1D), demonstrating a failure
of emotional task-induced modulation of this region. Finally, in
the ASD group, there was a high correlation (r ? 0.939) between
a clinical measure of social impairment and functional activity in
the ventral MPFC (Fig. 2). The subjects with higher social
impairment scores had less deactivation in the number vs. rest
contrast (i.e., the greater the behavioral abnormality, the greater
the neurofunctional abnormality) and vice versa. We should
emphasize, however, that we have only identified a correlative
relationship between abnormal social behavior and brain activ-
ity, rather than a causative one. Thus, although tantalizing, this
result cannot be used to argue that this neurofunctional abnor-
mality causes social impairment or that social impairment causes
this neurofunctional abnormality.
There are two possible reasons why the ASD group failed to
show the typical deactivation effect. One possibility is that
midline resting network activity during both rest and task
performance is high, and, thus, a subtraction between these
conditions would reveal no difference in activity levels. We
believe, however, that it is unlikely that high midline network
activity was maintained during the cognitively demanding num-
ber task in autism for several reasons. First, as mentioned
previously, behavioral performance was similar between control
and ASD groups. This result, however, would be unexpected if
(Inset) and score on the social subscale of the Autism Diagnostic Interview–
Revised in the ASD group.
Correlation between functional activity in a ventral MPFC cluster
Kennedy et al.
May 23, 2006 ?
vol. 103 ?
no. 21 ?
the ASD group were carrying out additional mental processing
that control subjects inhibit during cognitively demanding con-
ditions. Second, positron-emission tomography studies of au-
tism, which provide an absolute measure of brain metabolism,
have found reduced, as opposed to increased, glucose metabo-
lism in rACC and PCC (36) during task performance, as
compared with controls. Furthermore, one positron-emission
tomography study found that lower blood flow in MPFC and
rACC at rest was correlated with more severe social and
communicative impairments in subjects with autism (37), a
finding similar to our correlational results. Third, reduced
anatomical volumes and neurochemical deficiencies have con-
sistently been observed in MPFC?rACC in adults with autism
(reviewed in ref. 26), likely indicative of a reduced functioning
of these regions. Therefore, an alternative explanation, the one
to which we attribute the lack of deactivation, is that midline
activity is low during rest. We suggest, then, that the absence of
deactivation in this network indicates that the mental processes
that normally occur at rest are absent or abnormal in autism.
What are these mental processes that dominate during rest?
Evidence in the literature to date seems to suggest that tasks that
induce certain types of internal processing activate this resting
network. Examples of such tasks are self- and other-person
judgments (4, 6, 7, 19–22, 38–45), person familiarity judgments
(24, 25), emotion processing (15–17, 46), perspective-taking (22,
47), passive observation of social interactions vs. nonsocial
interactions (18), relaxation based on interoceptive biofeedback
(48, 49), conceptual judgments (based on internal knowledge
(51), among others [moral decision making (52), joint attention
experience (23), and pleasantness judgments (53)]. Therefore,
the activity in these regions at rest might simply reflect the extent
to which these types of internally directed thoughts are engaged
at rest. In fact, a particularly intriguing behavioral study found
that individuals with ASD report very different internal thoughts
than control subjects (54, 55), lending support to our interpre-
tation that an absence of this resting activity in autism may be
directly related to abnormal internal thought. Admittedly, this is
a speculative hypothesis but one that can be explicitly tested.
