Relationships between IQ and Regional
Cortical Gray Matter Thickness in Healthy
Katherine L. Narr1, Roger P. Woods2, Paul M. Thompson1,
Philip Szeszko3, Delbert Robinson3, Teodora Dimtcheva1,
Mala Gurbani1, Arthur W. Toga1,2and Robert M. Bilder2,4
1Laboratory of Neuro Imaging and2Ahmanson-Lovelace Brain
Mapping Center, Department of Neurology, Geffen School of
Medicine at the University of California, Los Angeles (UCLA),
Los Angeles, CA, USA,3Department of Psychiatry Research,
The Zucker Hillside Hospital, North-Shore Long Island Jewish
Health Systems, Glen Oaks, NY, USA and4Jane and Terry Semel
Institute for Neuroscience and Human Behavior, Geffen
School of Medicine at UCLA, Los Angeles, CA, USA
Priorstudiesshowpositive correlationsbetweenfull-scale intelligence
have addressed whether general intelligence is related to regional
variations in brain tissue and the associated influences of sex. Cortical
thickness may more closely reflect cytoarchitectural characteristics
than gray matter density or volume estimates. To identify possible
localized relationships, we examined FSIQ associations with cortical
thickness at high spatial resolution across the cortex in healthy young
adult (age 17--44 years) men (n 5 30) and women (n 5 35). Positive
matter but not cerebrospinal fluid volumes. Significant associations
with cortical thickness were evident bilaterally in prefrontal
(Brodmann’s areas [BAs] 10/11, 47) and posterior temporal cortices
(BA 36/37)and proximalregions.Sexinfluenced regionalrelationships;
women showed correlations in prefrontal and temporal association
cortices, whereas men exhibited correlations primarily in temporal--
occipital association cortices. In healthy adults, greater intelligence is
associated with larger intracranial gray matter and to a lesser extent
with white matter. Variations in prefrontal and posterior temporal
cortical thickness are particularly linked with intellectual ability. Sex
moderates regional relationships that may index dimorphisms in
cognitive abilities, overall processing strategies, or differences in
Keywords: cerebral cortex, cognition, frontal, intelligence,
magnetic resonance imaging, sex
The essence of human intelligence has been a topic of consider-
able interest for many centuries. With the advent of imaging
technologies and their advancement in recent decades, unique
opportunities have emerged to study the neurobiological
correlates of intellectual ability. To date, empirical evidence
from imaging data, that allows the in vivo assessment of brain
structure and function, has confirmed positive links between
brain size and general intellectual ability. Estimated population
correlations between brain size and general intelligence,
termed g by Spearman (1904), are approximately 0.33
(McDaniel 2005). These relationships persist in spite of age
(Reiss et al. 1996; McDaniel 2005) but are stronger in adults and
than in children (Wilke et al. 2003; McDaniel 2005). Correla-
tions also appear to be moderated by sex, although brain size--
intelligence associations are reported in both males and females
(Gur et al. 1999; McDaniel 2005). Relationships between brain
size and general intelligence, both separately identified as
heritable traits (Tramo et al. 1998; Baare et al. 2001; Posthuma
et al. 2002), have been shown to be almost entirely genetically
mediated (Thompson et al. 2001; Posthuma et al. 2002; Toga and
Thompson 2005). Because genes appear to influence both
phenotypes, a better understanding of how these traits are
linked may help elucidate the means of genetic transmission
that is not yet known.
For studies assessing associations with intracranial tissue
volumes, the majority of data suggests that higher standardized
intelligence scores are associated with larger cerebral gray
matter volumes (Andreasen et al. 1993; Reiss et al. 1996; Gur
et al. 1999). Positive relationships are also reported for white
matter volumes (Gur et al. 1999; Haier et al. 2004), but negative
findings exist (Andreasen et al. 1993). Although evidence
supports nontrivial associations between cerebral gray matter
and intelligence, few imaging studies have attempted to resolve
whether structural variation in specific brain regions are
associated with general intelligence and the regional specificity
of existing findings is mixed. Moreover, relationships between
intelligence and regional changes in brain tissue characteristics
appear to be influenced by brain maturation across develop-
ment. For example, in healthy children, associations between
intelligence scores and cerebral gray matter, which develop
with age (Wilke et al. 2003), have been localized in coarsely
defined prefrontal regions (Reiss et al. 1996) and within the
anterior cingulate in a study using voxel-based morphometry
(VBM) methods (Wilke et al. 2003). Another VBM investigation
of adolescents (age range: 12--21 years) again reported signifi-
cant positive associations between FSIQ and gray matter density
in cingulate cortices, whereas correlations in orbitofrontal
(Brodmann’s areas [BAs] 10, 11, and 47) and middle frontal
cortical regions were further identified (Frangou et al. 2004).
Volumetric regions of interest studies in healthy adults also
show correlations with full-scale intelligence quotients (FSIQs)
in prefrontal regions (Flashman et al. 1997), although correla-
tions are additionally reported with temporal lobe areas
(Andreasen et al. 1993; Flashman et al. 1997). In a VBM study
assessing correlations throughout the brain volume, Haier et al.
(2004) found significant associations between higher FSIQ
scores and increased gray matter density in prefrontal (BA 9,
10, 46), temporal (BA 21, 37, 22, 42), parietal (BA 43, 3), and
occipital (BA 19) cortices in healthy adults. Although results
concerning the spatial localization of gray matter--intelligence
relationships are disparate and existing data are sparse, associ-
et al. 2003; Frangou et al. 2004; Haier et al. 2004), temporal
(Andreasen et al. 1993; Flashman et al. 1997; Haier et al. 2004),
and, to a lesser extent, parietal and occipital (Haier et al. 2004)
regions are most replicated. Whereas structural variations of
thalamus and cerebellum appear involved at the subcortical
level (Andreasen et al. 1993; Frangou et al. 2004). Regional
Cerebral Cortex September 2007;17:2163--2171
Advance Access publication November 21, 2006
? The Author 2006. Published by Oxford University Press. All rights reserved.
