Neuropsychologia 62 (2014) 152–162

Contents lists available at ScienceDirect

Neuropsychologia
journal homepage: www.elsevier.com/locate/neuropsychologia

The architecture of the chess player's brain

Jürgen Hänggi a,n, Karin Brütsch a,b, Adrian M. Siegel c, Lutz Jäncke a,d,e,f,g,nn
a
Division Neuropsychology, Institute of Psychology, University of Zurich, Binzmuhlestrasse 14/Box 25, CH-8050 Zurich, Switzerland
b
Rehabilitation Center Affoltern am Albis, University Children's Hospital, Zurich, Switzerland
c
Division of Neurology, Zuger Kantonsspital, Switzerland
d
International Normal Aging and Plasticity Imaging Center, University of Zurich, Switzerland
e
Center for Integrative Human Physiology, University of Zurich, Switzerland
f
University Research Priority Program (URPP), Dynamic of Healthy Aging, University of Zurich, Switzerland
g
Department of Special Education, King Abdulaziz University, Jeddah, Saudi Arabia

art ic l e i nf o a b s t r a c t

Article history: The game of chess can be seen as a typical example for an expertise task requiring domain-specific
Received 3 April 2014 training and experience. Despite intensive behavioural studies the neural underpinnings of chess
Received in revised form performance and expertise are not entirely understood. A few functional neuroimaging studies have
15 June 2014
shown that expert chess players recruit different psychological functions and activate different brain
Accepted 17 July 2014
Available online 24 July 2014
areas while they are engaged in chess-related activities. Based on this functional literature, we predicted
to find morphological differences in a network comprised by parietal and frontal areas and especially the
Keywords: occipito-temporal junction (OTJ), fusiform gyrus, and caudate nucleus. Twenty expert chess players and
Neuroplasticity 20 control subjects were investigated using voxel-based and surface-based morphometry as well as
Structural magnetic resonance imaging
diffusion tensor imaging. Grey matter volume and cortical thickness were reduced in chess players
Neuromorphometry
compared with those of control men in the OTJ and precunei. The volumes of both caudate nuclei were
Grey matter
Cortical thickness not different between groups, but correlated inversely with the years of chess playing experience. Mean
Occipito-temporal junction diffusivity was increased in chess players compared with that of controls in the left superior longitudinal
Caudate nucleus fasciculus and the Elo score (a chess tournament ranking) was inversely related to mean diffusivity
White matter within the right superior longitudinal fasciculus. To the best of our knowledge we showed for the first
Mean diffusivity time that there are specific differences in grey and white matter morphology between chess players and
Superior longitudinal fasciculus control subjects in brain regions associated with cognitive functions important for playing chess.
Chess expertise
Whether these anatomical alterations are the cause or consequence of the intensive and long-term chess
Elo score
training and practice remains to be shown in future studies.
Deliberate practice
& 2014 Elsevier Ltd. All rights reserved.

1. Introduction that to reach grandmaster level in chess at least 10 years of intense

practice is required. Several recent studies support this notion
To perform chess requires the orchestration of several psycho- by showing a monotonic relationship between deliberate chess
logical functions among them attention, perceptual grouping, and practice and chess performance using longitudinal (De Bruin,
various memory functions (Bilalić, McLeod, & Gobet, 2009; Gobet, Smits, Rikers, & Schmidt, 2008) and retrospective study designs
1998; Gobet & Waters, 2003). Currently there is consensus among (Charness, Tuffiash, Krampe, Reingold, & Vasyukova, 2005). Thus,
cognitive psychologists that chess performance needs substantial the positive relation between accumulated amount of chess
practice. Thus, chess is seen as a typical example for an expertise practice and performance resembles the practice–performance
task requiring domain-specific experience. In this context Simon relation in other fields such as music, sports, and teaching
and Chase (1973) formulated the 10-year rule of expertise, stating (Ericsson, 2004). However, although practice is important in
determining chess performance, general intelligence, visuospatial
intelligence, and talent are also discussed as factors contributing to
n
Corresponding author. Tel.: þ 41 44 635 73 97; fax: þ 41 44 635 74 09. chess performance. Thus, some authors favour a kind of interac-
nn
Corresponding author at: Division Neuropsychology, Institute of Psychology, tion between practice and talent explaining chess performance
University of Zurich, Binzmuhlestrasse 14/Box 25, CH-8050 Zurich, Switzerland. (Howard, 2013).
Tel.: þ 41 44 635 74 01; fax: þ41 44 635 74 09.
Despite intensive behavioural studies the neural underpinnings
E-mail addresses: [email protected] (J. Hänggi),
[email protected] (K. Brütsch), [email protected] (A.M. Siegel), of chess performance and expertise are not entirely understood.
[email protected] (L. Jäncke). A few neuroimaging studies have been conducted in the last 20

http://dx.doi.org/10.1016/j.neuropsychologia.2014.07.019
0028-3932/& 2014 Elsevier Ltd. All rights reserved.
 J. Hänggi et al. / Neuropsychologia 62 (2014) 152–162 153

years examining brain activations in the context of chess using voxel-based (VBM) and surface-based morphometry (SBM) in the
different methods and experimental paradigms. Some studies context of structural MRI to delineate specific anatomical features
featured only novices (Atherton, Zhuang, Bart, Hu, & He, 2003; of the white matter (WM) and GM architecture. With the above-
Nichelli et al., 1994; Onofrj et al., 1995), examined experts during mentioned functional findings in mind, we aimed at investigating
chess-related-visual search tasks (Campitelli, Gobet, Head, whether brain regions differentially activated as a function of
Buckley, & Parker, 2007), compared experts with novices during chess expertise are also morphologically altered as a result of
chess-related and chess-unrelated visual perception tasks (Bilalić, structural neuroplasticity induced by intensive, long-term chess
Langner, Erb, & Grodd, 2010; Bilalić, Langner, Ulrich, & Grodd, training and extensive chess playing experience that included
2011), or measured brain activations in chess experts or novices international chess tournaments.
during chess moves or at rest (Amidzic, Riehle, Fehr, Wienbruch, & Our hypotheses can be divided into three broad classes:
Elbert, 2001; Duan et al., 2012a, 2012b). (1) predictions with respect to brain regions involved in chess-
Taken together, these studies have shown that expert chess related “higher cognitive functions” such as probed in the con-
players recruit different psychological functions and activate trasts reported above by the following studies (Amidzic et al.,
different brain areas while they are engaged in chess-related 2001; Atherton et al., 2003; Duan et al., 2012a, 2012b; Nichelli
activities. For example, in chess-related visual search tasks only et al., 1994; Wan et al., 2011). (2) Predictions with respect to those
chess-experts demonstrate bilateral activation in brain areas brain areas associated with chess-related “lower cognitive
involved in object perception (bilateral posterior temporal areas, mechanisms” such as the recognition of objects and their relations
fusiform gyrus, left inferior parietal lobe, and especially in the (Bilalić et al., 2011, 2010). (3) A third prediction is related to the
occipito-temporal junction, OTJ) or expertise-related pattern putative specific role of the caudate nucleus in chess playing
recognition (bilateral retrosplenial cortex and bilateral collateral (Duan et al., 2012a).
sulcus) (Bilalić et al., 2010, 2011). In addition, during chess games According to these three prediction classes we hypothesise that
only experts recruit brain areas involved in memory chunking, the chess player's brain will demonstrate specific anatomical
knowledge storage, and retrieval (frontal and parietal lobes) while features in (1) brain areas associated with higher cognitive func-
novices recruit more strongly the medial temporal cortex (e.g. the tions (e.g. memory and executive functions) in a distributed
hippocampus) that is more strongly involved in learning and cortical network including parietal, frontal, and temporal regions.
retrieving of new information (Amidzic et al., 2001). In a more (2) With respect to the second class of predictions we anticipate
recent study by Duan et al. the authors examined brain activations specific anatomical features in brain areas that have been shown
during the performance of a Chinese chess-based problem-solving to be involved in the recognition of objects and their relations (e.g.
task and identified that chess masters and grandmasters demon- the OTJ and fusiform gyrus). (3) Finally, we predict anatomical
strated increased activation in neural networks involved in the alterations in the caudate nucleus as suggested by Duan et al.
control of cognitive functions such as attention, executive func- (2012a).
tions, and problem-solving (Duan et al., 2012b). The new finding of Besides DTI that investigates differences in WM architecture,
this study was that the grandmasters and masters also showed a we used VBM and SBM techniques in order to measure morpho-
much stronger suppression of the default mode network (DMN) logical GM differences in the hypothesised brain areas. Although
during chess-related problem-solving than the novices. In addi- both techniques are based on the same T1-weighted MR images,
tion, when examining the resting-state functional connectivity in they measure different morphological features. With VBM we
the DMN they uncovered that, compared to novices, the DMN in examine probabilistic GM volume differences while the SBM
grandmasters and masters showed an increased functional con- technique provides us cortical thickness measures. We used both
nectivity between the caudate nucleus and the DMN network techniques in order to track down whether between-group differ-
hubs, suggesting an important role of the DMN-caudate nucleus ences potentially found in the VBM analysis are driven by
loop in chess expertise. differences in cortical thickness.
Taken together, there are substantial differences between chess
experts in terms of the involved psychological functions as well as
2. Materials and methods
substantial differences in terms of the underlying neurophysiolo-
gical systems. Most likely these differences have been established
2.1. Subjects
as a consequence of intensive chess practice. Since intensive
practice of cognitive and motor tasks is associated with substantial Twenty expert male chess players with a mean age of 28.9 years (standard
anatomical changes (Jancke, 2009; May, 2011; Zatorre, Fields, & deviation, SD 76.03 years) and 20 control male subjects matched for age, handed-
Johansen-Berg, 2012) it is most likely that those brain regions, ness, and education participated in the study (mean age7 SD; 27.2 7 6.32 years).
which are most strongly involved in controlling chess tasks, have The Elo score (Elo, 1978), an international chess tournament classification system,
was used to evaluate chess playing proficiency (Table 1). The sample of chess
been changed by neuroplastic mechanisms. However, little is
players included three grandmasters, seven international masters, and six FIDE
known so far with respect to the anatomical plasticity in chess (Fédération Internationale des Échecs) masters. The rest of the chess players did
players. To the best of our knowledge there is only one study, not have any international chess title. Handedness was evaluated according to the
which has examined structural differences in chess experts com- procedure proposed by Annett (1970). The participants had no history of neuro-
pared to chess novices (Duan et al., 2012b). In this study chess logical, neuropsychological, or psychiatric disease, and denied taking illegal drugs
or medication. The research reported in the present study was conducted according
experts compared with novices showed reduced grey matter (GM) to the principles expressed in the Declaration of Helsinki. The local ethics
volume bilaterally in the caudate nucleus. The authors associate committee of the canton Zurich approved the study and written informed consent
this GM volume reduction with the idea of synaptic pruning and was obtained from all participants.
the beneficial elimination of redundant and unused synapses
within the caudate nucleus allowing more efficient local computa- 2.2. Neuropsychological assessment
tions and more efficient integration of information from adjacent
brain regions. Several cognitive domains were assessed in the expert chess players and
In the present study, we conducted a comprehensive neuroa- control men. Fluid intelligence was measured with Raven's advanced progressive
matrices (Raven, 1998). Visuospatial abilities, especially mental rotation perfor-
natomical analysis in order to identify characteristic neuroanato- mance, were assessed with the mental rotation test version A (Peters et al., 1995).
mical differences between expert and novice chess players. For The immediate visuospatial block span was evaluated with the block-tapping test
this purpose we applied diffusion tensor imaging (DTI) as well as (similar to the block of Corsi) (Schellig, 1997). Due to logistical restrictions and time
154 J. Hänggi et al. / Neuropsychologia 62 (2014) 152–162

