www.elsevier.com/locate/ynimg

NeuroImage 25 (2005) 1175–1186

Voxel-based morphometry and stereology provide convergent

evidence of the importance of medial prefrontal cortex for fluid

intelligence in healthy adults

Qi-Yong Gong,a,b,c,* Vanessa Sluming,a,b Andrew Mayes,d Simon Keller,a,b Thomas Barrick,a

Enis Cezayirli,a and Neil Robertsa

aMagnetic Resonance and Image Analysis Research Centre (MARIARC), University of Liverpool, UKbDepartment of Medical Imaging, University of Liverpool, 2nd Floor, Johnston Building, P.O. Box 147, Liverpool L69 3BX, UKcClinical MR Research Centre (CMRRC), Department of Radiology, West China Hospital, Sichuan University, PR ChinadSchool of Psychological Sciences, University of Manchester, UK

Received 23 August 2004; revised 20 November 2004; accepted 17 December 2004

Available online 25 February 2005

We investigated whether a relationship exists between frontal lobe

volume and fluid intelligence as measured by both Cattell’s Culture

Fair test and the Wechsler Adult Intelligence Scale-Revised (WAIS-

R) Performance scale, but not with crystallized intelligence as

measured by the WAIS-R Verbal scale, in healthy adults, using two

well-established image analysis techniques applied to high-resolution

MR brain images. Firstly, using voxel-based morphometry (VBM),

we investigated whether a significant relationship exists between

gray matter concentration and fluid intelligence on a voxel-by-voxel

basis. Secondly, we applied the Cavalieri method of modern design

stereology in combination with point counting to investigate possible

relationships between macroscopic volumes of relevant brain

regions defined as dorsolateral, dorsomedial, orbitolateral, and

orbitomedial prefrontal cortex on the basis of neuroanatomical

landmarks, and fluid intelligence. We also examined the effect on

these relationships of normalizing regional brain volumes to

intracranial volume. VBM analysis revealed a positive correlation

between gray matter concentration in the medial region of

prefrontal cortex and Culture Fair scores (corrected for multiple

comparisons), and also WAIS-R Performance Intelligence sum of

scaled scores (SSS) (uncorrected for multiple comparisons before

controlling for age, and this converges with the stereological finding

of the positive correlation between volume of dorsomedial prefron-

tal cortex normalized to intracranial volume and Culture

Fair scores after controlling for age. WAIS-R Verbal Intelligence

SSS showed no correlations. We interpret our findings, from

independent analyses of both VBM and stereology, as evidence

1053-8119/$ - see front matter D 2004 Elsevier Inc. All rights reserved.

doi:10.1016/j.neuroimage.2004.12.044

* Corresponding author. Department of Medical Imaging, University of

Liverpool, 2nd Floor, Johnston Building, P.O. Box 147, Liverpool L69

3BX, UK. Fax: +44 151 794 5766.

E-mail address: qiyong.gong@liverpool.ac.uk (Q.-Y. Gong).

Available online on ScienceDirect (www.sciencedirect.com).

of the importance of medial prefrontal cortex in supporting fluid

intelligence.

D 2004 Elsevier Inc. All rights reserved.

Keywords:MRI; Voxel-based morphometry; Stereology; Intelligence; Brain

volume

Introduction

How the brain supports intelligence can be explored by

examining the effects on intelligence of focal brain lesions and by

examining which brain regions activate when intelligence is

exercised, but it can also be examined by determining whether

variations in the volume of any brain region correlate with

individual differences in intelligence in healthy individuals.

Whereas functional neuroimaging does not establish the causal

role of an activated region in intelligence, lesion evidence provides

strong support for this claim, and the bneophrenologicalQ approachdoes so as well, albeit more weakly and indirectly. It is assumed that

MRI-derived measures of the volume of a particular brain structure

in healthy brains are at least partially determined by the number and

size (hence, the complexity of the synaptic connections) of the

neurons that it contains so that greater volumes should mean that a

structure works more efficiently (Andreasen et al., 1993; Bigler et

al., 1995; Rushton and Ankney, 1995). Accordingly, it is plausible

to hypothesize that, unless greater volume primarily reflects

inadequate pruning of neurons in development (Howard et al.,

2000) and other factors, such as the efficiency of neurotransmission,

are fairly constant, individuals with larger brain regions should

perform the functions mediated by those regions better.

Although most leaders in the field of intelligence have refused

to give precise bmeaning-styleQ definitions of the term (Deary,

Q.-Y. Gong et al. / NeuroImage 25 (2005) 1175–11861176

2000), intelligence must depend on the cognitive abilities needed

to solve a range of problems, such as defining words, ordering a

related series of pictures, and indicating the odd one out among a

set of words or pictures. The psychometric approach uses

standardized tests to measure how well individuals solve problems

like these. A century of research has shown that individuals’ scores

on different intelligence tests inter-correlate positively, and factor

analysis has given rise to the notion that human intelligence is a

multilevel hierarchy of abilities with a general factor or g at the top

and more specific ability factors lower down the hierarchy (see

Neisser et al., 1996 for a discussion). Identification of whether

there are different, specific neural bases of the putative compo-

nents of intelligence will help confirm or indicate the need for

modifying the bunitary processQ interpretation of these statistical

conclusions.

Investigations of the neural bases of human intelligence have

generally supported the view that greater brain size correlates

positively with intelligence (Andreasen et al., 1993; Plomin and

Kosslyn, 2001; Tisserand et al., 2001), although not all studies

have provided positive results (Tramo et al., 1998). Even so, the

earlier studies did not provide consistent evidence for the appealing

notion that the volumes of specific brain regions correlate

positively with intelligence or specific intelligence factors (Flash-

man et al., 1998; MacLullich et al., 2002). Recently, however,

Thompson et al. (2001) identified a significant association between

total gray matter volume of frontal cortex and a proxy measure of

the general intelligence factor, g. This association is consistent with

evidence that patients with large frontal lobe lesions showed

marked impairments in those kinds of intelligence, sometimes

referred to as fluid intelligence, that load highly on g, and relate to

the ability to solve relatively novel problems which cannot be

solved through the routine use of heavily overlearned cognitive

routines (Duncan et al., 1995).

Fluid intelligence contrasts with crystallized intelligence or the

ability to solve those more familiar kinds of problem that can be

solved through the routine use of heavily overlearned cognitive

routines. Crystallized intelligence typically loads more weakly on

g (Gray and Thompson, 2004), and, unlike fluid intelligence,

which tends to decline with age, it is relatively stable as age

increases (Neisser et al., 1996). There is mixed evidence that fluid

and crystallized intelligence are supported by partially distinct

brain structures. Thus, although Duncan et al. (1995) found that

frontal lesions probably do not disrupt crystallized intelligence

whilst greatly disrupting fluid intelligence, the results of functional

imaging are harder to interpret (see Gray and Thompson, 2004 for

a discussion). As both kinds of intelligence load on g, they may

well depend on one or more common processes that involve

working memory and/or attentional control, although fluid

intelligence should rely on these processes more. However, fluid

intelligence may also depend on processes not needed for

crystallized intelligence. Thus, some frontal cortex regions may

be selectively critical for identifying what kind of more familiar

problem a novel problem resembles and others may be key for

monitoring the effectiveness of the routines consequently applied

so as to be prepared, if necessary, to adopt a more appropriate

analogy and its corresponding routines.