For permissions, please e-mail: email@example.com
by guest on June 12, 2013
discrepancies in findings may reflect differences in methodolog-
ical approaches (regions of interest vs. VBM approaches), cor-
rection strategies for addressing interindividual differences in
brain size, and differences in the demographic characteristics of
the samples studied. For example, the registration errors assoc-
iated with VBM, which are required to generate a measurable
changes in structural morphology (Bookstein 2001; Honea et al.
2005; Thacker 2005) and subsequent associations with intelli-
gence scores. Differences in the standardized measures used to
assess general intellectual ability, how strongly individual in-
telligence tests relate to the general factor of intelligence (g)
(Colom et al. 2006), and the level of intellectual ability within
study groups may also potentially impact the magnitude and
regional specificity of results.
The thickness of the cortex, ranging between 1.5 and 4.5 mm
in different cortical regions (Parent and Carpenter 1995),
reflects cytoarchitectural characteristics of the neuropil includ-
fibers. Measures of cortical thickness, although shown to relate
to other local measures of gray matter (Narr, Bilder, et al. 2005),
may more closely link with cognition and/or intellectual ability
than volumetric or intensity-based gray matter concentration
measures. Only one published study to date has examined cor-
relations between general intelligence and cortical thickness.
This investigation addressed the trajectory of change in the
thickness of the neocortex from early childhood to early adult-
hood and found that relationships between FSIQ and cortical
thickness were negative in early childhood (age range: 3.8--8.4
years) but that these relationships shifted toward positive
correlations predominantly in frontal and temporal cortical
regions in late childhood and in older subjects (age range: 8.6--
29 years) (Shaw et al. 2006).
To corroborate and extend previous findings, in this study, we
first set out to examine relationships between FSIQ scores and
brain tissue compartments in healthy adults of average intelli-
gence quotient (IQ). Our primary goal, however, was to newly
explore the patterns of relationships between FSIQ and regional
changes in cortical thickness in young adults across a relatively
Cortical pattern--matching methods were employed to spatially
align homologous regions of cortex across individuals, allowing
relationships between FSIQ and regional variations in cortical
thickness to be examined at high spatial resolution across the
cortical mantle (Thompson et al. 2001; Narr, Bilder, et al. 2005;
Notably, because brain size and age may influence the relation-
ships between FSIQ and cortical gray matter (Gur et al. 1999;
Haier et al. 2005), both variables were included in statistical
analyses. The effects of sex on the slopes of FSIQ--cortical thick-
ness associations were explicitly examined.
Study participants included 30 male (mean age: 27.9 ± 7.1 years) and 35
female (mean age: 28.5 ± 7.5 years) healthy individuals. Subjects
participated as healthy volunteers for a larger study aimed at examining
structural neuropathology in schizophrenia (Narr, Bilder, et al. 2005;
Narr, Toga, et al. 2005) and were recruited through local newspaper
advertisements and community word of mouth. All participants were
determined to have no history of psychiatric illness as assessed by
clinical interview using the Structured Clinical Interview for Axis I DSM-
IV Disorders, Non-patient Edition (SCID-NP). Study exclusion criteria
included serious neurological or endocrine disorders, any medical
condition or treatment known to affect the brain, or meeting Diagnostic
and Statistical Manual of Mental Disorders IV (DSM-IV) criteria for
mental retardation. The North Shore—Long Island Jewish Health System
Institutional Review Board (IRB) approved all procedures, and informed
written consent was obtained from all subjects. Additional approval for
image processing and analysis was received from the University of
California, Los Angeles (UCLA) IRB.
To assess general intellectual ability, we employed the Wechsler adult
intelligence scale (Wechsler 1981). This test, which demonstrates high
reliability and validity, quantifies intelligence according to age-based
norms that have been shown to correlate with academic and life
success, measures of work performance, and occupational level (Jensen
1998). FSIQ is a composite score obtained from 11 subtests in verbal
and performance categories where the FSIQ is standardized in a US
population sample to have a mean of approximately 100 and standard
deviation of approximately 15. In this investigation, FSIQ scores ranged
between 74 and 139 (mean FSIQ = 100.2 ± 11.7 for male subjects and
mean FSIQ = 100.0 ± 13.6 for female subjects).
Image Acquisition and Preprocessing
High-resolution 3-dimensional (3D) spoiled gradient recalled magnetic
resonance images were obtained on a 1.5-T scanner (General Electric,
(256 3 256 matrix, 0.86 mm 3 0.86 mm in-plane resolution). Image
volumes passed through a number of preprocessing steps that included
standard position of the ICBM-305 average brain (Mazziotta et al. 1995)
2002; Narr, Bilder, et al. 2005), removal of nonbrain tissue and the
cerebellum (interrater reliability for scalp editing procedures, rI= 0.99),
correction of intensity nonuniformity due to magnetic field inhomoge-
neities (Zijdenbos and Dawant 1994; Sled and Pike 1998), and tissue
correction method (Shattuck et al. 2001). The cortical surfaces of each
hemisphere, comprising of 65 536 surface points, were then extracted
identified on each surface rendering using previously validated anatom-
ical delineation protocols (Ballmaier et al. 2004; Sowell et al. 2004).
Interrater reliability estimates demonstrated less than a 2 mm root mean
on 6 test brains compared with a gold standard arrived at by a consensus
of raters (Narr, Bilder, et al. 2005; Narr, Toga, et al. 2005).
Cortical Pattern Matching
Previously detailed cortical pattern--matching methods were used to
cortical thickness to be estimated at spatially equivalent hemispheric
locations across individuals (Thompson et al. 2001; Sowell et al. 2004;
Narr, Bilder, et al. 2005; Narr, Toga, et al. 2005). Briefly, for matching
procedures, a surface-warping algorithm uses the manually derived
sulcal/gyral landmarks as anchors to compute a 3D vector deformation
field that records the amount of x, y, and z coordinate shift (or
deformation) required to match the same cortical surface locations in
each subject with reference to the average anatomical pattern of the
entire study group. These methods reparameterize (regrid) the hemi-
spheric surface without imposing any scaling, so that the same anatomy
Because the cortical pattern--matching algorithms spatially relate
homologous cortical surface locations between individuals, anatomi-
cally comparable measures of cortical thickness may be obtained and
FSIQ--Cortical Thickness Relationships
Narr et al.
by guest on June 12, 2013
compared at different regions across the cortex from each subject.
