© 2006 Nature Publishing Group
Intellectual ability and cortical development in
children and adolescents
P. Shaw1, D. Greenstein1, J. Lerch2, L. Clasen1, R. Lenroot1, N. Gogtay1, A. Evans2, J. Rapoport1& J. Giedd1
Children who are adept at any one of the three academic ‘R’s
(reading, writing and arithmetic) tend to be good at the others,
and grow into adults who are similarly skilled at diverse intellec-
tually demanding activities1–3. Determining the neuroanatomical
correlates of this relatively stable individual trait of general
intelligence has proved difficult, particularly in the rapidly devel-
oping brains of children and adolescents. Here we demonstrate
rather than cortical thickness itself, is most closely related to level
of intelligence. Using a longitudinal design, we find a marked
developmental shift from a predominantly negative correlation
between intelligence and cortical thickness in early childhood to a
positive correlation in late childhood and beyond. Additionally,
level of intelligence is associated with the trajectory of cortical
development, primarily in frontal regions implicated in the
maturation of intelligent activity4,5. More intelligent children
demonstrate a particularly plastic cortex, with an initial acceler-
ated and prolonged phase of cortical increase, which yields to
indicates that the neuroanatomical expression of intelligence in
children is dynamic.
Structural neuroimaging studies generally report a modest corre-
lation (r ¼ 0.3) between psychometric measures of intelligence and
total brain volume6. Links between intelligence and specific regions
of the brain may vary according to developmental stage: the anterior
cingulate in children7, the orbitofrontal and medial prefrontal cortex
in adolescents8, and the lateral prefrontal cortex in older adults9.
Most previous studies infer developmental processes from purely
cross-sectional data, an endeavour fraught with methodological
complications10. Only one longitudinal study has linked cortical
who gained more in a measure of verbal intelligence5. However, this
study was limited by its small sample size (n ¼ 45), narrow age
range (5–11yr), and consideration of only linear cortical change,
whereas brain development generally follows more complex growth
in a large group of typically developing subjects (n ¼ 307), the
majorityof who had prospectively acquired repeated neuroanatomic
scans (see the Methods). Subjects were stratified on the basis of
Wechsler intelligence scales, which give a standardized ‘intelligence
quotient’ (IQ) based on subtests assessing verbal and non-verbal
knowledge and reasoning12. We examined the thickness of the cortex
throughout the entire cerebrum, as it is a sensitive index of normal
brain development5,13, using a fully automated technique, and have
validated these measurements by expert manual determination of
cortical thickness and population simulations14,15. We reasoned that
the trajectory of cortical development in children stratified on the
basis of IQ would differ primarily in the prefrontal cortex, which has
both structural and functional correlations with intelligence. The
approved the research protocol, and written informed consent and
assent were obtained from parents and children, respectively.
We estimated Pearson’s correlations between IQ and cortical
thickness for all subjects (each subject contributing one scan), and
found modest positive correlations throughout most of the frontal,
parietal and occipital cortex, and similarly modest negative corre-
lations in the anterior temporal cortex (Fig. 1 and Supplementary
Table 1). Throughout most of the cerebral cortex, the correlations
were not significant at an unadjusted P , 0.05.
Dividing the sample into different age groups, however, revealed
notable age-related changes. A predominantly negative correlation
between IQ and cortical thickness in the early childhood group
contrasted with later positive correlations, which peaked in late
childhood, but were present in an attenuated form in the adolescent
and early adult groups. The change in the valence of the correlation
between IQ and cortical thickness was significant between the young
and late childhood groups throughout the prefrontal cortex, and the
Figure 1 | Correlations between IQ and cortical thickness. a, Pearson’s
correlations for all 307 subjects were generally positive and modest
(P . 0.05), with r between 0 and 0.10 (green/yellow), except in the anterior
temporal cortex (which showed a negative correlation, with r between 0 and
20.1; blue/purple). b, Correlations in different age groups showed that
negative correlations were present in the youngest group, indicating that
higher IQ was associated with a thinner cortex particularly in frontal and
temporal regions. The relationship reverses in late childhood, with most of
the cerebral cortex correlating positively with IQ.
