A key role for experimental task perform

Brain and Cognition 78 (2012) 14–27

Contents lists available at SciVerse ScienceDirect

Brain and Cognition
journal homepage: www.elsevier.com/locate/b&c

A key role for experimental task performance: Effects of math talent, gender
and performance on the neural correlates of mental rotation
Christian Hoppe a,⇑, Klaus Fliessbach a,b, Sven Stausberg a, Jelena Stojanovic a, Peter Trautner b,
Christian E. Elger a,b, Bernd Weber a,b
a
b

Department of Epileptology, University of Bonn Medical Centre, Germany
Department of NeuroCognition, Life & Brain Centre, Germany

a r t i c l e

i n f o

Article history:
Accepted 18 October 2011
Available online 15 November 2011
Keywords:
Mental rotation
Mathematical giftedness
Gender effects
Neural correlates of cognitive performance
Inferior parietal lobule
Functional magnetic resonance imaging

a b s t r a c t
The neurophysiological mechanisms underlying superior cognitive performance are a research area of
high interest. The majority of studies on the brain–performance relationship assessed the effects of capability-related group factors (e.g. talent, gender) on task-related brain activations while only few studies
examined the effect of the inherent experimental task performance factor. In this functional MRI study,
we combined both approaches and simultaneously assessed the effects of three relatively independent
factors on the neurofunctional correlates of mental rotation in same-aged adolescents: math talent
(gifted/controls: 17/17), gender (male/female: 16/18) and experimental task performance (median split
on accuracy; high/low: 17/17). Better experimental task performance of mathematically gifted vs. control
subjects and male vs. female subjects validated the selected paradigm. Activation of the inferior parietal

lobule (IPL) was identified as a common effect of mathematical giftedness, gender and experimental task
performance. However, multiple linear regression analyses (stepwise) indicated experimental task performance as the only predictor of parietal activations. In conclusion, increased activation of the IPL represents a positive neural correlate of mental rotation performance, irrespective of but consistent with the
obtained neurocognitive and behavioral effects of math talent and gender. As experimental performance
may strongly affect task-related activations this factor needs to be considered in capability-related group
comparison studies on the brain–performance relationship.
Ó 2011 Elsevier Inc. All rights reserved.

1. Introduction
Interindividual variance of performance in a given task (e.g.
accuracy, speed) is a ubiquitous psychological phenomenon. Any
functional or structural brain property which co-varies with task
performance can be addressed as a neural correlate of performance
(NCP) of the respective task. Searching NCPs is currently a highly
active field in neurocognitive research (Deary, Penke, & Johnson,
2010; Haier, 2009; Neubauer & Fink, 2009; Rypma & Prabhakaran,
2009).
Neuroimaging research on NCPs has largely focused on the effects of capability-related group factors (e.g. intelligence, talent,
or expertise) on task-related brain activations (Grabner, Neubauer,
& Stern, 2006; Lee et al., 2006; Singh & O’Boyle, 2004). Capability
effects have been reported since the very beginning of functional

neuroimaging (Charlot, Tzourio, Zilbovicius, Mazoyer, & Denis,
⇑ Corresponding author. Address: Department of Epileptology, University of Bonn
Medical Centre, Sigmund-Freud-Straße 25, 53105 Bonn, Germany. Fax: +49 228 287
90 16172.
E-mail address: christian.hoppe@ukb.uni-bonn.de (C. Hoppe).
0278-2626/$ - see front matter Ó 2011 Elsevier Inc. All rights reserved.
doi:10.1016/j.bandc.2011.10.008

1992). Since stimulation tasks which address the specific knowledge or skills of high-capability subjects are beyond reach for standard subjects, the applied tasks usually refer to elementary
cognitive abilities that presumably contribute to the respective
capability of interest. However, it is debatable whether a capability-related group effect on experimental task performance (i.e.
behavioral performance during scanning) validates the supposed
capability-task relationship or rather confounds the group factor
(Bell, Willson, Wilman, Dave, & Silverstone, 2006; Butler et al.,
2006; Jordan, Wustenberg, Heinze, Peters, & Jancke, 2002; Larson,
Haier, LaCasse, & Hazen, 1995; O’Boyle et al., 2005; Thomsen
et al., 2000; Unterrainer, Wranek, Staffen, Gruber, & Ladurner,
2000; Weiss et al., 2003a,b). In case of equal experimental task performance, neural correlates of the capability-related group factor
do not represent an NCP of the applied task (according to the above
definition) but rather indicate unknown group-specific neurocognitive factors which are irrelevant with regard to experimental task

performance (e.g. stress response).
Alternatively, studies on NCPs may focus more directly on the
effects of behavioral performance in an experimental task on the
brain activations which were elicited by this very task. Task

