Differential Patterns of Striatal Activation in Young
Children with and without ADHD
Sarah Durston, Nim T. Tottenham, Kathleen M. Thomas, Matthew C. Davidson,
Inge-Marie Eigsti, Yihong Yang, Aziz M. Ulug, and B.J. Casey
Background: Cognitive control, defined as the ability to
suppress inappropriate thoughts and actions, is compro-
mised in attention-deficit/hyperactivity disorder (ADHD).
This study examines the neural basis of this deficit.
Methods: We used a paradigm that incorporates a para-
metric manipulation within a go/nogo task, so that the
number of go trials preceding a nogo trial is varied to tax
the neural systems underlying cognitive control with
increasing levels of interference.
Results: Using this paradigm in combination with event-
related functional magnetic resonance imaging (fMRI), we
show that children without ADHD have increased suscep-
tibility to interference with increasing numbers of go trials
preceding a nogo trial, but children with ADHD have
difficulty even with a single go trial preceding a nogo trial.
In addition, children with ADHD do not activate frontos-
triatal regions in the same manner as normally developing
children, but rather rely on a more diffuse network of
regions, including more posterior and dorsolateral pre-
frontal regions.
Conclusions: Normal immature cognition may be char-
acterized as being susceptible to interference and sup-
ported by the maturation of frontostriatal circuitry. ADHD
children show a slightly different cognitive profile at 6 to
10 years of age that is paralleled by a relative lack of or
delay in the maturation of ventral frontostriatal circuitry.
Biol Psychiatry 2003;53:871–878 © 2003 Society of Bi-
ological Psychiatry
Key Words: ADHD, development, striatum, response
inhibition, event-related fMRI, parametric
Introduction
Attention-deficit/hyperactivity disorder (ADHD) is acommon and impairing neuropsychiatric disorder
with onset in childhood. It is thought to affect 3% to 5%
of all school-aged children and is characterized by age-
inappropriate symptoms of hyperactivity, inattentiveness,
and impulsivity (Buitelaar 2002; Kaplan et al 1994).
Converging evidence implies the involvement of dopami-
nergic frontostriatal circuitry in ADHD. Anatomical im-
aging studies using magnetic resonance imaging (MRI)
have demonstrated subtle reductions in volume in regions
of the basal ganglia and prefrontal cortex (e.g., Castellanos
et al 1996, 2001; Filipek et al 1997), whereas functional
studies have suggested that these regions may be hypoper-
fused (e.g., Lou et al 1984). Cognitive functioning is
mildly impaired in this disorder (for review, see Sergeant
et al [2002]). In particular, cognitive control, the ability to
inhibit inappropriate thoughts and actions, is affected.
Several studies have shown that this impairment is related
to the reduction in volume in frontostriatal regions (Casey
et al 1997a; Semrud-Clikeman et al 2000), and functional
studies have suggested that older children and adults with
ADHD may activate these regions less than controls
during tasks that require cognitive control (e.g., Bush et al
1999; Rubia et al 1999; Vaidya et al 1999).
We recently showed that development of this ability is
related to the maturation of ventral frontostriatal circuitry
in a sample of normally developing children relative to
adults (Durston et al 2002b). We manipulated task diffi-
culty within a go/nogo paradigm by parametrically vary-
ing the number of go trials preceding a nogo trial (Durston
et al 2002a). Such manipulations allow for comparisons
between groups on trials of similar performance. More
importantly, this manipulation allows one to test the extent
to which immature cognition is characterized by suscep-
tibility to interference by varying the salience of the
interfering information. Children demonstrated an in-
creased susceptibility to interference compared to adults,
as they made more errors overall. In addition, they
displayed an increase in the number of errors they made to
nogo trials as a function of the number of preceding go
From the Sackler Institute for Developmental Psychobiology (SD, NTT, KMT,
MCD, I-ME, AMU, BJC), Weill Medical College of Cornell University, New
York, New York, Department of Child and Adolescent Psychiatry (SD),
University Medical Center Utrecht, Utrecht, The Netherlands; Sackler Institute
for Developmental Psychobiology, Columbia University, New York, New
York (I-ME); and Department of Psychiatry (YY) and Department of Radiol-
ogy (AMU), Weill Medical College of Cornell University, New York, New
York.
