The Functional Neuroanatomy of Spatial
Attention in Autism Spectrum Disorder
Department of Psychiatry
University of California, San Diego
Maha Adamo and Marissa Westerfield
Department of Neurosciences
University of California, San Diego
Children’s Hospital Research Center
La Jolla, CA
Department of Neurosciences
University of California, San Diego
Department of Neurosciences
University of California, San Diego
This studyinvestigated the functional neuroanatomical correlates of spatial attention
impairments in autism spectrum disorders (ASD) using an event-related functional
magnetic resonance imaging (FMRI) design. Eight ASD participants and 8 normal
nation following a spatial cue that preceded target presentation by 100 msec (short
interstimulus interval [ISI]) or 800 msec (long ISI). The ASD group showed signifi-
sults showed a reduction in activity within frontal, parietal, and occipital regions in
ASD relative to the NC group, most notably within the inferior parietal lobule. ASD
ative to the NC group. ASD FMRI activity in the long ISI condition suggested that
DEVELOPMENTAL NEUROPSYCHOLOGY, 27(3), 425–458
Copyright © 2005, Lawrence Erlbaum Associates, Inc.
Requests for reprints should be sent to Frank Haist, Department of Psychiatry, University of Cali-
fornia, San Diego, 9500 Gilman Drive, La Jolla, CA 92093–0959. E-mail: firstname.lastname@example.org
the rudimentary framework of normal attention networks were engaged in ASD in-
cluding bilateral activation within the frontal, parietal, and occipital lobes. Notable
tex. No reliable activation was observed in the posterior cerebellar vermis in ASD
participants during either long or short ISI conditions. In addition, no frontal activa-
tion during short ISI and severely reduced frontal activation during long ISI was ob-
served in the ASD group. Taken together, these findings suggest a dysfunctional
cerebello-frontal spatial attention system in ASD. The pattern of findings suggests
that ASD is associated with a profound deficit in automatic spatial attention abilities
and abnormal voluntary spatial attention abilities. This article also describes a
method for reducing the contribution of physical eye movements to the blood-
oxygenation level dependent activity in studies of ASD.
Autism spectrum disorders (ASD), including autism and Asperger’s syndrome,
are pervasive developmental disorders defined behaviorally by deficits in com-
munication and reciprocal social interaction, restricted interests, and stereotyped
repetitive behaviors (American Psychiatric Association, 1994). In addition, ASD
are characterized by deficits in various cognitive functions. A characteristic
symptom of ASD, noted by both Kanner (1943) and Asperger (1991) in their
initial descriptions of autism, is aberrant attention. The attentional dysfunction
can be manifested in numerous ways (for review, see Allen & Courchesne,
2001). These include the failure to respond to auditory or visual stimulation, an
excessively intense focus on a single piece of information while the context is
ignored (Lovaas, Koegel, & Schreibman, 1979; Lovaas, Schreibman, Koegel, &
Rehm, 1971; Townsend & Courchesne, 1994), difficulty adjusting the spatial
scope of attention (Burack, 1994), and slowed shifting of attention between sen-
sory modalities (Casey, Gordon, Mannheim, & Rumsey, 1993; Courchesne et al.,
1994). Results from several studies have suggested a particular difficulty with
spatial attention function including difficulty disengaging attention from a spa-
tial focus (Wainwright & Bryson, 1996), slowed shifting among spatial locations
(Harris, Courchesne, Townsend, Carper, & Lord, 1999; Townsend et al., 1999;
Townsend, Harris, & Courchesne, 1996), and abnormal distributions of spatial
attention (Townsend & Courchesne, 1994).
In normal function, spatial attention can be oriented rapidly and automatically.
