Effects of Prenatal Alcohol Exposure on Brain Activation
During an Arithmetic Task: An fMRI Study
Priya Santhanam, Zhihao Li, Xiaoping Hu, Mary Ellen Lynch, and Claire D. Coles
Background: While behavioral studies have established that prenatal alcohol exposure (PAE)
can result in diminished arithmetic processing capability, the underlying neural correlates of this
deficit are still unclear. The aim of the present study was to use functional magnetic resonance
imaging to determine the effect of PAE on neuronal activation during a subtraction task.
Methods: Participants were young adults from a low socio-economic status population who
were identified prenatally; the sample consisted of healthy unexposed controls (n = 17) and PAE
who were subdivided based on the presence (n = 19) or absence of physical dysmorphic signs
(n = 18). Multiple regression analysis was used to determine extent of activation and percent sig-
nal change during arithmetic processing, using a letter-matching task as the baseline. Region of
interest analysis of activation was performed in the native space and normalized for each individ-
ual to compensate for the considerable variability in head size observed in the alcohol-exposed
Results: An exposure-dependent response was observed in task performance and neuronal acti-
vation. Dysmorphic PAE individuals showed significantly lower task-related performance and
activation in regions known to be associated with arithmetic processing, including left superior
and right inferior parietal regions and medial frontal gyrus, while the nondysmorphic PAE group
was generally intermediate but not significantly different from the control group in task perfor-
mance and activation.
Conclusions: Results indicate that there is a range of effects of PAE on arithmetic processing
and that the severity of this deficit may be dependent on degree of impairment demonstrated by
the exposed individual. Evidence of physical dysmorphia may be indicative of functional damage
to regions associated with arithmetic calculation, resulting in markedly impaired neuronal
Key Words: Prenatal Alcohol, fMRI, Arithmetic Processing.
be a prevalent social and health concern today. It has been
estimated that approximately 1 of 500 infants in the United
States is born affected by such exposure (Abel, 1995). The ter-
atogenic results for offspring of maternal alcohol consump-
tion during pregnancy are referred to as fetal alcohol
spectrum disorders, the most severe of which is fetal alcohol
syndrome (FAS). Currently, FAS is clinically characterized
by facial dysmorphia, diminished growth, and neurodevelop-
mental disorders including microcephaly (Jones and Smith,
1973). However, diagnosis of FAS can be challenging as there
is no one presenting symptom and often behavioral outcomes
appear similar to those associated with other neurocognitive
HE HAZARDS OF prenatal alcohol exposure (PAE)
have been documented for decades, yet it continues to
disorders (Coles, 2001; Coles et al., 1997; Nash et al., 2006).
Structural and functional effects are reported in individuals
exposed prenatally who lack the physical dysmorphia associ-
ated with FAS (Mattson et al., 1998).
Behavioral problems in individuals with a range of PAE
have been observed for decades and are reported to include
both neurocognitive deficits as well as social and adaptive
dysfunction (Mattson and Riley, 1998). In general, individu-
als with FAS have lower IQ, often accompanied by impaired
visuo-spatial, attentional, memory recall, and⁄or language
skills (Coles et al., 1997; Conry, 1990; Mattson and Riley,
1998; Olson et al., 1998; Streissguth et al., 1994b). Develop-
mental dyscalculia, the reduced ability to understand and⁄or
apply core mathematical processes due to teratogenic
damage, is also widely reported to be associated with PAE
(Goldschmidt et al., 1996; Streissguth et al., 1994b, 1989),
perhaps even more often than global and verbal deficits
(Streissguth et al., 1994b). In adolescents, math-related deficits
range from longer response interval for mental math calcula-
tions to an inability to do basic addition and subtraction
(Streissguth et al., 1994a). Additionally, a study in adults
found dysfunction in a number of math skills including the
ability to estimate efficiently (Kopera-Frye et al., 1996). As
math processing appears to be a specific deficit associated
From the Departments of Biomedical Engineering (PS, ZL, XH)
and Psychiatry and Behavioral Sciences (MEL, CDC), Emory
University School of Medicine, Atlanta, Georgia.
Received for publication August 27, 2008; accept June 3, 2009.
Reprint requests: Claire D. Coles, PhD, 1256 Briarcliff Road, NE,
Copyright ? 2009 by the Research Society on Alcoholism.
Alcoholism: Clinical and Experimental Research
Vol. 33, No. 11
Alcohol Clin Exp Res, Vol 33, No 11, 2009: pp 1901–19081901
with PAE, the underlying neurocognitive correlates of this
arithmetic processing impairment warrant the closer examina-
tion that can be provided through functional neuroimaging.