Given its large volume, which constitutes a substantial part of

the whole neocortex, it would not be surprising if different parts

of the frontal neocortex each supported distinct cognitive

functions relevant to novel problem solving. Efforts are therefore

being made to further localize a possible focus for fluid

intelligence within this large volume of prefrontal lobe. It has

been shown that greater volume in a region of medial prefrontal

cortex was associated with increased IQ in a healthy pediatric

population (Wilke et al., 2003). This is partially consistent with

the results of three PET imaging studies that found a significant

role for prefrontal cortex in performing tasks which depend

strongly on fluid intelligence (Duncan et al., 2000). Although,

relative to more routine tasks using matching materials, the three

different kinds of novel problem solving (loading highly on g)

only showed overlapping activations in the lateral frontal cortex,

solving novel visuospatial problems also activated medial

prefrontal cortex. More recently, Gray et al. (2003) showed, in

an event-related functional magnetic resonance imaging (fMRI)

study of 48 subjects, that the positive relationship between fluid

intelligence and performance on high versus low interference lure

trials of a three-back working memory task was mediated by the

greater neural activity in the lateral frontal cortex and parietal

cortex bilaterally produced by successful performance on the high

interference lure trials. As high interference trials demand high

levels of attentional control, a plausible explanation of the results

is that the use of fluid intelligence critically involves certain

kinds of attentional control, and that more efficient use of this

control, which leads to higher fluid intelligence, depends on

greater activation in key structures including the lateral frontal

cortex.

Gray and Thompson (2004) discuss some possible reasons why

Gray et al. (2003) and Duncan et al. (2000) partially conflict. One

reason is that functional imaging studies cannot directly establish

that activity in a brain region is causally critical for a cognitive

process. It is therefore of interest to investigate whether in healthy

adults individual differences in the size of lateral and/or medial

frontal regions as well as the parietal sites also identified by Gray

and his colleagues correlate with differences in fluid and crystal-

lized intelligence, or only with differences in fluid intelligence.

Finding frontal correlations only with fluid intelligence would

provide support for the view that activity in one or more of these

regions fairly selectively supports fluid intelligence as a function of

increasing size.

In the present study, we have applied two well-established brain

image analysis techniques to test the structural relationship between

gray matter in prefrontal cortex and performance on Cattell’s

Culture Fair test and the Wechsler Adult Intelligence Scale-Revised

(WAIS-R) measures of Performance and Verbal intelligence.

Cattell’s test is the best measure of fluid intelligence, the WAIS-

R Performance tests probably depend on both fluid and crystallized

intelligence, but the WAIS-RVerbal tests probably depend most on

crystallized intelligence (Duncan et al., 1995; Woodcock, 1990).

The Cattell and WAIS-R Performance tests directly tap the ability

to solve different kinds of visuospatial problems, but it is probable

that all three kinds of test tap verbal abilities to some extent because

these abilities are used to solve visuospatial as well as verbal

problems. Our primary aim was to determine whether fluid and

crystallized intelligence correlated differently with the size of brain

regions (particularly in the frontal lobes) after correcting for age-

related changes. To achieve this primary goal, we also had the

subsidiary aims of identifying: (i) the relationship between age and

fluid and crystallized intelligence; (ii) age-related changes in the

size of brain regions, particularly those in the frontal lobes; and (iii)

whether (voxel-based morphometry) VBM-based frontal gray

matter density correlations corresponded to stereology-based

prefrontal subfield volume correlations.

Q.-Y. Gong et al. / NeuroImage 25 (2005) 1175–1186 1177

The cross-validation of VBM with stereology that we attempted

to achieve was intended to give more confidence in the results as

VBM has been criticized (e.g., Bookstein, 2001). We therefore

used optimized VBM (Good et al., 2001a,b) that enables

quantification of group differences in neuroanatomy without any

a priori region-of-interest, and the Cavalieri method of modern

design stereology in combination with point counting on high-

resolution MR images. We have previously used VBM to

investigate relationships between gray matter concentration any-

where in the brain and cognitive abilities in neurologically healthy

subjects (Sluming et al., 2002) as well as the effect of temporal

lobe epilepsy (Keller et al., 2002). As our particular focus was the

frontal lobes, we used stereology to estimate volumes of brain

regions defined as dorsolateral, dorsomedial, orbitolateral, and

orbitomedial prefrontal cortex on the basis of previously identified

neuroanatomical landmarks (Howard et al., 2003). These estimates

were made both with and without normalization to intracranial

volumes. VBM and stereological analyses were also performed

with and without age correction.

Subjects and methods

Subjects and neuropsychological tests

Our analysis was applied to a group of healthy male and

female subjects who were recruited to a study to investigate the

relationship between brain changes and cognitive ability across

the life span (Cezayirli, 2000) and considers only those subjects

who were under 60 years of age. Subjects are all right handed

(as determined by the Edinburgh Handedness Inventory) and

underwent the following batteries of neuropsychological tests:

Wechsler Memory Scale-Revised (WMS-R), Warrington Recog-

nition Memory Test and a story recall test, SCOLP, STROOP,

Wechsler Adult Intelligence Scale-Revised (WAIS-R), and

Cattell’s Culture Fair Test (Cattell and Cattell, 1973). In the

present study of the neural bases of adult fluid versus crystallized

intelligence, we compared the structural neural correlates of

subjects’ scores on Cattell’s Culture Fair Test and WAIS-R

Performance Intelligence with the structural correlates of their

scores on WAIS-R Verbal Intelligence. In the original study, 16

subjects (8 male and 8 female) were recruited for each decade

between the ages of 20 and 80 years. Thus, data were potentially

available for 32 male and 32 female subjects. However, relevant

neuropsychological test scores were not available for all subjects

so that the study group comprised 25 males and 30 females

making a total of 55 subjects with a mean age of 40 years and

standard deviation of 12 years. The neuropsychological data are

presented in Table 1.

Table 1

Neuropsychological data for the subjects

Male (n = 25) Fem

Max Min Mean Std Max

Age 59 20 38 12 59

WAIS-R (Verbal SSS) 74 47 61 6 76

WAIS-R (Performance SSS) 66 40 56 7 65

WAIS-R (Full SSS) 137 97 117 10 141

Cattell’s Culture Fair Scores 43 18 34 6 41

SSS—sum of scaled scores; Std—the standard deviations.

It is important to note that in the present study, WAIS

Intelligence was measured with raw scaled scores. According to

the WAIS-R manual (Wechsler, 1981), the scaled scores for each of

the eleven subtests in the test battery are based on a reference

group in the standardization sample between the ages of 20 and 34.

Each subject’s Verbal Intelligence sum of scaled scores (SSS)

equals the sum of the scaled scores on the six Verbal tests, whereas

the Performance Intelligence SSS equals the sum of the scaled

scores on the five Performance tests. The Full Scale Intelligence

SSS, which is the sum of the Verbal SSS and the Performance SSS,

is the SSS on all eleven tests. The scaled score for each of the

eleven tests has a mean of 10 and a standard deviation of 3. Use of

the SSS ensures that different tests have an equally weighted

impact on composite intelligence scores. As noted above, these

scores were not age adjusted so as to facilitate examination of any

changes in intelligence as a function of age so that, if necessary, we

could control for these in examining the relationship between

structure and intelligence.

Data acquisition

High-resolution T1-weighted MR images were acquired on a

1.5-T SIGNA whole-body MR imaging system (General Electric,

Milwaukee, USA) with a 3D spoiled gradient echo (SPGR) pulse

sequence (TR = 34 ms, TE = 9 ms, flip angle = 308) using a

standard head coil. A Field of View (FOV) of 20 cm was used with

an acquisition matrix comprising 256 readings of 128 phase

encoding steps, producing 124 contiguous coronal slices with slice

thickness of 1.6 mm and in-plane resolution of 1 mm � 1 mm.