Table 1
Demographic, behavioural, and global brain measures of the chess players and control men.

Chess players (n¼20) Control men (n¼ 20) Significance

Measure Mean SD Min. Max. Mean SD Min. Max. p-value

Age (years) 28.9 6.02 19.7 41.2 27.2 6.32 17.4 42.8 0.38
Education (years) 16.6 2.09 13.0 21.0 17.6 4.29 12.0 28.0 0.33
Elo score 2366 107 2187 2560 – – – – –
Age at chess playing commencement (years) 7.9 3.08 4.0 14.0 – – – – –
Experience of chess playing (years) 19.7 6.58 8.0 31.0 – – – – –
Total chess playing (hours) 13231 14416 2496 52000 – – – – –
Raven's advanced progressive matrices 28.3 5.08 20.0 35.0 28.5 3.25 23.0 36.0 0.85
Mental rotation task – version A 12.8 5.07 5.0 23.0 13.1 3.23 9.0 21.0 0.83
Block tapping test (immediate block span) 6.3 0.97 5.0 8.0 6.1 0.95 5.0 8.0 0.37
Intracranial volume (cm3) 1271.1 100.5 1076.0 1473.6 1223.9 136.9 991.1 1517.6 0.22
Total left cortical grey matter volume (cm3) 262.3 20.0 238.0 312.7 255.4 19.8 228.5 301.8 0.28
Total right cortical grey matter volume (cm3) 265.8 21.0 239.8 316.5 257.9 20.7 228.1 310.1 0.24
Total left cortical surface area (cm2) 1105.7 57.97 1015.3 1249.3 1063.4 74.06 922.7 1226.5 0.052
Total right cortical surface area (cm2) 1110.6 60.05 1028.1 1240.0 1070.0 74.47 924.7 1235.2 0.065
Average left cortical thickness (mm) 2.578 0.103 2.427 2.732 2.607 0.111 2.367 2.847 0.40
Average right cortical thickness (mm) 2.591 0.109 2.423 2.775 2.617 0.118 2.376 2.830 0.47
Total left cortical white matter volume (cm3) 272.2 22.7 232.3 325.0 261.0 26.8 221.7 304.3 0.16
Total right cortical white matter volume (cm3) 273.4 23.5 234.1 330.2 262.1 26.2 225.0 307.2 0.16
Total left cortical white matter area (cm2) 925.9 46.9 835.9 1028.6 895.0 67.3 751.2 1003.4 0.10
Total right cortical white matter area (cm2) 932.4 51.7 841.0 1032.9 902.3 65.5 759.7 1023.8 0.11

Abbreviations: Max., maximum; Min., minimum; n, number of subjects; p-value, error probability; and SD, standard deviation.