The findings in the present study are supported by previous
functional imaging studies of autism, which have consistently
found abnormalities in these midline regions during a variety of
tasks. Although these functional abnormalities would be ex-
pected in socioemotional tasks [i.e., personally familiar face
processing (25), facial affect processing (28), theory of mind
(27)] because these resting network regions support these pro-
cesses, these abnormalities have also been seen using nonsocio-
emotional tasks [i.e., spatial working memory (29), visually cued
motor task (30), embedded figures task (31)]. Furthermore,
within both socioemotional and nonsocioemotional task do-
mains, the direction of abnormality has not been consistent in
that some studies report greater activity, whereas others report
reduced activity compared with control subjects. Unfortunately,
interpretation of these findings is not straightforward, because
different studies report results in different ways (e.g., individual
group results or direct group comparisons) and use different
control conditions. For instance, in our study, if we only reported
results from the direct comparison for control vs. ASD subjects
in the number vs. rest condition, it would seem like ASD subjects
had greater activity in resting network regions (Fig. 1A Right),
although such an oversimplified interpretation would be flawed
because it actually reflects an absence of deactivation in the
autism group. Therefore, although functional abnormality
within midline anterior and posterior regions has been repeat-
edly shown in ASD, the precise nature of and explanation for
these abnormalities have been unclear based on past studies. We
offer an encompassing explanation for these disparate imaging
results by suggesting that the seemingly inconsistent findings can
be attributed to a lack of presenting individual group data and
to a lack of a common reference to a resting baseline condition.
Although our discussion of resting network regions has been
limited to midline resting regions, other regions, including
superior temporal sulcus, temporal pole, and angular gyrus, are
known to show the same deactivation effect (5, 50). In fact, we
observed deactivation in these regions in control subjects in the
number vs. rest contrast, and, as expected, these regions failed
to deactivate in the ASD group (Table 1). However, functional
activity in these regions was not found to be significantly
different in our relatively conservative direct group comparison
(P ? 0.01, whole-brain corrected), likely due to the smaller size
and increased anatomic variability of these regions (particularly,
the superior temporal sulcus) as compared with midline regions.
Interestingly, several functional imaging studies have shown
abnormalities in these regions (27, 56–58). Follow-up studies
using region of interest approaches or sulcal mapping are needed
to determine the functionality of these other resting network
regions in autism.
It is particularly remarkable that the autism behavioral phe-
notype can be caused by a number of different factors (including
genetic, infectious, and environmental toxins), suggesting that
these factors are likely acting on a common neural substrate that
must be particularly vulnerable to developmental insult. In fact,
Raichle and colleagues (5) speculated that the PrC and PCC
might be highly susceptible to damage because of their high
regions are richly connected with numerous cortical and sub-
cortical regions (59–62), an early insult affecting any one of
these areas may have devastating, widespread consequences on
brain connectivity and subsequent functionality. Interestingly, a
recent diffusion tensor imaging study on autism identified white
matter abnormality in the region underlying the rACC?MPFC
(63). Furthermore, several fMRI studies have suggested reduced
functional connectivity in the autistic brain in a variety of tasks
(27, 64–66). Functional connectivity studies, which allow for
observation of brain activity during rest (67, 68) or naturalistic
viewing conditions (69), may be particularly useful in further
characterizing resting state abnormalities in autism.
Although purely speculative, it is interesting to hypothesize
about the nature of resting thoughts in ASD, if the typical
internally directed thoughts are absent or abnormal. One pos-
sibility is that their thoughts are directed toward their obsessive
interests and preoccupations (55), which are often of a concrete,
as opposed to abstract and subjective, nature (for instance,
calendars, maps, or schedules), or perhaps hypersensitivity to
their external environment constantly interrupts the full emer-
gence and elaboration of internally directed thoughts.
in ASD is abnormal, which suggests that individuals with ASD
might fail to engage in typical internally directed resting
thoughts. Furthermore, during an emotion processing task, both
behavioral performance and functional activity in a specific
region of this resting network (medial orbital frontal cortex) was
abnormal in ASD, suggesting that behavioral impairment might
the amount of functional abnormality in MPFC correlated with
experiments exploring resting network activity and organization
processes, such as self-reflection, may provide valuable insights
into the neurocognitive basis of the disorder.