Neural correlates of developmental and clinical dyscalculia
have been extensively documented and can serve as a guide to
atic activation of bilateral parietal, frontal, and precentral cor-
tices during arithmetic calculation (Fehr et al., 2007; Kazui
et al., 2000; Menon et al., 2000; Zhang et al., 2005). Lesion
studies and examinations of clinical populations indicate that
and colleagues (2003) point specifically to the horizontal inter-
confirmedthatthe intraparietal sulcus is activatedinbothchil-
dren and adults during both symbolic and nonsymbolic tests
verbal, spatial aspects of math processing with other areas of
the brain subsuming the verbal (symbolic) functions (Fehr
et al., 2007; Hubbard, 2005; Kong et al., 2005). To date, no
study has examined the effect of PAE on activation in these
A potential methodological confound in functional neuroi-
maging of individuals affected by PAE is head size differ-
ences. Reduced subregion and overall brain size have been
widely reported in alcohol-affected individuals (Riley et al.,
2004; Sowell et al., 2001) and is especially prevalent in those
with FAS (Archibald et al., 2001). The microcephaly common
to affected individuals has the potential to distort results of
imaging studies if not taken into account during activation
analysis. Previous studies of this clinical group have not
addressed this issue formally although it has been acknowl-
edged that normal spatial transformation may affect interpre-
tation of results (Bookheimer and Sowell, 2005).
In the present study we used functional magnetic resonance
imaging (fMRI) to examine the effects of PAE on brain acti-
vation during performance of a subtraction task. Previous
studies had demonstrated successful use of blood oxygenation
level-dependent (BOLD) signal as an indicator of arithmetic
functioning and deficiency (Cantlon et al., 2006; Delazer et al.,
2003; Fehr et al., 2007; Kong et al., 2005). Expected outcomes
included a significant difference between physically affected
(dysmorphic) PAE and control groups in task performance
and brain activation patterns in those regions previously asso-
ciated with general arithmetic calculation and specifically sub-
traction. In order to deal with the problem of potentially
confounding head size differences, we identified regions of
interest (ROI) on an individual basis, and normalized activa-
tion volumes based on the size of the identified subregion.
A second focus of the study was to examine the extent to
dysmorphic features would demonstrate similar patterns
of activation in comparison with socioeconomic status (SES)-
matched, nonexposed controls. If neurodevelopment is
equally affected in nondysmorphic individuals, we would
anticipate a similar pattern of dysfunction in nondysmorphic,
PAE individuals as that seen in the more physically affected
dysmorphic group. However, as we hypothesize that there is a
relationship between severity of physical effects of PAE and
the functional deficit associated with dyscalculia, we expect
that exposed but nondysmorphic individuals should show no
deficits or should be intermediate in performance between
those with dysmorphic features and nonexposed controls.
MATERIALS AND METHODS
Participants were 54 young adults, aged 20–26 years, whose prena-
tal exposure to alcohol was quantified prenatally through maternal
report. All were recruited from a longitudinal cohort, derived from a
predominantly African-American, low SES population first identified
between 1979 and 1986 when their mothers applied for prenatal care
(Smith et al., 1986). From this cohort, 3 groups were selected for par-
ticipation in the current study and recontacted. These included indi-
viduals who were (i) exposed to alcohol prenatally and exhibited
physical signs of such exposure, specifically facial dysmorphia
(n = 19); (ii) exposed without dysmorphia (n = 18) but with ability
scores (i.e., IQ £83) consistent with mean scores in Group 1; and (iii)
unexposed controls from the same low SES population (n = 17). All
participants were evaluated using a dysmorphia checklist (Fernhoff
et al., 1980), where characteristics associated with the disorder were
listed and weighted based on their saliency for the diagnosis (e.g.,
hypoplastic philtrum is a ‘‘3’’ while anteverted nares is a ‘‘1’’). The 30
items on the checklist were assessed either by a pediatric dysmophol-
ogist or a nurse trained and supervised by a dysmorphologist who
were blind to the participant’s exposure status. The weightings of
items checked were summed to yield a dysmorphology index. The
checklist has been evaluated repeatedly as part of longitudinal
research studies from birth to adolescence with individuals prenatally
exposed to alcohol receiving higher total scores in comparison with
nonexposed controls (Blackston et al., 2004; Fernhoff et al., 1980).
The criterion used to define the dysmorphic group was that the the
participant score had to be 1SD above the group mean at any one of
3 testing points (birth, age 7, mid-adolescence). Potential participants
who were left-handed, had some risk during the MRI procedure
(e.g., due to pregnancy or metal fragments), or those who were
uncomfortable with the procedure (e.g., claustrophobia) were not
Demographic and prenatal exposure characteristics are shown in
Table 1. While 74 volunteers were imaged originally, 20 were
excluded from data analysis and Table 1 includes information from
only participants included in the final analysis. Reasons for exclusion
were very poor behavioral task performance suggesting that the indi-
vidual was not able to perform the task (n = 2) and unusable data
due to head motion or artifact (n = 18). To assure that there was no
systematic bias associated with this participant loss, we compared
those who did and did not have usable data obtained during the
math protocol on the variables shown in Table 1 and found no signif-
icant differences on any of these measures. The mean ounces of abso-
lute alcohol consumption per week during pregnancy for the
dysmorphic and nondysmorphic groups were 13.5 (SD = 15.9) and
10.4 (SD = 18), respectively (Table 1).