1Child Psychiatry Branch, National Institute of Mental Health, Bethesda, Maryland 20182, USA.2Montreal Neurological Institute, McGill University, Montreal, Quebec H3A 2B4,
Vol 440|30 March 2006|doi:10.1038/nature04513
© 2006 Nature Publishing Group
left superior/middle temporal gyri. These age groups did not differ
in gender composition (x2¼ 2.76; P ¼ 0.62) or mean IQ
(F3,303¼ 1.58; P ¼ 0.19), and there was no significant gender
difference in the correlation between cortical thickness and IQ.
We further characterized the development of the relationship
between intelligence and cortical morphology using linear mixed-
models, which allowed inclusion of all 629 scans. In the determi-
nation of cortical thickness, there was a significant interaction
between IQ and age terms in the prefrontal cortex, suggesting that
the relationship between cortical thickness and IQ varies with age
(specifically cubic and quadratic age terms; see the Supplementary
To explore this interaction, the sample was split into three IQ
cortical points showing differences in cortical development between
the intelligence groups lay bilaterally within the superior frontal gyri
extending into the medial prefrontal cortex, and to a lesser extent
in the middle and orbitofrontal cortices (Fig. 2). In each of these
clusters, the trajectories for the local point of maximum trajectory
difference and for the entire cluster were similar: the superior
intelligence group started from a relatively thinner cortex, but then
showed a marked increase in cortical thickness peaking at ,11yr. In
contrast, the average intelligence group showed either a steady
decline in cortical thickness throughout the age period covered (in
orbitofrontal areas), or a short initial increase in cortical thickness
the high intelligence group followed an intermediate pattern, more
strongly resembling the pattern of the average intelligence group,
with no significant differences between these two groups in the
clusters shown in Fig. 2 (all P . 0.10).
Different developmental trajectories were also prominent in the
posterior left hemisphere between the superior and average intelli-
gence groups, specifically within the left middle prefrontal and
inferior temporal gyri, and to a lesser extent the angular gyrus. The
right hemisphere outside the frontal lobes showed trajectories of
cortical development that did not differ significantly between groups.
An overall decline in cortical thickness was noted in all groups,
present either throughout the age period covered (average intelli-
adolescence (superior intelligence). Velocity curves derived using a
first-order differential of the fitted cubic growth curvesillustrate that
the superior intelligence group had the most rapid rate of cortical
thinning, whereas the high and average intelligence groups had
similar, but slower, rates (Fig. 3). Thus, the relatively rapid increase
in cortical thickness in the superior intelligence group was followed
by a more rapid thinning.
Figure 2 | Trajectories of cortical change. The brain maps (centre panel)
showprominentclusterswherethe superior andaverageintelligencegroups
differ significantly in the trajectories of cortical development (t-statistic
maps show areas of significant interaction between these IQ groups and the
cubic age term). a, Graph showing the trajectories at the cortical point of
maximum trajectory difference in the right superior frontal gyrus (point
indicated in upper brain map). b–d, Graphs showing the trajectories of the
mean thickness of all cortical points in the other clusters. The graph in d
relates to the area indicated in the lower brain map. The age of peak cortical
thickness is arrowed and significance values of differences in shapes of
trajectories are given on the graphs. MNI, Montreal Neurological Institute.
Figure 3 | Rate of change in cortical thickness. The rate of change for the
cluster of cortical points in the right superior and medial frontal gyrus,
which showed a significant trajectory difference. Positive values indicate
increasing cortical thickness, negative values indicate cortical thinning. The
point of intersection on the x axis represents the age of maximum cortical
thickness (5.6yr for average, 8.5yr for high, and 11.2yr for the superior
NATURE|Vol 440|30 March 2006
© 2006 Nature Publishing Group
To illustrate the development of differences in cortical thickness
representing group differences in the height of the developmental
curves at each age were estimated from 7–16yr (Fig. 4). Initially, the
By 11yr, regions of thicker cortex became apparent in the superior
and middle frontal gyri, spreading to involve more posterior regions
of the right prefrontal cortex and the left superior and middle frontal
gyri. By late adolescence, the accelerated rate of cortical loss in the
most intelligent group leads to decreased regional differences.