The distribution of attentional resources in space is also subject to control by
slower, more effortful, voluntary attentional systems. Focus of attentional re-
sources to a spatial location enhances perceptual processing of information at that
location, resulting in faster and more accurate response (e.g., Hillyard, Vogel, &
Luck, 1998). A wealth of imaging, lesion, and single-unit studies has contributed
to understanding the anatomy of spatial attention systems. Taken together, these
frontal cortex (especially the frontal eye fields [FEF]), posterior parietal cortex,
HAIST ET AL.
occipital cortex, the cingulate gyrus, thalamus, superior colliculus, and striatum
(Corbetta, 1998; Corbetta et al., 1998; Coull & Nobre, 1998; Kastner & Unger-
leider, 2000; Martinez et al., 2001; Mesulam, 1981, 1999). Animal models and
functional magnetic resonance imaging (FMRI) studies have suggested that the
cerebellum is also involved in spatial attention networks (Coull & Nobre, 1998;
Crispino & Bullock, 1984; Goffart & Pélisson, 1994, 1997).
In autism, there are brain structural and functional abnormalities in regions
known to serve spatial attention. Postmortem and in vivo studies have consis-
tently reported abnormalities of the cerebellar vermis and hemispheres (Bailey et
al., 1998; Bauman & Kemper, 1994; Fehlow & Bernstein, 1993; Kemper &
Bauman, 1998; Ritvo & Freeman, 1986; Williams & Hauser, 1980; for review,
see Courchesne, 1999). The majority of quantitative MRI studies have also
found abnormally reduced size of cerebellar vermis or hemispheres (Courchesne
et al., 2001). Parietal volume loss in adults (Courchesne, Press, &
Yeung-Courchesne, 1993) has also been reported as has reduced thickness of the
posterior corpus callosum where parietal fibers pass (Egaas, Courchesne, &
Saitoh, 1995; Manes et al., 1999; Piven, Bailey, Ranson, & Arndt, 1997). Carper
and Courchesne (2000) reported an association between the size of the posterior
cerebellar vermis and abnormal enlargement of frontal lobes in young autistic
children. In the Carper and Courchesne study, children with the greatest cerebel-
lar hypoplasia had the greatest overgrowth of frontal cortex. This suggested that
autism is associated with growth dysregulation characterized by brain over-
growth during the first 5 postnatal years, particularly in frontal regions, followed
by abnormally slowed brain growth (Carper & Courchesne, 2000; Carper, Mo-
ses, Tigue, & Courchesne, 2002; Courchesne, 2002; Courchesne et al., 2001).
Abnormal growth patterns in turn suggest associated abnormal connectivity be-
tween the cerebellum and frontal cortex during critical periods of development
(Akshoomoff, Pierce, & Courchesne, 2002).
Imaging studies have shown abnormal patterns of brain electrical signal and
ing that suggest abnormalities of spatial attention networks in autism that include
impaired interactions among the cerebellum, frontal cortex, and parietal cortex
(Belmonte & Yurgelun-Todd, 2003b; Townsend et al., 2001). In normally devel-
oped adults, attention can be oriented to a location in space in less than 100 msec.
A series of tests in which high-functioning adults with autism were cued to a spa-
cation suggest that these individuals may require more than a second to execute
spatial attention shifts (Harris et al., 1999; Townsend et al., 1999; Townsend,
Courchesne, & Egaas, 1996). In these studies, attention orienting effects were in-
dependent of motor response and could not be explained by the abnormalities of
eye movements, which are prominent in autism. This bottleneck in resource allo-
cation severely compromises processing of new information.
larly slowed orienting in patients with acquired damage to the cerebellum and a
significant correlation between orienting deficits and cerebellar vermal volume
measures in normal controls and patients (Townsend et al., 1999). Electro-
physiological studies of spatial attention in autism and cerebellar stroke patients
lar vermal lobules and the onset of a potential recorded over frontal cortex that in-
attention orienting network that is disrupted in autism.