Participants had been seen during adolescence (Coles et al., 2002)
and when recontacted as adults, gave informed consent to continue
to participate in the research. To protect the confidentiality of their
mothers, who had originally given informed consent, no information
about exposure group status or maternal substance use was provided.
The informed consent procedure was consistent with the Declaration
SANTHANAM ET AL.
of Helsinki and was approved by the School of Medicine’s Institu-
tional Review Board. Study personnel provided transportation to
and from the University research site for data collection and imaging.
Experimental procedures, including neuropsychological testing and
functional neuroimaging, were carried out by staff blind to group sta-
tus. Participants were reimbursed for their time and effort.
The experimental paradigm, used previously by Connor (personal
communication, 2004) with adults affected by alcohol exposure,
allowed the evaluation of subtraction performance while using a let-
ter-matching task to control for baseline cognitive and motor
activity. The task was presented in blocks, alternating between the
letter-matching control task (10 consecutive presentations) and a sub-
traction task (10 consecutive presentations). Although problems were
repeated across blocks, the order of the problems was randomized.
Five blocks of each type of task were presented, with instructions
stating either ‘‘Name the letter’’ or ‘‘Subtract from 11’’ being shown
before each block. Both tasks had a similar visual presentation
(Fig. 1). Participants were asked to choose between the 2 letters or
numbers on the bottom half of the screen by pressing the left or right
button on a button response box (http://www.curdes.com). Paradigm
presentation and response collection including accuracy and reaction
time were carried out using E-prime (http://www.pstnet.com). It
should be noted that the subtraction task was of a type that required
estimation (Klahr, 1973) of quantity and the technique known as
‘‘borrowing’’ (McCloskey and Caramazza 1987) rendering it ‘‘com-
plex’’ rather than ‘‘simple’’ by the standards of previous studies of
arithmetic operation correlates (Fehr et al., 2007; Kong et al., 2005).
All fMRI data was acquired on a 3T Siemens Trio scanner (Sie-
mens Medical, Erlangen, Germany). The arithmetic study was only
one of several functional paradigms implemented, with a total scan
time of 39 minute 59 seconds. For the arithmetic task, single-shot
T2*-weighted echo-planar imaging (EPI) images were acquired, con-
sisting of 34 contiguous axial slices with 3-mm slice thickness. Pulse
sequence parameters, designed to minimize susceptibility to signal
loss, were TR (relaxation time)⁄TE (echo time)⁄flip angle (FA)⁄field
of view (FOV) of 3000 ms⁄32 ms⁄90ºcm⁄22 cm. The scan time was
5 minutes 6 seconds, with 102 time points collected. High-resolution,
T1-weighted, three-dimensional (3D) anatomical images were also
acquired with a 3D MPRAGE (magnetization prepared rapid gradi-
ent echo) sequence for all participants. The scan protocol, optimized
at 3T, used TR⁄TI (inversion time)⁄TE of 2600 ms⁄900 ms⁄3.93
ms,FAof8?,FOVof256 · 224 · 176 mm3,
256 · 224 · 176, corresponding to an isotropic resolution of 1 mm.
Scan time was 7 minutes 18 seconds.
AFNI (http://afni.nimh.nih.gov) was used to perform imaging data
analysis. After the data preprocessing steps (slice timing correction,
volume registration, signal normalization to percent change, and
5mm FWHM Gaussian blur), 3D + time fMRI datasets for each
individual were submitted to a multiple regression analysis. Using the
the boxcar stimulation functions with a standard impulse response
function [y = t^b·exp()t⁄c), where b and c are constants] (Cohen,
1997). In order to achieve a better modeling of the motion-related sig-
nal variation, the rigid body head motion parameters (x, y, z displace-
ments and roll, pitch, and yaw rotations) were included as six
additional regressors as well. The outcome of this multiple regression
analysis included statistical parametric maps which showed voxels
with a significant task effect (partial F-statistic) and regression coeffi-
proportional to the BOLD signal increase level in the arithmetic task
from the letter task (baseline). For each group, the statistical paramet-
ric maps of individuals were averaged after transforming the dataset
Fig. 1. Experimental paradigm.
Table 1. Background and Prenatal Exposure Characteristics for Participants by Exposure Group
(n = 17)a
(n = 18)a
(n = 19)a
Age at imaging, M (SD)
Monthly income—$ in past 30 days, M (SD), n = 53
Education completed—years, M (SD), n = 53
Full-scale IQ, M (SD) n = 53
Dysmorphia rating at adult visit, M (SD)
Adult head circumference—cm,bM (SD) n = 53
Amount of alcohol (AA per week) exposure during pregnancy, M (SD)
Cigarettes during pregnancy—% using, n = 53
Marijuana during pregnancy—% using, n = 49
Cocaine during pregnancy—% using, n = 44
p = 0.032
p = 0.001
p = 0.003
p = 0.001
p = 0.016
p < 0.001
aIf data for a variable are not available for some participants, the n used for the analysis is noted next to the variable name.
bTwo-way Group · Gender analysis of variance was completed for head circumference. No gender or interaction effects were significant.