The intelligence groups did not differ significantly in handed-
ness or gender composition, but did in socio–economic status
(F2;291¼ 14:1; P , 0.001), which was correlated with IQ
(r ¼ 20.35; P , 0.01). In the frontal clusters, where trajectories
were most closely tied to intelligence, none of these variables
contributed significantly to the final polynomial regression model
(all P values .0.1).
Thus, we have demonstrated that level of intelligence is related to
differing trajectories of cortical change are most prominent in the
prefrontal cortex, congruent with functional magnetic resonance
imaging (fMRI) studies showing that activation of the lateral
prefrontal cortex is common to a range of intelligence tests, and
that the magnitude of frontal cortical activation correlates highly
Our longitudinal structural MRI images provide adequate resolu-
tion to describe an in vivo change in cortical thickness, but the nature
of the underlying cellular events is largely unknown. Adeterminant of
cerebral lamination in utero and perinatally is the emergence and
afferents and their synapses18,19. Proliferation of myelin into the
peripheral cortical neuropil in childhood and adolescence is another
formation and usage-dependent selective elimination of synapses21,
which help to create and sculpt neural circuitry including those
supporting cognitive abilities22, may contribute to changing cortical
dimensions. The prefrontal cortex shows relatively late structural11
and metabolic23maturation, and the prolonged phase of prefrontal
cortical gain in the most intelligent might afford an even more
extended ‘critical’ period for the development of high-level cognitive
less grey matter at any one age. Rather, intelligence is related to
dynamic properties of cortical maturation.
Subjects. Three hundred and seven unrelated children and adolescents with no
(Supplementary Table 2). All subjects had age-appropriate versions of the
Weschler intelligence scales. In 220 subjects, full-scale IQ was estimated from
four subtests (vocabulary, similarities, block design and matrix reasoning), and
in 87 children two subtests were used (vocabulary and block design). For
longitudinal analyses, subjects were divided into three groups on the basis of
full-scale IQ with the primary constraint of attaining a roughly equal numberof
total scans in each group. The groups were: superior intelligence (IQ range 121–
149), high intelligence (IQ range 109–120) and average intelligence (IQ range
83–108). All subjects were scanned at least once; 178 participants (58%) had at
least two scans; 92 (30%) had three or more scans; the mean interscan interval
Neuroimaging. T1-weighted magnetic resonance images (1.5mm axial and
echo in the steady state on a 1.5-T Signa scanner (General Electric), were
registered into standardized space24and corrected for non-uniformity arte-
facts25. The inner and outer cortical surfaces were extracted from tissue-
segmented images using deformable models, and non-linearly aligned towards
a standard template surface26. Cortical thickness was measured in native space
millimetres using the linked distance between the pial white and grey matter
surfaces at 40,962 vertices throughout the cerebral cortex27(see Supplementary
Methods). In order to improve the ability to detect population changes, each
which respects anatomical boundaries and was chosen to maximize statistical
power while minimizing false positives15.
Figure 4 | Developing differences in cortical thickness between the
superior and average intelligence groups. Group differences are
represented by t-statistics (t . 2.6), and show that the superior intelligence
group has a thinner superior prefrontal cortex at the earliest age (purple
regions). There is then a rapid increase in cortical thickness (red, green and
yellow regions) in the superior intelligence group, peaking at age 13 and
waning in late adolescence.