VBM analysis

Optimized VBM was carried out using Statistical Parametric

Mapping software (SPM99, Welcome Department of Imaging

Neuroscience, London; available at http://www.fil.ion.ucl.ac.uk/

spm). The analysis procedure has been described in detail (Good

et al., 2001a,b).

Optimized VBM involves spatial transformation of all data to a

common stereotactic space by registering each image in native

space to the same template image prior to voxel-wise statistical

analysis. Gray matter was automatically segmented from the raw

MR images using tissue signal intensity values and a priori

information about the distribution of brain tissue type (the 148

normal data set of the Montreal Neurological Institute). An

automated brain extraction step was included to eliminate voxels

from non-gray matter structures, such as the dural venous sinuses,

scalp, cranial marrow, and diploic space. This avoids non-brain

voxels with similar intensities to gray matter voxels being

inherently included as brain tissue during standard segmentation.

ale (n = 30) Total (n = 55)

Min Mean Std Max Min Mean Std

20 41 12 59 20 40 12

51 63 7 76 47 62 7

36 53 8 66 36 55 7

88 117 13 141 88 117 11

18 31 5 43 18 32 6

Q.-Y. Gong et al. / NeuroImage 25 (2005) 1175–11861178

Gray matter partitions were spatially normalized (using a 12-

parameter affine transformation and 7 � 8 � 7 non-linear basis

functions, which are the default normalization parameters in

SPM99) to a customized gray matter template, which was

constructed from the normalized, segmented, and smoothed gray

matter data sets of all 55 subjects. The deformation parameters

obtained from the normalization process were applied to the

original raw images (in native space) of all participants to create

optimally normalized whole-brain images, which were recursively

segmented and brain tissue extracted. The optimally processed

images were smoothed with an isotropic Gaussian kernel with full-

width half-maximum of 8 mm.

Voxel-by-voxel regression analysis of gray matter concentration

was performed taking total brain gray matter as a confound and

Culture Fair score, or WAIS-R Performance or Verbal Intelligence

SSS in turn as covariates of interest, within the framework of the

general linear model in SPM99. Analyses were performed both

with and without including age as a nuisance variable. Further-

more, a statistical model was formulated to examine [analogous to

a study of possible sex by age interactions] whether the increase in

gray matter concentration with fluid intelligence measures in males

was greater than those in females, or vice versa, for all 55 subjects.

For all analyses, contrasts were defined to detect those voxels of

brain tissue in one group which had a greater or lesser probability

of being gray matter in homologous voxels in the other group. The

output for each comparison is a statistical parametric map of the t

statistic (SPM{t}), which was transformed to a normal distribution

(SPM{z}). Results from analyses were thresholded at a P value of

less than 0.05 (corrected for multiple comparisons), and after

partialling out age or sex, were thresholded at P b 0.0001

(uncorrected).

Stereological brain volume estimation

The Cavalieri method of modern design stereology was applied

in conjunction with point counting to estimate the volume of

several sub-regions of frontal cortex (Cruz-Orive, 1989; Gundersen

and Jensen, 1987; Gundersen et al., 1999). This technique has been

applied in studies of a range of biological structures (Gong et al.,

1999, 2000, 2004). We specifically estimated the prefrontal cortical

subfield volumes using a previously established parcellation

technique (Gur et al., 2000). To facilitate the analysis, the 3D data

set was parcellated based on macroanatomical landmarks. The

parcellation of anatomical subfields for the prefrontal cortex, the

computer packages employed, as well as the detailed procedure for

image processing can be found in an earlier report (Howard et al.,

2003). The procedure involved resizing of the data set to isotropic

voxels, reformatting the image, and orienting it to a standardized

sagittal plane orthogonal to the bicommissural plane. Within each

hemisphere, prefrontal cortex was divided into four subfields,

namely, dorsolateral, dorsomedial, orbitolateral, and orbitomedial

consistent with Damasio’s anatomical demarcations (Damasio,

1995). These subfields are relevant to the functional subdivision of

prefrontal cortex supported by neuropsychological, neuropsychi-

atric, and neuroimaging studies (Gur et al., 2000; Gusnard et al.,

2001; Postle et al., 2000). Total intracranial volumes were also

estimated and used as a basis for normalizing the regional brain

volumes. The stereological volume estimates of the prefrontal

subfields were correlated with neuropsychological measures of

fluid intelligence (i.e., Culture Fair scores and WAIS-R Perfor-

mance or Verbal Intelligence SSS) by performing a stepwise

multiple regression analyses. Prior to measurement being obtained,

the reproducibility and repeatability of the method was assessed

(Howard et al., 2003). Predicted coefficients of error for individual

volume estimates were less than 5%, and regional subfield volume

estimates for prefrontal cortex derived from point counting

methods were reproducible between raters [Intraclass Correlation

Coefficients (ICC) ranged from 0.92 to 0.95] and repeatable by the

same rater (ICC ranged from 0.93 to 0.99).

Results

Aging effect

A significant negative correlation (P b 0.05) was observed

between age and Culture Fair score (upper left panel of Fig. 1, total

group: r = �0.64; female group: r = �0.67; male group: r =

�0.60), and WAIS-R Full Intelligence SSS (upper right panel of

Fig. 1, total group: r = �0.49; female group: r = �0.50; male

group: r = �0.49). With respect to the WAIS-R subscales, a

negative correlation was observed for Performance Intelligence

SSS (lower left panel of Fig. 1, total group: r = �0.58; female

group: r = �0.58; male group: r = �0.57) but not for Verbal

Intelligence SSS (P N 0.05, lower right panel of Fig. 1, total group:

r = �0.18; female group: r = �0.28; male group: r = �0.12).

There was no statistically significant difference between the

correlations of age with Culture Fair score and age with Perform-

ance Intelligence SSS (P N 0.05). However, both the correlations

of age with Culture Fair score and age with Performance

Intelligence SSS were significantly stronger than the correlation

of age with Verbal Intelligence SSS (both P b 0.01).

In the VBM analysis, we observed small discrete areas in

which gray matter concentration declines with age, and these

regions were mainly scattered throughout the parietal lobes,

corrected for multiple comparison (P b 0.05, Fig. 2). The mean

volumes and the standard deviations for regional prefrontal

subfield estimated from the Cavalieri method of modern designed

stereology are presented in Table 2. Multiple regression analysis

revealed that stereological volume estimates for left dorsolateral

prefrontal cortex declined with age both before (b = �0.59, P =

0.01) and after normalization to intracranial volumes (b = �0.56,

P = 0.00) in the whole group.

Correlation of gray matter concentration with fluid and

crystallized intelligence

Optimized VBM analysis detected a single cluster of gray

matter in the medial region of frontal cortex (see Fig. 3, mainly left

anterior cingulate gyrus and dorsomedial prefrontal cortex,

Brodmann region 24/32/8/9/10), that showed a significant positive

correlation with Culture Fair scores (P b 0.05 corrected for

multiple comparisons, Table 3) and remains after controlling for

age (P b 0.0001, uncorrected for multiple comparison) (see Fig.

3). A significant positive correlation was also observed between

gray matter concentration and WAIS-R Performance Intelligence

SSS for the same but smaller region of medial prefrontal cortex

(Brodmann regions 32/10/11) before controlling for age (P b

0.001, uncorrected for multiple comparison, Fig. 4) although this

relationship did not survive age correction (P N 0.05). No

significant relationships were found for the WAIS-R Verbal

Intelligence SSS regardless of whether age was controlled for

Fig. 1. Graphs illustrate relationships between age and Culture Fair Scores (upper left); WAIS-R Full Intelligence Sum of Scaled Score (SSS) (upper right);

WAIS-R Performance Intelligence SSS (lower left) and WAIS-R Verbal Intelligence SSS (lower right). Solid and dotted lines represent the linear regression fit

for Female (o) and Male (.) subgroups, respectively.