C. Hoppe et al. / Brain and Cognition 78 (2012) 14–27

performance effects on functional brain activation (and also connectivity) were repeatedly reported and actually challenge any
too simplistic approach to functional brain mapping (Rypma
et al., 2006; Tagaris et al., 1996b, 1997; Unterrainer et al., 2000,
2005). Evidently, this approach makes full use of the available
behavioral and neurophysiological data obtained by functional
neuroimaging. In addition, NCPs of a given task can be assessed
in non-indicated standard subjects, for example by contrasting retrospectively identified high vs. low experimental task performers
(Rypma et al., 2006).
In the present fMRI study, we combined these two research
strategies to allow their evaluation and comparison. Referring to
several previous studies from other groups (O’Boyle et al., 2005;
Unterrainer et al., 2000, 2004, 2005), we simultaneously examined
the effects of math talent, gender and experimental task performance on brain activations during mental rotation. Mental rotation

is one of the best studied paradigms in both experimental psychology and cognitive neuroscience since its introduction by Shepard
and Metzler (1971). On a behavioral level, both gender (Collins &
Kimura, 1997; Kimura, 1996; Linn & Petersen, 1985; Lippa, Collaer,
& Peters, 2010; Lubinski & Humphreys, 1990; Masters & Sanders,
1986, 1993; Moore & Johnson, 2008; Peters, 2008; Quinn & Liben,
2008; Voyer & Hou, 2006; Voyer, Voyer, & Bryden, 1995) and math
talent (Casey, Nuttall, & Benbow, 1995; Hyde, 2005; O’Boyle, Benbow, & Alexander, 1995; Spelke, 2005) have shown reliable effects
on mental rotation performance. Mental rotation reliably activates
the posterior parietal cortices (PPC) which also play a key role for
working memory and general intellectual functioning (Champod
& Petrides, 2007, 2010; Jung & Haier, 2007; Zacks, 2008).
Unfolding the idea of specific neural mechanisms underlying
better cognitive performance of a given task, we tested the following hypotheses:
(I) Mathematically gifted vs. control subjects and male vs.
female subjects show better mental rotation performance.
(II) Effect of the experimental task performance factor: Activation
of the PPC is obtained as an NCP, i.e. a positive neural correlate of mental rotation performance.
(III) Effects of capability-related factors: Math talent and gender
yield activations of the PPC similar to the effects of experimental task performance as both are associated with better
experimental task performance.

(IV) As the elicited neurocognitive activations are more inherently related to the experimental task, most of the variance
of PPC activations can be explained by the experimental task
performance factor.
2. Materials and methods
This study was approved by the Ethical Review Board of the
Medical Faculty at the University of Bonn (No. 039/06). The study
was carried out in accordance with The Code of Ethics of the World
Medical Association (Declaration of Helsinki) for experiments
involving humans.
2.1. Subjects
The study included 17 adolescent mathematically gifted subjects (MATH) and 20 same-aged control subjects (CON) between
the ages of 15 and 18 without mathematical talent. Mathematical
talent was assigned if the student was matriculated for Mathematics at the University of Bonn while attending high school (which
relies on the recommendation of their schools) or if a subject recently participated in the ‘Mathematical Olympiad’ at a federal
state level (state of North-Rhine Westphalia; total in 2007:
N = 350 out of 16,000 participants on the community level). During

15

subject recruitment, we additionally aimed at an equal distribution

of male and female subjects in both samples to implement gender
as an independent second capability-related group factor. Due to
technical artifacts in the MRI data, three control subjects had to
be excluded from the final analysis. Table 1 lists the characteristics
of the included subjects. All subjects had normal or corrected-tonormal vision. Subjects were reimbursed for participation (10€/h)
and travel costs. All subjects and their parents gave written informed consent according to the rules of good scientific practice.

2.2. Task
The original Shepard–Metzler (SM) paradigm shows a pair of
drawings of quasi-3D-figures, each of which is constructed out of
10 cubes rendered in two dimensions (‘‘3D figures’’) and requires
a matching decision (identical vs. mirrored). In contrast, the Vandenberg–Kuse (VK) paradigm (Vandenberg & Kuse, 1978) which
was applied in the majority of neuroimaging studies (Zacks,
2008) simultaneously shows one Shepard–Metzler figure as the
target and four additional figures as probes and requires the subjects to select the one probe which matches the target though
being spatially rotated (i.e. 4-alternatives forced choice; Peters &
Battista, 2008; Peters et al., 1995). In this paradigm, identical stimuli (i.e. 0° angular disparity condition), scrambled dot patterns derived from the figures, or black and white bars serve as the control
stimuli.
We propose a modified preparation rotation paradigm (JansenOsmann & Heil, 2007): (i) To avoid the reportedly high error rates
(>40%) of established paradigms (O’Boyle et al., 2005; Weiss et al.,

2003a,b); (ii) to separate mental rotation proper from both the
encoding of the stimulus and the matching test; and (iii) to improve experimental control over the task difficulty in terms of both
cognitive load and speed demands for future studies. The modified
paradigm is shown in Fig. 2. A two-dimensional rendering of a
three-dimensional stair-like figure composed of three cubes (i.e.
a fragment of the original Shepard–Metzler items; Fig. 1) was used
as the stimulus. Rotations of this object had to be performed at 90°
(instead of 15°) angles within one of the three spatial planes (horizontal, sagittal, frontal) resulting in 12 possible positions of the
object. At the beginning of each trial, the stimulus was presented
for 2 s in a randomly selected position. Subjects were then
prompted to perform four continuous mental rotations of the object starting from the initial position as indicated by four serially
presented arrows (duration: 13 s, i.e. 3.25 s per rotation; each arrow presented for 0.813 s). Pilot studies in adolescent control subjects (N = 30) confirmed the appropriateness of these speed
demands. In the task condition, the rotation plane was changed
either for one time or for three times: In Fig. 2, the task condition
shows three rotation plane changes (upward/sagittal, right/horizontal, clockwise/frontal, downward/sagittal). In the control condition (active low level task), the arrows indicated back-and-forth
left and right rotating of the object within the horizontal plane
(Fig. 2). The subjects were instructed that this condition only required the maintenance of the figure in its initial position. Each
trial was completed by a matching decision task presenting the object in one of the twelve possible positions as a probe. Subjects
pressed a button to indicate if the probe matched the figure’s position after performing the instructed rotations (response latency:
3 s). A total of 30 control and 30 task trials were presented alternately allowing mental rotation-related brain activations to return

