Mental rotation performance has been found to produce one of the largest sex differences in cognition accompanied by sex differences in functional cerebral asymmetry. Although sex differences in mental rotation performance can be reliably demonstrated as early as age 5 years old, that is, long before puberty, no data exist as to whether preschooler's mental rotation performance is accompanied by sex differences in functional cerebral asymmetry. Based on the electrophysiological brain correlates of mental rotation, we observed a bilateral parietal brain activity for preschool boys whereas the preschool girls' brain activity was clearly lateralized toward the left hemisphere if and only if mental rotation was needed to solve the task. Thus, sex differences in functional cerebral asymmetry during mental rotation do not require hormonal changes that occur during puberty.
Sex differences in spatial cognition, especially in mental rotation, that is, imagining an object turning around (Heil, Bajric, Rösler, & Hennighausen, 1997; Shepard & Metzler, 1971), constitute an empirical fact (Linn & Petersen, 1985), although the effect size depends upon the stimulus material (Voyer, Voyer, & Bryden, 1995) and whether accuracy or speed data are used as dependent variables (Jansen-Osmann & Heil, 2007c). The causes for the sex effect, however, are still far from being understood. Both “psychosocial” as well as “biological–neuronal” explanations can quote empirical support and are by no means mutually exclusive. In this article, discussion will be restricted to sex differences in functional cerebral asymmetry.
Hemispheric lateralization of brain activity during mental rotation as a function of sex was reported rather consistently (Hugdahl, Thomsen, & Ersland, 2006; Seurinck, Vingerhoets, de Lange, & Achten, 2004; Jordan, Wustenberg, Heinze, Peters, & Jäncke, 2002), with—as a rule—a larger right parietal activity observed in men and a larger left parietal activity in women. Compelling data from diverse fields (neuroendocrinology, psychology, neurology, genetics, neurochemistry, and neuroanatomy) and numerous species including humans indicate that the brain is a sexually differentiated organ. Sex hormones have organizational effects on brain structure and function and also have subsequent activational effects on the brain (Carne, Vogrin, Litewka, & Cook, 2006; Güntürkün & Hausmann, 2003; Williams & Meck, 1991). The latter has repeatedly been demonstrated by the observation that the continuous steroid fluctuations in normally cycling women lead to concomitant changes of sex differences both in performance (Hausmann et al., 2006; Halari et al., 2005; Hausmann, Slabbekoorn, Van Goozen, Cohen-Kettenis, & Güntürkün, 2000) and in cerebral asymmetries (Schöning et al., 2007; Hausmann, 2005). Evidence for the former statement, however, is less conclusive (Güntürkün & Hausmann, 2003). Moreover, it is still completely unknown whether the continuous steroid fluctuations in cycling women cause or modify these sex differences. In the former case, one would expect no sex differences before puberty whereas in the latter case these sex differences should already exist in early childhood. The existence of such effects long before puberty would constitute indirect evidence for organizational effects of pre- and postnatal sex hormones on cerebral asymmetry—although alternative explanations exist, see below—but such data are simply absent.
Given a sufficient statistical power, male superior performance in mental rotation can already be observed in children as young as 5 or 6 years old (Casey et al., 2008; Levine, Huttenlocher, Taylor, & Langrock, 1999). Whether these early sex effects are or are not accompanied by a differential hemispheric lateralization, however, has not been reported yet at all. Therefore, we analyzed mental rotation-related amplitude effects of ERPs extracted from the EEG in 5-year-old children. The advantages of this approach are that ERP recordings are completely noninvasive and required sample sizes to obtain sufficient statistical power are realizable. In fact, n = 96 participants are needed assuming a large sex effect of d = .75 and levels of α = β = .05. Moreover, the amplitude effects used here are highly specific for mental rotation (Heil, 2002); for example, the presence of the amplitude effect depends upon the presence of mental rotation (Heil, Bajric, Rösler, & Hennighausen, 1996) but is independent from the presence of an overt response (Heil, Rauch, & Hennighausen, 1998). Heil and Rolke (2002) provided evidence that not only the process of mental rotation but also the amplitude modulation itself was delayed, either when the perceptual quality of the stimulus was reduced or when character discrimination was more difficult. In addition, the amplitude effect was already successfully used for measuring functional parietal asymmetry in children (Heil & Jansen-Osmann, 2007; Jansen-Osmann & Heil, 2007a). Finally, the ERPs in the present study were recorded in a preschool setting, thus avoiding the unfamiliar and probably intimidating lab environment. Taken together, at least in the case of sex differences in preschoolers during mental rotation, these advantages outweigh the limited spatial resolution of ERPs compared with, for example, fMRI.