Q.-Y. Gong et al. / NeuroImage 25 (2005) 1175–1186 1179

(P N 0.05). No negative relationships were observed for any of the

above correlation analyses (P N 0.05).

Correlation of prefrontal subfield stereological volumes with fluid

and crystallized intelligence

Positive relationships between subfield volume estimates and

fluid intelligence scores (i.e., Culture Fair scores and WAIS-R

Fig. 2. Standardized sagittal (left), transaxial (middle), and coronal (right) glass b

gray matter concentration and age for whole group with correction for multiple com

matter concentration with age ( P b 0.05).

Performance Intelligence SSS) were revealed in medial prefrontal

cortex (Table 4, Fig. 5). In particular, simple regression analysis

showed that Culture Fair score was highly correlated with volume

of left dorsomedial prefrontal cortex (b = 0.50, P b 0.05). After

controlling for age, multiple regression analyses revealed signifi-

cant correlations of prefrontal cortex subfield volumes (corrected

for intracranial volumes) with Culture Fair scores in left

dorsomedial (b = 0.37, P b 0.05) and left orbitomedial subfields

rain images illustrate statistical parametric analysis of relationship between

parison. There were discrete but significant regions showing decreased gray

Table 2

Mean volumes [est V] and the standard deviations (Std) for regional

prefrontal subfield estimated from the Cavalieri method of modern designed

stereology

Side Structure Mean [est V] Std

R OM 17.27 4.16

L OM 16.71 3.89

R DM 22.89 5.18

L DM 21.77 4.61

R OL 14.34 3.75

L OL 14.83 3.67

R DL 24.88 5.43

L DL 23.45 5.02

Abbreviations: R, right; L, left; OM, orbitomedial; DM, dorsomedial; OL,

orbitolateral; DL, dorsolateral.

Q.-Y. Gong et al. / NeuroImage 25 (2005) 1175–11861180

(b = 0.32, P b 0.05). Similar findings were also observed for

WAIS-R Performance Intelligence SSS in left orbitomedial

prefrontal cortex irrespective of controlling for age and intra-

cranial volume (P b 0.05). No significant relationships at all were

found for the WAIS-R Verbal Intelligence SSS regardless of

Fig. 3. Statistical parametric maps showing a positive correlation between gray mat

single cluster in medial region of frontal cortex (mainly left anterior cingulate gyru

between Culture Fair scores and gray matter concentration ( P b 0.05 corrected

significant cluster appears larger if uncorrected for multiple comparisons (upper rig

in the similar medial region of prefrontal cortex ( P b 0.0001, uncorrected for multi

corrected for multiple comparison). The red cursors (b) point to the locations of

whether we controlled for age and corrected for intracranial

volumes (P N 0.05).

Discussion

The present study has shown that fluid intelligence, as

measured by Culture Fair scores and by WAIS-R Performance

Intelligence SSS, declined with age whereas crystallized intelli-

gence, as measured by WAIS-R Verbal Intelligence SSS, did not.

The VBM analysis also showed that gray matter concentration

declined with age in a discrete region of the parietal lobes whereas

the stereology analysis showed that the left dorsolateral subfield of

the prefrontal cortex, both before and after normalization with

respect to intracranial volume, declined with age. Furthermore, in

the stereological analysis, prefrontal cortex subfield volume was

positively correlated with fluid intelligence [i.e., left dorsomedial

cortex (Culture Fair scores) and left orbitomedial cortex (both

Culture Fair Scores and Performance Intelligence SSS)] (Table 4).

For the VBM analysis, the relationship between brain structure and

fluid intelligence was most evident for the medial frontal lobe

when taking the Culture Fair Score as the covariate of interest

ter concentration and Culture Fair scores. Top row: Without age correction, a

s and dorsomedial prefrontal cortex) shows a significant positive correlation

for the multiple comparisons) (upper left panel). Note the extent of the

ht panel). Bottom row: After controlling for age, the cluster remains evident

ple comparison) (lower right panel) but no significant findings were found if

the global maxima.

Table 3

Regions showing positive correlation between gray matter concentration and Culture Fair scores, and WAIS-R Performance sum of scaled score (SSS)

Brain regions Cluster size (mm3) T score Z score Peak coordinates (x,y,z) Significance level

Culture Fair scores before controlling for age (corrected for multiple comparisons)

Medial PFC, mainly left anterior cingulate gyrus and

DM PFC

28 5.80 5.08 �5,49,8 P = 0.017

Culture Fair scores after controlling for age (uncorrected for multiple comparisons)

Medial PFC, mainly bilateral anterior cingulate gyrus, DM

and OM PFC (BA24/32/8/9/10)

3904 4.04 3.75 1,55,�13 P b 0.0001

WAIS-R performance SSS before controlling for age (uncorrected for multiple comparisons)

Medial PFC, mainly bilateral anterior cingulate gyrus, DM

and OM PFC (BA32/10/11)

688 4.47 4.10 �3,46,�3, P b 0.0001

Listed are coordinates corresponding to the voxels with maximum (peak) effect sizes defined in Montreal Neurological Institute (MNI) space. Brain regions (in

parentheses) defined as the approximate Brodmann areas (BA) are in accordance with Talairach and Tournoux (1988).

Abbreviations: PFC, prefrontal cortex; OM, orbitomedial; DM, dorsomedial.

Q.-Y. Gong et al. / NeuroImage 25 (2005) 1175–1186 1181

before age correction (corrected P b 0.05) (upper left panel of Fig.

3) at the level of correction for multiple comparisons, but after age

correction, the relationship for this brain region was only

significant before correction for multiple comparisons (uncorrected

P b 0.0001) (lower right panel in Fig. 3). The correlation was less

significant when taking WAIS-R Performance Intelligence SSS as

covariate of interest. WAIS-R Verbal Intelligence SSS was not

found to correlate significantly with the volume of any brain region

regardless of whether we used VBM or stereology, controlled for

age, or corrected for brain volume.

This study included only subjects who were under 60 years of

age in order to minimize the influence of age-related changes in

fluid intelligence. However, although the decline in fluid intelli-

Fig. 4. Statistical parametric maps showing a positive correlation between gray

scores (SSS). Note significant positive correlation was only observed without c

controlling for age. The red cursors (b) point to the locations of the global maxim

gence accelerates after the age of 60, we still observed a significant

negative correlation (P b 0.05) between age and fluid intelligence

measures, which was more pronounced for the Culture Fair score

(Fig. 1), agreeing well with a previous cross-sectional study (Bigler

et al., 1995). Our study was cross-sectional and it is possible that

some of the fluid intelligence decline shown by age 60 was a

cohort effect related to the average gain of about three IQ points

per decade in many technologically advanced countries, known as

the bFlynn effectQ (for example, see Deary, 2000). However, some

of the decline may have related to cerebral atrophy associated with

aging. Thus, although VBM analyses and stereology did not yield

convergent findings about the correlations between age and frontal

region volumes, there is a suggestion that not only parts of the

matter concentration and WAIS-R Performance Intelligence sum of scaled

orrections for multiple comparison ( P b 0.001, upper right panel) before

a.