Address reprint requests to Sarah Durston, Sackler Institute for Developmental
Psychobiology, Weill Medical College of Cornell University, 1300 York
Avenue, Box 140, New York NY 10021.
Received August 16, 2002; revised November 14, 2002; accepted November 19, 2002.
© 2003 Society of Biological Psychiatry 0006-3223/03/$30.00
doi:10.1016/S0006-3223(02)01904-2
trials as adults did, demonstrating that their ability to
inhibit an action was sensitive to the preceding context;
however, whereas in adults activation in ventral prefrontal
regions increased with increasing interference, this cir-
cuitry appeared to be maximally activated in children
when suppressing a behavioral response regardless of the
number of preceding responses. This pattern of ventral
frontostriatal activity correlated with both age and perfor-
mance, suggesting that changes in susceptibility to inter-
ference are paralleled by maturation in underlying fron-
tostriatal circuitry (Durston et al 2002b).
In the present study, we used our go/nogo paradigm in
combination with event-related functional magnetic reso-
nance imaging (fMRI) to investigate the effect of increas-
ing interference on both behavior (the suppression of the
button press involved in the go response) and magnetic
resonance (MR) signal in young children with ADHD and
group-matched controls. We predicted that children with
ADHD would demonstrate increased susceptibility to
interference compared to control children, as the develop-
ment of this ability is thought to be impaired in these
children. Therefore, we hypothesized that children with
ADHD would make more errors overall to nogo trials than
control children. Second, we hypothesized that these
differences would be paralleled by less efficient or focal
patterns of frontostriatal activity than in children without
ADHD.
Methods and Materials
Subjects
A total of 14 right-handed children completed the study1,
including 7 healthy controls (mean age � 8.68 [1.51] years,
range 6.2–10.3 years, 1 female) and 7 children meeting DSM-IV
criteria for ADHD (mean age � 8.55 [1.59] years, range
6.6–10.8 years, 1 female; 3 inattentive subtype, 4 combined
subtype) (American Psychiatric Association 1994). A structured
interview was used to establish diagnosis in the ADHD children
and to confirm absence of diagnosis for the controls (Schedule
for Affective Disorders and Schizophrenia for School-Age Chil-
dren–Parent Version) (Orvaschel 1994). ADHD children with
comorbid disorders other than opposition defiance disorder
(ODD) and conduct disorder (CD) were excluded. The Conners
rating scale (Conners 1969) and the Child Behavior Checklist
(CBCL) (Achenbach and Edelbrock 1983) were used to inven-
tory ADHD symptomatology. Subjects were screened for any
contraindications for MRI. All ADHD subjects were on stimulant
medication, which they withheld from taking the day of the MRI.
We obtained written assent from all subjects before scanning and
written consent from a parent or legal guardian. The procedure
was approved by the institutional review board at Weill Medical
College of Cornell University. Ten children were excluded from
the imaging study after having participated in a scanning session
(including 7 subjects with ADHD) due to excessive motion in the
scanner or technical problems; however, we were able to analyze
the behavioral data for the larger sample of children (14 children
with ADHD and 10 control children).
Paradigm
The subject’s task was to press a button in response to visually
presented stimuli but to avoid responding to a rare nontarget. The
task consisted of five runs, which lasted 3 minutes and 56
seconds each. Each run contained a total of 57 trials, with 75%
go trials, resulting in a total of 70 nogo trials, including 20 of
each type (with 1, 3, or 5 preceding go trials) per subject. Foil
trials (nogo trials after 2 or 4 go trials) were included to prevent
learning of the pattern, but these trials were not included in the
analysis. The order of presentation of the different types of nogo
trials was pseudorandomized. To make the task more interesting
for children, characters from the Pokemon cartoon series were
used as stimuli. Stimulus duration was 500 milliseconds and the
interstimulus interval was 3500 milliseconds (total trial length �
4000 milliseconds).