constraints, it was not possible to test the expert chess players cognitively in more 2011), superior parietal lobule (Atherton et al., 2003), inferior parietal lobule
detail and therefore we also did not assess long-term memory performance. comprised by the angular gyrus, supramarginal gyrus (anterior and posterior part),
and the parietal operculum (Amidzic et al., 2001; Atherton et al., 2003; Bilalić et al.,
2010, 2011; Nichelli et al., 1994), occipito-temporal junction (Bilalić et al., 2010,
2.3. MRI data acquisition 2011) parieto-occipital junction (Nichelli et al., 1994), precuneus (Wan et al., 2011),
occipital pole and intracalcarine cortex (Nichelli et al., 1994), fusiform gyrus
Magnetic resonance imaging (MRI) scans were acquired on a 3.0 T Philips (temporal and occipital parts) (Bilalić et al., 2010, 2011), and the premotor cortex
Achieva whole body scanner (Philips Medical Systems, Best, The Netherlands) (Atherton et al., 2003; Nichelli et al., 1994). The anterior cingulate cortex was
equipped with a transmit-receive body coil and a commercial eight-element head investigated in addition due to its prominent role in error detection, a cognitive
coil array capable for sensitivity encoding (SENSE). Three volumetric 3D T1- function also important for the game of chess. The masks of these regions of
weighted gradient echo sequence (fast field echo) scans were obtained from all interest were derived from the Juelich histological atlas (premotor cortex) (https://
40 participants. Slices were acquired in the sagittal plane with a measured and www.jubrain.fz-juelich.de/apps/cytoviewer/cytoviewer-main.php), the Harvard-
reconstructed spatial resolution of 0.94  0.94  1.00 mm3 (matrix 256  256 pix- Oxford subcortical (caudate nucleus) and cortical (all other ROIs) structural atlas
els, 160 slices). Further imaging parameters were: field of view (http://www.cma.mgh.harvard.edu/fsl_atlas.html) as implemented in FSL.
FOV ¼240  240 mm2, echo-time TE ¼3.7 ms, repetition-time TR¼ 8.06 ms, flip- Although the caudate nucleus is included in the VBM analysis, which operates
angle α ¼81, and SENSE factor SF¼ 2.1. Scan time was about 8 min/scan. The three at the voxel level, we also investigated the caudate nucleus as a whole using the
T1-weighted MRI scans were realigned and averaged to obtain a single image with fully automated subcortical segmentation procedure implemented in the Free-
an increased signal-to-noise and contrast-to-noise ratio. Surfer software suite (http://surfer.nmr.mgh.harvard.edu/fswiki) that provides the
Two identical diffusion-weighted spin-echo echo-planar imaging sequences total volume of the caudate nucleus. To rule out that local brain differences are
were applied to all 40 participants' brain. Slices were acquired in the transversal confounded by global brain differences, we also computed global brain measures
plane with a measured and reconstructed spatial resolution of 2.0  2.0  2.0 mm3 such as intracranial volume, hemispheric GM and WM volumes and surface area as
(matrix 112  112 pixels, 75 slices). Further imaging parameters were: field of view well as cortical thickness using the standard procedure of FreeSurfer.
FOV ¼224  224 mm2, echo-time TE ¼ 55 ms, repetition-time TR ¼13,472 ms, flip-
angle α¼ 901, and SENSE factor SF¼ 2.1. Diffusion was measured in 32 non-collinear
directions with a b-value of b ¼1000 s/mm2 preceded by a non-diffusion-weighted 2.5. Analysis of fractional anisotropy and mean, axial and radial diffusivity
volume (reference volume). Scan time was about 10 min/sequence. The scans from
the two DTI sequences were realigned and averaged to obtain diffusion-weighted Here we applied tract-based spatial statistics (TBSS) (Smith et al., 2006) using
data with an increased signal-to-noise ratio. FSL (FMRIB Software Library; http://www.fmrib.ox.ac.uk/fsl/) (Smith et al., 2004)
tools such as the FDT (FMRIB Diffusion Toolbox) (Behrens et al., 2003) to create
fractional anisotropy (FA), mean, axial and radial diffusivity maps. FA, a marker for
2.4. Voxel-based morphometry – analysis of probabilistic grey matter volume the integrity of WM, is a measure of the degree of directional preference of water
diffusion (Basser, Mattiello, & LeBihan, 1994). Mean diffusivity is the average of the
Between-group differences in GM volume were evaluated by using voxel-based diffusion tensor's three eigenvalues (λ1 þλ2þ λ3)/3 and represents the amount of
morphometry (VBM) (Ashburner & Friston, 2000; Good et al., 2001a). T1-weighted diffusion independent of the diffusion direction. Axial diffusivity, a parameter
MRI scans were analysed with the FSL-VBM tool (Douaud et al., 2007) (http://fsl. reflecting the principal direction of diffusion in WM (Song et al., 2002), was
fmrib.ox.ac.uk/fsl/fslwiki/FSLVBM), an optimised VBM protocol (Good et al., 2001b) estimated using the first eigenvalue (λ1) of the diffusion tensor. Axial diffusivity
that was carried out with FMRIB Software Library (FSL) version 5.0.1 (Smith et al., reflects rather the properties of the axonal membrane than the properties of
2004) (www.fmrib.ox.ac.uk/fsl). First, structural images were brain-extracted and myelin. The second and third eigenvalues were averaged and referred to as radial
segmented into different tissue classes before being registered to the MNI 152 diffusivity (λ2þλ3)/2 (Basser et al., 1994; Song et al., 2002), rather a marker for the
standard space using non-linear registration. The resulting images were averaged properties of the myelin sheets than the properties of the axonal membrane. The
and flipped along the x-axis to create a left–right symmetric, study-specific GM following steps were realised: Eddy current and head movement corrections were
template. Second, all native GM images were non-linearly registered to this study- applied using the EDDY_CORRECT tool of FDT. An individual binary brain mask was
specific template and “modulated” to correct for local expansion (or contraction) created on the non-diffusion weighted images using BET (Brain Extraction Tool).
due to the non-linear component of the spatial transformation. The modulated GM Tensors were fitted to the data using the DTIFIT tool of FDT after the b-vectors were
images were then smoothed with an isotropic Gaussian kernel with a sigma of adjusted for the rotations introduced by head movement correction. Linear and
5.0 mm (corresponding to a kernel with a full width at half-maximum of about non-linear spatial registrations of the FA map into a standard stereo-tactic space
11.5 mm). These GM maps were then subjected to statistical analyses (see below). (MNI space represented by the FMRIB58-FA template) were applied using FSL's
We restricted the statistical analysis to cortical and subcortical structures, registration tools FLIRT and FNIRT, respectively. These transformations were then
which turned out to be specific, or at least, differentially involved in chess playing. applied to the mean, axial and radial diffusivity maps using the TBSS_NON_FA tool.
Structures included the caudate nucleus (Duan et al., 2012a, 2012b; Wan et al., Next, the mean of all subjects' aligned FA images is created, and then ‘thinned’
 J. Hänggi et al. / Neuropsychologia 62 (2014) 152–162 155