Although FMRI studies have detailed the brain networks that contribute to
spatial attention, to date, none have specifically addressed differences in
early automatic (bottom-up) attention and later voluntary (top-down) pro-
cesses. The single FMRI study (Belmonte & Yurgelun-Todd, 2003b) of spa-
tial attention in autism published to date was also not designed to examine
this difference. Evidence from a number of studies shows that although ASD
participants show little or no use of spatial attention during the early auto-
matic stages, over time, their attentional and behavioral responses are more
normal. This suggests that spatial attention deficits in autism may affect
largely the rapid early automatic attentional processes, whereas the later vol-
untary processes may be intact. The results we report here are from an
event-related FMRI study designed to compare early automatic and later vol-
untary spatial attention networks in ASD.
Eight male participants with an ASD were tested. Six
individuals were diagnosed with autistic disorder because they met criteria for
autism on the Autism Diagnostic Interview–Revised (ADI–R; Lord, Rutter, &
Le Couteur, 1994) and the Autism Diagnostic Observation Schedule (ADOS;
Lord, Rutter, DiLavore, & Risi, 2001) and the Diagnostic and Statistical Manual
of Mental Disorders (4th ed. [DSM–IV]; American Psychiatric Association,
1994) criteria for autistic disorder. Two individuals were diagnosed with
Asperger’s disorder because they met criteria for autism on the ADI–R and au-
tism or ASD on the ADOS and met DSM–IV criteria for Asperger’s disorder.
None of the participants had a history of seizures or additional psychiatric or
neurologic disorders. All ASD participants were nonretarded, obtaining Full
Scale IQ scores of 80 or higher as determined from IQ testing using the Wechs-
ler Adult Intelligence Scale–Revised (WAIS–R; Wechsler, 1981), WAIS–III
HAIST ET AL.
(Wechsler, 1997), or Wechsler Abbreviated Scale of Intelligence (WASI; Wechs-
ler, 1999). Five of the ASD participants were unmedicated at the time of testing.
One participant was being treated with Paroxetine (Paxil®) and lithium, one par-
ticipant with dextroamphetamine/amphetamine (Adderall®) and fluoxetine
(Prozac®), and one with sertraline HCl (Zoloft®). Table 1 shows data for the in-
dividual ASD participants.
Normal comparison (NC) participants.
Eight healthy male volunteers
with no reported history of major medical illness, neurological or psychiatric dis-
order, head trauma, or substance abuse were tested. Due to time constraints, IQ
testing using the WASI was conducted in four randomlyselected participants. The
NC participants ranged in age from 19 to 32 years (see Table 1).
versity of California, San Diego approved this study. Participants provided in-
formed consent prior to participation in the experiment.
Characteristics of the Participants
Group Diagnosis Age VIQPIQFSIQ SocVerbBehav
ASD Aut4386 11510022 196
ASD Aut38 98 114104 21 22 10
ASDAut21 105101 10323 186
ASDAut18 86 109 9827 2212
ASD Aut 16 7689 8027 197
ASD Aut14108 10810826 196
ASDAsp 19 1119910621 207
ASD Asp14 102 117109 18 116
M 23.496.5 106.5101.0 23.125 18.75 7.5
SD11.4 12.59.6 9.33.3 3.4 2.3
M 25.6107.2 115.2 112.2
SD3.8 11.5 10.78.8
Note. NC mean IQ estimates based on scores from 4 of the 8 participants. Separate independent
sample t tests showed no significant differences between the two groups for age, t(14) = 0.67, p > .10;
VIQ, t(10) = 1.44, p > .10; and PIQ, t(10) = 1.44, p > .10; although the difference betwwen groups for
FSIQ approached significance, t(10) = 2.01, p < .10. ADI–R = Autism Diagnostic Inventory–Revised;
VIQ = Verbal IQ; PIQ = Performance IQ; FSIQ = Full Scale IQ; Soc = social communication subscale
(minimum cutoff = 10); Verb = verbal communication subscale (minimum cutoff = 8); Behav = Re-
strictedBehaviorsubscale(minimum cutoff=3);ASD=autism spectrum disorders;Aut=autism;Asp
= Asperger’s syndrome; NC = normal comparison group.