In the present study, blocks with upright versus blocks with rotated pairs of animal drawings were presented, and the children had to decide whether the two drawings were the same or mirror reversed. RTs, error rates, and average ERP amplitudes were recorded. Hemispheric lateralization was determined by comparing the mental rotation-related amplitude modulations over the left versus the right parietal electrode leads (P3 vs. P4).
We recorded ERPs in 97 children in two different types of trials: In upright trials, subjects had to decide whether two upright animal drawings were the same or mirror reversed. In rotated trials, the right animal additionally was rotated in the picture plane, and subjects had to rely on mental rotation to solve the task.
In five preschools in the area of Düsseldorf, Germany, 104 children (age range = 4.8 to 6.1 years) participated in the study after their parents signed informed consent. Data of seven children had to be excluded because of less than 60 artifact-free and correctly answered trials per upright and rotated condition (five children) or because of technical problems during EEG acquisition (two children). The final sample comprised 47 girls (mean age = 5.3 years) and 50 boys (mean age = 5.4 years).
The experiment was run on a PC laptop with a 15-in. monitor located approximately 50 cm in front of the children. The experimental stimuli consisted of colored drawings of 12 different animals (camel, crocodile, dog, donkey, elephant, grizzly, lion, pig, rhino, sheep, turtle, and zebra; taken from Rossion & Pourtois, 2004). In each trial, two drawings of the same animal were presented together. The left drawing was presented always upright facing either to the left or to the right. The right drawing was either identical or mirror reversed. In the upright blocks, the right drawing was also presented upright. In the rotated blocks, the right drawing was rotated 90°, 180°, or 270°. Participants responded “identical” by pressing the left mouse button and “mirror reversed” by pressing the right one. This simpler response assignment (see, e.g., Jansen-Osmann & Heil, 2007b) was the same for all participants. In addition, participants were free to use two fingers of the left hand or of the right hand or to use one finger from each hand, and in fact, children varied their response assignment during the course of the experiment. This reduction of experimental control constituted a concession to the fact that 5-year-olds completed 240 trials.
Procedure and Task
Participants were tested individually in a quiet room of their particular preschool. Before the experiment was started, the required decision (identical vs. mirror reversed) was explained in the following way: First, articles with two upright drawings of the same animal either facing into the same or into opposite directions were presented, and the children were instructed that their task was to decide whether the two drawings of the animal were facing into the same direction. Thereafter, sheets of article with one upright animal drawing on the left and one (moveable) rotated drawing on the right were presented. Now children had the opportunity to manually rotate the right drawing to understand the required decision, that is, whether the two drawings of the animal were facing into the same direction when standing on their feet.
In each trial, a pair of drawings was presented 1 sec after presentation of a fixation point. The drawings remained visible until a button press response and were then replaced by a “+” or “−” for 500 msec, indicating the correctness of the response. Trials were separated by randomly varying intervals of 2–3 sec. In addition, after every eight trials, participants were free to take a break. Two hundred forty trials were presented in four blocks. Blocks 1 and 4 consisted of 48 trials, each with upright stimuli only [12 (animals) × 2 (left one facing left vs. right) × 2 (right one identical vs. mirror reversed)]. Blocks 2 and 3 consisted of 72 trials each with rotated stimuli only [12 (animals) × 2 (right one identical vs. mirror reversed) by the three different angular disparities; the assignment of facing of the left animal to Block 2 versus Block 3 was determined randomly for each subject and for each of the 72 combinations mentioned above]. Each block was preceded by 16 practice trials with animal drawings not used in the test sessions to familiarize the children with the respective task. The whole experiment, including electrode application (see below) and removal, lasted less than 1.5 hr.