Cortical thickness was measured by referencing tissue-classified brain
volumes using an implementation of the 3D Eikonal equation (Sapiro
2001). The thickness of the cortex was defined as the shortest distance
in 3D, without crossing CSF voxels, from the cortical white--gray matter
boundary to the outer gray--CSF hemispheric surface at all hemispheric
surface points (Sowell et al. 2004; Thompson et al. 2004; Narr, Bilder,
et al. 2005; Narr, Toga, et al. 2005). Notably, when gray matter surfaces
are adjacent, the algorithm progressively codes distance values until
meeting a voxel already assigned a distance code. Thus, although op-
posing gray matter banks not separated by CSF may receive the same
thickness value, a spatial filter of radius 8 mm was applied to the coded
distance values so that no practical difference could result from
misattributing the thickness to one side of the sulcus or the other.
For descriptive purposes, analyses of variance were employed to
characterize sex and age effects for intracranial tissue volumes. Linear
regression analyses were used to examine relationships between FSIQ
and overall intracranial volumes, intracranial gray matter, white matter,
and CSF while removing the variance associated with sex and age from
the data. For each tissue compartment (intracranial gray matter, white
matter, and CSF), analyses were additionally performed after including
overall intracranial volume as a nuisance variable.
The same statistical model above was used to examine relationships
between FSIQ and regional variations in cortical thickness using the
statistical package R (http://www.r-project.org/), again with and with-
showed that the regional specificity of FSIQ--gray matter relationships
differs inmen and women (Haieret al. 2005), differences in the slopes of
these relationships were examined using sex as the independent
variable. To follow up the presence of interaction effects, FSIQ--cortical
thickness relationships were mapped within male and female groups
Notably, for the main effects examined here (i.e., mapping regional
relationships between FSIQ and cortical thickness), testing the null
hypothesis that the slope (b) is zero is equivalent to testing the null
hypothesis that the correlation coefficient (r) is zero (Swinscow and
Campbell 2001). That is, there is a one-to-one mathematical mapping
between the P value for a nonnegative slope and the r value that
quantifies the correlation. For within-sex mapping of FSIQ--cortical
thickness relationships, regional partial correlation coefficients (r
values) and the corresponding probability maps (P values) are both
shown. In the presence of interactions (e.g., the differential effect
between males and females on the relationship between FSIQ and
cortical thickness), this mathematical equivalence no longer holds so
that 2 distinct hypotheses can potentially be addressed. The first of
these hypotheses is that the slopes of the relationship between
that is readily addressed in the context of the linear regression model
described above. The second hypothesis is that the magnitude of the
correlations (i.e., the proximity of the points to a straight line) differs
between the 2 groups. Although conceptually straightforward, this
second hypothesis is computationally problematic because the corre-
lations at each of the 65 536 points analyzed are correlated with one
another. This invalidates the assumptions of the Fisher z-test that would
otherwise be used for testing for differences in correlations statistically.
Although alternative statistical methodologies that take into accountthe
fact that the correlations are correlated exist (Olkin and Finn 1990;
Bilker et al. 2004), they are only suitable for circumstances where the
number of such correlated correlations is small and are rarely used. In
the current context, these methods are computationally impractical.
Therefore, rather than additionally addressing the issue of whether the
magnitude of correlations differ by sex using formal statistical signifi-
cance testing, we have provided in Figure 3 (bottom left) maps that
directly show the differences in the correlations.
For all regional analyses of FSIQ--cortical thickness relationships,
statistical mapping results were projected onto the 3D group-averaged
hemispheric surface models where significant results are indexed in
color. An uncorrected 2-tailed alpha level of P <0.05 was determined as
the threshold for interpreting statistical mapping results. However,
because regression analyses were performed at thousands of homolo-
gous cortical surface coordinate points, permutation testing was used to
test the overall significance of regional FSIQ--cortical thickness relation-
ships and to confirm regional sex interaction effects. Because we
predicted a priori that FSIQ--cortical thickness relationships would be
present primarily in frontal and temporal neocortical regions, as based
on prior results using different morphometric measures, permutation
testing was performed only within these regions, whereas other
regional effects were treated as exploratory. Frontal and temporal
regions of interest were constructed using an average anatomical atlas.
The number of cortical points showing significant FSIQ--cortical
thickness relationships at a statistical threshold of P < 0.01 using the
reduced model (i.e., controlling for sex and age, or sex, age, and
intracranial volume) was then compared with the number of significant
surface points that occurred by chance within each region of interest in
a thousand new randomized analyses. Similarly, residuals from the
reduced model were used to permute sex interaction effects for the
slopes of FSIQ--cortical thickness relationships while controlling for
the other terms in the model (Anderson and Legendre 1999; Anderson
and Braak 2003).
Intracranial Tissue Volumes
The means and standard deviations for intracranial tissue
of sex (males larger) and age (decreases with age) for intra-
cranial gray (F1,64= 21.19, P < 0.001 and F1,64= 14.20, P <
0.001, respectively) and white matter volumes (F1,64= 17.21,
P <0.001 and F1,64= 4.74, P <0.03). Effects of sex (males larger),
but not age, were observed for intracranial CSF volumes (F1,64=
8.28, P < 0.005) and overall intracranial volumes (F1,64= 29.32,
P <0.001). After correcting for overall intracranial volume, sex
effects were absent for all intracranial tissue compartments,
although significant age effects were observed for intracranial
P <0.001), and CSF (F1,64= 3.95, P <0.05).