Design and Procedure
Each experiment session consisted of behavioral training
tive test runs during functional magnetic resonance imaging (FMRI) data collec-
tion. Each of the three experimental runs consisted of 132 trials, each 2,800 msec
in duration, which included 64 active and 68 “null” trials. A graphical depiction of
the active trials is shown in Figure 1. All stimuli were white presented against a
HAIST ET AL.
spatial attention task. On each trial, the lines of one of the two flanking boxes thickened, which
produced the visual effect of the box brightening. Following either a short (100 msec) or long
(800 msec) interstimulus intervals (ISI), a block letter “E” target appeared either in the box that
brightened (valid trials) or in the box that remained unchanged (invalid trials). The orientation
of the letter E was counterbalanced across trials. A pattern mask followed 50 msec after target
presentation. Valid cues appeared on 75% of trials.
black background. Throughout each experiment run, a box to the left and to the
right of a central fixation marker was shown on the screen (i.e., static display). At
the beginning of an active trial, the lines of one of the two boxes thickened to cue
or down). Two cue-to-target interstimulus intervals (ISI) were used. The short ISI
condition used a delaybetween the onset of the cue and the presentation of the tar-
mained on the screen for 50 msec and was then replaced by a pattern mask (see
Figure 1). Of the 64 active trials, the cue was presented on the same side as the tar-
get stimulus 75% of the time (48 trials) and was on the opposite side of the target
stimulus 25% of the time (16 trials). These are hereafter referred to as “valid” and
“invalid” cue trials, respectively. During null trials, there was no change in the
static display. Each run began and ended with two null stimuli. Otherwise, the or-
der of active and null trials, the side of target display, and the orientation of the tar-
get were counterbalanced within each experiment run. The different pseudo-
random order of active and null trials was used in each of the three runs. The three
runs were presented in the same order for all participants. Each run lasted 6 min
and 9.6 sec.
and to covertly move their attention to the target when it was presented. When the
target was presented, they were instructed to indicate which way the letter E was
facing as quickly and accurately as possible via a custom-made, three-button de-
vice held in their right hand. The open side of the letter E defined the direction.
They pressed one of three buttons if the letter E was facing left, right, or up. They
were instructed to make no button response if the E faced down. Participants were
told that one of the boxes would change prior to target presentation and that often,
but not always, the target would appear in that box; and thus, the best strategy was
to maintain fixation in the center of the screen to allow them to quicklyshift atten-
tion to the target. Participants were given thorough instructions prior to entering
the MR scanner and received a brief example experimental run to insure compre-
hension of the instructions. Once in the MR scanner, participants were given one
full-length practice experimental run during the acquisition of the structural MRI
scan. The presentation of the three experimental runs began immediately after the
MRI Data Acquisition
Participants lay supine within the MR scanner during data collection. Their heads
port to reduce motion. A PC-compatible laptop computer using Presentation soft-
ware (Woods, 2003) controlled stimulus presentation and behavioral response ac-
quisition. Stimuli were back-projected onto a screen located at the foot of the
ipants viewed the stimuli using a 90° mirror attached to the head coil above their
eyes and indicated their responses using a custom-designed, three-button mouse
Imaging data were acquired at the Thornton Hospital at the University of Cali-
fornia, San Diego using a 1.5 T Siemens Symphony MR scanner (Erlangen, Ger-
many) equipped with the standard clinical head coil. During each of the three test
runs, 132 whole-brain T2*-weighted axial images were acquired using a sin-
mm slab; TR = 2,800 msec; TE = 34 msec; flip angle = 90°; field of view [FOV] =
256 mm; matrix = 64 × 64; in-plane resolution = 4 mm2). A high-resolution
three-dimensional (3D) magnetization prepared-rapid gradient echo (MP-RAGE)
structural scan was acquired for anatomical localization (TR = 11.08 msec; TE =
4.3 msec; flip angle = 7°; FOV = 256 mm; matrix 256 × 256; 180 sagittal slices;
resolution = 1 mm3).