ERP Recording and Analysis
The EEG was recorded from nine Ag/AgCl electrodes placed at F3, Fz, F4, C3, Cz, C4, P3, Pz, and P4, respectively, based on the international 10–20 system (Jasper, 1958). All electrodes were attached directly to the scalp and were referenced to digitally averaged earlobes. Impedance was kept below 10 kΩ. To control for eye movement artifacts, vertical and horizontal electrooculogram was recorded. A ground electrode was placed at the right mastoid. Digitization rate was 250 Hz and an on-line band-pass was set from DC to 100 Hz. A digital off-line filter was set from 0.53 to 40 Hz. Continuous EEG was segmented into epochs of 1500 msec length starting 200 msec before stimulus presentation. The EEG was corrected for blinks and eye movements (Gratton, Coles, & Donchin, 1983). Automatic artifact rejection excluded epochs if voltage steps between two sampling points exceeded 80 μV, if absolute amplitude exceeded ±175 μV, if absolute voltage difference within a segment exceeded 225 μV, or if activity was less than 5 μV for a time epoch longer than 150 msec. ERPs were extracted by averaging single trials with correct responses separately for subjects and upright versus rotated trials. On the basis of previous studies (e.g., Heil & Jansen-Osmann, 2007), the mental rotation-related amplitude effect was quantified as the mean voltage in the time window 500–800 msec after stimulus onset referenced to a prestimulus baseline of 200 msec. The pattern of significant results, however, did not change when a more fine grained temporal analysis or a more prolonged one was used.
With error rate as the dependent variable, an ANOVA was calculated with the between-subjects factor “sex” and the within-subjects factors “upright versus rotated trials” and “first versus second half of the experiment.” Boys outperformed girls with respect to accuracy only when rotated pairs of drawings were presented, resulting in a significant interaction of sex and upright versus rotated trials, F(1,95) = 4.64, p < .05, d = .44, see Figure 1. No other main effect or interaction turned out to be significant, F(1,95) < 1.5, p > .30.
Only trials with correct responses were used for RT analysis. Before the analysis, RT data were trimmed for outliers. RTs more than 1.5 SDs above or below the mean per condition and per subject were excluded (12.9% of the data, on average). One ANOVA was based on the inverse of the slope of the regression line, calculated separately for each participant, relating RT to angular disparity, expressed as degrees per second. A second ANOVA was based on the intercept of the regression line. ANOVAs were calculated with the between-subjects factor “sex” and the within-subjects factor “first versus second half of the experiment.” Neither the mental rotation speed (M = 138.7°/sec, SE = 17.4°/sec) nor the intercept (M = 2145.7 msec, SE = 64.2 msec) was affected by sex, F(1,95) = 1.14 and 0.98, respectively, or by the interaction of sex and half, F(1,95) = 1.04 and 0.46, all p > .30, all d < .25. This pattern of effects did not change when untrimmed RT data were used for the regression analysis, F(1,95) = 0.84 and 0.62, respectively, both p > .30, d < .20.
With respect to brain imaging data, an ANOVA for upright trials with the between-subjects factor “sex” and the within-subjects factors caudality (frontal vs. central vs. parietal electrodes) and laterality (left vs. right electrodes) revealed only a significant main effect of caudality, F(2,190) = 10.48, p < .05. Neither sex, F(1,95) = 1.06, nor laterality, F(1,95) = 0.69, both p > .30, affected ERP amplitudes for upright stimuli.
A second ANOVA was calculated with the additional electrical brain activity observed when mental rotation was required, that is, with the difference between rotated and upright trials. In addition to a significant two-way interaction of sex by laterality, F(1,95) = 7.32, p < .01, a significant three-way interaction of sex by laterality by caudality was obtained, F(2,190) = 3.47, p < .05. The increase in brain activity due to mental rotation differed as a function of sex and laterality at parietal leads, F(1,95) = 10.62, p < .01, see Figure 2. The sex effect of this hemispheric lateralization amounted to d = .65. Sex-dependent hemispheric lateralization was absent at frontal leads, F(1,95) = 0.48, p > .30, d = .11, and missed significance at central leads, F(1,95) = 3.09, p > .08, d = .36. At parietal leads, boys revealed a more bilateral pattern of brain activity, F(1,49) = 2.71, p > .10, with only a numerically higher right parietal activation. Girls' brain activity, however, was clearly lateralized toward the left parietal hemisphere, F(1,46) = 8.03, p < .01, see Figure 3.