EFFECT OF PAE ON ARITHMETIC PROCESSING
into the Talairach space (Talairach and Tournoux,1988)and normal-
izing the F values into Z-scores. Voxels that followed the general lin-
ear model of letter task baseline and arithmetic block activation were
considered to reflect the so-called ‘‘arithmetic effect.’’ Thresholded
activation maps of this arithmetic effect, averaged from individual
datasets, are shown for PAE and control groups in Fig. 2. Voxel-wise
group t-test maps were also created to determine activation difference
between control and both exposure groups (Fig. 3). To account for
multiple comparisons, voxel-wise thresholding (p < 0.05 for Fig. 2;
p < 0.01 for Fig. 3) with cluster thresholding of 4 contiguous voxels
was applied. Monte Carlo simulation revealed that these thresholds
corresponded to afalse-positive discoveryrate(alpha)oflessthan 1%
for Fig.2 and lessthan0.1%forFig.3.
In order to quantitatively compare brain activities between groups,
we defined ROIs in the Talairach space based on an atlas provided
by AFNI. ROIs were chosen based on activation maps of arithmetic
effect for the sample as a whole (regions with higher activation in the
arithmetic task vs. control task). The following brain areas were iden-
tified: left and right superior and inferior parietal regions, superior
frontal, medial frontal, middle frontal, and inferior frontal gyri. It
was noted that all these ROIs were also implicated in previous studies
of arithmetic processing. ROI-associated functional activation vol-
umes and corresponding regression coefficients were then calculated
in native space for each individual. The location and extent of each
ROI was defined anatomically in native space by applying the inverse
warping from the Talairach transformation onto the ROI mask and
using the functional dataset as a template. Activation extent (number
of active voxels) was subsequently determined for each ROI and then
normalized to the voxel size of the entire ROI for each individual.
Normalized activation volumes were compared between exposure
Fig. 2. Arithmetic effect for (A) controls, (B) nondysmorphic prenatal alco-
hol exposure (PAE), and (C) dysmorphic PAE groups. D indicates location
of chosen axial slices (z = +46 to +53).
Fig. 3. Subtraction map of the arithmetic effect in control subjects minus
arithmetic effect in the (A) nondysmorphic prenatal alcohol exposure (PAE)
subjects or (B) dysmorphic PAE subjects. C indicates location of chosen
axial slices (z = +46 to +53).
SANTHANAM ET AL.
groups by t-test. Additionally, the unthresholded regression coeffi-
cient was calculated for each ROI in native space, converted to per-
cent BOLD signal change, and compared between exposure groups
Using ‘‘number correct’’ as the outcome measure, there
was a significant difference in performance on the arithmetic
task between control subjects and dysmorphic PAE subjects
(Table 2), with lower accuracy in the dysmorphic group.
When participants gave no response (so-called ‘‘skips’’), it
was counted as an incorrect response in determining accuracy.
Participants who skipped more than 50% of arithmetic
responses were excluded from analysis (n = 2), and of the
remaining participants, the number of skips was comparable
among groups. Accuracy was at or near 100% on the letter-
matching control task for all groups, and there was not a sig-
nificant difference in reaction time for either task between any
of the groups.
Figure 2 shows the arithmetic effect (arithmetic task minus
control task) in selected slices for each group. Specifically,
robust activation was seen in bilateral parietal lobe, medial
frontal gyrus, and bilateral middle frontal gyrus in the control
group, while activation in the exposed individuals was sparser
and primarily on the right side of the middle frontal and pari-
etal regions. Figure 2 also indicates more overall activation in
the control group when compared with the 2 PAE groups.
Regions of interest (ROI), listed in Table 3, were identified
based on these activation maps. Selected slices from group
difference maps of arithmetic effect-related activation are
shown in Fig. 3. Greater activation in controls when com-
pared with nondysmorphic PAE in the middle frontal and
parietal regions was notable in Fig. 3A. Figure 3B shows sig-
nificantly more activation in control subjects when compared
with dysmorphic PAE subjects in bilateral parietal, middle
frontal, and medial frontal gyri.