NATURE|Vol 440|30 March 2006
© 2006 Nature Publishing Group Download full-text
Statistical analysis. Pearson’s correlations between IQ and cortical thickness
were estimated at each cortical point. Each subject contributed only one scan to
was covered. Developmental effects were explored by dividing the sample
equally into four age groups (called early childhood (age range 3.8–8.4yr),
(17–29yr)). Correlations for each of 56 brain subregions were Z-transformed,
and the difference between the Z scores for each age group, and its significance,
was calculated. To correct for the large numberof comparisons, a false discovery
rate of 0.05 was applied28. Gender effects were examined for the entiresample in
a similar manner.
person, missing data, and irregular intervals between measurements, thereby
increasing statistical power while controlling for within-individual variation29.
and a cubic model found to provide the best fit, with the exception of anterior
temporal cortices where a linear model was appropriate. A cubic model was
therefore used to model age effects in the analyses presented. We first examined
whether the relationship between IQ and cortical thickness differs with age by
regressing cortical thickness at every vertex against IQ, age terms, and the
interaction of IQ and age terms. For further exploration of the interaction, we
divided the subjects into three IQ groups. This approach loses some power by
readily interpretable, allowing comparisons between highly intelligent and less
intelligent groups. The resulting statistical maps were thresholded to control for
multiple comparisons using the false discovery rate (FDR) procedure with
q ¼ 0.05 (refs 28, 30). An FDR threshold was determined for the statistical
model using all P values pooled across all effects included in the model. At every
cortical point, t-statistics were visualized through projection onto a standard
brain template (the map shows the results of the interaction between the cubic
age term and IQ groups). Such visualization showed clusters of cortical points
that had a significant difference between the intelligence groups in the trajectory
of cortical growth. The longitudinal analyses selected and averaged all cortical
points within each of these clusters. Graphs illustrating the trajectories were
generated using fixed-effects parameter estimates.
intelligence groups at different ages, linear mixed-models were run at different
centredages.Forexample,foragesevenyears, sevenwassubtractedfrom theage
at scan acquisition, and this value entered as the age term. t-statistics represent-
ing the differences in cortical thickness between the two intelligence groups at
each age were projected onto brain templates. This analysis represents group
differences at each age based on values estimated from developmental curves
modelled on all data.
Received 25 October; accepted 29 November 2005.
1.Spearman, C. ‘General intelligence’ objectively determined and measured. Am.
J. Psychol. 15, 201– -293 (1904).
Gottfredson, L. S. Why g matters: The complexity of everyday life. Intelligence
24, 79– -132 (1997).
Deary, I. J., Whalley, L. J., Lemmon, H., Crawford, J. R. & Starr, J. M. The
stability of individual differences in mental ability from childhood to old age:
Follow-up of the 1932 Scottish Mental Survey. Intelligence 28, 49– -55 (2000).
Booth, J. R. et al. Neural development of selective attention and response
inhibition. Neuroimage 20, 737– -751 (2003).
Sowell, E. R. et al. Longitudinal mapping of cortical thickness and brain growth
in normal children. J. Neurosci. 24, 8223– -8231 (2004).
McDaniel, M. Big-brained people are smarter. Intelligence 33, 337– -346 (2005).
Wilke, M., Sohn, J. H., Byars, A. W. & Holland, S. K. Bright spots: correlations of
gray matter volume with IQ in a normal pediatric population. Neuroimage 20,
202– -215 (2003).
Frangou, S., Chitins, X. & Williams, S. C. Mapping IQ and gray matter density in
healthy young people. Neuroimage 23, 800– -805 (2004).
Haier, R. J., Jung, R. E., Yeo, R. A., Head, K. & Alkire, M. T. Structural brain
variation and general intelligence. Neuroimage 23, 425– -433 (2004).
10. Kraemer, H. C., Yesavage, J. A., Taylor, J. L. & Kupfer, D. How can we learn
about developmental processes from cross-sectional studies, or can we? Am.
J. Psychiatry 157, 163– -171 (2000).
Giedd, J. N. et al. Brain development during childhood and adolescence: a
longitudinal MRI study. Nature Neurosci. 2, 861– -863 (1999).
12. Wechsler, D. Manual for the Wechsler Intelligence Scale for Children—Revised
(The Psychological Corporation, New York, 1974).
13. O’Donnell, S., Noseworthy, M. D., Levine, B. & Dennis, M. Cortical thickness of
the frontopolar area in typically developing children and adolescents.
Neuroimage 24, 948– -954 (2005).