to the baseline. Matching decisions were recorded as correct or
false responses. The individual mental rotation accuracy score,
MRX, was defined by MRX = (1 (number of errors during task trials/number of task trials number of errors during control trials/
number of control trials)) which obtains a positive indicator of

16

C. Hoppe et al. / Brain and Cognition 78 (2012) 14–27

Table 1
Subject characteristics and test scores: means (SD) and frequencies.
Math talent

Gender m/f
Math talent yes/no
Handedness RH/LH/MH
Handedness Laterality Indexc
Age (years)
Extracurricular math activities (min/day)
Math graded

Total average graded
CFT-20 R (IQ)e
LPS 7 (C score)f
Pre-scanning: mental rotation level
Mental rotation score (MRX)

Gender

Talented subjects n = 17

Control subjects n = 17

p

8/9

8/9

1.000a

14/1/2
0.7 (0.6)
16.7 (1.1)
64.1 (49.7)
1.3 (0.6)
1.9 (0.5)
118.2 (13.1)
8.5 (1.2)
7.0 (2.4)
0.88 (0.09)

15/0/2
0.9 (0.4)
16.5 (0.9)
14.3 (12.2)
2.3 (1.1)
2.3 (0.5)
110.1 (14.5)
7.9 (1.0)
5.5 (1.0)
0.72 (0.20)

0.596a
0.540b
0.518b
0.001b
0.005b
0.085b
0.106b
0.160b
0.031b
0.018b

Male subjects n = 17

Female subjects n = 17

p

8/8
13/0/3
0.8 (0.5)
16.5 (1.0)
52.3 (56.6)
1.6 (1.0)
2.1 (0.6)
115.5 (15.7)
27.4 (4.3)
7.2 (2.4)
0.86 (0.14)

9/9
16/1/1
0.9 (0.5)
16.7 (0.9)
40.9 (34.1)
1.9 (1.0)
2.1 (0.5)
113.0 (13.1)
25.7 (5.2)
5.4 (1.0)
0.74 (0.18)

1.000a
0.333a
0.528b
0.463b
0.940b
0.347b
0.772b
0.772b
0.403b
0.017b
0.030b

Behavioral performance: means (SD).
a
Chi-square test.
b
Mann–Whitney test.
c
Oldfield (1971).
d
German grades: 1 = very good (A grade), 6 = insufficient (F grade).
e
Weiß (2006).
f
Horn (1983), C score: [mean = 5, SD = 2].

Fig. 1. Object for mental rotation. The figure is constructed out of three small cubes
in contrast to the original Shepard and Metzler (1971) objects which are
constructed out of ten cubes.

mental rotation proper performance controlled for maintenance
performance.
2.3. Adjunctive behavioral measures
Nonverbal intelligence was estimated by the CFT 20-R, part 1
(Weiß, 2006), a German measure which is well-established for
individual diagnosis and research on intellectual giftedness. The
test comprises four subtests on visuospatial logical reasoning similar to Raven’s progressive matrices (test duration: 14 min.); a selfdeveloped computerized protocol version showing copies of the
original item sheets were used. Handedness was assessed by the
Edinburgh Handedness Inventory (Oldfield, 1971). Mental rotation
performance was measured using the Letter Rotation subtest LPS
7 from the Leistungs-Prüf-System (Horn, 1983), a well-established
paper–pencil test for mental plane rotation of single letters (test
duration: 2 min). In addition, criteria of math talent, math and
mean school grades (German grades: 1 = very good, 6 = not sufficient), the daily time spent on extracurricular math activities and
other hobbies were documented.
2.4. Procedure
After the subjects were enrolled in the study, had their personrelated data recorded and handedness surveyed, they practiced the

Fig. 2. Modified mental rotation paradigm. Shown are the control condition (norotation or pure maintenance condition) and the task condition (rotation axis
changing). The task condition was announced to the subject each time before the
starting stimulus was presented.

experimental mental rotation task outside of the scanner for
15 min (PC version; Borland Delphi 6.0). The program allowed
the subjects to explore the task, the properties of the stimulus
and the meaning of arrows (Fig. 2). For example, they could actively rotate the object in the diverse directions by mouse-clicks
to become familiar with the different positions and the effects of
90° rotations on the figure. The speed demands (i.e. the number
of required mental rotations during the mental rotation phase)
could be controlled to find out the maximum speed demand level
which was subjectively experienced as still convenient. For further
exploration, the stimulus was made available as a white cardboard
object (120  120  60 mm3). During familiarization, subjects received feedback on the correctness of their responses and the rotation plane was changed three times during one trial, i.e. after each