Table 4

Correlation of prefrontal cortex subfield volumes with intelligence measures by multiple regression analyses

Fluid intelligence Regression analysis R-OM L-OM R-DM L-DM R-OL L-OL R-DL L-DL

Culture Fair scores Simple correlation – – – 0.50 – – – –

P = 0.000

Covariate with age – 0.32 – 0.37 – – – –

P = 0.03 P = 0.01

Covariate with age – 0.32 0.37 – – – –

(corrected for ICV) P = 0.03 P = 0.01

WAIS-R Simple correlation – 0.38 – – – – – –

Performance SSS P = 0.008

Covariate with age – 0.35 – – – – – –

P = 0.02

Covariate with age – 0.26 – – – – – –

(corrected for ICV) P = 0.08

Abbreviations: R, right; L, left; OM, orbitomedial; DM, dorsomedial; OL, orbitolateral; DL, dorsolateral; SSS—sum of scaled scores; ICV—intracranial

volume. Dash means statistically insignificant ( P N 0.05).

Q.-Y. Gong et al. / NeuroImage 25 (2005) 1175–11861182

parietal lobes, but also the left dorsolateral prefrontal lobe atrophy

appreciably by the age of 60, and these declines may partially

explain the reduction in fluid intelligence by this age. The

possibility that the age-associated dorsolateral prefrontal cortex

volume reduction may partially underlie the age-related decline in

fluid intelligence is interesting because of two of the studies

considered in the Introduction (Duncan et al., 2000; Gray et al.,

2003), which found evidence that the structures we found to

atrophy may support fluid intelligence. Although these studies used

different designs, both found that activation of the left dorsolateral

prefrontal cortex was associated with the exercise of fluid

intelligence. Duncan et al. (2000) found that this region was the

only one that consistently activated when subjects performed

different kinds of fluid intelligence task. It may, therefore, mediate

functions that play a role in all domains of fluid intelligence. Like

us, Gray et al. (2003) used an individual difference approach and

found that change in activation in several brain regions in response

to performing the high interference three-back lure condition of a

working memory task was associated with subjects’ fluid

intelligence (as measured by Raven’s Advanced Progressive

Matrices Test). One of these regions was the dorsolateral prefrontal

cortex, but there were also effects bilaterally in parietal cortex. As

our VBM and stereology results related to aging were not mutually

Fig. 5. Stereological subfield analysis (Howard et al., 2003) for left dorsomedial (D

prefrontal subfield volumes expressed as percentage fractions of intracranial volu

P b 0.05) (right panel).

reinforcing, further work (with both techniques) needs to explore

whether the age-related decline of one or all of these structures

causes the decline of fluid intelligence that also occurs with age.

Although we only found evidence for age-related atrophy of

parts of the parietal lobes and of the left dorsolateral prefrontal

lobe, we controlled for age when making the fluid intelligence–

brain volume correlations because fluid intelligence correlated with

age and it is known that the volumes of other brain regions reduce

with age although perhaps less dramatically than the parietal and

left dorsolateral prefrontal regions. This may explain why the

positive correlation between fluid intelligence and medial pre-

frontal cortex was weaker after controlling for age. Both VBM and

stereology converged in finding that the volume of this brain

region was positively associated with fluid intelligence (as

measured by the Culture Fair Test) and that it was the only brain

region to show such an association. The association with the

medial prefrontal region seemed to be weaker for WAIS-R

Performance Intelligence SSS, which is probably less dependent

on fluid intelligence in adults than is the Culture Fair Test. No

significant associations at all were found for the WAIS-R Verbal

Intelligence SSS, which is generally viewed as primarily a test of

crystallized intelligence even though it loads quite strongly on g

(see Deary, 2000). It, therefore, is plausible that the medial

M) prefrontal cortex (left panel) and multi-covariate regression analysis of

me (ICV) against Culture Fair Scores with fitted regression line (b = 0.50,

Q.-Y. Gong et al. / NeuroImage 25 (2005) 1175–1186 1183

prefrontal cortex is specifically associated with fluid intelligence in

adults.

The results are broadly convergent with those of a structural

imaging study of intelligence in a healthy pediatric population,

which used VBM and found a positive correlation between

intelligence and gray matter volume in an almost identical region

of medial prefrontal cortex (see Fig. 4a in Wilke et al., 2003). In

the pediatric study, however, only a so-called compact IQ (that

seems to have been based on the WAIS) was analyzed. For

children, the WAIS may contain relatively novel problems and

hence may tap a similar kind of fluid intelligence to the Culture

Fair test in adults, but this is a speculation. A functional imaging

study using MRI also found selective activation of the same medial

prefrontal region when participants engaged in a task that involved

processing high levels of relational complexity (see Fig. 5 in

Kroger et al., 2002). Another functional MRI study found that

increasing difficulty of the inspection time task (which correlates

with fluid intelligence) was associated with increasing medial

frontal cortex activation (Deary et al., 2004). These findings

suggest that our study may have shown that the level of fluid

intelligence is linked to the quantity of gray matter in the very

frontal regions where the results from the above functional imaging

study exhibit focal activation when complex reasoning was

performed.

Two other functional neuroimaging studies (Duncan et al.,

2000; Gray et al., 2003), however, raise interpretative problems for

our finding that the volume of the medial prefrontal region

correlates with the level of fluid intelligence. Neither study found

evidence that the medial prefrontal region activated in a way that

related to fluid intelligence. Thus, the PET imaging study of

Duncan et al. (2000) found a significant common activation in the

lateral prefrontal cortex (BA 6, 8, 45–47) with three different fluid

intelligence tasks, but only Culture Fair test performance activated

the medial regions of prefrontal cortex (BA 8). As the Culture Fair

test taps fluid intelligence, these results are consistent with those of

Gray et al. (2003), who used multiple regression to show that there

was virtually nothing about the relationship between subjects’ fluid

intelligence and their successful performance on the high

interference lure condition of a three-back working memory task

that was not mediated by lateral prefrontal (BA 9, 10, 44–46) and

bilateral parietal cortex (BA 31, 40) activations produced by the

latter task. Significant mediator effects were not found in the

medial prefrontal cortex even though activity in this region (BA

24) correlated with fluid intelligence during lure trials.

One speculative interpretation of our results is that they show

an association between the volume of the medial frontal cortex and

fluid intelligence whereas those of Duncan, Gray, and their

colleagues (Duncan et al., 2000; Gray et al., 2003) show an

association between the lateral frontal (and possibly parietal) cortex

and g. As indicated in the Introduction, tests of both fluid and

crystallized intelligence may load on g because they depend on

certain common processes. These processes may activate the lateral

frontal (and possibly) parietal cortex. However, fluid intelligence

may depend on some processes not required for crystallized

intelligence. Our results are consistent with the medial frontal

cortex supporting these processes. At present, it is unclear whether

this region underpins processing that supports all kinds of fluid

intelligence or whether there are different kinds of fluid intelli-

gence supported by non-overlapping (as well as common)

processes with the medial frontal cortex supporting just those

kinds of fluid intelligence that we tested. Duncan et al.’s (2000)

findings that different fluid intelligence tasks (which probably

would have inter-correlated and loaded heavily on g) do not

produce identical frontal activations are consistent with the region

having a more specific functional role. Even if verbal processing

was used to help solve our visuospatial tasks, the kind of fluid

intelligence tapped may still have differed in certain important

respects from that applied to verbal, perceptual, and other kinds of

problem, and have called on the medial frontal cortex to a much

greater degree. These speculations raise two puzzles. The first

puzzle is why only variations in the volume of this medial frontal

cortex region correlated with fluid intelligence when, given the

structure’s relationship to fluid intelligence, the volume of the

lateral prefrontal cortex should do so as well. The second puzzle is

why the fluid intelligence-related processing that the medial

prefrontal region performs should be relevant for the solution of

novel visuospatial problems, but much less so for the solution of

novel non-visuospatial problems.