FMRI data analysis.
FMRI analyses were conducted using the Analysis of
Functional Neuroimagespackage (AFNI;Version2.5;see
http://afni.nimh.nih.gov/afni; Cox & Hyde, 1997). Motion correction and 3D reg-
istration were performed using 3dvolreg, an automated alignment program that
coregistered each volume in the time series to the fourth volume acquired in that
series (Cox & Jesmanowicz, 1999). All of the volumes in the three test runs were
then registered to the fourth volume of the first run. The three runs were concate-
nated into a single time-series file of 396 volumes and smoothed with a Gaussian
filter (FWHM = 8 mm). The first two acquisitions from each of the runs were cen-
sored from the analyses to compensate for T1 equilibration effects.
fMRI analyses of individual participants.
The fMRI data from individual
participants were analyzed using a deconvolution approach (3dDeconvolve). A
Dirac delta impulse function was used to derive the impulse response function
ing the sum of scaled and time-delayed versions of the stimulus time series. Six
time series were used that coded stimuli in each condition according to behavioral
response. These included
1. Correct responses to short ISI (100 msec) valid cue targets.
2. Correct responses to long ISI (800 msec) valid cue targets.
3. Incorrect responses to short and long ISI valid targets.
4. Correct responses to the combined short and long ISI invalid cue targets.
HAIST ET AL.
6. All down facing target trials.
placement motion effects from the activation maps for motion within each run (6
ing the AFNI hand land-marking procedure (resampled volumes = 3 mm3).
FMRI group analyses.
Three separate one-way analyses of variance
(ANOVAs) were conducted using the positive Z scores obtained from the
deconvolution analyses as the dependent variable and group as the between-sub-
jects factor. The three ANOVAs compared the BOLD activity of the ASD and NC
groups during the short ISI condition, the long ISI condition, and for eye move-
ment measured throughout the experiment. Corrections for multiple comparisons
were established using a voxel-cluster threshold technique (Forman et al., 1995)
for an overall corrected alpha level of 0.05. For the comparisons of mean activity
for each group to the baseline (null trial) condition, a voxelwise threshold of p <
.001, t(14) ≥ 4.14 was used, which required a minimum cluster of 24 contiguous
voxels (648 µl) (full-width, half-maximum [FWHM] autocorrelation estimate = 8
the ASD and NC groups, a voxelwise threshold of p < .001, t(14) ≥ 4.14) with a
minimum cluster of 6 contiguous voxels (162 µl) was used.
Functional conjunction analysis figures (FCA).
The results from each of
cluded regions where task related activity was significantly greater than the base-
line condition for the ASD and NC groups, regions where the activityfor the ASD
group was significantly greater than that observed in the NC group, and regions
where the activity of the NC group was significantly greater than that observed in
the ASD group. The regions identified in the four maps were converted to binary
mask images, separately coded using a power of two coding system (e.g., 1, 2, 4,
and 8), and summed together. The values in the resulting single activation map,
termed a functional conjunction analysis map, specified the separate effects con-
tributing to activity observed within each voxel. These separate effects were color
coded to produce the maps displayed in Figure 2 (see insert).
Behavioral data analysis.
Behavioral data for accuracy and reaction time
were analyzed using separate 2 × 2 repeated measures ANOVA with group (ASD
vs. NC) and ISI (100 msec vs. 800 msec) as factors. Accuracy was defined as the
percent correct target detection excluding down-facing stimuli for which no re-
sponse was made. The analyses were conducted with SPSS Version 11.5. Post hoc
level of 0.05 was used for all tests.
between the participants with a diagnosis of autism and those with a diagnosis of
Asperger’s syndrome. Specifically, the mean accuracyand reaction time measures
across all of the ISI and validity conditions (e.g., short ISI valid cue targets, long
ISI invalid cue targets) for the 2 participants with Asperger’s syndrome fell within
the 95% confidence intervals established from the performance of the 6 partici-
pants with an autism diagnosis. Thus, the results from the participants with
Asperger’s disorder and those with autism are presented as a single group of ASD
The mean accuracyand reaction time results for the ASD and NC groups at the
short ISI (100 msec) and long ISI (800 msec) conditions are shown in Figure 3.