C. Hoppe et al. / Brain and Cognition 78 (2012) 14–27

single rotation. Subjects were informed about the control condition
and the alternating sequence of true tasks and control tasks applied during scanning. The total time in the scanner was about
35 min, including the preceding localizer and the subsequent T1weighted structural brain scan. The assessment was completed
by applying the CFT-20 R part 1 and the LPS 7 after scanning. The
total duration of the examination was about 75 min.
2.5. Magnetic resonance image acquisition
Magnetic resonance image scanning was performed on a 1.5T
MRI Scanner (Siemens Avanto, Erlangen, Germany) using a TIM
8-channel standard head coil. We acquired 500 T2-weighted, gradient echo planar imaging (EPI) scans including three initial dummy scans that were discarded in order to achieve steady-state
magnetization with the following parameters: slice-thickness = 3 mm; interslice gap = .3 mm; matrix size = 64  64; field
of view = 192  192 mm2; echo time (TE) = 40 ms; repetition time
(TR) = 2910 ms. Thirty-five transversal slices were acquired which
covered the whole cerebral cortex but only the upper part of the
cerebellum. In addition, we obtained a sagittal T1-weighted 3Dmprage sequence with 160 slices.
Scanning was comprised of 30 trials in the control condition
and 30 trials in the task condition. The inter-block interval was jittered and varied randomly between 1.5 and 2.5 s. The scanning
procedure always followed the sequence: localizer (1 min), mental
rotation (60  22 s = 22 min), MPRAGE (8 min). The task was presented via video goggles (Nordic NeuroLab, Bergen, Norway) using
PresentationÓ software (NeuroBehavioural Systems Inc., Albany/
California, USA; monitor resolution: 1024  768 pixels). Subjects
indicated their answers with the help of response grips (NordicNeuroLab, Bergen, Norway).
2.6. Behavioral data analysis
The effects of math talent and gender on MRX were tested using
a bifactorial ANOVA and post hoc T-tests for independent samples.
Nonparametric tests (Mann–Whitney U-test, v2-test) were used to
test group differences of non-normally distributed adjunctive measures (e.g. school grades). Correlations of MRX, other performancerelated parameters and brain activation parameter estimates were
analyzed by Pearson’s product-moment correlation. The significance level was set to a = .05. Behavioral data were analyzed by
SPSS (Version 17.0.1., German release).
2.7. Neuroimaging data analysis
The preprocessing was performed by FSL software version 4.1.2
(FMRIB’s Software Library, www.fmrib.ox.ac.uk/fsl). Preprocessing
included realignment with unwarping; slice timing correction
using Fourier-space time-series phase-shifting; motion correction
using MCFLIRT; grand-mean intensity normalization of the entire
4D dataset by a single multiplicative factor; registration to standard space an EPI-template (resampled voxel size after registration: 3  3  3 mm3); and smoothing with a 8-mm Gaussian
kernel. The fMRI statistical analysis was done using Statistical
Parametric Mapping 5 (SPM5, www.fil.ion.ucl.ac.uk/spm/). The
hemodynamic response to each block was modeled by a canonical
hemodynamic response function. The onset was defined by the
occurrence of the starting stimulus and the modeled block comprised the entire mental rotation phase, except for the final test
(duration: 17 s). For each subject, parameter images for the contrasts of each condition were generated and subjected to a second-level group effects analysis using a one-way ANOVA (within
subject) with group membership as a between subject factor. In order to identify mental rotation related activation irrespective of the

17

group factors, we calculated the contrast of the task condition vs.
the control condition (‘‘task effect’’). To identify effects of the group
factors, the task effect was contrasted between the respective
groups (i.e. group  task interaction, ‘‘group effects’’).
To exclude potentially confounding effects of error-related processes, the correct trials were modeled separately from the error
trials in the first-level analysis and the second-level analysis included correct trials only. Error trials were modeled as an additional regressor of no interest. To further control for possible
confounding effects of different amounts of error-related processing, the experimental task performance score was included as a
covariate in all group effects analyses on the a priori group factors,
i.e. math talent and gender. All analyses were conducted with a
threshold of p < .001, uncorrected, and an extent threshold of
k = 10 adjacent voxels. Anatomical labeling of peak activation voxels was done by the Masked Contrast Images (mascoi) tool for SPM
(version 2.11; Reimold, Slifstein, Heinz, Mueller-Schauenburg, &
Bares, 2006) with a secondary p < .001. Specific brain regions were
analyzed by the MARSeille Boîte À Région d’Intérêt (MarsBaR)
extension of SPM (Brett, Anton, Valabregue, & Poline, 2002;
www.nitrc.org/projects/marsbar/).
According to the findings of a meta-analysis of neuroimaging
studies on mental rotation (Zacks, 2008), our main focus was on
the bilateral PPC, i.e. inferior (IPL) and superior parietal lobule
(SPL). With regard to the frontoparietal axis of general intellectual
functioning (Jung & Haier, 2007), we also wanted to examine group
effects on activations in the dorsolateral prefrontal cortices
(DLPFC). Therefore, we defined bilateral prefrontal and parietal
search volumes and extracted individual mean beta values from
all voxels of each search volume as parameter estimates for the
task-related activations in this area. The search volumes were
masked by task effects from the total sample restricting this analysis to task-related brain regions (Wake Forest University PickAtlas
tool for SPM, release 2.4; Maldjian, Laurienti, & Burdette, 2004;
Maldjian, Laurienti, Kraft, & Burdette, 2003). Multivariate analyses
of covariance (MANCOVA) on the activation parameter estimates
were performed with the experimental task performance score,
MRX, as a covariate and math talent, gender and experimental task
performance as the group factors. In addition, correlation and
regression analyses were performed on activation parameter estimates and MRX. To evaluate the relative impact of the three group
factors on task-related brain activations in the selected brain areas,
multiple linear regression analyses (method: stepwise) were performed on the activation parameters from the search volumes.
Analyses including brain activation parameters were also performed by SPSS (Version 17.0.1., German release).