14. Kabani, N., Le Goualher, G., MacDonald, D. & Evans, A. C. Measurement of
cortical thickness using an automated 3-D algorithm: a validation study.
Neuroimage 13, 375– -380 (2001).
15. Lerch, J. P. & Evans, A. C. Cortical thickness analysis examined through power
analysis and a population simulation. Neuroimage 24, 163– -173 (2005).
16. Gray, J. R., Chabris, C. F. & Braver, T. S. Neural mechanisms of general fluid
intelligence. Nature Neurosci. 6, 316– -322 (2003).
17. Duncan, J. et al. A neural basis for general intelligence. Science 289, 457– -460
18. Kostovic, I., Judas, M., Rados, M. & Hrabac, P. Laminar organization of the
human fetal cerebrum revealed by histochemical markers and magnetic
resonance imaging. Cereb. Cortex 12, 536– -544 (2002).
19. Kostovic, I. & Rakic, P. Developmental history of the transient subplate zone in
the visual and somatosensory cortex of the macaque monkey and human
brain. J. Comp. Neurol. 297, 441– -470 (1990).
20. Yakovlev, P. I. & Lecours, A. R. in Regional Development of the Brain in Early Life
(ed. Minokowski, A.) (Blackwell Scientific, Oxford, 1967).
21. Huttenlocher, P. R. & Dabholkar, A. S. Regional differences in synaptogenesis in
human cerebral cortex. J. Comp. Neurol. 387, 167– -178 (1997).
22. Hensch, T. K. Critical period regulation. Annu. Rev. Neurosci. 27, 549– -579
23. Chugani, H. T., Phelps, M. E. & Mazziotta, J. C. Positron emission tomography
study of human brain functional development. Ann. Neurol. 22, 487– -497
24. Collins, D. L., Neelin, P., Peters, T. M. & Evans, A. C. Automatic 3D intersubject
registration of MR volumetric data in standardized Talairach space. J. Comput.
Assist. Tomogr. 18, 192– -205 (1994).
25. Sled, J. G., Zijdenbos, A. P. & Evans, A. C. A nonparametric method for
automatic correction of intensity nonuniformity in MRI data. IEEE Trans. Med.
Imaging 17, 87– -97 (1998).
26. Zijdenbos, A. P., Forghani, R. & Evans, A. C. Automatic “pipeline” analysis of
3-D MRI data for clinical trials: application to multiple sclerosis. IEEE Trans.
Med. Imaging 21, 1280– -1291 (2002).
27. MacDonald, D., Kabani, N., Avis, D. & Evans, A. C. Automated 3-D extraction
of inner and outer surfaces of cerebral cortex from MRI. Neuroimage 12,
340– -356 (2000).
28. Benjamini, Y. & Hochberg, Y. Controlling the false discovery rate: a practical
and powerful approach to multiple testing. J. R. Stat. Soc. B 57, 289– -300
29. Pinheiro, J. C. & Bates, D. M. Mixed-effects Models in S and S-PLUS (Springer,
New York, 2000).
30. Genovese, C. R., Lazar, N. A. & Nichols, T. Thresholding of statistical maps in
functional neuroimaging using the false discovery rate. Neuroimage 15,
870– -878 (2002).
Supplementary Information is linked to the online version of the paper at
Acknowledgements This research was supported by the Intramural Research
Program of the National Institutes of Health. We acknowledge the statistical
advice of G. Chen and technical assistance from T. Nugent III. The authors thank
the children who participated in the study and their families.
Author Contributions P.S. designed and wrote the study with J.R. and J.G., and
conducted neuroimaging analyses. J.G. and J.R. directed the project. D.G.
conducted longitudinal analyses. L.C. was data manager, and R.L. and N.G.
advised on interpretation and analysis. J.L. and A.E. developed cortical thickness
analytic tools and J.L. developed software for longitudinal neuroimaging
Author Information Reprints and permissions information is available at
npg.nature.com/reprintsandpermissions. The authors declare no competing
financial interests. Correspondence and requests for materials should be
addressed to P.S. (firstname.lastname@example.org).
NATURE|Vol 440|30 March 2006