In relation to the first puzzle, given our interpretation of Duncan

et al.’s (2000) results, it would be predicted that the volume of the

medial prefrontal region would correlate with the ability to solve

novel visuospatial problems, but not with the ability to solve novel

verbal and perceptual problems. However, this prediction remains

to be tested. More importantly, it is unclear why the volume of the

lateral prefrontal region does not correlate positively with fluid

intelligence, although it may do so with the ability to solve novel

verbal and perceptual problems. If future work confirms the lack of

positive correlation with the Culture Fair Test and WAIS-R

Performance Intelligence SSS and it is confirmed that performing

these tasks consistently activates the lateral frontal cortex, then it

will become important to find an explanation. One speculative

possibility is that this brain region mediates executive/working

memory functions that play an important role in all kinds of novel

problem solving, but that, provided circuitry is well organized and

undamaged/not atrophied, the number of neurons/synapses it

contains has little effect on its efficiency. In contrast, metabolic/

biochemical factors and/or the richness of the region’s white matter

connections may have a major effect (the reverse would presumably

be true for the medial prefrontal region). Support for the first kind of

general possibility comes from previous studies using Magnetic

Resonance Spectroscopy (MRS), which have shown that differ-

ences between healthy subjects in intellectual and neuropsycho-

logical performance relates to measurements of brain metabolite

concentration (i.e., N-acetylaspartate and choline) in occipitopar-

ietal white matter (Jung et al., 1999, 2000). MRS in these studies

provides a measure of glial and neuronal metabolism so future

studies should identify frontal areas of interest to identify whether

metabolic factors within the dorsolateral prefrontal region are much

more powerful predictors of various kinds of fluid intelligence than

are metabolic factors in the medial prefrontal region. It will

therefore be interesting to study prefrontal cortex using MRS as

long as effective transmit and receive coils can be developed for this

region.

If the results of a VBM study by Haier et al. (2004) are reliable,

however, then the volume of the lateral prefrontal cortex as well as of

a number of other prefrontal cortex structures including the medial

prefrontal cortex correlate with intelligence in healthy adult subjects

so the above puzzle might not exist. Unlike us, however, Haier and

his colleagues used VBM to determine the relationship between

subjects’ brain volumes and their performance on the WAIS-R Full

Scale IQ Test. This study was concordant with ours in so far as it

found strong effects with prefrontal cortex correlations. But it also

Q.-Y. Gong et al. / NeuroImage 25 (2005) 1175–11861184

found correlations between IQ and posterior cortex volumes. Indeed,

in its younger sample, the statistically strongest effects were actually

found in left temporal rather than frontal lobe. One possibility is that

these posterior cortex effects relate to crystallized intelligence rather

than fluid intelligence. However, not only did we not find posterior

neocortex effects, which Haier and his colleagues did with age-

corrected Full Scale IQ scores, but we also completely failed to find

that either WAIS-RVerbal or Performance Intelligence SSS showed

correlations with the size of any brain structure following age

correction. Our results are therefore, to some extent, in tension with

those of Haier et al. (2004). Although the tension can only be

resolved by future work, we would argue that it is important to check

VBM results with stereological measures, to use raw intelligence

scores and correct for age if some subjects are older than their mid-

30s, and preferably to use larger groups than those of Haier et al.

(2004).

The second puzzle is how the functions supported by the medial

prefrontal region might be relevant for visuospatial fluid intelli-

gence, but not for verbal and perceptual forms of fluid intelligence.

We found that the volume of both the dorsal and ventral medial

prefrontal region correlated with visuospatial fluid intelligence.

This region included dorsal and ventral parts of both BA 10 and the

anterior cingulate cortex (ACC). Ramnani and Owen (2004) have

suggested that BA 10 plays a critical role in the coordination of

multiple related cognitive operations. This is exactly the kind of

function that is likely to be central to the solution of difficult kinds

of novel problem, i.e., for fluid intelligence. BA 10 is well suited to

this role because it receives input largely, if not entirely, from other

supramodal cortex: other pre-frontal cortex regions, ACC, and the

temporal pole so the information that it processes is likely to be

abstract. The very high density of dendritic spines and low density

of cell bodies of BA 10 neurons is also compatible with it playing a

role in integrating its multiple inputs. Paus (2001) has noted that

the ACC seems to come into play when rehearsed routines are

insufficient to guide behavior as applies when fluid intelligence is

exercised. As he indicates, this region has been associated both

with error detection and conflict evaluation, which are functions

likely to be valuable in addressing novel problems. Rushworth et

al. (2004) have related ACC to a role in linking actions (which

could include mental actions) to their consequences, i.e., monitor-

ing the implications of ongoing thought processes to determine

whether it is worth continuing with them.

Although these ideas relate these medial frontal regions to fluid

intelligence, they neither distinguish between different kinds of

fluid intelligence nor between dorsal and ventral regions. The latter

distinction can be made most clearly for the ACC, which divides

into dorsal (BA 24c and 32) and ventral (BA 24a, 24b, 25) regions.

The ventral region is strongly connected with limbic regions such

as the amygdala and ventral striatum that are linked to emotion,

and to thalamic and monoaminergic nuclei linked to arousal and

motivation whereas the dorsal region is connected to frontal sites

such as the lateral frontal cortex that are linked to cognition. Dorsal

ACC and lateral frontal cortex activations are often associated and

these activations tend to be greater for more cognitively demand-

ing and anxiety provoking tasks. In contrast, blood flow decreases

are noted in the ventral ACC as cognitive demands increase, which

suggests that the two regions, although extensively connected,

have distinct functions (see Paus, 2001).

An fMRI study of Gusnard et al. (2001) found evidence that

dorsal and ventral BA 10 as well as dorsal and ventral ACC are

functionally distinct. This study found evidence that the dorsal

medial prefrontal cortex (particularly BA 10 although the

activation extended into the anterior cingulate cortex) is associated

with self-referential mental activity whereas the ventral medial

prefrontal cortex is associated with emotional processing. Duncan

et al. (2000) may have found dorsal medial prefrontal cortex

activation because solving a novel visuospatial problem involves

rapidly forming an impression of what is the correct answer and

then checking this. Forming an impression involves deciding what

feels right to you, i.e., self-referencing. Being good at this means

that your impressions tend to be correct and may be associated with

having a large dorsal medial prefrontal cortex. It seems likely that

this region supports, not just self-referencing, but a range of

executive functions that underlie efficient impression formation.

The strategy used for solving novel verbal and perceptual problems

may be different so that subjects do not often try and decide what

feels like the correct solution to them, and consequently do not

activate the dorsal medial prefrontal region. However, it could be

that the difference in activation patterns relates to the finding that

both dorsal and medial prefrontal regions show default activations

in baseline (or less demanding) states because of the mental state

subjects typically adopt when relatively unoccupied. If verbal and

perceptual novel problem solving does not rely on self-referencing

impression formation to quite the same degree as visuospatial

novel problem solving, then no activation will be found in the

medial prefrontal cortex region relative even to a low-level

baseline where some degree of default activation will be occurring.

Although our finding that the ventral medial prefrontal region is

involved with visuospatial fluid intelligence is supported by not

only VBM, but also by stereology, it is more of a problem because

of the association of this region with emotional processing. It is

also unclear whether activation produced by any task that tapped

fluid intelligence spread into this region in the PET studies of

Duncan et al. (2000) because the coordinates of the most

significant relevant voxel fell in the dorsal medial frontal cortex.

However, as with the dorsal medial prefrontal region, it is likely

that the ventral region performs several functions only partially

captured by the notion of emotional processing. For example,

Bechara et al. (1997) have found support for the suggestion that the

ventromedial cortices hold nondeclarative dispositional knowledge

related to an individual’s previous emotional experiences of similar

situations and that this knowledge facilitates the cognitive

processes of evaluation and reasoning. Bechara’s findings (Bechara

et al., 1997) suggest that, in normal individuals, covert biases

precede overt reasoning about the available facts so as to boost the

efficiency of the reasoning processes required for conscious

decisions.