using standard image processing techniques to create a mean FA skeleton that distribution by permuting the subject's group membership randomly. Chess players
represents the centres of major tracts common to the group of subjects. Each and controls were contrasted with respect to local probabilistic GM volume (VBM)
subject's aligned FA data is then projected (perpendicular to the local tract and local FA, mean, axial, and radial diffusivity (DTI). Within the chess players, the
direction) onto this skeleton so that the projected FA values are taken from the Elo score, age of chess playing commencement, years of chess playing experience,
centres of the tracts in the original FA image (Smith et al., 2007). These maps were and lifetime hours of chess training have been correlated with the local brain
then subjected to statistical analyses (see below). measures (derived from the VBM and DTI analyses). As described above, we
To restrict the statistical analysis to the fibre tracts, which interconnect the GM restricted the statistical analyses to a priori defined regions of interest. The peak
regions associated with chess playing (see above), we used probability maps of the coordinates of the clusters reported are in Montreal Neurological Institute (MNI)
superior longitudinal fasciculus (arcuate and temporal part) and the cingulum stereotactic space.
(cingulate and hippocampal part) derived from the JHU (Johns Hopkins University)
WM tractography atlas (Hua et al., 2008) implemented in FSL. The probability maps
2.7.2. Surface-based morphometry
were neither thresholded nor combined into a single mask.
For the exploratory and confirmatory whole brain analysis of the SBM data, a
GLM based on parametric statistics was applied within FreeSurfer (MRI_GLMFIT
2.6. Surface-based morphometry – analysis of cortical thickness tool) without applying the TFCE technique that has been used in the VBM and DTI
analyses. Although the SBM data reported in the present study are not corrected for
Cortical surface reconstruction, cortical parcellation, and subcortical volumetric multiple comparisons, we indicated the clusters that survive a Monte Carlo
segmentation were performed with the FreeSurfer software (version 5.3.0), which simulation based on the cluster extent and 5000 permutations using FreeSurfer's
is documented and available online (http://surfer.nmr.mgh.harvard.edu/). The MRI_GLMFIT-SIM tool. Here, synthetic random z-fields were generated, smoothed,
technical details of these procedures are described in prior publications (Dale, and then statistically tested to assess how many times the resulting randomly
Fischl, & Sereno, 1999; Fischl & Dale, 2000; Fischl et al., 2001, 2004). The 3D produced clusters reach or exceed the size of the true clusters. Error probability
structural T1-weighted MRI scans were used to construct models of each subject's was set at po 0.05 in both the uncorrected and corrected analyses. In the
cortical surface in order to measure cortical thickness and cortical surface area. This uncorrected analysis, we additionally applied a cluster extent threshold that only
fully automated procedure comprised segmentation of the cortical WM (Dale et al., considers clusters larger than 200 mm2 in size. The size of this cluster threshold
1999), tessellation of the GM/WM junction, inflation of the folded surface tessella- was determined arbitrarily. Combining a height threshold with a cluster extent
tion patterns (Fischl, Sereno & Dale, 1999) and automatic correction of topological threshold helps guarding against spurious findings because false positives do not
defects in the resulting manifold (Fischl et al., 2001). This surface was then used as cluster in space.
starting point for a deformable surface algorithm designed to find the GM/WM and
pial (GM/cerebrospinal fluid (CSF)) surfaces with sub-millimetre precision (Fischl & 2.7.3. Demographic, behavioural, and global brain characteristics
Dale, 2000). The procedures for measuring cortical thickness have been validated For the comparison of demographic, behavioural, and global brain measures
against histological analysis (Rosas-Cholula et al., 2013) and manual measurements between groups (independent t-tests) as well as for the correlation between the Elo
(Kuperberg et al., 2003; Salat et al., 2004). This method uses both intensity and score and total hours of training (Spearman's correlation) within chess players IBM
continuity information from the surfaces in the deformation procedure in order to SPSS statistics version 20 was used (SPSS, an IBM company, Armonk, New York).
interpolate surface locations for regions in which the MR image is ambiguous Error probabilities for the comparisons of the demographic, behavioural, and global
(Fischl & Dale, 2000). For each subject, cortical thickness of the cortical ribbon was brain measures were corrected for multiple comparisons using Bonferroni correc-
computed on a uniform grid (comprised by vertices) with 1 mm spacing across tion. Error probability is indicated two-tailed if not otherwise stated. Instead of
both cortical hemispheres, with the thickness being defined by the shortest Pearson's correlation, Spearman's correlation was used where parametric assump-
distance between the GM/WM and pial surface models. The thickness maps tions were not given (see below).
produced are not limited to the voxel resolution of the image and thus sensitive
for sub-millimetre differences between groups (Fischl & Dale, 2000). The way in
which the resolution of the cortical thickness maps goes beyond the resolution of
the original acquisition is conceptually similar to a (conventional) partial volume
3. Results
correction procedure. The cortex is smooth at the spatial scale of several
millimetres, which is imposed as constraint by FreeSurfer to estimate the location 3.1. Demographic, behavioural, and global brain characteristics
of the surface with subvoxel accuracy. For instance, if a given voxel is darker than
its neighbouring GM it probably contains more CSF and so the surface model is at a
Demographic, behavioural and global brain characteristics are
slightly different position than if the neighbouring voxels were brighter and
therefore contain probably more WM. Cortical thickness, surface area, and volume summarised in Table 1. There were no significant differences
measures were mapped to the inflated surface of each participant's brain recon- between expert chess players and controls with respect to age,
struction; thus allowing visualisation of data across the entire cortical surface (gyri education, and performance in the Raven's advanced progressive
and sulci) without the data being obscured by cortical folding. Data were re- matrices test (Raven, 1998), mental rotation test version A (Peters
sampled for all subjects and rendered onto a common spherical coordinate system
(Fischl, Sereno, Tootell & Dale, 1999). Then surface-based vertex-wise cortical
et al., 1995), and block-tapping test (Schellig, 1997) as well as with
thickness maps were computed for each hemisphere and participant. For the respect to intracranial volume, left and right total cortical GM and
whole-brain vertex-wise analysis, the data were smoothed on the surface tessella- WM volume and surface area, and average cortical thickness as
tion using an iterative nearest-neighbour averaging procedure with 139 iterations revealed by t-tests for independent samples.
on the left hemisphere and 138 iterations on the right hemisphere, corresponding
With respect to the weekly and lifetime hours of chess training,
to a 2D surface-based diffusion smoothing kernel with a FWHM of about 15 mm.
These maps were subjected to statistical whole brain analyses (see below). four expert chess players showed extreme values. Therefore, a
procedure for rank ordered data (Spearman's correlation) was
2.7. Statistical analyses applied for correlations involving theses two variables because
outliers do not bias nonparametric tests. The Elo score was weakly
2.7.1. Voxel-based morphometry and diffusion tensor imaging but positively correlated with the amount of time spent for chess
For the VBM and DTI data, voxel-wise general linear models (GLM) were training per week (Spearman's correlation r ¼0.36, p¼ 0.061, one-
applied using permutation-based non-parametric statistical procedures (Hayasaka tailed) as well as correlated with lifetime hours of chess training
& Nichols, 2003; Nichols & Holmes, 2002) that also correct for multiple compar-
(Spearman's correlation r ¼ 0.29, p ¼ 0.108, one-tailed), but these
isons across space (FSL's randomise tool). The threshold free cluster enhancement
(TFCE) technique was used in addition (Smith & Nichols, 2009). The TFCE technique correlations did not reach statistical significance (see Section 4).
addresses problems of smoothing, threshold dependence and localisation in cluster The Elo score of the chess players neither significantly correlated
inference. Many image enhancement and thresholding techniques make use of with age at chess playing commencement (Pearson's correlation
spatial neighbourhood information to boost the validity in extended areas of signal r ¼0.05, p ¼0.42, one-tailed) nor with the years of chess playing
(Smith & Nichols, 2009). The most common such approach in neuroimaging is
experience (Pearson's correlation r ¼0.23, p¼ 0.16, one-tailed).
cluster-based thresholding, which is often more sensitive than voxel-wise thresh-
olding. However, a limitation is the need to define the initial cluster-forming
threshold (Smith & Nichols, 2009). The TFCE technique avoids this problem and it is 3.2. Group differences in probabilistic grey matter volume (VBM)
recommended to use this technique when analysing VBM and DTI data with FSL's and white matter diffusivity (DTI)
randomise tool.
Error probability was set at p o 0.05 corrected for multiple comparisons using
5000 permutations in all analyses. This permutation-based non-parametric proce- GM volume was reduced in chess players compared with that
dure that also corrects for multiple comparisons across space derives the null of control men in a cluster located in the left OTJ (Fig. 1B, Table 2).
156 J. Hänggi et al. / Neuropsychologia 62 (2014) 152–162

Fig. 1. Regions with altered grey and white matter architecture in chess players. Shown are clusters with increased mean diffusivity (red–yellow) in the left superior
longitudinal fasciculus (SLF, A) and decreased probabilistic grey matter volume (blue–lightblue) in the left occipito-temporal junction (OTJ, B) in chess players compared with
control men. The regions of interest subjected to the statistical analysis are shown in green–lightgreen. Colour bars indicate alpha error probability that is corrected for
multiple comparisons. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

Table 2
Regions with increased mean diffusivity and decreased probabilistic grey matter volume in chess players compared with control men.

Name Letter in Fig. 1 Number of voxels Error probability MNI coordinates

x y z

Increased mean diffusivity

Left superior longitudinal fasciculus (below supramarginal gyrus) A 729 0.035  38  43 24

Decreased grey matter volume

Left occipito-temporal junction (posterior middle temporal gyrus) B 44 0.041  52  44 4

p-values are corrected for multiple comparisons using permutation-based non-parametric testing across space (FSL's randomise tool with 5000 permutations).
Abbreviations: Fig., figure; MNI, Montreal Neurological Institute.

There was no cluster with increased GM volume in chess players mean diffusivity in chess players compared with that of controls.
compared with that of controls. A whole-brain analysis did not There were also no significant differences between the groups
reveal any additional statistically significant cluster with GM with respect to FA and axial and radial diffusivities. Whole brain
volume changes (when applying correction for multiple compar- analyses did not reveal any additional statistically significant
isons) and the cluster in the left OTJ (see above) showed only a (when applying correction for multiple comparisons) cluster with
slightly increased error probability of p ¼0.072 in the whole-brain changed diffusivity values.
compared with p¼ 0.041 in the ROI approach. It should be noted
that when applying a whole-brain analysis, a lot of brain regions,
for which we did not have any a priori hypothesis, would be 3.3. Correlations between chess-related parameters and probabilistic
subjected to the statistical analysis, which would reduce the grey matter volume (VBM) and white matter diffusivity (DTI)
statistical power.
Mean diffusivity was increased in chess players compared with Within the regions of interest, there was no cluster with
that of controls in a cluster located in the left superior longitudinal significant (either positive or negative) correlations between
fasciculus (Fig. 1A, Table 2). There was no cluster with reduced probabilistic local GM volume on one hand and the Elo score,
 J. Hänggi et al. / Neuropsychologia 62 (2014) 152–162 157

age of chess playing commencement, years of chess training, and not related to the age of chess playing commencement (r ¼  0.07,
lifetime hours of chess training and gaming on the other hand. p¼ 0.76 and r ¼  0.06, p ¼0.80, respectively). However, the left
However, when regressing these behavioural measures against and right caudate nucleus volumes were inversely related to the
the DTI-derived parameters (FA, mean, axial, and radial diffusivity) years of chess playing experience (r ¼  0.67, p¼ 0.001 and
several clusters with negative correlations were found. The Elo r ¼ 0.62, p¼ 0.003, respectively), revealing that more years of
score was inversely related to mean diffusivity in two clusters experience goes with smaller caudate nucleus volumes.
within the right superior longitudinal fasciculus (Fig. 2A, Table 3), Because age and years of chess training and playing experience
both clusters located in the WM underlying the right supramar- are heavily confounded (r ¼0.868, p ¼7.15E-7) it remains unclear
ginal gyrus. Within the same region, axial diffusivity was also whether the effect reported is driven by chess experience or rather
inversely related to the Elo score (Fig. 2B, Table 3), suggesting that by age. However, we were able to show that there is an effect
differences in mean diffusivity are rather driven by differences in related to chess experience that goes beyond the age effect. To
axial than radial diffusivity. The same cluster that revealed an illustrate this, we first computed the association between age and
inverse relationship between the Elo score and mean diffusivity caudate volume for each group separately. For chess players, age
within the right superior longitudinal fasciculus (Fig. 2A, Table 3) was strongly negatively related to the left caudate volume
showed also an inverse relation among lifetime hours of chess (r ¼  0.787, p¼ 0.000038) and also to the right caudate volume
training and gaming and mean diffusivity (Fig. 2C, Table 3), (r ¼  0.774, p ¼0.000061), whereas in control subjects, age was
although the Elo score and lifetime hours of chess training and not significantly related to caudate volumes (left caudate:
playing are only moderately correlated. r ¼ 0.263, p¼ 0.262; right caudate: r ¼  0.325, p¼ 0.162).
There were no positive correlations between mean or axial Therefore, we computed whether the differences in the corre-
diffusivity on one hand and the Elo score, age of chess playing lation coefficients between chess players and control men were
commencement, years of chess training, and lifetime hours of statistically significantly different using Fisher's r-to-z-transforma-
chess training and gaming on the other hand. No significant tion. Both differences in the correlation coefficients are significant
(neither positive nor negative) correlations were found between (left caudate volume: z ¼ 2.32, p¼ 0.02, two-tailed; right caudate
these behavioural measures on one hand and FA and radial volume: z¼  2.02, p ¼0.043, two-tailed) revealing that there is an
diffusivity on the other hand. No whole brain analyses were effect of the years of chess experience on the caudate volume that
conducted for the correlations. goes beyond the effect of age.