The analysis of the accuracy data showed a reliable Group × Cue interaction, F(1,
in their judgments for valid cue targets than for invalid cue targets (M = 82.1% vs.
70.2%, standard error of measurement [SEM] = 7.3 vs. 6.9, respectively), t(7) =
3.92, p < .025, whereas the ASD group performed similarly for valid and invalid
cuetargets(M =56.5% vs.57.0%, SEM =7.6vs.6.9,respectively), t(7)=0.13, ns.
The Group × Cue × ISI three-way interaction approached significance, F(1, 14) =
3.98, p < .07. This suggested a trend for the NC group to show a greater accuracy
for target judgments following valid cues compared to invalid cues in the short ISI
group were similarly accurate for valid and invalid cue targets in the long ISI con-
HAIST ET AL.
dition (M = 83.8% vs. 78.2%, SEM = 8.0 vs. 6.8, respectively). The ASD group
shortISI(M=52.6%vs.53.2%, SEM =7.4vs.7.6,respectively)andlongISIcon-
ditions (M = 60.4% vs. 60.7%, SEM = 8.0 vs. 6.8, respectively). The ISI main ef-
ISI condition than in the short ISI condition (M = 70.8% vs. 62.1%, SEM = 5.0 vs.
5.2, respectively), F(1, 14) = 22.29, p < .001. Accuracy averaged over ISI and cue
conditions, the group main effect, showed that the NC group performed overall
only marginally better than the ASD group (M = 76.1% vs. 56.7%, SEM = 7.1 vs.
7.1, respectively), F(1, 14) = 3.76, p < .08.
Speed of response was not emphasized in the instructions to participants as a
critical requirement of performance, although they were encouraged to respond
promptly to insure that they would be prepared for subsequent trials. Thus, there
the marginal main effect of group suggested a trend for the NC group to respond
more quicklyoverall than the ASD group (M = 714.8 msec vs. 857.3 msec, SEM =
or interactions were significant, ps > .10.
dicted the target location or predicted the incorrect location. In contrast, the NC
Behavioral results from the spatial attention task for the autism spectrum disor-
ders (ASD; closed circles) and normal control (NC; open circles) groups. The left panel shows
the accuracy scores, as measure by percent correct target identification, from the short (100
msec) and long (800 msec) cue-to-target interstimulus interval (ISI) conditions following a
valid or invalid cue. The right panel shows the reaction time to make correct target identifica-
tions. Error bars indicate standard deviations.
cation relative to an invalid spatial cue. This difference in accuracy between valid
and invalid cues in the NC group was greatest in the short ISI condition.
Short ISI condition (100 msec).
Figure 2A displays selected brain regions
that showed significant BOLD activity in response to correct target detection fol-
termined from the cluster threshold corrected ANOVA. The figure is color coded
cantly greater than the baseline condition (null trials) for both the ASD and NC
groups, was significantly above baseline for either the ASD or the NC group, and
group. There were no regions observed where the ASD group showed greater
in Table 2.
Consistent with their overall superior accuracy in target detection and related
advantage for target detection at the validly cued location, BOLD activity in the
NC group was substantially more extensive than that observed in the ASD group.