Activation volumes and percent signal change for each
ROI are indicated in Table 3. To verify that activation differ-
ences were reflective of impairment and not lack of task
engagement, activation was also examined excluding subjects
performing below chance (50%) on the arithmetic task
(ExLS: n = 17 for nondysmorphic PAE group; n = 13 for
dysmorphic PAE group). None in the control group scored
below chance. Left and right superior and right inferior parie-
tal regions and medial frontal gyrus showed an exposure-
dependent response, with the dysmorphic PAE group having
the lowest amount of activation. Furthermore, dysmorphic
PAE subjects had significantly less (p < 0.05) activation
when compared with the control group in all of these regions
except the right superior parietal area. When low-scoring sub-
jects were excluded, the same ROIs remained significantly dif-
ferent from controls, and no other ROIs had significantly
different activation volumes. The percent BOLD signal
change is also indicated in Table 3 for each ROI. Average per-
cent signal change trended in the same direction as activation
volumes (correlation with average activation volume was
r = 0.97), with marginally significant differences (p < 0.10)
between the control and dysmorphic groups in the right
Table 2. Performance on the Arithmetic Task for Each Exposure
Group, With Accuracy Determined as Percent of Questions
Correctly Answered (of 60)
(mean ± SEM)p-Value
(mean ± SEM), msp-Value
72.6 ± 3.8
65.3 ± 4.2
60.1 ± 4.4
1174 ± 46
1161 ± 38
1121 ± 48
aIndicates significantly different from control group.
Table 3. Comparison of Average Activation Volume and Percent Signal Change in Selected Regions of Interest (ROI)
[x, y, z]
Normalized cluster volumePercent signal change
[27, 57, 53]
[)27, 57, 53]
[48, 41, 39]
[)48, 41, 39]
[±19, )40, 27]
[±9, )24, 35]
[±37, )29, 26]
[±44, )24, 2]
TAL, Talairach; Dys, dysmorphic.
aTAL coordinate is center coordinate in Talairach space for each ROI.
bExLS indicates exposure group excluding low-scoring subjects (below chance).
*Indicates significantly different from control group (p < 0.05).
?Indicates marginally significant difference from control group (p < 0.10).
EFFECT OF PAE ON ARITHMETIC PROCESSING
inferior parietal and medial frontal gyri. With the exclusion of
the low-scoring subjects, these ROI differences were still mar-
ginally significant. No significant correlation was found
between task performances and either activation volume or
percent signal change (r < 0.5 for all groups and all ROIs).
Given that PAE has been reported to cause deficits in arith-
response in task performance and in brain regions previously
associated with arithmetic calculation, with significantly dif-
ferent activation patterns between the dysmorphic PAE group
and controls. As predicted, in the present study, dysmorphic
PAE individuals showed significantly diminished ability to
perform a subtraction task while activation differences were
noted in regions known to be associated with arithmetic pro-
cessing. Activation in the left superior parietal regions, right
inferior parietal region, and medial frontal gyrus during the
task reflected an exposure-dependent response, with dysmor-
phic PAE individuals having significantly less activity. It
should be noted that excluding those subjects with task per-
formance below chance level still resulted in less activation in
the dysmorphic PAE group in the same ROIs which verified
that reduced activation volume was reflective of exposure-
based deficit as opposed to lack of engagement in the task. In
general, the nondysmorphic PAE group had both intermedi-
ate activation and task performance although they were not
significantly different in performance from either group. Fur-
thermore, the trend of less activation in exposed groups than
controls by volume measure was also reflected in percent sig-
It should be noted that the control group had greater
activation volume in all ROIs when compared with the
dysmorphic PAE group, with the exception of superior
frontal gyrus, although the difference was not always sig-
impaired in the nondysmorphic group, and was actually
comparable or higher when compared with controls in the
inferior parietal region and medial⁄inferior frontal gyri.
While activation volume differences might appear sizeable
for some ROIs, they were not always significant. This lack
of significance might be due to the considerable intersub-
ject variability within each group. We also noted that
while percent signal change correlated with activation vol-
ume in this study, it was only marginally statistically sig-
performance on the subtraction task by alcohol-affected
individuals was consistent with previous reports that PAE
was associated with diminished arithmetic processing in
children and adolescents. As noted in the introduction, a
number of studies have reported such effects. Streissguth
and colleagues (1989, 1994b) showed significant effects in
children asked to perform arithmetic-based tasks at several
stages of academic development. This longitudinal study
additionally noted that 91% of the PAE children who
showed arithmetic deficiency at 7 years of age, continued
to show deficits at 14 years of age as opposed to only
45% in the control group (Streissguth et al., 1994b).
The control and PAE groups in this study were not IQ-
matched, raising the question of whether task performance
was influenced by IQ differences. However, it should be noted
that while both PAE groups had significantly lower IQ when
compared with the controls, only the dysmorphic PAE group
had significantly poorer task performance. Furthermore, a
study of learning deficits in this cohort (Howell et al., 2006)
revealed that while PAE groups specifically demonstrated
arithmetic deficits, a low-IQ ‘‘special-education’’ contrast
group had deficits in reading and spelling in addition to arith-
metic. This finding suggests that the contrast group may have
global damage more closely tied to their low IQ whereas the
PAE groups have specific problems with math resulting from
This fMRI study found significant differences in activation
in bilateral parietal regions as well as the medial frontal
region, which are known to be associated with arithmetic pro-
cessing (Dehaene et al., 2004, 2003; Menon et al., 2000).