3.4. Group differences in cortical thickness (SBM)

4. Discussion and conclusion
The exploratory and confirmatory whole brain analysis of
cortical thickness differences revealed only clusters with reduced The aim of this study is to identify specific anatomical features
cortical thickness in chess players compared with that of control of the chess player's brain. Our analysis was guided by three
men mainly in occipito-temporal and parietal regions including specific hypotheses based on previous neuroimaging studies con-
the OTJ. Most interesting clusters are those found in the right and ducted with chess experts and chess novices. We hypothesised
left OTJ, in the left and right precunei (PrCun) as well as the one in that the chess player's brain will demonstrate specific anatomical
the left supramarginal gyrus (SMG) (Fig. 3). The cluster in the right features in (1) brain areas associated with higher cognitive func-
OTJ is the only cluster of all clusters presented in Fig. 3 that tions in a distributed cortical network including parietal, frontal,
survives cluster extent error correction for multiple comparisons. and temporal regions, (2) in brain regions known to be involved in
Further clusters with reduced cortical thickness in chess the recognition of objects and their relations (e.g. the OTJ and the
players compared with control men were located within the fusiform gyrus), and (3) finally in the caudate nucleus. In order to
middle temporal gyrus, superior and inferior temporal sulci, identify these specific features we applied three different neuroi-
temporal pole, superior frontal gyrus, cuneus, middle occipital maging techniques: 1.) VBM to delineate probabilistic GM volume
gyrus, subcentral gyrus, and posterior segment of the Sylvian differences, 2.) SBM to identify differences in terms of cortical
fissure of the left hemisphere. In the right hemisphere, clusters thickness, and 3.) DTI to investigate differences in the WM
were located within the superior and inferior temporal sulci, architecture.
superior temporal gyrus, pars opercularis, occipital areas, and
the marginal part of cingulate sulcus (Fig. 3). Because we applied 4.1. Structurally altered brain regions in chess players
a whole brain analysis uncorrected for multiple comparisons in
our SBM approach and did not had any a priori hypothesis with In our group analysis we only identified a very few differences
respect to brain areas outside the regions of interest (see above), between chess players and control subjects. In chess players
we do not discuss these regions further. Future analyses are clearly compared to control subjects there were reduced GM volume
necessary to confirm the significance of these regions before any and cortical thickness bilaterally in the OTJ. In addition, chess
sound conclusions can be drawn. players compared to controls demonstrated increased mean diffu-
sivity within two clusters of the left superior longitudinal fascicu-
3.5. Volume of the caudate nucleus lus. Beside these differences there were no further anatomical
features that were different between chess players and control
The mean volumes of the left caudate nucleus (mean/standard subjects even when the statistical threshold was lowered. Thus,
deviation: chess players 4082.5/415.9 mm3, control men 4074.6/ we did not identify substantial anatomical differences in the
578.6 mm3) and that of the right caudate nucleus (chess players distributed cognitive network including several brain areas in
4181.3/432.7 mm3, control men 4136.6/597.2 mm3) were not sig- the frontal and parietal regions. We also did not identify differ-
nificantly different between groups (left caudate: p ¼0.33, right ences between chess players and controls in the caudate nucleus.
caudate: p¼ 0.48, analysis of covariance corrected for total GM Thus we could not replicate the findings of Duan et al. (2012a)
volume). There was no significant correlation between the Elo who identified reduced caudate volumes in chess players com-
score of the chess players and their left (Pearson's correlation pared with control subjects. However, the volumes of the left and
r ¼  0.06, p ¼0.80) or right caudate nucleus volume (r ¼  0.05, right caudate nucleus were inversely related to the years of chess
p ¼0.83) and the left and right caudate nucleus volumes were also playing experience revealing that more years of experience goes
158 J. Hänggi et al. / Neuropsychologia 62 (2014) 152–162

Fig. 2. Associations between mean/axial diffusivity and the Elo score as well as lifetime hours of chess training. Shown are clusters with negative correlations (blue–lightblue)
between the Elo score and mean diffusivity in the right superior longitudinal fasciculus (SLF, A) and between the Elo score and axial diffusivity in the right SLF (B), in the same
region as the cluster in A. Lifetime hours of chess training correlated negatively with mean diffusivity in the right SLF (C), in the same region as the clusters in A and B.
The regions of interest (left and right SLF) subjected to the statistical analysis are shown in green–lightgreen. The colour bar indicates alpha error probability that is corrected for
multiple comparisons. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)
 J. Hänggi et al. / Neuropsychologia 62 (2014) 152–162 159

Table 3
Regions with negative correlations between mean/axial diffusivity and the Elo score as well as lifetime hours of chess training.

Correlations between Letter in Fig. 2 Number of voxels Error probability MNI coordinates

x y z

Elo score and mean diffusivity

Right superior longitudinal fasciculus (below supramarginal gyrus) A 298 0.027 39  31 33
Right superior longitudinal fasciculus (below supramarginal gyrus) A 170 0.033 30  33 41

Elo score and axial diffusivity

Right superior longitudinal fasciculus (below supramarginal gyrus) B 26 0.031 40  37 33

Lifetime hours of chess training and mean diffusivity

Right superior longitudinal fasciculus (below supramarginal gyrus) C 793 0.015 39  32 33

p-values are corrected for multiple comparisons using permutation-based non-parametric testing across space (FSL's randomise tool with 5000 permutations).
Abbreviations: Fig., figure; MNI, Montreal Neurological Institute.

Fig. 3. Regions with reduced cortical thickness in chess players revealed in an exploratory whole brain analysis. Shown are clusters with reduced cortical (blue–lightblue) in
the left and right occipito-temporal junction (OTJ), precuneus (PrCun), and supramarginal gyrus (SMG) besides other regions in chess players compared with control men.
Note there is only one cluster with increased cortical thickness (red–yellow) in chess players. Colour bars indicate alpha error probability uncorrected for multiple
comparisons. The right OTJ cluster is the only cluster that survives cluster extent error corrections for multiple comparisons. (For interpretation of the references to colour in
this figure legend, the reader is referred to the web version of this article.)

with smaller caudate nucleus volumes. These negative correlations and during shape perception (Ben-Shachar, Dougherty, Deutsch, &
are in line with findings reported by Duan et al. (2012a). Wandell, 2007). However, the common denominator of these
The OTJ for which we identified the only substantial anatomical perceptual processes is the association of visual information with
difference between chess players and novices in terms of GM and higher cognitive stimulus properties. The OTJ is partly overlapping
cortical thickness is known to be involved in the recognition of with the posterior superior temporal sulcus (STS) region, which
objects and their relations. This region has repeatedly been shown has also been recognised as a brain area devoted to integrate
to be activated when chess players process visually presented information from different modalities (Hein and Knight, 2008).
chess game positions (Bilalić et al., 2010, 2011). Bilalic et al. relate The precuneus (medial parietal lobe) seems to play an impor-
these activations to the advantage of chess experts in object tant role in chess playing (Wan et al., 2011). Although we did not
perception utilised during the perception of chess-specific stimuli find any difference in the precuneus in our VBM analysis when
(e.g. chess configurations). The OTJ is a brain area, which is not correcting for multiple comparisons, we found evidence for
only involved in the perception of chess-related stimuli, but is structural differences in the precuneus in our uncorrected SBM
rather a brain area that is involved in many different perceptional analysis where cortical thickness was reduced in the left and right
processes, which are not directly related to chess-specific object precunei in chess players compared with that of control men
perception, e.g. watching at signed narratives (Courtin et al., 2011) (Fig. 3). The importance of the precuneus for playing chess on a
160 J. Hänggi et al. / Neuropsychologia 62 (2014) 152–162