Overall, the NC group showed significant BOLD activity within a volume of ap-
proximately 239.5 cl of the brain versus 3.5 cl for the ASD group. Pairwise com-
parisons revealed that NC BOLD activation significantly exceeded that observed
in the ASD group in five regions within the frontal lobe, one in the left hemisphere
parietal lobe, and the right hemisphere insula. NC frontal lobe activation was sig-
nificantly greater than ASD activity in left and right hemisphere precentral gyrus
= 8, z = 26), t(14) = 4.15, p < .001; right (x = –52, y = 5, z = 27), t(14) = 4.16, p <
.001; and right hemisphere BA 10 within the middle frontal gyrus (x = –29, y = 43,
z = 11), t(14) = 4.25, p < .001. Although both groups produced activation within
the left hemisphere inferior parietal lobule (BA 40), the activity for the NC group
was greater than ASD activity in the inferior aspects of this region near the border
with the supramarginal gyrus (x = 36, y = –40, z = 40), t(14) = 5.62, p < .001. NC
activation was also greater than ASD activity within the right hemisphere insula
(BA 13; x = –25, y = 16, z = 8), t(14) = 5.05, p < .001. Although not significantly
greater than the activation shown in the ASD group, the NC group showed signifi-
cant BOLD activation relative to the baseline condition in numerous regions that
sulcus encompassing the superior and inferior parietal lobules (BA 7 and 40, re-
spectively), precuneus (BA 7), the frontal and supplementary eye fields and sup-
HAIST ET AL.
FMRI BOLD Activation to Correct Responses From the Short ISI (100 msec)
Valid Cue Condition
LocationHemisphereBAxyz Z Score
ASD and NC > baseline
Postcentral gyrusL2 48–30 542.67
Inferior parietal lobuleLa
40 40 –3645 3.01
Supramarginal gyrusL 40 39 –45 372.28
Fusiform gyrusL 1935–71–20 1.38
Cerebellar hemisphere VIL 35–71–23 1.62
NC only > baseline
6 45 –2293.05
6 –412 283.06
L4 38–1960 4.13
Middle frontal gyrusL6 29–10 473.38
R6 –31–3 47 2.50
L9 3926 26 1.80
Middle frontal gyrusL 46 453025 1.79
10–32 4611 2.30
Superior frontal gyrusR 10 –32 52 12 1.66
Inferior frontal gyrusLa
9388 27 1.64
9 –52–5 272.11
Medial frontal gyrusL6416 432.28
R6 –4 15 432.18
Cingulate gyrusR 24/32–9 1132 1.49
L24/32 1414 30 0.89
Postcentral gyrusL1 42 –29603.90
R2 –51–23 371.77
Inferior parietal lobuleR 40–50 –4229 2.27
Superior parietal lobuleL7 24–66 47 2.58
PrecuneusL7 19 –71 452.49
R7 –9 –62452.26
Supramarginal gyrusR40 –50 –46371.77
Superior temporal gyrusL 2256–47 14 1.78
L 21 59–47 101.75
R 21 –51–41 101.76
Middle occipital gyrusL 19 35 –856 1.20
L 37 44 –65–92.46
Fusiform gyrusR 37 –38 –59–11 2.30
R19 –32 –66 –101.55
InsulaL 30 12 102.02
R–13 –310 2.04
plementary motor area of the middle and medial frontal gyri (BA 6), the
dorsolateral prefrontal cortex (BA 9 and 10 and left hemisphere BA 46), and
extrastriate cortex (BA 19 and 37). Many of these activations were observed bilat-
erally, consistent with the bilateral presentation of target stimuli (see Table 2).
bly, the anterior and posterior vermis of the cerebellum (lobules IV, V, and VII).
Significant BOLD activity relative to the baseline condition for the ASD group
was observed in only six regions, all located in the posterior left hemisphere. Five
group (see Table 2) and included the postcentral gyrus, the inferior parietal lobule
(BA 40), supramarginal gyrus (BA 40), fusiform gyrus (BA 19), and lateral cere-
bellar hemisphere (lobule VI). ASD activation was significantly greater than the
baseline condition in the posterior fusiform gyrus (BA 19), a region not observed
as active in the NC group. Notably lacking was ASD group activation within any
frontal regions or the cerebellar vermis.
Long ISI condition (800 msec).