Recently, Fehr and colleagues (2007) used fMRI to compre-
hensively identify brain areas related to a number of simple
arithmetic operation (e.g., addition, subtraction, etc.). One
specific finding was that, among other regions, medial frontal
and bilateral inferior parietal regions were significantly more
activated during a ‘‘complex’’ subtraction task when com-
pared with a ‘‘simple’’ arithmetic task. Kong and colleagues
(2005) also recently examined the neural correlates associated
with simple and complex arithmetic operations using fMRI.
Complex subtraction was defined by the authors as involving
‘‘borrowing,’’ using tasks similar to those in the present study.
They too found involvement of medial frontal gyrus, among
other regions, for the complex arithmetic tasks. Furthermore,
left superior and right inferior parietal cortices were identified
as the 2 subregions of the parietal lobe specifically associated
with subtraction. It was further shown that all regions
recruited in performing addition tasks were also required for
subtraction. The association of these 2 subregions with sub-
traction calculation specifically supported our finding that the
dysmorphic PAE group had less activation during the sub-
traction task in the left superior and right inferior parietal cor-
tices. In the current study, differences in activation in these
regions could reflect a deficiency on the part of the dysmor-
phic PAE group in recruiting the neuronal arithmetic net-
work. Specifically, bilateral parietal region differences could
indicate dyscalculia or the inability to perform the subtraction
itself, while medial frontal gyrus differences could signify poor
recruitment of a region needed for complexity (‘‘borrowing’’).
This component is believed to be involved in the working
memory aspect of the task (Hampson et al., 2006). Dysmor-
phic alcohol-affected individuals may therefore have neuronal
recruitment problems in both the regions activated by all
types of arithmetic function and those unique to subtraction
operation calculation. Such a deficiency could also account
for the poorer task performance by the dysmorphic group.
SANTHANAM ET AL.
There have been few other studies that utilize fMRI to
examine neurocognitive deficits associated with PAE. Malisza
and colleagues (2005) reported functional differences in brain
regions in individuals with fetal alcohol spectrum disorder
(FASD) during a spatial working memory task. In both chil-
dren and adults, the authors found increased activation in
FASD individuals in inferior-middle frontal lobe and greater
activation in control individuals in superior frontal and parie-
tal lobes. Additionally, adults had less overall activation when
compared with children and FASD groups had lower activa-
tion overall versus controls.
Another very recent fMRI study on FASD children
(Meintjes et al., 2007) reported increased activity in controls
when compared with FASD in the left HIPS and left supe-
rior frontal region during an exact addition task. The chil-
dren also performed a proximity judgment task, in which
increased activation in controls in left and right HIPS and
frontal areas was noted, along with greater activation in
FASD in the anterior cingulate and left angular gyrus. As
the task in the present study mirrored exact addition more
than proximity judgment, our findings were consistent with
the report that FASD children had diminished neuronal
As we have noted, one challenge when using the fMRI
method on a prenatally alcohol exposed population was the
smaller head size that resulted from perturbed neurodevelop-
ment and characterized this group. In the present study, for
example, whole brain size was found to be significantly differ-
ent between both PAE groups and the control group
(p = 0.0048 for nondysmorphic and p = 0.0007 for dysmor-
phic). Bookheimer and Sowell (2005) pointed out that
because of microcephaly, apparent increases in activation vol-
ume in the FASD population could be a result of structural
abnormality or improper image registration. Therefore, stud-
ies in which anatomical images were normalized to common
space might be distorting the activation differences. In this
study, we wished to control for this potential methodological
issue. We verified that whole brain activation differences
between nondysmorphic and dysmorphic PAE groups and
controls were not significant when normalized to whole brain
anatomical size (p = 0.35 and p = 0.39, respectively). There-
fore, for our activation volume measurements, we utilized a
warping method in which ROI were chosen by Talairach atlas
in common space and their masks were warped with the
inverse matrix back into original space for each individual.
The activation volumes in each ROI were then normalized to
the size of the whole ROI for each individual. In this way the
ROI analyzed were uniquely sized and standardized for each
individual, making the number of active voxels in the region
As noted in the Results, while the dysmorphic PAE group
performed more poorly overall on the subtraction task, no
correlation was found between this behavioral performance
and activation. The use of different strategies by different sub-
jects (e.g., rote memorization, counting) is a possible explana-
tion for the general lack of association between activation
and task performance. However, several studies have shown
activation patterns in the parietal lobe varying with arithmetic
competency (Delazer et al., 2003; Fehr et al., 2007; Grabner
et al., 2007), including degree of automaticity and efficient
functioning with task (Ischebeck et al., 2006) and these results
suggest that further research is needed to evaluate the rela-
tionship between performance and activation.