Materials and Methods
Participants. Fifteen participants with ASD (10 with high-
functioning autism, 3 with Asperger’s syndrome, and 2 with
pervasive developmental disorder–not otherwise specified; 2
were left-handed and 1 was ambidextrous) and 14 healthy
www.pnas.org?cgi?doi?10.1073?pnas.0600674103Kennedy et al.
controls (three were left-handed) participated in this experi-
ment. Three participants with ASD (two with autism and one
with pervasive developmental disorder–not otherwise specified)
were excluded from all analyses because of uncorrectable head
motion. All participants or their legal guardians gave informed
written consent and were monetarily compensated for partici-
pation in the experiment. The protocol was approved by the
Institutional Review Board of the University of California at San
Diego, and Children’s Hospital at San Diego. All ASD partic-
ipants were diagnosed by a clinical psychologist with the Autism
Diagnostic Interview–Revised (70) and the Autism Diagnostic
Observation Schedule (71) and administered the Wechsler
Adult Intelligence Scale or Wechsler Adult Intelligence Scale–
Revised. The mean ages of ASD participants (25.49 ? 9.61) and
control participants (26.07 ? 7.95) were not significantly differ-
ent [t(24) ? ?0.169, P ? 0.87]. Clinical data for individual ASD
subjects are reported in Table 3, which is published as supporting
information on the PNAS web site.
Stimuli. While in the scanner, subjects performed a counting
Stroop task with three conditions of interest (34). This task was
based on the classic color–word Stroop (72) but adapted to be
more suitable for the fMRI environment (34). Instead of
naming the color of a word, subjects were instructed to count
the number of words that appeared on the screen and respond
as quickly and accurately as possible by pressing the button on
the response device corresponding to either two, three, or four
words. Subjects were presented with three different types of
words: emotional, neutral, and number words. Number words
were always incongruent with the number of words that
appeared on the screen (e.g., ‘‘two’’ written three times). The
emotional words were chosen from a set of negatively valenced
emotional words that had been rated by a separate set of
participants for degree of emotional arousal. Ten words with
the highest ratings of emotional arousal were included in the
study. Examples of these words include ‘‘murder,’’ ‘‘torture,’’
and ‘‘blood.’’ The 10 neutral words, each naming a household
item, included words like ‘‘table,’’ ‘‘curtain,’’ and ‘‘desk.’’
Stimuli were presented for 1.5 s each in a block design, with
each block containing 20 presentations of either emotional,
neutral, or number words, for a total of 30 s per block. There
were 12 task blocks, interlaced with three 21-s rest periods,
wherein subjects were simply instructed to passively view a
Behavioral Data Acquisition and Analysis. Stimuli were presented
using the PRESENTATION software package (Neurobehavioral
Systems, Albany, CA), whereby reaction time and number of
after scanning, each participant was asked to complete a surprise
word recognition test outside of the scanner. The test consisted
of two columns; one with 20 neutral words describing household
items and one with 20 negatively valenced emotional words.
Subjects were instructed to identify the words they recognized
from the scanning session.
Subject responses were scored for average reaction time
(from stimulus onset until subject response) and percent
correct for each condition. Data from the word recognition
test were scored for hits, misses, and false alarms and then
converted into a percent correct score. All behavioral analyses
were conducted with SPSS 12.0 statistical software package
(SPSS, Chicago, IL). Two repeated-measures ANOVAs were
run with group (autism, control) as the between-subjects factor
and reaction time or percent correct for each of the three
conditions (emotional, neutral, and number) as the within-
subjects factor. A third repeated-measures ANOVA was run
on the posttest word recognition data with group as the
between-subjects variable and percent correct for each con-
dition (emotional words or neutral words) as the within-
Functional Imaging Data Acquisition and Analysis. Images were
acquired on a Symphony 1.5 Tesla Scanner (Siemens, Iselin,
NJ) at the University of California at San Diego Hillcrest
Medical Center. Whole-brain axial slices were collected with
a gradient-recalled echo-planar imaging pulse sequence [rep-
etition time, 3,000 ms; echo time, 35 ms; flip angle, 90°; field
of view, 256 mm; matrix, 64 ? 64 (4-mm2, in-plane resolution);
slice thickness, 4 mm; number of slices, 30; number of volumes,
148]. A T1-weighted anatomical image using an MPRAGE
sequence was collected during each scan session for coregis-
tration with the functional images. Anatomical scans were
collected in the sagittal plane [field of view, 256 mm; matrix,
256 ? 256 (1-mm2, in-plane resolution); slice thickness, 1 mm;
number of slices, 180].