high level has further been elaborated by Duan et al. who showed The reason for this, at a first glance, heterogeneous picture
a broader task-induced deactivation of the DMN in a chess depends on the fact that GM volume and cortical thickness are
problem-solving task in chess experts compared with novices influenced by several microscopic features that can be influenced
(Duan et al., 2012b) as well as enhanced integration (functional at different stages during ontogeny. Meanwhile it is known that
connectivity) between the caudate nucleus and the regions of the GM volume and cortical thickness as measured by MRI methods
DMN including the precuneus in the resting-state in chess experts are influenced by early sensory influences, maturation processes,
compared with novices (Duan et al., 2012a). The deactivation of age-dependent influences, as well as use-dependent influences.
the DMN has been previously explained as the reallocation of Each of these factors most probably influences GM volume and
cognitive resources in order to focus more on the task and cortical thickness differently. For example, it has been suggested
suppress unrelated and irrelevant thoughts (Fox et al., 2005). that sensory deprivation during early childhood may cause a
There is no doubt that chess playing benefits from such a neural substantial reduction of pruning of the exuberant cortico-cortical
mechanism. Furthermore, it has also been shown that individual and/or cortico-thalamo-cortical connections that exist during early
differences in resting-state functional connectivity predict task- infancy. This reduced pruning due to sensory deprivation will
induced BOLD activity (Mennes et al., 2010). result in a greater survival of these exuberant connections and a
Taken together, chess experts demonstrate specific anatomical thicker visual cortex. This has been suggested for early or con-
features in at least three brain areas (OTJ, caudate nucleus, and genitally blind subjects who demonstrate a thicker visual cortex
precuneus) that are obviously important nodes in the neural than sighted control subjects (Anurova, Renier, De Volder, Carlson,
network activated during chess playing. & Rauschecker, 2014). Changes in GM volume or cortical thickness
in adulthood due to training or practice are mostly likely caused by
axonal sprouting and/or an increase in spine density in brain
4.2. Is bigger better or is less smarter? structures critical for a certain task or performance. But which of
these microscopic aspects are different in the chess players will be
But why are these regions, which are obviously that strongly examined in our study? Our chess experts have commenced
involved in identifying chess positions, smaller in terms of GM volume playing chess between 4 and 14 years with a mean age of 8 years.
and demonstrate reduced cortical thickness in chess players? Actually, Thus, it is possible that most of our chess players have received
this is difficult to explain on the basis of the current knowledge since it substantial chess-related stimulation during early childhood. From
is not that clear how cortical thickness and GM volume are related to neurophysiological studies we know that the rapid phase of
the performance in psychological tasks. synaptogenesis in early childhood is followed by subsequent
Mostly, it is argued that measures of cortical morphology (GM longer period of pruning during which synapses are eliminated
volume or cortical thickness) are positively related to cognitive by about 40% to reach near-mature levels at 11 years of age
performance. This “bigger-is-better” or “thicker-is-better” assump- (Huttenlocher, de Courten, Garey, & Van der Loos, 1982;
tion is reflected for example in findings of training-related Huttenlocher & de Courten, 1987; Huttenlocher, 1984, 1990). Thus,
enhancement of performance and associated increases in GM it could be possible that the chess-specific stimulation of the OTJ
volume or cortical thickness (Jancke, 2009; May, 2011). There are might have caused a stronger pruning of cortico-cortical and/or
also studies demonstrating a reduction in GM volume or cortical thalamo-cortical connections in chess experts than in control
thickness as a consequence of non-use (Langer, Hänggi, Muller, subjects since the chess experts might have used the OTJ to cope
Simmen, & Jäncke, 2012). Other studies identified a positive with the chess tasks. We are aware of the fact that our interpreta-
correlation between cortical thickness as well as GM volume and tion is partly speculative, but we believe that it is highly plausible
psychological performance measures (Karama et al., 2011; Engvig in the context of the sensory-deprivation studies. An interesting
et al., 2010). On the other hand, cortical thickness generally strategy to directly test our hypothesis would be to compare
declines with ageing and this goes hand in hand with a decline chess-experts commenced playing chess very early life with chess
of cognitive functions (Rathi et al., 2013). Thus, it seems to be quite experts who have started substantially later (e.g. older than 18
straightforward to argue that increased cortical thickness and GM years).
volume would be associated with better performance in psycho- We also identified increased mean diffusivity in the left super-
logical tasks. ior longitudinal fasciculus in chess players compared with that of
In fact, the picture is a bit more complicated since some studies control men. In addition, the Elo score and lifetime hours of chess
have shown that deficits in psychological functions can also be training were inversely related to mean diffusivity in the right
associated with increased cortical thickness. For example, indivi- superior longitudinal fasciculus. The superior longitudinal fascicu-
duals with congenital amusia compared to that of healthy control lus connects parieto-temporal brain regions with frontal brain
subjects show enhanced cortical thickness and reduced GM areas and is a major fibre tract of the dorsal visual stream and also
volume in the auditory cortex (Hyde et al., 2007). Other studies the dorsal attentional control system (Capotosto, Babiloni, Romani,
have shown that it is the rate of increase in cortical thickness & Corbetta, 2009). Thus, this fibre tract is necessary for attentional
rather than thickness itself that correlates with better cognitive control and the transformation of visual information to the frontal
performance (here with psychometric intelligence) (Shaw et al., executive system. Thus, it is plausible to assume that this fibre
2006). Studies combining morphometric with neurophysiological tract is involved in the control of psychological operations used
measures also revealed partly inconclusive results. For example, during chess playing.
studying normally developing children revealed a negative corre- Increased MD (and decreased FA) reflects more radial diffusion
lation between blood oxygenation level dependent (BOLD) (perpendicular to the fibres) than axial diffusion (along the fibres)
responses and cortical thickness in fronto-parietal regions in an or both. In clinical populations, increased MD and/or decreased FA
orthographic task (Lu et al., 2009). A more recent study from our are commonly interpreted as reduced WM integrity and reduced
own lab identified that a thinner auditory cortex was associated fibre coherence (Assaf & Pasternak, 2008; Sexton, Mackay, &
with larger N1 amplitudes to acoustic stimuli (measured with Ebmeier, 2009). Increased FA and/or decreased MD in diseases
electroencephalography) reflecting a more efficient neurophysio- are rarely reported in the DTI literature, although the loss of fibres
logical processing in the auditory cortex that is associated with a that turn, cross, and/or twist within a single voxel should be
thinner auditory cortex (Liem, Zaehle, Burkhard, Jancke, & Meyer, accompanied by an increase in FA and a decrease in MD. Denser
2012). connected components within a network, a potential structural
 J. Hänggi et al. / Neuropsychologia 62 (2014) 152–162 161