Regression analyses showed significant positive correlations
between FSIQ and overall brain volume, r = 0.36, degrees of
61, P <0.004; and intracranial white matter volume, r = 0.26, df =
r = 0.12, df = 61, P > 0.05. As expected, after removing the
variance associated with brain size, correlations between FSIQ
and intracranial gray matter, r = 0.09, df = 60, P >0.05, and white
matter, r = 0.02, df = 60, P > 0.05, were no longer significant.
between overall brain size and all brain tissue compartments
(intracranial gray matter: r = 0.87, P <0.0001; intracranial white
matter: r = 0.78, P < 0.0001; and intracranial CSF: r = 0.55, p<
0.0001). Because overall brain size and intracranial tissue
Means and standard deviations of demographic variables and intracranial tissue volumes
Females (n 5 35),
Males (n 5 30),
Percent dextral (%)
Intracranial volume (cm3)
Intracranial gray matter (cm3)
Intracranial white matter (cm3)
Intracranial CSF (cm3)
aHandedness information was not available for 4 male subjects.
Cerebral Cortex September 2007, V 17 N 9 2165
by guest on June 12, 2013
compartments are highly correlated, in order to determine the
stepwise regression analyses were performed. After modeling
the combinations of terms shown to significantly associate with
FSIQ (overall intracranial volume, intracranial white, and gray
matter), the greatest amount of variance was explained by
including only intracranial gray matter in the model (r = 0.33,
serve to improve the model using a liberal exclusion criteria of
The slopes of FSIQ--tissue volume relationships were not
shown to differ by sex (gray matter: F1,64= 0.74, P >0.05; white
matter: F1,64= 1.67, P > 0.05; and CSF: F1,64= 0.08, P > 0.05).
Regression plots in Figure 1 show associations between FSIQ
scores and intracranial volumes after partialling out sex and age.
Statistical Mapping Results
Statistical maps in the left panel of Figure 2 show significant
correlations between FSIQ and regional variations in cortical
thickness after removing the partial effects of sex and age. The
right panel (shaded) shows results after additionally including
overall intracranial volume as a nuisance variable. Probability
values associated with significant positive or negative partial
correlations are indexed in color. Significant positive associa-
tions between FSIQ and cortical thickness were evident in
prefrontal (anterior--ventral prefrontal and frontopolar cortices;
BA 10/11 and 47) and temporal cortices (inferior temporal,
fusiform, and parahippocampal cortices; BA 20, 37, and 36)
bilaterally. Associations were more spatially diffuse in the right
hemisphere including anterior middle temporal (BA 21) and
extrastriate occipital(BA 19) corticalregions. Onlysmall and very
localized associations were observed within cingulate cortices.
Results were more spatially concentrated and less pronounced
after controlling for overall brain volume (Fig. 2, right panel),
although significant correlations were identified within the same
regions. No brain region exhibited significant negative correla-
tions between FSIQ and cortical thickness. Permutation testing
confirmed the significance of FSIQ--cortical thickness relation-
partial effects of sexand age (correctedP values: P <0.03 and P <
0.02,respectively). However, permutationtestingdidnotconfirm
regional results when intracranial volume was additionally in-
cluded in the model (corrected P values: P < 0.09 and P < 0.11).
Figure 3 (top left) shows statistical mapping results compar-
ing the slopes of FSIQ--cortical thickness relationships between
males and females after removing the partial effects of age. Sex
effects were observed within frontopolar (rostral superior and
middle frontal gyral regions; BA 10/11), caudal middle temporal
Figure1. Partialregression plots, controllingfor sex and age,showing relationships betweenFSIQ andoverall intracranial volumes(cm3) (top left), intracranialgray matter volumes
(cm3) (top right), intracranial White volumes (cm3) (bottom left), and intracranial CSF volumes (cm3) (bottom right).
FSIQ--Cortical Thickness Relationships
Narr et al.
by guest on June 12, 2013
(BA 21), occipitotemporal (BA 37), and inferior parietal (BA 7)
gyral regions. Statistical mapping results additionally including
overall intracranial volume in the model are not shown,
although the regional pattern of effects were extremely similar
to those observed before intracranial size correction. Permuta-
tion testing confirmed the significance of interaction effects
within frontal and temporal regions both with and without
partialling out overall intracranial volume (corrected P values:
P < 0.04 and P < 0.02 before and P < 0.03 and P < 0.04 after
intracranial volume correction for frontal and temporal regions,
To follow up interaction effects, FSIQ--cortical thickness
associations were mapped within females and males separately
after removing the partial effects of age (Fig. 3, top right).
Statistical maps revealed highly significant FSIQ--cortical thick-
ness relationships in females within frontopolar and bordering
prefrontal regions (BA 10/11 and 47) and within inferior
temporal (BA 20), lateral occipitotemporal, and fusiform (BA
37 and 36) gyral regions. In male subjects, localized correlations
were less pronounced in temporal cortices and predominantly
observed in medial temporo-occipital BA 37 (bordering BA 19
and 30), whereas significant correlations in prefrontal cortical
regions appeared absent. Negative correlations between FSIQ
and cortical thickness were observed in a discrete area of the
left inferior parietal cortex (BA 7) in male subjects. The spatial
patterns of regional significance were similar within each sex
when intracranial volume was additionally included as a nui-
sance variable (results not shown).
Maps showing the magnitudes of correlation coefficients
within female and male subjects are shown in the bottom right
panels of Figure 3 that correspond to the female and male
probability maps shown directly above. Sex differences in the
magnitude of correlations between FSIQ and cortical thickness
were assessed by subtracting regional correlation coefficients
in male subjects from correlation coefficients in females at
homologous cortical locations (Fig. 3, bottom left). Hot colors
indicate regions where females show larger partial correlation
coefficients compared with males, and cool colors reflect
regions where males show larger correlations with respect to
magnitudes, as opposed to the slopes, of relationships between
FSIQ and cortical thickness, regional difference profiles were
similar in spatial pattern to interaction effects (top left above).