All fMRI analyses were conducted with the AFNI 2.56 (Anal-
ysis of Functional Neuroimages) statistical software package
(73). Motion correction and three-dimensional registration of
each participant’s functional images were performed with an
automated alignment program (3DVOLREG), which coregis-
tered each volume in the time series to the middle volume
acquired in that run using an iterative process. Brief periods of
motion in nine ASD subjects that were uncorrectable by
3DVOLREG were removed from analysis. The percent of the run
that was removed ranged from 2.7% to 14.9% (mean, 7.13%).
Images were then smoothed with a Gaussian filter (full-width
half-maximum, 6 mm).
Individual data were analyzed using the program 3DDECON-
VOLVE. This program first estimates an impulse response
function based on the measured fMRI signal data and input
stimulus functions. These nine functions included the three
word conditions (number, neutral, and emotional) and six
motion parameters derived from the output of 3DVOLREG
(intrarun motion in the x, y, and z and roll, pitch, and yaw
planes). The impulse response function was then convolved
with the input stimulus time series and multiple regressions
were run to determine a goodness-of-fit coefficient (or linear
contrast weight). The global mean and linear trend were
included in the regression to remove their effects from the
analysis. For each condition, the linear contrast weight was
determined for 0, 3, 6, and 9 s after stimulus presentation.
These weights were summed for each condition to give an
overall best fit for each condition compared with the rest
condition. Additionally, an a priori contrast of emotional vs.
neutral was determined for each individual.
Group Analysis. To average data across participants, the func-
tional images were then transformed into Talaraich space. t
tests were conducted to determine whether linear contrast
weights for the autism and control groups were significantly
different from zero for each of the four main comparisons of
interest (emotional vs. rest, number vs. rest, neutral vs. rest,
and emotional vs. neutral). An additional set of t tests was then
run to determine whether the contrast weights were signifi-
cantly different between the autism and control groups for the
four comparisons. All data were intensity-thresholded at P ?
0.01 and cluster-thresholded at a voxelwise ?-level of P ? 0.05.
The contrasts of any condition vs. rest reveal regions of
activation and deactivation and are displayed together in each
Exploratory Correlation Analysis. Because we predicted that func-
tional activity at rest is related to social behavioral impairment
in ASD, we ran a whole-brain regression analysis between the
number vs. rest contrast and the social subscale score on the
Autism Diagnostic Interview–Revised. Because results did not
Kennedy et al.
May 23, 2006 ?
vol. 103 ?
no. 21 ?
survive correction for multiple comparisons based on cluster
termed this analysis ‘‘exploratory.’’
Additional Analyses. Analyses of individual differences in deac-
tivation and activation patterns in ASD and control subjects
were carried out, and these methods are published as Sup-
We thank the subjects who graciously gave of their time to participate in
this study. This research was supported by National Institutes of Health
Grant R01 MH36840 (to E.C.).
1. Gusnard, D. A. & Raichle, M. E. (2001) Nat. Rev. Neurosci. 2, 685–694.
2. Maddock, R. J. (1999) Trends Neurosci. 22, 310–316.
3. Northoff, G. & Bermpohl, F. (2004) Trends Cognit. Sci. 8, 102–107.
4. Ochsner, K. N., Beer, J. S., Robertson, E. R., Cooper, J. C., Gabrieli, J. D.,
Kihsltrom, J. F. & D’Esposito, M. (2005) NeuroImage 28, 797–814.