3. Results
3.1. Performance
As shown in Table 1, math talent and gender were stochastically
independent. In addition, both group factors showed no effect on
age, handedness distribution, nonverbal intelligence (CFT-20 R)
and two-dimensional mental letter rotation performance (LPS 7).
MATH spent significantly more time per day on extracurricular
mathematical activities (Mann–Whitney U-test, p < .001), had better math grades (p < .005), and a tendency towards better total
average grades (non-significant trend, p < .085). No gender effects
on adjunctive behavioral measures were obtained.
Fig. 3 shows the group effects on the mental rotation performance score, MRX. Bifactorial univariate ANOVA confirmed main
effects of the factors math talent [F(1, 30) = 10.57, p < .003,
g2 = 0.22] and gender [F(1, 30) = 6.72, p < .015, g2 = 0.14] with no
interaction effect [F(1, 30) = 1.22, p = .279] indicating that MATH

18

C. Hoppe et al. / Brain and Cognition 78 (2012) 14–27

shows a superposition of these effects (for anatomic labels of peak
activation voxels see Table 2). Behaviorally superior subjects (i.e.
MATH, male subjects, and high performers) consistently showed
higher activations in the left IPL (SMG, not AG) when compared
to behaviorally inferior subjects. In addition, mathematically gifted
subjects showed increased activation in the left SPL and the right
postcentral gyrus. Additionally, male subjects showed activations
in the left precuneus, left precentral and right postcentral gyrus.
High performers also showed activations in the right IPL and the
right middle frontal gyrus. In contrast, behaviorally inferior as
compared to superior groups showed more diverse patterns of relatively increased activations either in left frontal (math talent, gender) or temporoparietal (experimental task performance) regions.
3.4. Activation parameters [Fig. 6 and 7]
Fig. 3. Experimental task performance score, MRX (group means, bars in graph are
SEs of the mean). Bifactorial univariate ANOVA obtained main effects of math talent
[F(1, 30) = 10.57, p < .003, g2 = 0.22] and gender [F(1, 30) = 6.7, p < .015, g2 = 0.14]
indicating higher accuracy in the talented subjects and in the male subjects; no
‘math talent  gender’ interaction effect was obtained, F(1, 30) = 1.22, p < .279.

vs. CON and male vs. female subjects showed better mental rotation performance. Regarding the control task condition, the bifactorial univariate ANOVA obtained a main effect for math talent
[F(1, 30) = 5.16, p < .030, g2 = 0.15] indicating higher accuracy of
MATH during the maintenance condition; no main effect of gender
(p = .836) and no ‘‘talent  gender’’ interaction effect (p = .653)
were revealed. MRX was correlated with the pre-scanning self-estimate of maximum mental rotation speed (r = 0.41, p < .017), mental letter rotation performance (r = 0.41, p < .016), math grades
(r = 0.57, p < .001), and, in a non-significant trend, nonverbal
intelligence (r = 0.32, p = .062).
The stimulation task performance factor was retrospectively defined by a median split of the total sample based on MRX (cutoff = 0.85, high/low performers: 17/17). The performance factor
was significantly correlated with math talent [v2(1) = 5.765,
p < .016] (5/17 MATH were low performers, 5/17 CON were high
performers) but not with gender [v2(1) = 1.889, p = .169] (6/16
male subjects were low performers, 7/18 female subjects were
high performers).
For the total sample, error rates from the active control condition were significantly lower than from the task condition
(Mean ± SD:
baseline = 0.06 ± 0.07;
task = 0.26 ± 0.19)
[T(33) = 6.85, P < .001]. As a non-significant trend, mean error
rates from task trials with one change of the rotation axis were
lower than for trials with three changes of the rotation axis
[T(33) = 1.977, p = .056] thus indicating that the modified mental
rotation paradigm allows the manipulation of task difficulty in
terms of cognitive load.
3.2. Task effects
Mental rotation activated widespread parietal and frontal regions in the cerebral hemispheres, midbrain and also cerebellum
(Table 2). The peak activation for the ‘‘task vs. control condition’’
contrast was in the left IPL (Montreal Neurological Institute
[MNI] template coordinates for the peak activation voxel:
X = 42; Y = +39; Z = +42). More specifically, the activated inferior
parietal region comprised the supramarginal gyrus (SMG, Brodmann area/BA 40/7) but not the angular gyrus (AG, BA 39).
3.3. Group effects
The effects of the two capability-related group factors, math talent and gender, and the retrospectively defined task performance
factor on task-related brain activations are shown in Fig. 4; Fig. 5