A next step in understanding the relationship between
structural damage induced by PAE exposure and its effects
on the functional brain activation is to obtain a more direct
correlation between performance and brain activity for cogni-
tive tasks. Using a simpler task could decrease the high vari-
ance in activation measures and elucidate a quantifiable
relationship between arithmetic calculation and neuronal acti-
vation in alcohol affected and exposed individuals. It should
also be noted that as the brain regions affected in the present
study are associated specifically with subtraction, a paradigm
consisting of several different arithmetic operations could elu-
cidate the extent of dyscalculia in the affected population.
The behavioral and imaging results of this study suggested
that PAE was associated with diminished arithmetic process-
ing capabilities and such deficits were the result of functional
damage to regions known to be associated with mathematical
calculation. Specifically, the dysmorphic PAE group appears
to have marked impairment in recruiting neurons from bilat-
eral parietal and medial frontal regions for arithmetic process-
ing. Given prior characterization of the neural correlates of
arithmetic operations, more heavily exposed alcohol-affected
individuals may have difficulty with both the operation itself
and its complexity. Furthermore, that the nondysmorphic
PAE group did not have significant activation or performance
problems implied a range of responses to the teratogenic
exposure that required further study to delineate. Overall, the
findings of this study further supported the direct relationship
between PAE and functional brain damage, specifically eluci-
dating a neurological basis for observed arithmetic deficit.
This work was supported by Georgia Research Alliance
and by the National Institute of Alcohol and Alcoholism
(NIAAA) RO1 AA014373. There are no conflicts of interest,
including specific financial interests and relationships and
affiliations relevant to this manuscript. We thank Sharron
Paige-Whitaker for her work corresponding with and
recruiting participants. We would also like to thank the study
participants and their families for their continuing coopera-
tion with this research.
Abel EL (1995) An update on incidence of FAS: FAS is not an equal opportu-
nity birth defect. Neurotoxicol Teratol 17:437–443.
Archibald SL, Fennema-Notestine C, Gamst A, Riley EP, Mattson SN,
Jernigan TL (2001) Brain dysmorphology in individuals with severe prenatal
alcohol exposure. Dev Med Child Neurol 43:148–154.
EFFECT OF PAE ON ARITHMETIC PROCESSING
Blackston RD, Coles CD, Kable JA, Seitz R (2004) Reliability and validity of
the Dysmorphia Checklist: relating severity of dysmorphia to cognitive and
behavioral outcomes in children with prenatal alcohol exposure. Poster
presented at the American Society of Human Genetics Annual Meeting,
Bookheimer SY, Sowell ER (2005) Brain imaging in FAS: commentary on the
article by Malisza et al. Pediatr Res 58:1148–1149.
Cantlon JF, Brannon EM, Carter EJ, Pelphrey KA (2006) Functional imaging
of numerical processing in adults and 4-y-old children. PLoS Biol 4:e125.
Cohen MS (1997) Parametric analysis of fMRI data using linear systems
methods. Neuroimage 6:93–103.
Coles CD (2001) Fetal alcohol exposure and attention: moving beyond
ADHD. Alcohol Res Health 25:199–203.
Coles CD, Platzman KA, Lynch ME, Freides D (2002) Auditory and visual
sustained attention in adolescents prenatally exposed to alcohol. Alcohol
Clin Exp Res 26:263–271.
Coles CD, Platzman KA, Raskind-Hood CL, Brown RT, Falek A, Smith IE
(1997) A comparison of children affected by prenatal alcohol exposure and
attention deficit, hyperactivity disorder. Alcohol Clin Exp Res 21:150–161.
Conry J (1990) Neuropsychological deficits in fetal alcohol syndrome and fetal
alcohol effects. Alcohol Clin Exp Res 14:650–655.
Dehaene S, Molko N, Cohen L, Wilson AJ (2004) Arithmetic and the brain.
Curr Opin Neurobiol 14:218–224.
Dehaene S, Piazza M, Pinel P, Cohen L (2003) Three parietal circuits for num-
ber processing. Cognitive Neuropsychology 20:487–506.
Delazer M, Domahs F, Bartha L, Brenneis C, Lochy A, Trieb T, Benke T
(2003) Learning complex arithmetic: an fMRI study. Brain Res Cogn Brain
Fehr T, Code C, Herrmann M (2007) Common brain regions underlying dif-
ferent arithmetic operations as revealed by conjunct fMRI-BOLD activa-
tion. Brain Res 1172:93–102.
Fernhoff PM, Smith IE, Falek A (1980) Dysmorphia Checklist. Document
available through the Maternal Substance Abuse and Child Development
Project. Department of Psychiatry and Behavioral Sciences, Emory Univer-
sity School of Medicine, Georgia.
Goldschmidt L, Richardson GA, Stoffer DS, Geva D, Day NL (1996) Prena-
tal alcohol exposure and academic achievement at age six: a nonlinear fit.
Alcohol Clin Exp Res 20:763–770.
Grabner RH, Ansari D, Reishofer G, Stern E, Ebner F, Neuper C (2007) Indi-
vidual differences in mathematical competence predict parietal brain activa-
tion during mental calculation. Neuroimage 38:346–356.