Scan Acquisition
Echo planar imaging (EPI) blood oxygenation level dependent
(BOLD) images were acquired in 24 axial slices on a 1.5 T GE
Signa scanner (Advanced NMR, Wilmington, MA), covering
most of the brain (repetition time [TR] � 2000, echo time [TE]
� 40, 64 � 64, 4-mm slice thickness, 3.125 � 3.125 mm
in-plane resolution). Anatomical spin echo images were also
collected (TR� 500, TE�min, 256� 256, field of view [FOV]
� 20, 4-mm slice thickness) in the same locations as the
functional slices. Stimuli were presented using the integrated
functional imaging system (IFIS) (MRI Devices Corporation,
Waukesha, WI) that uses a liquid crystal display (LCD) video
display in the bore of the MR scanner and a fiberoptic response
collection device. Scanning sessions lasted no longer than 1 hour.
The functional images were collected in 20–25 minutes, while
the anatomical images were collected within a similar time
frame. The participants were shown cartoons during the anatom-
ical scans to prevent boredom and restlessness.
Analysis
Automated Image Registration (AIR) version 3.08 (Woods et al
1992) was used for motion correction, image smoothing (2 mm),
and cross registration of data. Cross registration was checked by
visual inspection of an overlay of each subject’s brain with the
brain chosen as the standard. There were no differences in
variance between groups in MR signal. Foil trials were not
included in the analyses. NeuroImaging Software (Laboratory for
Clinical Cognitive Neuroscience, University of Pittsburgh and
Princeton University) was used to perform a series of voxelwise,
multifactorial analyses of variance (ANOVAs), beginning with a
14 (subjects) � 2 (condition: go vs. nogo) design, averaging
1 At least three runs of the paradigm with less than one voxel of movement. Runs
with too much motion were replaced with runs from the same subject to keep
the number of runs per subject constant. At least one run was replaced for three
subjects.
872 S. Durston et alBIOL PSYCHIATRY
2003;53:871–878
across all runs, to look for common patterns of activation
between groups. Secondly, a 14 (subjects) � 2 (group) � 2
(condition: go vs. nogo) ANOVA was performed to look for
differences in pattern of activation between groups. Post hoc
analyses were performed for each group separately. For each trial
type, two 2-second scans were included in the analyses taken at
the peak of the hemodynamic response (4 and 6 seconds after
stimulus presentation), yielding 40 data points per trial type per
subject. In each analysis, regions of three or more contiguous
voxels (p � .05 for each voxel in the cluster) were identified
(Forman et al 1995). To separate clusters that had more than one
maximum with contiguous voxels, analyses were also performed
at more conservative p-values, ranging from p � .005 to p � .01
to examine whether these regions were behaving similarly by
comparing their response over go trials and the three levels of
nogo trials. If they did not display the same pattern, they were
treated as separate clusters. Images were warped into stereotaxic
space using Analysis of Functional Neuroimages (AFNI) (Cox
1996) to localize regions of activity, based on the coordinate
system of the Talairach atlas (Talairach and Tournoux 1988).
Only correct trials were analyzed. A post hoc scan by scan
analysis similar to one we have used previously (Casey et al
2000; Durston et al 2002b) was performed on brain regions
identified as having significant signal change by the omnibus
group � condition ANOVA to test for a differential response
between groups on the three different types of nogo trials.
Correlations between MR signal change, age, and accuracy were
calculated for regions that differentiated the two groups.
Results
Behavioral Results
Across all subjects that participated in the behavioral
study, children with ADHD made significantly more
errors on nogo trials (mean accuracy � 90.4 [8.1]% for
controls; 79.1 [14.4]% for ADHD; t � 2.43, p � .032).