marker of increased network efficiency, might consist of an 320030-120661 and 4-62341-05 both from the Swiss National
increased number of fibres turning, crossing, and/or twisting that Science Foundation (SNF) to L.J.
would be reflected by increased MD and decreased FA. Although,
there is consensus that the axonal membranes are the most
important hindrance for water diffusion in fibre tracts (Beaulieu,
2002) and that myelin is not needed for anisotropic diffusion and References
only modulates the diffusion tensor by about 20% (Gulani, Webb,
Duncan, & Lauterbur, 2001), the actual and individual contribu- Amidzic, O., Riehle, H. J., Fehr, T., Wienbruch, C., & Elbert, T. (2001). Pattern of focal
tions of the axonal membrane, myelin, neurofilaments, and micro- gamma-bursts in chess players. Nature, 412, 603.
Annett, M. (1970). A classification of hand preference by association analysis. British
tubuli are still a matter of dispute (Beaulieu, 2002). DTI Journal of Psychology, 61, 303–321.
neuroplasticity studies suggested that better motor or cognitive Anurova, I., Renier, L. A., De Volder, A. G., Carlson, S., & Rauschecker, J. P. (2014).
functioning is associated with increased FA/decreased MD as well Relationship between cortical thickness and functional activation in the early
blind. Cerebral Cortex.
as associated with decreased FA/increased MD (Bengtsson et al.,
Ashburner, J., & Friston, K. J. (2000). Voxel-based morphometry — the methods-
2005; Hänggi, Koeneke, Bezzola, & Jäncke, 2010; Schmithorst & NeuroImage, 11, 805–821.
Wilke, 2002). Assaf, Y., & Pasternak, O. (2008). Diffusion tensor imaging (DTI)-based white matter
mapping in brain research: a review. Journal of Molecular Neuroscience, 34(1),
51–61.
4.3. Limitations Atherton, M., Zhuang, J., Bart, W. M., Hu, X., & He, S. (2003). A functional MRI study
of high-level cognition. I. The game of chess. Cognitive Brain Research, 16(1),
26–31.
Several limitations of the present study are worth mentioning.
Basser, P. J., Mattiello, J., & LeBihan, D. (1994). MR diffusion tensor spectroscopy and
First, chess involves an abstract skill-set, which unquestionably imaging. Biophysical Journal, 66(1), 259–267.
draws upon a multitude of cognitive functions and their interac- Beaulieu, C. (2002). The basis of anisotropic water diffusion in the nervous system –
tions. Using a univariate localizationalist approach in looking at a technical review. NMR in Biomedicine, 15(7–8), 435–455.
Behrens, T. E., Woolrich, M. W., Jenkinson, M., Johansen-Berg, H., Nunes, R. G., Clare,
the complex skills involved in chess is an inherent shortcoming of S., et al. (2003). Characterization and propagation of uncertainty in diffusion-
the study. A topological approach of both structural and functional weighted MR imaging. Magnetic Resonance in Medicine, 50(5), 1077–1088.
whole-brain connectivity networks as opposed to a topographical Ben-Shachar, M., Dougherty, R. F., Deutsch, G. K., & Wandell, B. A. (2007).
Differential sensitivity to words and shapes in ventral occipito-temporal cortex.
approach should be applied in future studies that aim at investi- Cerebral Cortex, 17(7), 1604–1611.
gating the neural correlates of chess expertise. Second, although Bengtsson, S. L., Nagy, Z., Skare, S., Forsman, L., Forssberg, H., & Ullén, F. (2005).
differences in diffusion parameters were found between the Extensive piano practicing has regionally specific effects on white matter
development. Nature Neuroscience, 8(9), 1148–1150.
groups, these results should not be considered as direct evidence Bilalić, M., Langner, R., Erb, M., & Grodd, W. (2010). Mechanisms and neural basis of
for enhanced or reduced structural connectivity. Diffusion para- object and pattern recognition: a study with chess experts. Journal of Experi-
meters derived from DTI data cannot provide such information. mental Psychology General, 139, 728–742.
Bilalić, M., Langner, R., Ulrich, R., & Grodd, W. (2011). Many faces of expertise:
Third, due to logistical restrictions and time constraints it was not
fusiform face area in chess experts and novices. The Journal of Neuroscience: the
possible to assess the cognitive profile of the chess players in Official Journal of the Society for Neuroscience, 31, 10206–10214.
greater detail. Forth, although the sample size of the present study Bilalić, M., McLeod, P., & Gobet, F. (2009). Specialization effect and its influence on
memory and problem solving in expert chess players. Cognitive Science, 33,
was of moderate size (n¼ 40), further morphological differences
1117–1143.
that would survive corrections for multiple comparisons might be Campitelli, G., Gobet, F., Head, K., Buckley, M., & Parker, A. (2007). Brain localization
revealed if using larger samples. Last, due to the fact that we of memory chunks in chessplayers. International Journal of Neuroscience, 117
investigated a cross-sectional sample, we cannot infer the direc- (12), 1641–1659.
Capotosto, P., Babiloni, C., Romani, G. L., & Corbetta, M. (2009). Frontoparietal cortex
tion of causality. Longitudinal studies are needed to track down controls spatial attention through modulation of anticipatory alpha rhythms.
whether the morphological alterations revealed in the chess The Journal of Neuroscience, 29, 5863–5872.
player's brain are the cause (nature, genetic disposition for a Charness, N., Tuffiash, M., Krampe, R., Reingold, E., & Vasyukova, E. (2005). The role
of deliberate practice in chess expertise. Applied Cognitive Psychology, 19(2),
particular trait) or the consequence (nurture, structural neuro- 151–165.
plastic adaptations) of the intensive and long-term chess training Courtin, C., Jobard, G., Vigneau, M., Beaucousin, V., Razafimandimby, A., Herve, P. Y.,
and experience. A longitudinal study design also permits to et al. (2011). A common neural system is activated in hearing non-signers to
process French sign language and spoken French. Brain Research Bulletin, 84(1),
investigate the dynamics of the morphological alterations, i.e. 75–87.
whether they mainly occur early in chess training course or Dale, A. M., Fischl, B., & Sereno, M. I. (1999). Cortical surface-based analysis. I.
whether these changes are continuously established across the Segmentation and surface reconstruction. NeuroImage, 9(2), 179–194.
De Bruin, A., Smits, N., Rikers, R., & Schmidt, H. (2008). Deliberate practice predicts
entire course of chess training.
performance over time in adolescent chess players and drop-outs: a linear
mixed models analysis. British Journal of Psychology, 99, 473–497.
4.4. Conclusions Douaud, G., Smith, S., Jenkinson, M., Behrens, T., Johansen-Berg, H., Vickers, J., et al.
(2007). Anatomically related grey and white matter abnormalities in
adolescent-onset schizophrenia. Brain, 130(Pt 9), 2375–2386.
We showed to the best of our knowledge for the first time that Duan, X., He, S., Liao, W., Liang, D., Qiu, L., Wei, L., et al. (2012a). Reduced caudate
the grey and white matter architecture is altered in the brain of volume and enhanced striatal-DMN integration in chess experts. NeuroImage,
60, 1280–1286.
expert chess players compared with that of control subjects. Our
Duan, X., Liao, W., Liang, D., Qiu, L., Gao, Q., Liu, C., et al. (2012b). Large-scale brain
structural findings complement and extent the results of the networks in board game experts: insights from a domain-related task and task-
functional imaging studies investigating brain activity associated free resting state. PloS One, 7, e32532.
with chess-relevant cognition. Structural and functional investiga- Elo, A. E. (1978). The rating of chess players past and present. .
Engvig, A., Fjell, A. M., Westlye, L. T., Moberget, T., Sundseth, O., Larsen, V. A., et al.
tions on chess expertise in particular might help to shed light on (2010). Effects of memory training on cortical thickness in the elderly. Neuro-
the neural substrate underlying human problem-solving mechan- Image, 52(4), 1667–1676.
isms in general. Ericsson, K. A. (2004). Deliberate practice and the acquisition and maintenance of
expert performance in medicine and related domains. Academic Medicine:
Journal of the Association of American Medical Colleges, 79, S70–S81.
Fischl, B., & Dale, A. M. (2000). Measuring the thickness of the human cerebral
Acknowledgements cortex from magnetic resonance images. Proceedings of the National Academy of
Sciences of the United States of America, 97(20), 11050–11055.
Fischl, B., Liu, A., & Dale, A. M. (2001). Automated manifold surgery: constructing
The authors thank all volunteers for their participation in the geometrically accurate and topologically correct models of the human cerebral
study. The present work was supported by the Grant numbers cortex. IEEE Transactions on Medical Imaging, 20(1), 70–80.
162 J. Hänggi et al. / Neuropsychologia 62 (2014) 152–162