Correlations were greater in magnitude within frontopolar (BA
10/11), caudal middle temporal (BA 21), occipitotemporal (BA
37), and inferior parietal (BA 7) gyral regions in females
compared with males. In males, correlations appeared larger
only in spatially discrete medial temporo-occipital regions.
Our data show that larger brain volumes are associated with
greater intellectual ability, in line with established findings
(McDaniel 2005). As further confirmed, positive relationships
extend to intracranial gray matter (Andreasen et al. 1993; Reiss
et al. 1996; Gur et al. 1999; Haier et al. 2004) and white matter
tissue compartments (Gur et al. 1999; Haier et al. 2004),
whereas associations between FSIQ and intracranial CSF appear
minimal or absent (Andreasen et al. 1993). Overall intracranial
size and tissue compartment measures, however, are them-
selves correlated, and when overall intracranial size is taken into
account, relationships between FSIQ and intracranial gray or
white matter are below the threshold of significance. The
genetic factors contributing to brain size, which is known to
Figure 2. Statistical maps show the regional significance of relationships between FSIQ and cortical thickness after removing the partial effects of sex and age (left) and after
additionally removing the partial effects of overall intracranial volume (right, shaded). Positive probability values shown in hot colors (red, pink) on the color bar encode the
significance of positive relationships. Negative probabilities in cold colors (blues) on the color bar encode the significance of negative relationships, although negative relationships
between FSIQ and cortical thickness were not observed in any region in the above analyses.
Cerebral Cortex September 2007, V 17 N 9 2167
by guest on June 12, 2013
be highly heritable (Posthuma et al. 2002), overlap with the
genetic factors accounting for variations in gray and white
matter volumes (Baare et al. 2001). Therefore, genes expressed
in gray and white matter, which make up approximately 80--
90% of total brain volume in humans (Zhang and Sejnowski
2000), account to a large extent for variations observed in
overall intracranial size. Our post hoc stepwise regression
analyses indicate that when including overall intracranial, gray,
and white matter volumes in the model, gray matter is the best
predictor of variations in FSIQ scores. Thus, because gray matter
contains neurons, axons, and dendritic trees and spines that act
as the units of brain function and sites of information transfer,
respectively, relationships between intracranial volume and
FSIQ may reflect primary relationships with gray matter and
to a slightly lesser extent with white matter properties.
Functional imaging studies demonstrate that distributed brain
regions linked by cortical networks are activated during higher
cognitive functioning. Overlap of brain activity in some specific
cortical regions, particularly in prefrontal, parietal, and/or
temporo-occipital networks, has been shown to relate to
general intelligence (Duncan and Owen 2000; Gray et al.
2003; Haier et al. 2003; Lee et al. 2006). Although relationships
between intelligence and gray matter volume are well repli-
cated, fewer investigations have focused on localizing the
Figure 3. Statistical maps in the top left corner show significant interactions between males and females in the slopes of regional FSIQ--cortical thickness relationships (after
removing the partial effects of age from the data). Positive probability values (hot colors) indicate greater slopes in females with respect to males, whereas negative probability
values (cool colors) indicate greater slopes in males. Top center and top right statistical maps show the regional significance of FSIQ--cortical thickness relationships within female
and male groups separately. Positive probability values (hot colors) encode positive relationships. Negative probabilities (cool colors) encode negative relationships. Bottom center
and bottom left statistical maps show the magnitude of partial correlation coefficients between FSIQ and cortical thickness (after partialling out age effects) within females and
males separately, which correspond to the same sex probability maps shown directly above. Positive r values (hot colors) index positive FSIQ--cortical thickness relationships;
negative r values (cool colors) index negative FSIQ--cortical thickness relationships. Statistical maps in the bottom left show differences in the magnitude of correlations between
males and females where r values in males are subtracted from r values in females at homologous cortical locations. Positive values (hot colors) index larger correlations in females
with respect to males; negative values (cool colors) index larger correlations in males compared with females.
FSIQ--Cortical Thickness Relationships
Narr et al.
by guest on June 12, 2013
structural homologues of general intelligence. In spite of
discrepancies in findings that may relate to differences in image
analysis methods and the demographic characteristics of
subjects studied, gray matter volume or density increases in
prefrontal (Flashman et al. 1997; Frangou et al. 2004; Haier et al.
2004), medial and lateral temporal (Andreasen et al. 1993;
Flashman et al. 1997; Haier et al. 2004), and in parietal (Haier
et al. 2004) cortical regions appear most linked with improved
performance on standardized intelligence tests in adolescent
and adult samples. To date, only one published study has ex-
amined relationships between FSIQ and regional variations in
cortical thickness specifically (Shaw et al. 2006). This investiga-
tion, focused on identifying cross-sectional and longitudinal
changes in FSIQ--cortical thickness relationships across devel-
opment, reported positive correlations between intelligence in
cortical thickness after the age of 8 years, peaking in late
childhood, that remained evident in prefrontal and temporal
cortical regions in subjects within the adolescent--early adult
age range. This prior investigation also showed, via longitudinal
assessment, that more intelligent children demonstrate more
dynamic changes in cortical thickness over development par-
ticularly in prefrontal cortices.
The cortex consists of layers of cells, also organized into
columns that vary in cortical depth (~1 to 4.5 mm) as dependent
size and density. The cytoarchitectural characteristics defining
and Parker 1929) and may represent a more regionally relevant
survey of structural integrity than gross measures of gray matter
volume. Because laminar thickness varies across the cortical
mantle, associations can be compared across individuals only at
matching methods to align anatomy across subjects, we found
significant relationships between FSIQ and cortical thickness
37 and 36) cortical regions where findings remained present
irrespective of brain size corrections. These results largely agree
with a previous study using VBM methods to correlate in-
telligence with gray matter density throughout the brain in
a smaller, high-IQ adult sample (Haier et al. 2004). Our findings
are also consistent with the cross-sectional results of Shaw et al.
(2006) reporting significant correlations between greater in-
telligence and increased cortical thickness predominantly fron-
tal and temporal cortices within the late adolescent--early
adulthood age group studied.