Differences in reaction time (RT) and accuracy on go trials
did not reach significance (mean RT � 677 [112] milli-
seconds for controls; 758 [102] milliseconds for ADHD; t
� 1.87, p � .09; mean accuracy � 97.8 [1.8]% for
controls; 95.8 [3.7]% for ADHD; t � 1.71, p � .12). For
the group of children with usable imaging data, there were
no differences between the children with and without
ADHD in reaction time (mean RT � 678 [140] millisec-
onds for controls; 719 [133] milliseconds for ADHD; t �
.02, p � .99) or accuracy (for go trials: mean accuracy �
97.8 [1.8]% for controls; 95.8 [3.7]% for ADHD; t � .71,
p � .49; for nogo trials: mean accuracy � 86.4 [8.8]% for
controls; 74.9 [9.6]% for ADHD; t � 1.19, p � .26.);
however, the number of errors made on the nogo trials
increased as a function of the number of preceding go
trials for the control children (10.2%, 17.2%, 17.9%, r �
.90, p � .02) but not for the children with ADHD (15.6%,
22.7%, 21.7%, r � .79, p � .07). Thus, ADHD children’s
performance on nogo trials preceded by only one go trial
was more similar to control children’s performance on
nogo trials preceded by as many as three to five go trials
(see Figure 1). Average motion was no different between
groups used in the final analysis (x � .02 mm, y � .00
mm, and z � .42 mm for controls; x � .23 mm, y � .01
mm, and z � 1.10 mm for ADHD; t � 1.3, p � .22 for x;
t � .1, p � .92 for y; t � 1.91, p � .09 for z).
fMRI Results
EFFECTS OF CONDITION (GO TRIALS VS. NOGO TRI-
ALS). Significant regions of activation for the ANOVAs
comparing go to nogo trials across groups are summarized
in Table 1. The most robust activation for the 14 (subjects)
� 2 (condition) ANOVA was in the left primary motor
cortex (Brodmann Area [BA] 4), where MR signal in-
creased for go trials (motor response) compared to nogo
trials (no motor response). MR signal increased for nogo
trials compared to go trials for regions in the right inferior
parietal lobe and bilateral posterior cingulate gyrus and
posterior hippocampus (see Table 1).
EFFECTS OF GROUP BY CONDITION. In the 14 (sub-
jects) � 2 (group) � 2 (condition) ANOVA, only one
region in the left caudate nucleus displayed more activa-
tion for control children than for children with ADHD. In
contrast, children with ADHD displayed more activation
in a number of regions, including the right superior frontal
gyrus (BA 10), the right middle frontal gyrus (BA 9/46),
the right inferior parietal lobe (BA 40), the bilateral
posterior cingulate gyrus (BA 31), the bilateral precuneus
(BA 7), the right superior temporal gyrus (BA 22), and the
bilateral occipital cortex (BA 18) (see Table 1).
EFFECTS OF CONDITION FOR CONTROL CHILDREN.
The results from the individual group ANOVAs are
summarized in Table 2. Control children showed an
increase in MR signal in the left primary motor cortex for
go trials compared to nogo trials. For nogo trials, MR
Figure 1. Percentage of errors on nogo trials as a function of the
number of preceding go trials (mean � SE). ADHD, attention-
deficit/hyperactivity disorder.
Differential Patterns of Striatal Activation 873BIOL PSYCHIATRY
2003;53:871–878
signal increased relative to go trials in the inferior frontal
gyrus (BA 44/45), caudate nucleus, right globus pallidus,
anterior cingulate gyrus (BA 32), right middle frontal
gyrus (BA 8), and right inferior parietal lobe (BA 40) (see
Table 2 and Figure 2).
EFFECTS OF CONDITION FOR CHILDREN WITH
ADHD. Children with ADHD showed an increase in MR
signal in the left primary motor cortex for go trials
compared to nogo trials similar to the controls. For the
comparison of nogo trials relative to go trials, they showed
an increase in activation in the right superior frontal gyrus
(BA 10), right inferior parietal cortex (BA 40), bilateral
posterior cingulate gyrus (BA 31), bilateral precuneus (BA
7), and bilateral occipital cortex (BA 18) (see Table 2 and
Figure 3).