Our study failed to find any evidence of brain region volume/

visuospatial fluid intelligence correlations that involved posterior

neocortical regions, such as parts of the parietal lobes, believed to

process visuospatial information. Fluid visuospatial intelligence

must depend both on planning and monitoring abilities that may

mainly depend on prefrontal cortex, but also on basic visuospatial

processing that is believed to involve the parietal cortex. For

example, Duncan et al.’s study (Duncan et al., 2000) indicates that

trying to solve Culture Fair Test items that depend strongly on

visuospatial fluid intelligence activates certain bilateral parietal

cortex regions more than trying to solve similar tasks that depend

less on fluid intelligence. Our failure to find that variations in the

volume of parts of the parietal lobe correlate with visuospatial fluid

intelligence may relate to the possibility that the efficiency of the

visuospatial processing parts of this region depends mainly on

Q.-Y. Gong et al. / NeuroImage 25 (2005) 1175–1186 1185

metabolic/biochemical factors and/or white matter connections

rather than on the number and size of its neurons. However, there

were some limitations to the present study and it is possible that

future work that avoids these limitations and uses more sophisti-

cated volumetric procedures may succeed in showing posterior

cortex/visuospatial intelligence correlations.

A limitation of the present investigation was the relatively

small size of the study group. A considerably larger group would

have been needed to provide sufficient statistical power to

examine whether there are sex differences in the way in which

the volume of prefrontal (and possibly other brain structures)

relate to different kinds of fluid intelligence. Another limitation

was the age range of the group we used, given that both fluid

intelligence and the volume of certain brain structures was less in

our older than in our younger subjects so that we had to control

for age in order to identify the brain volume–fluid intelligence

relationship. Other limitations involved techniques used in this

study. The prefrontal subfield volume estimated with stereology

represents the global structural volume containing both gray and

white matter whereas VBM provides a measure of gray matter

concentration. Criticisms have also been made of VBM (Book-

stein, 2001; Crum et al., 2003) so it is desirable to confirm VBM

findings with stereology as we have done. But, if stereology is to

cross-validate VBM appropriately, then ideally it needs to make

volume estimates of gray matter selectively rather than in

combination with white matter. In order to facilitate the focusing

of our image analysis studies on the gray matter compartment, at

the data acquisition stage we are considering using the 3D T1-

weighted MDEFT sequence designed by Deichmann et al. (2004)

to improve contrast between gray and white matter, and also

obtaining multi-slice co-registered series of T1-, T2-, and PD-

weighted images to support improved tissue classification. We are

also in the process of transferring studies to a new 3.0-T MR

system which offers increased signal to noise ratio.

Future work, therefore, should examine a larger group of young

healthy subjects of both sexes, perhaps between 20 and 30 years of

age (so as to avoid the need to control for age), on several different

kinds of fluid and crystallized intelligence test using matching

materials (minimally including visuospatial and verbal tests). This

would allow derivation, using factor analytic methods, of good

estimates of g as well as of different kinds of fluid and crystallized

intelligence. The convergent evidence we have obtained from

VBM and stereology studies represents important methodological

cross-validation of the significance of medial prefrontal cortex in

fluid intelligence. We recommend that our findings are also cross-

validated with functional neuroimaging studies performed on

selected subjects using similar materials to determine to what

degree activated sites and sites that show volume–intelligence

correlations correspond. If our medial prefrontal region findings

are confirmed and their generality across different kinds of fluid

intelligence determined, it will be important to utilize the

developments in MR image acquisition proposed above so as to

potentially identify more precisely those brain regions which are

involved.

Acknowledgments

The authors gratefully acknowledge the technical support of

Dr. J. Brooks. We also wish to express our thanks for the

constructive suggestions of two anonymous reviewers. This

research was supported by grant G9300193 from the Medical

Research Council of the United Kingdom awarded to AM and

NR, and two HERG Research Grants awarded to QYG/NR and

VS/NR, respectively. QYG is a member of the State Key Lab of

Biotherapy and is also supported by the Program for New

Century Excellent Talents in University (NCET) from the

Ministry of Education, PR China.

References

Andreasen, N.C., Flaum, M., Swayze II, V., O’Leary, D.S., Alliger, R.,

Cohen, G., Ehrhardt, J., Yuh, W.T., 1993. Intelligence and brain

structure in normal individuals. Am. J. Psychiatry 150, 130–134.

Bechara, A., Damasio, H., Tranel, D., Damasio, A.R., 1997. Deciding

advantageously before knowing the advantageous strategy. Science 275

(5304), 1293–1295.

Bigler, E.D., Johnson, R., Jackson, C., Blatter, D.D., 1995. Aging, brain

size, and IQ. Intelligence 21, 109–119.

Bookstein, F.L., 2001. bVoxel-based morphometryQ should not be used withimperfectly registered images. NeuroImage 14, 1454–1462.

Cattell, R.B., Cattell, H.E.P., 1973. Measuring Intelligence with the Culture

Fair Tests. Institute for Personality and Ability Testing, Champaign, IL.

Cezayirli, E., 2000. Age related changes in the hippocampal volume, T2

relaxation time, and memory performance in healthy subjects aged from

20–80 years. PhD thesis, University of Liverpool, Liverpool, UK.

Crum, W.R., Griffin, L.D., Hill, D.L., Hawkes, D.J., 2003. Zen and the art

of medical image registration: correspondence, homology, and quality.

NeuroImage 20, 1425–1437.

Cruz-Orive, L.M., 1989. On the precision of systematic sampling: a review

of Matheron’s transitive methods. J. Microsc. 153, 315–333.

Damasio, H., 1995. Human Brain Anatomy in Computerized Images.

Oxford Univ. Press, New York.

Deary, I.J., 2000. Looking Down on Human Intelligence. Oxford Univ.

Press, New York.

Deary, I.J., Simonotto, E., Meyer, M., Marshall, A., Marshall, I., Goddard,

N., Wardlaw, J.M., 2004. The functional anatomy of inspection time: an

event-related fMRI study. NeuroImage 22, 1466–1479.

Deichmann, R., Schwarzbauer, C., Turner, R., 2004. Optimisation of the 3D

MDEFT sequence for anatomical brain imaging: technical implications

at 1.5 and 3 T. NeuroImage 21, 757–767.

Duncan, J., Burgess, P., Emslie, H., 1995. Fluid intelligence after frontal

lobe lesions. Neuropsychologia 33, 261–268.

Duncan, J., Seitz, R.J., Kolodny, J., Bor, D., Herzog, H., Ahmed, A.,

Newell, F.N., Emslie, H., 2000. A neural basis for general intelligence.

Science 289, 457–460.

Flashman, L.A., Andreasen, N.C., Flaum, M., Swayze, V.W., 1998.

Intelligence and regional brain volumes in normal controls. Intelligence

25, 149–160.

Gong, Q.Y., Tan, L.T., Romaniuk, C.S., Jones, B., Brunt, J.N., Roberts, N.,

1999. Determination of tumour regression rates during radiotherapy for

cervical carcinoma by serial MRI: comparison of two measurement

techniques and examination of intraobserver and interobserver varia-

bility. Br. J. Radiol. 72 (853), 62–72.

Gong, Q.Y., Phoenix, J., Kemp, G.J., Garcia-Finana, M., Frostick, S.P.,

Brodie, D.A., Edwards, R.H., Whitehouse, G.H., Roberts, N., 2000.