Individual task-related activation parameter estimates (beta
values) were extracted from the following frontal and parietal
search volumes: left DLPFC corresponding to BA 9 (L.DLPFC, 60
voxels); right DLPFC (R.DLPFC, 29 voxels); left IPL (L.IPL, 383 voxels); right IPL (R.IPL, 217 voxels); left SPL (L.SPL, 199 voxels); and
right SPL (R.SPL, 171 voxels). The shown voxel numbers refer to
the search volumes after masking by the task effect.
Activations in the left and right prefrontal search volumes
showed no group effects [math talent: L.DLPFC: T(32) = 1.471,
p = .151; R.DLPFC: T(32) = 0.739; p = .465; gender: L.DLPFC:
T(32) = 0.953, p = .348; R.DLPFC: T(32) = 0.640, p = .527; experimental task performance: L.DLPFC: T(32) = 0.736, p = .467;
R.DLPFC: T(32) = 1.426, p = .163].
Fig. 6 shows the group means of the parameter estimates extracted from the parietal cortical search volumes for MATH and
CON (panel A), male and female subjects (B), and high and low
mental rotation performers (C). Repeated measures MANOVA
including the a priori group factors, math talent and gender, and
the four parietal activation parameter estimates (L.IPL, R.IPL,
L.SPL, R.SPL) as dependent variables obtained no multivariate main
or interaction effects (p > .196 for all effects). Post hoc univariate
testing yielded a main effect of math talent on L.IPL activation [F(1, 30) = 5.518, p < .026, g2 = 0.14] and near-significant
trends towards main effects of math talent on L.SPL activation [F(1, 30) = 3.469, p = .072, g2 = 0.092] and of gender on L.IPL
[F(1, 30) = 3.999, p = .055, g2 = .101] and L.SPL activation
[F(1, 30) = 4.126, p = .051, g2 = 0.109]; no interaction effect was
indicated. The univariate effects of talent and gender were omitted
if MRX was included as a covariate into the model (p > .130 for all
effects). In contrast to the a priori group factors, the task performance group factor yielded a multivariate main effect on the parietal activation parameters [Wilks lambda = 0.702, F(4, 29) = 3.079,
p < .031]. In addition, post hoc univariate testing obtained effects
of the performance factor on activation parameter estimates from
each included parietal search volume [L.IPL: F(1, 32) = 8.398,
p < .007, g2 = .021; R.IPL: F(1, 32) = 9.742, p < .004, g2 = 0.23; L.SPL:
F(1, 32) = 4.362, p < .045, g2 = 0.12; R.SPL: F(1, 32) = 5.084, p < .031,
g2 = 0.14].
MRX showed positive correlations with the activation parameter estimates from the parietal search volumes (L.IPL: r = 0.41,
p < .016; R.IPL: r = 0.40, p < .018; R.SPL: r = 0.39, p < .023) except
for L.SPL the activation of which showed a near-significant trend
(r = 0.34, p = .051). The scatter plots in Fig. 7 show the correlation
of MRX and task-related activations of the bilateral IPL; math talent
and gender are also indicated. L.IPL activation was also positively
correlated with the pre-scanning self-estimate of mental rotation
speed (r = 0.363, p < .035).
To evaluate the relative contributions of the three performance
factors to task-related brain activation, we performed stepwise
multiple linear regression analyses on the activation parameters

C. Hoppe et al. / Brain and Cognition 78 (2012) 14–27

19

Table 2
Total sample analyses: task effects and task by group effects.
Contrast

Location

N

Region

Task effects
Task > control
condition

[ 48; +6; +30]

9.61

1872

[ 42; 39; +42]
[+24; 54; +66]
[ 24; 3; +60]
[+30; 51; 27]
[ 48; 66; 12]
[+51; 63; 6]
[ 12; 27; +15]
[+54; +9; +27]
[+33; +21; +6]
[ 6; 15; 15]

12.42
9.87
11.48
8.07
8.32
7.07
5.79
7.19
5.18
4.06

1601
1360
1283
683
347
272
135
114
43
15

L. inferior frontal gyrus, lentiform nucleus, caudate (head), putamen, middle frontal gyrus,
precentral gyrus, Thalamus, Insula, anterior cingulate, lateral and third ventricle
L. inferior parietal lobule, superior parietal lobule (BA 7/40), precuneus, postcentral gyrus
R. inferior parietal lobule, superior parietal lobule (BA 7/40), precuneus, precentral and postcentral gyrus
L. precentral gyrus, medial frontal, middle frontal, superior frontal gyrus, cingulate gyrus
R. cerebellum (culmen)
L. inferior temporal gyrus, middle temporal gyrus, inferior and middle occipital gyrus (BA 37)
R. inferior temporal gyrus, middle temporal gyrus, inferior and middle occipital gyrus (BA 37)
L. thalamus (pulvinar), ventricle
R. inferior frontal gyrus (BA 9)
R. insula, inferior frontal gyrus (BA 13)
L. midbrain

Task  group effects
Talented > control
[ 36; 30; +33]
subjects
[+60; 24; +45]
[ 42; 48; +57]
[ 33; 60; +57]