POST HOC ANALYSES. The post hoc scan by scan
analysis revealed no effect of preceding context on MR
signal in any region for either group. Correlations between
MR signal change, age, and accuracy were not significant
for any region in this sample.
Discussion
In this study, we show that, behaviorally, children with
and without ADHD are susceptible to behavioral interfer-
ence during performance of a go/nogo task, with children
Table 1. Regions Activated in Comparisons Including All Subjects
Area Brodmann Area Side Talairach Maximum F
Condition (go versus nogo) for All Subjects
Primary Motor 4 left (�35, �15, 49) 40.37
Posterior Hippocampus bilateral (21, �46, 0) (�20, �47, 0) 9.98
Inferior Parietal Lobe 40 right (53, �32, 26) 8.20
Posterior Cingulate Gyrus 31 bilateral (8, �49, 35) (�8, �46, 35) 6.60
Group by Condition (go versus nogo)
More Activation in Control Subjects
Caudate Nucleus left (�10, 25, �5) 8.73
More Activation in ADHD Subjects
Superior Frontal Gyrus 10 right (11, 57, 21) 23.92
Occipital Cortex 18 bilateral (�20, 84, 3) (14, 87, �1) 17.95
Inferior Parietal Lobe 40 right (28, �42, 24) 16.89
Precuneus 7 bilateral (5, �53, 48) (�3, �55, 48) 13.89
Superior Temporal Gyrus 22 right (38, �35, 4) 12.68
Middle Frontal Gyrus 46 right (41, 30, 25) 9.73
Posterior Cingulate Gyrus 31 bilateral (19, �43, 28) (�7, �49, 26) 9.61
Middle Frontal Gyrus 9/46 right (53, 17, 25) 9.49
Activation is greater for nogo trials than for go trials in all areas except primary motor cortex, where activation is greater on go trials.
ADHD, attention-deficit/hyperactivity disorder.
Table 2. Regions Activated in Comparisons for Individual Groups
Area Brodmann Area Side Talairach Maximum F
Condition (go versus nogo) for Control Children
Primary Motor 4 left (�25, 22, 54) 180.04
Inferior Parietal Lobe 40 right (60, �25, 27) 39.06
Middle Frontal Gyrus 8 right (33, 35, 39) 34.43
Caudate Nucleus right (8, 25, �5) 29.00
Inferior Frontal Gyrus 44/45 left (�43, 23, 38) 18.27
Caudate Nucleus left (�15, 4, 16) 12.66
Globus Pallidus right (25, �6, �6) 11.58
Anterior Cingulate Gyrus 32 medial (�2, 2, 25) 6.58
Condition (go versus nogo) for ADHD Children
Superior Frontal Gyrus 9/10 right (17, 50, 22) 78.48
Posterior Cingulate Gyrus 31 bilateral (19, �35, 31) (�16, �43, 35) 33.37
Primary Motor 4 left (�32, 16, 54) 26.71
Inferior Parietal Lobe 40 right (48, �45, 25) 17.08
Occipital Cortex 18 bilateral (18, �82, �1) (�13, 82, 6) 16.45
Precuneus 7 bilateral (12, �40, 45) (�11, �38, 42) 10.21
Activation is greater for nogo trials than for go trials in all areas except primary motor cortex, where activation is greater on go trials.
ADHD, attention-deficit/hyperactivity disorder.
874 S. Durston et alBIOL PSYCHIATRY
2003;53:871–878
with ADHD being more susceptible to our parametric
manipulation. Using the same task, we previously showed
that children and adults differ in their susceptibility to
interference (with children making more errors overall),
but that both age groups showed a behavioral effect of
preceding context, with a greater number of false alarms
on nogo trials that were preceded by more go trials
(D