5. Raichle, M. E., MacLeod, A. M., Snyder, A. Z., Powers, W. J., Gusnard, D. A.
& Shulman, G. L. (2001) Proc. Natl. Acad. Sci. USA 98, 676–682.
6. Gusnard, D. A., Akbudak, E., Shulman, G. L. & Raichle, M. E. (2001) Proc.
Natl. Acad. Sci. USA 98, 4259–4264.
7. Kjaer, T. W., Nowak, M. & Lou, H. C. (2002) NeuroImage 17, 1080–1086.
8. Mazoyer, B., Zago, L., Mellet, E., Bricogne, S., Etard, O., Houde, O., Crivello,
F., Joliot, M., Petit, L. & Tzourio-Mazoyer, N. (2001) Brain Res. Bull. 54,
9. Cavanna, A. E. & Trimble, M. R. (2006) Brain 129, 564–583.
J. Cognit. Neurosci. 15, 394–408.
11. Shulman, G. L., Fiez, J. A., Corbetta, M., Buckner, R. L., Miezin, F. M.,
Raichle, M. E. & Peterson, S. E. (1997) J. Cognit. Neurosci. 9, 648–663.
12. Menon, V., Leroux, J., White, C. D. & Reiss, A. L. (2004) Proc. Natl. Acad. Sci.
USA 101, 3615–3620.
13. Lustig, C., Snyder, A. Z., Bhakta, M., O’Brien, K. C., McAvoy, M., Raichle,
M. E., Morris, J. C. & Buckner, R. L. (2003) Proc. Natl. Acad. Sci. USA 100,
14. Greicius, M. D., Srivastava, G., Reiss, A. L. & Menon, V. (2004) Proc. Natl.
Acad. Sci. USA 101, 4637–4642.
15. Cato, M. A., Crosson, B., Gokcay, D., Soltysik, D., Wierenga, C., Gopinath, K.,
Himes, N., Belanger, H., Bauer, R. M., Fischler, I. S., et al. (2004) J. Cognit.
Neurosci. 16, 167–177.
16. Maddock, R. J., Garrett, A. S. & Buonocore, M. H. (2003) Hum. Brain Mapp.
17. Whalen, P. J., Bush, G., McNally, R. J., Wilhelm, S., McInerney, S. C., Jenike,
M. A. & Rauch, S. L. (1998) Biol. Psychiatry 44, 1219–1228.
18. Iacoboni, M., Lieberman, M. D., Knowlton, B. J., Molnar-Szakacs, I., Moritz,
M., Throop, C. J. & Fiske, A. P. (2004) NeuroImage 21, 1167–1173.
19. Fletcher, P. C., Happe, F., Frith, U., Baker, S. C., Dolan, R. J., Frackowiak,
R. S. & Frith, C. D. (1995) Cognition 57, 109–128.
20. Gallagher, H. L., Jack, A. I., Roepstorff, A. & Frith, C. D. (2002) NeuroImage
21. Gallagher, H. L. & Frith, C. D. (2003) Trends Cognit. Sci. 7, 77–83.
22. Vogeley, K., Bussfeld, P., Newen, A., Herrmann, S., Happe, F., Falkai, P.,
Maier, W., Shah, N. J., Fink, G. R. & Zilles, K. (2001) NeuroImage 14, 170–181.
23. Williams, J. H., Waiter, G. D., Perra, O., Perrett, D. I. & Whiten, A. (2005)
NeuroImage 25, 133–140.
24. Maddock, R. J., Garrett, A. S. & Buonocore, M. H. (2001) Neuroscience 104,
25. Pierce, K., Haist, F., Sedaghat, F. & Courchesne, E. (2004) Brain 127,
26. Courchesne, E., Redcay, E. & Kennedy, D. P. (2004) Curr. Opin. Neurol. 17,
27. Castelli, F., Frith, C., Happe, F. & Frith, U. (2002) Brain 125, 1839–1849.
28. Wang, A. T., Dapretto, M., Hariri, A. R., Sigman, M. & Bookheimer, S. Y.
(2004) J. Am. Acad. Child Adolesc. Psychiatry 43, 481–490.