Specific links between FSIQ and cortical thickness within
prefrontal BA areas 10 and 11 (and bordering association area
47), spanning the rostral portions of the inferior and superior
frontal gyri extending medially, are not surprising given that
these regions are sometimes collectively referred to as the
center of biological intelligence and are involved in higher
cognitive functions like problem solving, planning, sequencing,
reasoning, and judgment. In the temporal lobe, cortical associ-
ation area 37, spanning portions of the inferior temporal gyrus
laterally and fusiform gyrus medially, participates in the analysis
of visual form, motion, and the representation of objects. Obser-
vations of specific links with FSIQ in this region and within
proximal visual association area 36 suggest variations in laminar
thickness influence visual analysis abilities that are central to
most cognitive processes, especially those assessed with stan-
dardized intelligence testing. Interestingly a positron emission
tomographic study showed that brain activations associated
with a nonreasoning condition (passively viewing videos) were
greater within temporo-occipital regions (BA 37/19) in subjects
with higher scores on a g-loaded test and that connectivity was
increased between this region and prefrontal regions (Haier
et al. 2003). Furthermore, a functional magnetic resonance
imaging study showed positive correlations between FSIQ and
verb-generation task-associated brain activity in BA 19 as well as
in frontal and temporal areas in children (Schmithorst and
As universally reported in the literature, men exhibited larger
mean brain volumes than women. Because sex accounts for
a large amount of the variance in brain size, after controlling for
intracranial volume, sex effects for gray and white tissue
volumes were no longer significant. Relationships between
men and women. That is, although women possess smaller mean
intracranial volumes and tissue contents, the slopes of FSIQ
relationships were approximately parallel in both sexes. Al-
though FSIQ represents abilities across several cognitive do-
by degree to general intelligence (g) by factor analysis (Neisser
et al. 1996). The few studies relating separate intelligence
subtests or factors to brain structure have shown little consis-
tency, especially with regard to sex differences. For example,
female-biased correlations were observed between brain size
and verbal IQ, whereas performance IQ showed male-biased
correlations (Andreasen et al. 1993). Another study, however,
showed relationships between performance IQ, but not verbal
IQ, in frontal, temporal, and parietal regions but failed to detect
sex differences (Flashman et al. 1997). In contrast, postmortem
data revealed relationships between verbal IQ and cerebral
volume in women and in right-handed men, and exclusively in
investigated whether crystallized (abilities dependent on ac-
quired knowledge) or fluid intelligence (reasoning and problem
solving), which is occasionally isolated as a separate factor, link
more closely with structural brain variations with the evidence
coming down on the side of fluid intelligence (Gray and
Thompson 2004). Discrepancies in these results may stem
from studying broad-based intelligence measures (e.g., verbal
ing to how closely they relate to general intelligence or g.
Notably, investigators have recently shown that individual
subtests with greater g loadings (i.e., tests that indicate greater
average correlations with other tests in an intelligence battery)
are associated with greater gray matter volumes throughout the
brain and,for particular tests, with increased gray matterdensity
in frontal, temporal, and parietal regions (Colom et al. 2006).
Although the slopes of the relationships between intracranial
in our investigation, the slopes of FSIQ--cortical thickness
relationships were significantly different within prefrontal and
posterior temporal neocortical regions and to a lesser extent in
of regional correlation coefficients were observed within the
same regions (Fig. 3, bottom left). When FSIQ--cortical thickness
relationships were examined separately within each sex, only
females showed significant relationships between FSIQ and
cortical thickness in prefrontal cortices. Correlations were also
more spatially diffuse in temporal cortices in females; males
Cerebral Cortex September 2007, V 17 N 9 2169
by guest on June 12, 2013
showed significant correlations primarily in posterior medial
temporal cortices (BA 37) and bordering regions. These findings
contrast with VBM findings of strong FSIQ correlations in frontal
and parietal regions in male subjects. Female-specific associa-
tions, however, were observed in prefrontal cortices (BA 10) as
consistent with our results, but associations were not detected
in temporal association regions (Haier et al. 2005). Men and
women possess some regional differences in cortical thickness
distributions where sex differences have been shown to be
Thus, discrepancies in results may stem from differences in
spatial normalization approaches, smoothing, and the gray
matter indices (gray matter density vs. cortical thickness)
examined. Shaw et al. (2006) did not report significant sex
differences in cross-sectional examination of intelligence--
cortical thickness relationships, although sex differences were
not assessed exclusively within the adolescent--early adulthood
Although men and women are of similar intelligence, on
average, men are reportedly better at visuospatial tasks, fitting
with our observations of regionally specific FSIQ--cortical
thickness relationships in temporo-occipital association areas
that are important for the analysis of visual form (Kimura 1996).
Women, however, generally perform better in tasks requiring
verbal processing and memory involving frontal--temporal
cortical networks. Spatially distinct relationships between
FSIQ and cortical thickness in men and women may thus reflect
sexual dimorphisms in some processing abilities. Alternatively,
sex differences in regional relationships may reflect differences
in processing strategies that do not depend on competence or
reflect differences in neural circuitry and brain structural
organization. Interestingly, although performance was similar
across sexes, the patterns of brain activity associated with
a variety of cognitive tasks were shown to differ between men
and women (Bell et al. 2006). The focus of our study was to
examine only the moderating influences of sex on global and
regional relationships between brain tissue measures and
general intellectual ability, thereby also avoiding the potential
for inflated Type I error by excluding analyses of subtest scores
that are shown to correlate with general intellectual ability to
varying degrees. Future studies, however, may elucidate
whether specific information processes, especially those shown
to exhibit sex differences, are linked with variations in cortical
thickness. On a final note, because there are many factors, such
as differences in socio-economic and cultural backgrounds that
cannot be dissociated from racial--ethnic class, potential influ-
ences of race toward the relationships between brain size and
intelligence were not examined here.