4.53
4.46
4.03
3.85

53
23
18
11

L. inferior parietal lobule (BA 40), postcentral gyrus, subgyral
R. postcentral gyrus
L. inferior parietal lobule (BA 40)
L. superior parietal lobule (BA 7)

Tmax

Control > talented
subjects

[
[
[
[
[

39; +33; 15]
42; +21; 9]
24; +45; +30]
6; 63; +36]
6; +27; +33]

4.83
4.03
4.20
3.68
4.07

30
24
16
16
13

L.
L.
L.
L.
L.

Male > female
subjects

[ 42; 36; +51]
[ 30; 45; +57]
[ 27; 15; +60]
[+36; 36; +54]
[ 12; 72; +42]
[+24; 39; +15]
[ 33; +18; +51]

4.65
4.22
5.10
4.75
3.63
3.75
4.08

67
53
52
51
24
19
20

L. inferior parietal lobule (BA 40), postcentral gyrus
L. inferior parietal lobule (BA 7/40)
L. precentral gyrus (BA 6)
R. postcentral gyrus
L. precuneus (BA 7)
R. extranuclear, lateral ventricle
L. middle frontal gyrus (BA 8)

High > low
performers

[ 45; 36; +48]
[+36; 42; +51]
[+30; 54; +42]
[+24; +0; +51]

4.32
4.06
3.99
3.91

115
80
18
15

L. inferior parietal lobule (BA 40)
R. inferior parietal lobule (BA 40)
R. parietal, subgyral
R. middle frontal gyrus (BA 6)

Low > high
performers

[ 60; 15; 9]
[+60; 48; +24]

4.69
3.66

81
10

L. middle temporal gyrus (BA 21)
R. superior temporal gyrus

Female > male
subjects

inferior frontal gyrus (BA 47)
inferior frontal gyrus (BA 47)
superior frontal gyrus (BA 10)
precuneus
cingulate gyrus (BA 32)

Peak activations. Location: Montreal Neurological Institute template coordinates; Tmax: T-value of peak activation voxel; N: number of clustering voxels that survived the
significance threshold of p < .001, uncorrected (cluster size). The anatomic region of the peak activation voxel is written in bold (tentative Brodmann areas in parentheses).
The experimental task performance score (MRX) was included as a covariate in the group effects analysis on math talent and gender. The T-maps are shown in Figs. 4 and 5.

from each of the four parietal search volumes. For all parietal
activation parameters, the models only included the task
performance factor and excluded the capability-related factors,
math talent and gender (model parameters: L.IPL: b = 0.456,
corrected R2 = 0.183, F(1,32) = 8.398, p < .007; R.IPL: b = 0.483, corrected
R2 = 0.209, F(1, 32) = 9.742, p < .004; L.SPL: b = 0.346, corrected
R2 = 0.092, F(1,32) = 4.362, p < .045; R.SPL: b = 0.370, corrected R2 =
0.110, F(1, 32) = 5.084, p < .031). The models explained only
a small percentage of the variance of the task-related brain
activations.
Multiple linear regression analyses for the prediction of MRX included math talent (model parameter estimate: b = 0.475) and
gender (b = 0.370) but none of the activation parameters as regressors [corrected R2 = 0.321, F(2, 31) = 8.813, p < .001]. A second model which exclusively considered the four parietal activation
parameters only included L.IPL activation as a regressor
(b = 0.411, corrected R2 = 0.143, F(1, 32) = 6.487, p < .016).
4. Discussion
To examine the neural mechanisms underlying better cognitive
performance, we combined the two established approaches to research on the brain–performance relationship by evaluating the effects of math talent and gender as capability related group factors
and experimental task performance on the brain activations during

a modified mental rotation task. As predicted, math talent and
male gender were associated with higher mental rotation accuracy
(hypothesis I). Mental rotation activated the bilateral IPL, among
other parietal, frontal and temporal areas, and the left IPL was
identified as a convergent zone of the effects of math talent, gender
and task performance on task-related brain activations (hypotheses II–III). Activations in the left and right IPL and SPL were mostly
explained by the effects of task performance, whereas math talent
and gender were excluded from the multiple linear regression
models (hypothesis IV). In the following discussion, we refer these
findings to the neuroimaging evidence on mental rotation and to
reported effects of math talent, gender and task performance.
The discussion will be completed by relating NCP research to the
research on neural efficiency.
4.1. Task effects
We proposed a novel mental rotation paradigm for NCP research. Mental rotation proper was separated from both the preceding encoding of the initial position of the figure and from the
final target-probe stimulus comparison test according to the preparation rotation paradigm (Bethell-Fox & Shepard, 1988; JansenOsmann & Heil, 2007; Lamm, Windischberger, Moser, & Bauer,
2007). In this paradigm response latency no longer represents a
meaningful measure of mental rotation performance; conse-

20

C. Hoppe et al. / Brain and Cognition 78 (2012) 14–27

Fig. 4. Group effects analyses on task-related brain activations. T-maps are shown as glass brains. Panel A: math talent; panel B: gender; panel C: experimental task
performance factor (post hoc). Significance threshold: p < .001, uncorrected (height threshold: T(95) = 3.178248, N = 34); extent threshold: k = 10 clustering voxels. MNI
template coordinates and Tmax-values of the peak activation voxel, number of clustering voxels surviving the significance threshold, and anatomic regions are shown in
Table 2. A superposition of these effects (left column) is shown in Fig. 5.