Estimation of body composition in muscular dystrophy by MRI and

stereology. J. Magn. Reson. Imaging 12, 467–475.

Gong, Q.Y., Eldridge, P.R., Brodbelt, A.R., Garcia-Finana, M., Zaman, A.,

Jones, B., Roberts, N., 2004. Quantification of tumour response to

radiotherapy. Br. J. Radiol. 77 (917), 405–413.

Good, C.D., Johnsrude, I., Ashburner, J., Henson, R.N., Friston, K.J.,

Frackowiak, R.S., 2001a. Cerebral asymmetry and the effects of sex and

handedness on brain structure: a voxel-based morphometric analysis of

465 normal adult human brains. NeuroImage 14, 685–700.

Good, C.D., Johnsrude, I.S., Ashburner, J., Henson, R.N., Friston, K.J.,

Q.-Y. Gong et al. / NeuroImage 25 (2005) 1175–11861186

Gazzaniga, M.S., 2001b. A voxel-based morphometric study of ageing

in 465 normal adult human brains. NeuroImage 14, 21–36.

Gray, J.R., Thompson, P.M., 2004. Neurobiology of intelligence: science

and ethics. Nat. Rev., Neurosci. 5 (6), 471–482.

Gray, J.R., Chabris, C.F., Braver, T.S., 2003. Neural mechanisms of general

fluid intelligence. Nat. Neurosci. 6, 316–322.

Gundersen, H.J., Jensen, E.B., 1987. The efficiency of systematic sampling

in stereology and its prediction. J. Microsc. 147 (Pt. 3), 229–263.

Gundersen, H.J., Jensen, E.B., Kieu, K., Nielsen, J., 1999. The efficiency of

systematic sampling in stereology—reconsidered. J. Microsc. 193,

199–211.

Gur, R.E., Cowell, P.E., Latshaw, A., Turetsky, B.I., Grossman, R.I.,

Arnold, S.E., Bilker, W.B., Gur, R.C., 2000. Reduced dorsal and orbital

prefrontal gray matter volumes in schizophrenia. Arch. Gen. Psychiatry

57, 761–768.

Gusnard, D.A., Akbudak, E., Shulman, G.L., Raichle, M.E., 2001. Medial

prefrontal cortex and self-referential mental activity: relation to a default

mode of brain function. Proc. Natl. Acad. Sci. U. S. A. 98, 4259–4264.

Haier, R.J., Jung, R.E., Yeo, R.A., Head, K., Alkire, M.T., 2004. Structural

brain variation and general intelligence. NeuroImage 23, 425–433.

Howard, M.A., Cowell, P.E., Boucher, J., Broks, P., Mayes, A., Farrant, A.,

Roberts, N., 2000. Convergent neuroanatomical and behavioural

evidence of an amygdala hypothesis of autism. NeuroReport 11,

2931–2935.

Howard, M.A., Roberts, N., Garcia-Finana, M., Cowell, P.E., 2003. Volume

estimation of prefrontal cortical subfields using MRI and stereology.

Brain Res. Brain Res. Protoc. 10, 125–138.

Jung, R.E., Yeo, R.A., Chiulli, S.J., Sibbitt Jr., W.L., Weers, D.C., Hart,

B.L., Brooks, W.M., 1999. Biochemical markers of cognition: a proton

MR spectroscopy study of normal human brain. NeuroReport 10,

3327–3331.

Jung, R.E., Yeo, R.A., Chiulli, S.J., Sibbitt Jr., W.L., Brooks, W.M., 2000.

Myths of neuropsychology: intelligence, neurometabolism, and cogni-

tive ability. Clin. Neuropsychol. 14, 535–545.

Keller, S.S., Mackay, C.E., Barrick, T.R., Wieshmann, U.C., Howard,

M.A., Roberts, N., 2002. Voxel-based morphometric comparison of

hippocampal and extrahippocampal abnormalities in patients with left

and right hippocampal atrophy. NeuroImage 16, 23–31.

Kroger, J.K., Sabb, F.W., Fales, C.L., Bookheimer, S.Y., Cohen, M.S.,

Holyoak, K.J., 2002. Recruitment of anterior dorsolateral prefrontal

cortex in human reasoning: a parametric study of relational complexity.

Cereb. Cortex 12, 477–485.

MacLullich, A.M., Ferguson, K.J., Deary, I.J., Seckl, J.R., Starr, J.M.,

Wardlaw, J.M., 2002. Intracranial capacity and brain volumes are

associated with cognition in healthy elderly men. Neurology 59,

169–174.

Neisser, U., Boodoo, G., Bouchard, T.J., Boykin, A.W., Brody, N., Ceci,

S.J., Halpern, D.F., Loehlin, J.C., Perloff, R., Stermberg, R.J.,

Urbina, S., 1996. Intelligence: knowns and unknowns. Am. Psychol.

51, 77–101.

Paus, T., 2001. Primate anterior cingulate cortex: where motor control,

drive and cognition interface. Nat. Rev., Neurosci. 2 (6), 417–424.

Plomin, R., Kosslyn, S.M., 2001. Genes, brain and cognition. Nat.

Neurosci. 4, 1153–1154.

Postle, B.R., Berger, J.S., Taich, A.M., D’Esposito, M., 2000. Activity in

human frontal cortex associated with spatial working memory and

saccadic behavior. J. Cogn. Neurosci. 12 (Suppl. 2), 2–14.

Ramnani, N., Owen, A.M., 2004. Anterior prefrontal cortex: insights into

function from anatomy and neuroimaging. Nat. Rev., Neurosci. 5 (3),

184–194.

Rushton, J.P., Ankney, C.D., 1995. Brain size matters: a reply to Peters.

Can. J. Exp. Psychol. 49, 562–569 (discussion 570–566).

Rushworth, M.F., Walton, M.E., Kennerley, S.W., Bannerman, D.M., 2004.

Action sets and decisions in the medial frontal cortex. Trends Cogn. Sci.

8 (9), 410–417.

Sluming, V., Barrick, T., Howard, M., Cezayirli, E., Mayes, A., Roberts, N.,

2002. Voxel-based morphometry reveals increased gray matter density

in Broca’s area in male symphony orchestra musicians. NeuroImage 17,

1613–1622.

Talairach, J., Tournoux, P., 1988. Co-plannar Stereotaxic Atlas of the

Human Brain. Thieme, New York.

Thompson, P.M., Cannon, T.D., Narr, K.L., van Erp, T., Poutanen, V.P.,

Huttunen, M., Lonnqvist, J., Standertskjold-Nordenstam, C.G., Kaprio,

J., Khaledy, M., Dail, R., Zoumalan, C.I., Toga, A.W., 2001. Genetic

influences on brain structure. Nat. Neurosci. 4, 1253–1258.

Tisserand, D.J., Bosma, H., Van Boxtel, M.P., Jolles, J., 2001. Head size

and cognitive ability in nondemented older adults are related.

Neurology 56, 969–971.

Tramo, M.J., Loftus, W.C., Stukel, T.A., Green, R.L., Weaver, J.B.,

Gazzaniga, M.S., 1998. Brain size, head size, and intelligence quotient

in monozygotic twins. Neurology 50, 1246–1252.

Wechsler, D., 1981. Wechsler Adult Intelligence Scale—Revised. The

Psychological Corporation, New York.

Wilke, M., Sohn, J.H., Byars, A.W., Holland, S.K., 2003. Bright spots:

correlations of gray matter volume with IQ in a normal pediatric

population. NeuroImage 20, 202–215.

Woodcock, R.W., 1990. Theoretical foundations of the WJ-R measures of

cognitive ability. J. Psychoeduc. Assess. 8, 231–258.