Fischl, B., Sereno, M. I., & Dale, A. M. (1999). Cortical surface-based analysis. II: Lu, L. H., Dapretto, M., O0 Hare, E. D., Kan, E., McCourt, S. T., Thompson, P. M., et al.
inflation, flattening, and a surface-based coordinate system. NeuroImage, 9(2), (2009). Relationships between brain activation and brain structure in normally
195–207. developing children. Cerebral Cortex, 19(11), 2595–2604.
Fischl, B., Sereno, M. I., Tootell, R. B. H., & Dale, A. M. (1999). High-resolution May, A. (2011). Experience-dependent structural plasticity in the adult human
intersubject averaging and a coordinate system for the cortical surface. Human brain. Trends in Cognitive Science, 15, 475–482.
Brain Mapping, 8, 272–284. Mennes, M., Kelly, C., Zuo, X. N., Di Martino, A., Biswal, B. B., Castellanos, F. X., &
Fischl, B., van der Kouwe, A., Destrieux, C., Halgren, E., Segonne, F., Salat, D. H., et al. Milham, M. P. (2010). Inter-individual differences in resting-state functional
(2004). Automatically parcellating the human cerebral cortex. Cerebral Cortex, connectivity predict task-induced BOLD activity. NeuroImage, 50, 1690–1701.
14(1), 11–22. Nichelli, P., Grafman, J., Pietrini, P., Alway, D., Carton, J. C., & Miletich, R. (1994).
Fox, M. D., Snyder, A. Z., Vincent, J. L., Corbetta, M., Van Essen, D. C., & Raichle, M. E. Brain activity in chess playing. Nature, 369(6477), 191.
(2005). The human brain is intrinsically organized into dynamic, anticorrelated Nichols, T. E., & Holmes, A. P. (2002). Nonparametric permutation tests for
functional networks. Proceedings of the National Academy of Sciences of the functional neuroimaging: a primer with examples. Human Brain Mapping, 15,
United States of America, 102, 9673–9678. 1–25.
Gobet, F. (1998). Expert memory: a comparison of four theories. Cognition, 66(2), Onofrj, M., Curatola, L., Valentini, G., Antonelli, M., Thomas, A., & Fulgente, T. (1995).
115–152.
Non-dominant dorsal-prefrontal activation during chess problem solution
Gobet, F., & Waters, A. J. (2003). The role of constraints in expert memory. Journal of
evidenced by single photon emission computerized tomography (SPECT).
experimental psychology. Learning, memory, and cognition, 29, 1082–1094.
Neuroscience Letters, 198(3), 169–172.
Good, C. D., Ashburner, J., & Frackowiak, R. S. (2001b). Computational neuroanat-
Peters, M., Laeng, B., Latham, K., Jackson, M., Zaiyouna, R., & Richardson, C. (1995). A
omy: new perspectives for neuroradiology. Revue Neurologique (Paris), 157(8–9
redrawn Vandenberg and Kuse mental rotations test: different versions and
Pt 1), 797–806.
Good, C. D., Johnsrude, I. S., Ashburner, J., Henson, R. N., Friston, K. J., & Frackowiak, factors that affect performance. Brain and Cognition, 28(1), 39–58.
R. S. (2001a). A voxel-based morphometric study of ageing in 465 normal adult Rathi, Y., Pasternak, O., Savadjiev, P., Michailovich, O., Bouix, S., Kubicki, M., et al.
human brains. NeuroImage, 14, 21–36. (2013). Gray matter alterations in early aging: a diffusion magnetic resonance
Gulani, V., Webb, A. G., Duncan, I. D., & Lauterbur, P. C. (2001). Apparent diffusion imaging study. Human Brain Mapping.
tensor measurements in myelin-deficient rat spinal cords. Magnetic Resonance Raven, J. C. (1998). Advanced progressive matrices. Oxford: Oxford Psychology Press.
in Medicine, 45(2), 191–195. Rosas-Cholula, G., Ramirez-Cortes, J. M., Alarcon-Aquino, V., Gomez-Gil, P., Rangel-
Hänggi, J., Koeneke, S., Bezzola, L., & Jäncke, L. (2010). Structural neuroplasticity in Magdaleno, J. D. J., & Reyes-Garcia, C. (2013). Gyroscope-driven mouse pointer
the sensorimotor network of professional female ballet dancers. Human Brain with an EMOTIVs EEG headset and data analysis based on empirical mode
Mapping, 31(8), 1196–1206. decomposition. Sensors (Basel, Switzerland), 13, 10561–10583.
Hayasaka, S., & Nichols, T. E. (2003). Validating cluster size inference: random field Salat, D. H., Buckner, R. L., Snyder, A. Z., Greve, D. N., Desikan, R. S., Busa, E., et al.
and permutation methods. NeuroImage, 20, 2343–2356. (2004). Thinning of the cerebral cortex in aging. Cerebral Cortex, 14, 721–730.
Hein, G., & Knight, R. T. (2008). Superior temporal sulcus—it's my area: or is it? Schellig, D. (1997). Block-tapping-testFrankfurt: Swets Test Services.
Journal of Cognitive Neuroscience, 20(12), 2125–2136. Schmithorst, V. J., & Wilke, M. (2002). Differences in white matter architecture
Howard, R. W. (2013). Practice other than playing games apparently has only a between musicians and non-musicians: a diffusion tensor imaging study.
modest role in the development of chess expertise. British journal of Psychology Neuroscience Letters, 321(1–2), 57–60.
(London, England: 1953), 104, 39–56. Sexton, C. E., Mackay, C. E., & Ebmeier, K. P. (2009). A systematic review of diffusion
Hua, K., Zhang, J., Wakana, S., Jiang, H., Li, X., Reich, D. S., et al. (2008). Tract tensor imaging studies in affective disorders. Biological Psychiatry, 66(9),
probability maps in stereotaxic spaces: analyses of white matter anatomy and 814–823.
tract-specific quantification. NeuroImage, 39(1), 336–347. Shaw, P., Greenstein, D., Lerch, J., Clasen, L., Lenroot, R., Gogtay, N., et al. (2006).
Huttenlocher, P. R. (1984). Synapse elimination and plasticity in developing human Intellectual ability and cortical development in children and adolescents.
cerebral cortex. American Journal of Mental Deficiency, 88(5), 488–496. Nature, 440(7084), 676–679.
Huttenlocher, P. R. (1990). Morphometric study of human cerebral cortex devel- Simon, H. A., & Chase, W. G. (1973). Skill in chess. American Scientist, 61, 394–403.
opment. Neuropsychologia, 28(6), 517–527. Smith, S. M., Jenkinson, M., Johansen-Berg, H., Rueckert, D., Nichols, T. E., Mackay, C.
Huttenlocher, P. R., & de Courten, C. (1987). The development of synapses in striate E., et al. (2006). Tract-based spatial statistics: voxelwise analysis of multi-
cortex of man. Human Neurobiology, 6(1), 1–9. subject diffusion data. NeuroImage, 31(4), 1487–1505.
Huttenlocher, P. R., de Courten, C., Garey, L. J., & Van der Loos, H. (1982). Smith, S. M., Jenkinson, M., Woolrich, M. W., Beckmann, C. F., Behrens, T. E.,
Synaptogenesis in human visual cortex–evidence for synapse elimination
Johansen-Berg, H., et al. (2004). Advances in functional and structural MR
during normal development. Neuroscience Letters, 33(3), 247–252.
image analysis and implementation as FSL. NeuroImage, 23(Suppl. 1),
Hyde, K. L., Lerch, J. P., Zatorre, R. J., Griffiths, T. D., Evans, A. C., & Peretz, I. (2007).
S208–S219.
Cortical thickness in congenital amusia: when less is better than more. The
Smith, S. M., Johansen-Berg, H., Jenkinson, M., Rueckert, D., Nichols, T. E., Miller, K.
Journal of Neuroscience, 27(47), 13028–13032.
L., et al. (2007). Acquisition and voxelwise analysis of multi-subject diffusion
Jancke, L. (2009). The plastic human brain. Restorative Neurology and Neuroscience,
27(5), 521–538. data with tract-based spatial statistics. Nature Protocols, 2(3), 499–503.
Smith, S. M., & Nichols, T. E. (2009). Threshold-free cluster enhancement: addres-
Karama, S., Colom, R., Johnson, W., Deary, I. J., Haier, R., Waber, D. P., et al. (2011).
Cortical thickness correlates of specific cognitive performance accounted for by sing problems of smoothing, threshold dependence and localisation in cluster
the general factor of intelligence in healthy children aged 6 to 18. NeuroImage, inference. NeuroImage, 44(1), 83–98.
55(4), 1443–1453. Song, S. K., Sun, S. W., Ramsbottom, M. J., Chang, C., Russell, J., & Cross, A. H. (2002).
Kuperberg, G. R., Broome, M. R., McGuire, P. K., David, A. S., Eddy, M., Ozawa, F., et al. Dysmyelination revealed through MRI as increased radial (but unchanged axial)
(2003). Regionally localized thinning of the cerebral cortex in schizophrenia. diffusion of water. NeuroImage, 17(3), 1429–1436.
Archives of General Psychiatry, 60(9), 878–888. Wan, X., Nakatani, H., Ueno, K., Asamizuya, T., Cheng, K., & Tanaka, K. (2011). The
Langer, N., Hänggi, J., Muller, N. A., Simmen, H. P., & Jäncke, L. (2012). Effects of limb neural basis of intuitive best next-move generation in board game experts.
immobilization on brain plasticity. Neurology, 78, 182–188. Science (New York, N.Y.), 331, 341–346.
Liem, F., Zaehle, T., Burkhard, A., Jancke, L., & Meyer, M. (2012). Cortical thickness of Zatorre, R. J., Fields, R. D., & Johansen-Berg, H. (2012). Plasticity in gray and white:
supratemporal plane predicts auditory N1 amplitude. Neuroreport, 23(17), neuroimaging changes in brain structure during learning. Nature Neuroscience,
1026–1030. 15, 528–536.