quently, the classical mental rotation effect of a positive correlation between angular disparity of the stimulus to be rotated and
the response latency can no longer be assessed (Shepard & Metzler,
1971). As the figure is physically absent during mental rotation, the
modified paradigm addresses the maintenance (control task) and
manipulation component (mental rotation proper) of visuospatial
working memory without perceptual support. As intended, the obtained mean error rate during mental rotation proper in this study
was moderate (26%) but still allowed superior subjects to excel (i.e.
no floor/ceiling effect). Even in the subgroup with the worst performance (i.e. non-gifted female subjects), the mean error rate (36%)
was below the mean error rates reported by others (>40%; O’Boyle
et al., 2005; Weiss et al., 2003a,b).
Many fMRI studies demonstrated a specific role of the PPC for
mental rotation (Alivisatos & Petrides, 1997; Booth et al., 2000; Cohen et al., 1996; Gill, O’Boyle, & Hathaway, 1998; Halari et al.,
2006, 2000; Harris & Miniussi, 2003; Hattemer et al., 2011; Jaga-

roo, 2004; Just, Carpenter, Maguire, Diwadkar, & McMains, 2001;
Podzebenko, Egan, & Watson, 2002; Podzebenko, Egan, & Watson,
2005; Suchan, Botko, Gizewski, Forsting, & Daum, 2006; Weiss
et al., 2003a,b; Zacks, 2008). These findings are in accordance with
neuropsychological evidence (Farah, 1989). Current neurocognitive
models for mental rotation consistently assign a key role to the PPC
(Ecker, Brammer, & Williams, 2008; Jordan et al., 2002; Zacks,
2008). PPC is generally activated during the manipulation of information in working memory (Champod & Petrides, 2007, 2010). In
addition, current models coherently include motor areas (precentral G., premotor area I, supplementary motor area/SMA) indicating
either preparatory motor processes or a motor imagery component
during mental rotation.
In keeping with these models, we obtained extended bilateral
PPC activations during mental rotation. However, in our paradigm
the peak activations were in the IPL instead of the SPL. SPL activation was shown to be correlated with angular disparity in standard

C. Hoppe et al. / Brain and Cognition 78 (2012) 14–27

21

Fig. 5. Superposition of effects of math talent (blue), gender (yellow) and experimental task performance (green) on task-related brain activation (behaviorally superior vs.
inferior subjects). For further details see Fig. 4. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

mental rotation paradigms (Gogos et al., 2010) but we constantly
used a rather small rotation angle of 90°. Several studies, including
the first fMRI study on this paradigm (Cohen et al., 1996), discussed the possibility that IPL activation during mental rotation
in three-dimensional space reflects manual action imitation (motor imagery) rather than visual imagery. Accordingly, a study
employing repetitive transcranial magnetic stimulation provided
evidence for a causal role of the IPL in motor imagery but not visual
imagery (Pelgrims, Andres, & Olivier, 2009). Also, no IPL activation
was obtained during a two-dimensional letter plane rotation
which was presumed to comprise no motor imagery component
(Podzebenko et al., 2002). In our own study, applying a mental action imitation strategy might have been suggested to the subjects
by offering them the real cardboard object for evaluation during
the pre-scan familiarization phase. In addition, our sample was
comprised of non-standard subjects with a specific math talent
and markedly higher mental rotation performance; both of these
factors were related to increased activation of the IPL.
Consistent with one of the models (Ecker et al., 2008), we observed bilateral activations of the inferior temporal gyri indicating
a role for higher-order visual object recognition. No SMA activation
as a possible indicator of motor imagery was obtained, which is in
line with two of the models (Ecker et al., 2008; Jordan et al., 2002).
Contrasting with other recent studies, we did not find activations
in hV5/MT+ or other occipital or temporooccipital regions (Seurinck, de Lange, Achten, & Vingerhoets, 2010). Notwithstanding
the models, we obtained an extended activation of the left inferior
frontal gyrus which might indicate a role for language-related processing (e.g. subvocal self-instruction).
4.2. Effects of math talent
Mental rotation performance was reported to be correlated
with mathematical aptitude but other factors (e.g. gender) are

known to modulate this relation (Casey et al., 1995; Hyde, 2005;
Nuttall, Casey, & Pezaris, 2004; Spelke, 2005). For example, Casey
et al. (1995) reported that mental rotation performance was a better predictor of mathematical abilities (Scholastic Aptitude Test,
SAT-M) than verbal performance (SAT-V) in all female and two of
four male subsamples but not in talented males and male college
students. However, one meta-analysis failed to confirm a relevant
relation between visuospatial abilities and mathematical aptitude
(Lubinski & Humphreys, 1990). Talent effects on mental rotation
performance depend on the appropriate scaling of the task difficulty (no floor/ceiling effects) and might also be affected by the test
modalities (e.g. paper–pencil vs. computerized test). For example,
O’Boyle et al. (2005) reported an unexpected equally moderate
mental rotation accuracy of mathematically gifted and control subjects during scanning (accuracy