Although the concept of general intelligence or ‘‘g’’ is disputed
by some, the common view remains that this trait represents
a major dimension of mental competence across a diverse range
of cognitive abilities, which unequivocally serves to predict
many aspects of life success. This study is the first to explore the
regional specificity of relationships between cortical thickness
and general intellectual ability in healthy adults of normal IQ
within a relatively narrow age range, as well as the moderating
effects of sex. Our results suggest that a common biological
substrate influences both general intelligence and intracranial
gray and white matter volumes, where variation in the thickness
of prefrontal and temporal association cortices is particularly
relevant to intellectual ability. Variations in intracranial tissue
characteristics, particularly in prefrontal regions (Thompson
et al. 2001), may thus prove useful as endophenotypes that may
help to determine the overlapping genetic factors accounting
for general intelligence. However, sex is shown to influence the
regional relationships between general intelligence and cortical
thickness. Women show significant associations in prefrontal
and temporal association cortices, whereas men show associa-
tions primarily in temporal--occipital association cortices. It
remains to be determined if sex-specific competences account
for these regional differences or vice versa and/or whether
regional differences in relationships between men and women
represent differences in processing strategies or underlying
This work was generously supported by research grants from the
National Center for Research Resources (P41 RR13642), the National
Institute of Mental Health (RO1 MH60374), the National Institutes of
Health (NIH) Roadmap Initiative (P20 RR020750), the National Library
of Medicine (R01 LM05639), the NIH Roadmap for Medical Research,
Grant U54 RR021813 entitled Center for Computational Biology,
a National Alliance for Research on Schizophrenia and Depression
Young Investigator Award and Career Development Award (K01
MH073990-01A1, to KLN). Algorithm development was also supported
by grants R21 EB01651, R21 RR019771, and AG016570 (to PT). Conflict
of Interest: None declared.
Address correspondence to Dr Katherine L. Narr, Laboratory of
Neuro Imaging, Department of Neurology, UCLA School of Medicine,
Neuroscience Research Building, Suite 225, 635 Charles E. Young Drive
South, Los Angeles, CA 90095-7334, USA. Email: firstname.lastname@example.org.
Anderson MJ, Braak CJ. 2003. Permutation tests for multi-factorial
analysis of variance. J Stat Comput Simul. 73:85--113.
Anderson MJ, Legendre P. 1999. An empirical comparison of permuta-
tion methods for tests of partial regression coefficients in a linear
model. J Stat Comput Simul. 62:271--303.
Andreasen NC, Flaum M, Swayze V 2nd, O’Leary DS, Alliger R, Cohen G,
Ehrhardt J, Yuh WT. 1993. Intelligence and brain structure in normal
individuals. Am J Psychiatry. 150:130--134.
Baare WF, Hulshoff Pol HE, Boomsma DI, Posthuma D, de Geus EJ,
Schnack HG, van Haren NE, van Oel CJ, Kahn RS. 2001. Quantitative
genetic modeling of variation in human brain morphology. Cereb
Ballmaier M, O’Brien JT, Burton EJ, Thompson PM, Rex DE, Narr KL,
McKeith IG, DeLuca H, Toga AW. 2004. Comparing gray matter loss
profiles between dementia with Lewy bodies and Alzheimer’s
Bell EC, Willson MC, Wilman AH, Dave S, Silverstone PH. 2006. Males and
females differ in brain activation during cognitive tasks. Neuroimage.
Bilker WB, Brensinger C, Gur RC. 2004. A two factor ANOVA-like test for
correlated correlations: CORANOVA. Multivar Behav Res. 39:
Bookstein FL. 2001. ‘‘voxel-based morphometry’’ should not be used
with imperfectly registered images. Neuroimage. 14:1454--1462.
Colom R, Jung RE, Haier RJ. 2006. Distributed brain sites for the g-factor
of intelligence. Neuroimage. 31(3):1359--1365.
Duncan J, Owen AM. 2000. Common regions of the human frontal lobe
recruited by diverse cognitive demands. Trends Neurosci. 23:
Economo C, Parker S. 1929. The cytoarchitectonics of the human
cerebral cortex. London: Humphrey Milford, Oxford University
FSIQ--Cortical Thickness Relationships
Narr et al.
by guest on June 12, 2013
Flashman LA, Andreasen NC, Flaum M, Swayze I, Victor W. 1997. Download full-text
Intelligence and regional brain volumes in normal controls. In-
Frangou S, Chitins X, Williams SC. 2004. Mapping IQ and gray matter
density in healthy young people. Neuroimage. 23:800--805.
Gray JR, Chabris CF, Braver TS. 2003. Neural mechanisms of general fluid
intelligence. Nat Neurosci. 6:316--322.
Gray JR, Thompson PM. 2004. Neurobiology of intelligence: science and
ethics. Nat Rev Neurosci. 5(6):471--482.
GurRC,Turetsky BI, MatsuiM, YanM, BilkerW, HughettP, GurRE. 1999.
Sex differences inbrain grayand whitematterinhealthyyoung adults:
correlations with cognitive performance. J Neurosci. 19:4065--4072.
Haier RJ, Jung RE, Yeo RA, Head K, Alkire MT. 2004. Structural brain
variation and general intelligence. Neuroimage. 23:425--433.
Haier RJ, Jung RE, Yeo RA, Head K, Alkire MT. 2005. The neuroanatomy
of general intelligence: sex matters. Neuroimage. 25:320--327.
Haier RJ, White NS, Alkire MT. 2003. Individual differences in general
intelligence correlatewith brain functionduringnonreasoning tasks.
Honea R, Crow TJ, Passingham D, Mackay CE. 2005. Regional deficits in
brain volume in schizophrenia: a meta-analysis of voxel-based
morphometry studies. Am J Psychiatry. 162:2233--2245.
Jensen AR. 1998. The g factor: the science of mental ability. Westport
Kimura D. 1996. Sex, sexual orientation and sex hormones influence
human cognitive function. Curr Opin Neurobiol. 6:259--263.
Lee KH, Choi YY, Gray JR, Cho SH, Chae JH, Lee S, Kim K. 2006. Neural
correlates of superior intelligence: stronger recruitment of posterior
parietal cortex. Neuroimage. 29(2):578--586.