Figure 2B displays selected brain regions
that showed significant BOLD activity in response to correct target detection fol-
HAIST ET AL.
TABLE 2 (Continued)
LocationHemisphere BAxyz Z Score
CaudateR –16 17 101.44
PutamenL 21 –5 101.35
Cerebellar anterior vermis lobules IV
–2 –52 –161.86
Cerebellar posterior vermis lobules VII
Cerebellar hemisphere lobules IV to VIL 40 –59–28 2.15
Cerebellar hemisphere lobules III to VIR –25 –51–25 3.15
ASD only > baseline
Fusiform gyrusL 1931 –80–16 1.45
Note. No regions were found where ASD activation significantly exceeded that observed in the NC
group. The description of NC only > baseline and ASD only > baseline includes regions that are distinctly ac-
tive rather than regions surrounding commonly activated areas. The x, y, and z coordinates of Talairach space
are defined with positive indicating the millimeters to the left, anterior, and superior of the anterior
commissure, respectively. The Z scores reported for the ASD and NC > baseline are the mean of the ASD and
NC Z scores at the voxel location indicated. All region labels and BAs were determined using the Analysis of
Functional Neuroimages (AFNI) package implimentation of the Talairach daemon (Lancaster et al., 2000).
Cerebellar nomenclature according to Schmahmann, Doyon, Toga, Petrides, and Evans (2000). FMRI = func-
tional magnetic resonance imaging; BOLD = blood-oxygenation level dependent; ISI = interstimulus interval;
BA=Broadmann’sarea;ASD=autism spectrum disorder;NC=normalcomparison;L=lefthemisphere;R=
aNC > ASD in pairwise comparison (p < .001).
lowing valid cues in the long ISI condition (800 msec cue-to-target interval) as de- Download full-text
BOLD activation effects is provided in Table 3.
BOLD activitythan observed in the ASD group (NC = 243.3 cl vs. ASD = 46.4 cl,
respectively). Whereas the overall activity in the NC group was similar to that ob-
served in the short ISI condition in both extent and spatial distribution, the overall
vation. This had the effect of reducing the number of regions where NC activity
tion onlyin the right middle frontal gyrus in BA 6 (x = –45, y = 2, z = 407), t(14) =
5.36, p < .001 and the left precentral gyrus in BA 6 (x = 58, y = –1, z = 31), t(14) =
5.48, p < .001, both areas within the general region of the FEF. More regions were
sphere inferior parietal lobule (BA 40), supramarginal gyrus (BA 40), fusiform
gyrus (BA 19), and cerebellar hemisphere (lobule VI). In addition, in the long ISI
condition, both groups produced activation within the left hemisphere middle and
bilateral middle occipital gyrus (BA 31), and bilateral fusiform gyrus (BA 37). In
addition, both groups produced subcortical activation in the thalamus bilaterally
and the right hemisphere putamen.
As can be seen in Figure 2A and 2B, the overall increase in the activation ob-
the short ISI condition rather than the recruitment of substantially different brain
regions in the long ISI condition. Notable differences between the two ISI condi-
tions for the NC group included the recruitment of bilateral areas in the superior
occipital and inferoposterior parietal region including the middle occipital gyrus
and cuneus (BA 19), enhanced right hemisphere extrastriate cortex activation in-
cluding the fusiform gyrus (BA 37), and enhanced activity in the right cerebellar
hemisphere. In contrast, activityin the left hemisphere dorsolateral prefrontal cor-
was reduced relative to the short ISI condition. In summary, the NC group showed
tivation from the short to the long ISI condition.
For the ASD group, several regions were activated relative to the baseline con-
dition in the long ISI condition that were not observed in the short ISI condition.
These included the right medial frontal gyrus (BA 6) in the region of the supple-
bilateral intraparietal sulcus regions including the superior parietal lobule and
precuneus (BA 7), bilateral middle occipital gyrus (BA 19), bilateral extrastriate