29. Luna, B., Minshew, N. J., Garver, K. E., Lazar, N. A., Thulborn, K. R., Eddy,
W. F. & Sweeney, J. A. (2002) Neurology 59, 834–840.
30. Muller, R.A., Pierce, K., Ambrose, J. B., Allen, G. & Courchesne, E. (2001)
Biol. Psychiatry 49, 665–676.
31. Ring, H. A., Baron-Cohen, S., Wheelwright, S., Williams, S. C., Brammer, M.,
Andrew, C. & Bullmore, E. T. (1999) Brain 122, 1305–1315.
32. Eskes, G. A., Bryson, S. E. & McCormick, T. A. (1990) J. Autism Dev. Disord.
33. Ozonoff, S. & Jensen, J. (1999) J. Autism Dev. Disord. 29, 171–177.
34. Bush, G., Whalen, P. J., Rosen, B. R., Jenike, M. A., McInerney, S. C. & Rauch,
S. L. (1998) Hum. Brain Mapp. 6, 270–282.
35. Beversdorf, D. Q., Anderson, J. M., Manning, S. E., Anderson, S. L., Nordgren,
R. E., Felopulos, G. J., Nadeau, S. E., Heilman, K. M. & Bauman, M. L. (1998)
J. Neurol. Neurosurg. Psychiatry 65, 685–692.
36. Haznedar, M. M., Buchsbaum, M. S., Wei, T. C., Hof, P. R., Cartwright, C.,
Bienstock, C. A. & Hollander, E. (2000) Am. J. Psychiatry 157, 1994–2001.
37. Ohnishi, T., Matsuda, H., Hashimoto, T., Kunihiro, T., Nishikawa, M., Uema,
T. & Sasaki, M. (2000) Brain 123, 1838–1844.
38. D’Argembeau, A., Collette, F., Van der Linden, M., Laureys, S., Del Fiore, G.,
Degueldre, C., Luxen, A. & Salmon, E. (2005) NeuroImage 25, 616–624.
39. Fossati, P., Hevenor, S. J., Graham, S. J., Grady, C., Keightley, M. L., Craik,
F. & Mayberg, H. (2003) Am. J. Psychiatry 160, 1938–1945.
40. Johnson, S. C., Baxter, L. C., Wilder, L. S., Pipe, J. G., Heiserman, J. E. &
Prigatano, G. P. (2002) Brain 125, 1808–1814.
41. Kelley, W. M., Macrae, C. N., Wyland, C. L., Caglar, S., Inati, S. & Heatherton,
T. F. (2002) J. Cognit. Neurosci. 14, 785–794.
42. Lou, H. C., Luber, B., Crupain, M., Keenan, J. P., Nowak, M., Kjaer, T. W.,
Sackeim, H. A. & Lisanby, S. H. (2004) Proc. Natl. Acad. Sci. USA 101,
43. Macrae, C. N., Moran, J. M., Heatherton, T. F., Banfield, J. F. & Kelley, W. M.
(2004) Cereb. Cortex 14, 647–654.
44. Mitchell, J. P., Banaji, M. R. & Macrae, C. N. (2005) NeuroImage 28, 757–762.
45. Mitchell, J. P., Macrae, C. N. & Banaji, M. R. (2005) NeuroImage 26, 251–257.
46. Adolphs, R. (1999) Trends Cognit. Sci. 3, 469–479.
47. Vogeley, K., May, M., Ritzl, A., Falkai, P., Zilles, K. & Fink, G. R. (2004) J.
Cognit. Neurosci. 16, 817–827.
48. Critchley, H. D., Melmed, R. N., Featherstone, E., Mathias, C. J. & Dolan, R. J.
(2001) Brain 124, 1003–1012.