Luders E, Narr KL, Thompson PM, Rex DE, Woods RP, Deluca H, Jancke
L, Toga AW. 2006. Gender effects on cortical thickness and the
influence of scaling. Hum Brain Mapp. 27:314--324.
MacDonald D, Avis D, Evans AC. 1994. Multiplesurface identification and
matching in magnetic resonance imaging. Proc Soc Photo Opt
Instrum Eng. 2359:160--169.
Mazziotta JC, Toga AW, Evans A, Fox P, Lancaster J. 1995. A probabilistic
atlas of the human brain: theory and rationale for its development
(The International Consortium for Brain Mapping; ICBM). Neuro-
McDaniel. 2005. Big-brained people are smarter: a meta-analysis of the
relationship between in vivo brain volume and intelligence. In-
Narr KL, Bilder RM, Toga AW, Woods RP, Rex DE, Szeszko PR, Robinson
D, Sevy S, Gunduz-Bruce H, Wang YP, et al. 2005. Mapping cortical
thickness and gray matter concentration in first episode schizophre-
nia. Cereb Cortex. 15:708--719.
Narr KL, Cannon TD, Woods RP, Thompson PM, Kim S, Asunction D, van
Erp TG, Poutanen VP, Huttunen M, Lonnqvist J, et al. 2002. Genetic
contributions to altered callosal morphology in schizophrenia. J
Narr KL, Toga AW, Szeszko P, Thompson PM, Woods RP, Robinson D,
Sevy S, Wang Y, Schrock K, Bilder RM. 2005. Cortical thinning in
cingulate and occipital cortices in first episode schizophrenia. Biol
Neisser U, Boodoo G, Bouchard TJ, Boykin AW, Brody N, Ceci SJ, Halpern
DF, Loehlin JC, Perloff R, Sternberg RJ, et al. 1996. Intelligence:
knowns and unknowns. Am Psychol. 51:77--101.
Olkin I, Finn J. 1990. Testing correlated correlations. Psychol Bull.
Parent A, Carpenter MB. 1995. Human neuroanatomy. Baltimore (MD):
Williams & Wilkins.
Posthuma D, De Geus EJ, Baare WF, Hulshoff Pol HE, Kahn RS, Boomsma
DI. 2002. The association between brain volume and intelligence is
of genetic origin. Nat Neurosci. 5:83--84.
Reiss AL, Abrams MT, Singer HS, Ross JL, Denckla MB. 1996. Brain
development, gender and IQ inchildren. Avolumetricimagingstudy.
Brain. 119 (5):1763--1774.
Sapiro G. 2001. Geometric partial differential equations and image
analysis. Cambridge (UK): Cambridge University Press.
Schmithorst VJ, Holland SK. 2006. Functional MRI evidence for disparate
developmental processes underlying intelligence in boys and girls.
Shattuck DW, Sandor-Leahy SR, Schaper KA, Rottenberg DA, Leahy RM.
2001. Magnetic resonance image tissue classification using a partial
volume model. Neuroimage. 13:856--876.
Shaw P, Greenstein D, Lerch J, Clasen L, Lenroot R, Gogtay N, Evans A,
Rapoport J, Giedd J. 2006. Intellectual ability and cortical develop-
ment in children and adolescents. Nature. 440:676--679.
Sled JG, Pike GB. 1998. Standing-wave and rf penetration artifacts caused
by elliptic geometry: an electrodynamic analysis of MRI. IEEE Trans
Med Imaging. 17:653--662.
Sowell ER, Thompson PM, Leonard CM, Welcome SE, Kan E, Toga AW.
2004. Longitudinal mapping of cortical thickness and brain growth
in normal children. J Neurosci. 24:8223--8231.
Spearman C. 1904. General intelligence objectively determined and
measured. Am J Psychol. 15:201--293.
Swinscow TDV, Campbell MJ. 2001. Statistics at square one. 10th ed.
London: BMJ Books.
Thacker NA. 2005. A critical analysis of voxel-based morphometry.
Technical Report. http://www.tina-vision.net/docs/memos/2003-
011.pdf, Tina Memo No 2003-011. p. 1--10.
Thompson PM, Cannon TD, Narr KL, van Erp T, Poutanen VP, Huttunen
M, Lonnqvist J, Standertskjold-Nordenstam CG, Kaprio J, Khaledy M,
et al. 2001. Genetic influences on brain structure. Nat Neurosci.
Thompson PM, Hayashi KM, Sowell ER, Gogtay N, Giedd JN, Rapoport JL,
de Zubicaray GI, Janke AL, Rose SE, Semple J, et al. 2004. Mapping
cortical change in Alzheimer’s disease, brain development, and
schizophrenia. Neuroimage. 23(Suppl 1):S2--S18.
Toga AW, Thompson PM, 2005. Genetics of brain structure and
intelligence. Annu Rev Neurosci. 28:1--23.
Tramo MJ, Loftus WC, Stukel TA, Green RL, Weaver JB, Gazzaniga MS.
1998. Brain size, head size, and intelligence quotient in monozygotic
twins. Neurology. 50:1246--1252.
Wechsler D. 1981. Wechsler adult intelligence scale, revised: manual.
New York: Harcourt Brace Jovanovich.
Wilke M, Sohn JH, Byars AW, Holland SK. 2003. Bright spots: correlations
of gray matter volume with IQ in a normal pediatric population.
Witelson SF, Beresh H, Kigar DL. 2006. Intelligence and brain size in 100
postmortem brains: sex, lateralization and age factors. Brain.
Zhang K, Sejnowski TJ. 2000. A universal scaling law between gray
matter and white matter of cerebral cortex. Proc Natl Acad Sci USA.
Zijdenbos AP, Dawant BM. 1994. Brain segmentation and white
matter lesion detection in mr images. Crit Rev Biomed Eng. 22:
Cerebral Cortex September 2007, V 17 N 9 2171
by guest on June 12, 2013