Hampson M, Driesen NR, Skudlarski P, Gore JC, Constable RT (2006) Brain
connectivity related to working memory performance. J Neurosci 26:13338–
Howell KK, Lynch ME, Platzman KA, Smith GH, Coles CD (2006) Prenatal
alcohol exposure and ability, academic achievement, and school functioning
in adolescence: a longitudinal follow-up. J Pediatr Psychol 31:116–126.
Hubbard TL (2005) Representational momentum and related displacements
in spatial memory: a review of the findings. Psychon Bull Rev 12:822–851.
Ischebeck A, Zamarian L, Siedentopf C, Koppelstatter F, Benke T, Felber S,
Delazer M (2006) How specifically do we learn? Imaging the learning of
multiplication and subtraction Neuroimage 30:1365–1375.
Jones KL, Smith DW (1973) Recognition of the fetal alcohol syndrome in
early infancy. Lancet 2:999–1001.
Kazui H, Kitagaki H, Mori E (2000) Cortical activation during retrieval of
arithmetical facts and actual calculation: a functional magnetic resonance
imaging study. Psychiatry Clin Neurosci 54:479–485.
Klahr D (1973) Quantification Processes. Academic Press, Inc., New York,
Kong J, Wang C, Kwong K, Vangel M, Chua E, Gollub R (2005) The neural
substrate of arithmetic operations and procedure complexity. Brain Res
Cogn Brain Res 22:397–405.
Kopera-Frye K, Dehaene S, Streissguth AP (1996) Impairments of number
processing induced by prenatal alcohol exposure. Neuropsychologia
Malisza KL, Allman AA, Shiloff D, Jakobson L, Longstaffe S, Chudley AE
(2005) Evaluation of spatial working memory function in children and
adults with fetal alcohol spectrum disorders: a functional magnetic reso-
nance imaging study. Pediatr Res 58:1150–1157.
Mattson SN, Riley EP (1998) A review of the neurobehavioral deficits in chil-
dren with fetal alcohol syndrome or prenatal exposure to alcohol. Alcohol
Clin Exp Res 22:279–294.
Mattson SN, Riley EP, Gramling L, Delis DC, Jones KL (1998)
McCloskey M, Caramazza A (1987) Cognitive mechanisms in normal and
impaired number processing, in Mathematical Disabilities: A Cognitive
Neuropsychological Perspective (Deloche G, Seron X eds), pp. 201–219.
Lawrence Erlbaum Associates, Hillsdale, NJ.
Meintjes EM, Jacobson SW, Jacobson JL, Gatenby JC, Cannistraci CJ,
Molteno CD, Warton C, Dehaene S, Gore JC (2007) Abstract 232. Organi-
zation for Human Brain Mapping, Chicago, IL.
Menon V, Rivera SM, White CD, Glover GH, Reiss AL (2000) Dissociating
prefrontal and parietal cortex activation during arithmetic processing.
Nash K, Rovet J, Greenbaum R, Fantus E, Nulman I, Koren G (2006) Identi-
fying the behavioural phenotype in fetal alcohol spectrum disorder:
sensitivity, specificity and screening potential. Arch Womens Ment Health
Olson HC, Feldman JJ, Streissguth AP, Sampson PD, Bookstein FL (1998)
Neuropsychological deficits in adolescents with fetal alcohol syndrome:
clinical findings. Alcohol Clin Exp Res 22:1998–2012.
Smith IE, Coles CD, Lancaster J, Fernhoff PM, Falek A (1986) The effect of
volume and duration of prenatal ethanol exposure on neonatal physical and
behavioral development. Neurobehav Toxicol Teratol 8:375–381.
Sowell ER, Thompson PM, Mattson SN, Tessner KD, Jernigan TL, Riley
EP, Toga AW (2001) Voxel-based morphometric analyses of the brain in
children and adolescents prenatally exposed to alcohol. Neuroreport
Streissguth AP, Barr HM, Olson HC, Sampson PD, Bookstein FL, Burgess
DM (1994a) Drinking during pregnancy decreases word attack and arithme-
tic scores on standardized tests: adolescent data from a population-based
prospective study. Alcohol Clin Exp Res 18:248–254.
Streissguth AP, Barr HM, Sampson PD, Bookstein FL (1994b) Prenatal alco-
hol and offspring development: the first fourteen years. Drug Alcohol
Streissguth AP, Bookstein FL, Sampson PD, Barr HM (1989) Neurobehavior-
al effects of prenatal alcohol: Part III. PLS analyses of neuropsychologic
tests. Neurotoxicol Teratol 11:493–507.
Talairach J, Tournoux P (1988) Co-planar Stereotaxic Atlas of the Human
Brain. New York: Thieme Medical Publishers, Inc.
Zhang YT, Zhang Q, Zhang J, Li W (2005) Laterality of brain areas associ-
ated with arithmetic calculations revealed by functional magnetic resonance
imaging. Chin Med J (Engl) 118:633–638.
SANTHANAM ET AL.