49. Nagai, Y., Critchley, H. D., Featherstone, E., Trimble, M. R. & Dolan, R. J.
(2004) NeuroImage 22, 243–251.
50. Binder, J. R., Frost, J. A., Hammeke, T. A., Bellgowan, P. S., Rao, S. M. & Cox,
R. W. (1999) J. Cognit. Neurosci. 11, 80–95.
51. Andreasen, N. C., O’Leary, D. S., Cizadlo, T., Arndt, S., Rezai, K., Watkins,
G. L., Ponto, L. L. & Hichwa, R. D. (1995) Am. J. Psychiatry 152, 1576–1585.
52. Greene, J. D., Sommerville, R. B., Nystrom, L. E., Darley, J. M. & Cohen, J. D.
(2001) Science 293, 2105–2108.
53. Paulus, M. P. & Frank, L. R. (2003) NeuroReport 14, 1311–1315.
54. Hurlburt, R. T., Happe, F. & Frith, U. (1994) Psychol. Med. 24, 385–395.
55. Frith, U. & Happe, F. (1999) Mind Lang. 14, 1–22.
56. Baron-Cohen, S., Ring, H. A., Wheelwright, S., Bullmore, E. T., Brammer,
M. J., Simmons, A. & Williams, S. C. (1999) Eur. J. Neurosci. 11, 1891–1898.
57. Gervais, H., Belin, P., Boddaert, N., Leboyer, M., Coez, A., Sfaello, I.,
Barthelemy, C., Brunelle, F., Samson, Y. & Zilbovicius, M. (2004) Nat.
Neurosci. 7, 801–802.
58. Pelphrey, K. A., Morris, J. P. & McCarthy, G. (2005) Brain 128, 1038–1048.
59. Carmichael, S. T. & Price, J. L. (1995) J. Comp. Neurol. 363, 615–641.
60. Kobayashi, Y. & Amaral, D. G. (2003) J. Comp. Neurol. 466, 48–79.
61. Ongur, D. & Price, J. L. (2000) Cereb. Cortex 10, 206–219.
62. Parvizi, J., Van Hoesen, G. W., Buckwalter, J. & Damasio, A. (2006)Proc. Natl.
Acad. Sci. USA 103, 1563–1568.
63. Barnea-Goraly, N., Kwon, H., Menon, V., Eliez, S., Lotspeich, L. & Reiss, A. L.
(2004) Biol. Psychiatry 55, 323–326.
64. Horwitz, B., Rumsey, J. M., Grady, C. L. & Rapoport, S. I. (1988) Arch. Neurol.
65. Just, M. A., Cherkassky, V. L., Keller, T. A. & Minshew, N. J. (2004) Brain 127,
66. Koshino, H., Carpenter, P. A., Minshew, N. J., Cherkassky, V. L., Keller, T. A.
& Just, M. A. (2005) NeuroImage 24, 810–821.
67. Fox, M. D., Snyder, A. Z., Vincent, J. L., Corbetta, M., Van Essen, D. C. &
Raichle, M. E. (2005) Proc. Natl. Acad. Sci. USA 102, 9673–9678.
Sci. USA 100, 253–258.
69. Bartels, A. & Zeki, S. (2005) NeuroImage 24, 339–349.
70. Lord, C., Rutter, M. & Le Couteur, A. (1994) J. Autism Dev. Disord. 24,
71. Lord, C., Risi, S., Lambrecht, L., Cook, E., Jr., Leventhal, B., DiLavore, P.,
Pickles, A. & Rutter, M. (2000) J. Autism Dev. Disord. 30, 205–223.
72. Stroop, J. R. (1935) J. Exp. Psychol. 18, 643–662.
73. Cox, R. W. (1996) Comput. Biomed. Res. 29, 162–173.
www.pnas.org?cgi?doi?10.1073?pnas.0600674103Kennedy et al.