The well-know amplitude modulation primarily at parietal leads as a function of mental rotation consistently observed with adults (e.g., Bajric, Rösler, Heil, & Hennighausen, 1999) was replicated in the present study with children as young as 5 or 6 years. Although in the time range of the late positive complex (500–800 msec poststimulus), this effect should not be understood as a P300 modulation. Instead, existing evidence strongly suggests that the effect is caused by a modulation of a slow negativity that should be taken as a direct physiological correlate of the mental rotation process itself, as originally suggested by Wijers, Otten, Feenstra, Mulder, and Mulder (1989; for a review of the evidence, see Heil, 2002). Negativity in the slow ERP usually is understood as an increase of neural activity in the underlying cortex (e.g., Rösler, Heil, & Röder, 1997).
Most importantly, the present study extends the findings of a hemispheric lateralization during mental rotation as a function of sex in adults toward children as young as 5 years. Thus, although menstrual cycle–related activating effects of sex hormones on sex differences in hemispheric lateralization were reported (see, e.g., Schöning et al., 2007; Hausmann, 2005), a fundamental sex difference in hemispheric lateralization of spatial cognition exists long before puberty. Thus, menstrual cycle effects in postpuberty samples modulate sex differences in hemispheric lateralization that already existed before puberty-related sex hormone activation. The main finding of the present study thus is straightforward, its theoretical consequences, however, are far from being understood. One might claim that the finding, admittedly indirectly, suggests the organizing influence of early—probably postnatal rather than prenatal—differential exposure to sex hormones (Falter, Arroyo, & Davis, 2006; Hines et al., 2003), but this interpretation is not obligatory.
Given the theoretical impetus these results eventually might have regarding early hormonal effects, a number of qualifications have to be discussed: First of all, although the pattern of results can be understood as sex differences in hemispheric lateralization, one might also argue that the main result is a sex difference in left hemisphere activation during mental rotation. Therefore, it would be interesting to see whether the (in the present study nonsignificant) trend toward a larger right than left hemisphere activation for boys might be replicated in further studies. If not, one might suggest that the left-lateralized brain activity in girls might be understood as a consequence of girls' language development usually being ahead of boys'. As an indirect test of this assumption, we calculated the correlation between children's age and the hemispheric asymmetry of the amplitude modulation as a function of stimulus orientation for the whole sample as well as separately for both sex. In fact, none of these three correlations turned out to be significantly different from zero (r < .15, p > .25). Thus, we conclude that the available data from our experiment do not support the assumption that the age-correlated level of language development affects sex differences in hemispheric lateralization during mental rotation. In addition, the absence of significant correlations between age and hemispheric lateralization does also indirectly address the possibility that the mature profiles of laterality (left with female and right with male participants) emerge earlier for girls than for boys. We cannot exclude this possibility, but at least in the restricted age range realized in the present study (4.8 to 6.1 years), no evidence is observed for this interpretation.
The most important theoretical problem, however, is the fact that the causal link between sex differences in mental rotation and sex differences in hemispheric lateralization, although plausible indeed (see, e.g., Bayer, Kessler, Güntürkün, & Hausmann, 2008), still has to be fully validated. Especially, the question arises as to whether sex differences in hemispheric lateralization are the cause of or an effect of sex differences in mental rotation. If sex-specific strategies of solving mental rotation tasks exist, then the neuroscientific approach that suggests a causal link between brain activation and performance might be replaced by a cognitive approach suggesting a causal link between cognitive processing strategies and performance with brain activation being understood as an effect instead. In fact, some evidence exists suggesting that men generally tend to use more holistic strategies whereas women tend to prefer analytic strategies (e.g., Heil & Jansen-Osmann, 2008; Thomsen et al., 2000) that in turn might cause sex-specific lateralization patterns. This evidence, however, originates from studies with adults and with complex stimuli and is based on sex differences in mental rotation speed. It is completely open as to whether these sex-related strategy differences exist at all in preschoolers tested with relatively simple stimuli. In the present study, we obtained no evidence in favor of sex-related strategy differences because mental rotation speed was not affected by sex. New ideas and creative experimental designs are needed to learn more about the causes and effects of sex differences in hemispheric lateralization. Our results at least provide the opportunity to investigate these questions independent of modulating menstrual cycle effects.
The question arises, however, whether sex differences in functional hemispheric lateralization are based upon sex differences in structural asymmetry. Recent evidence by Hänggi et al. (2008) suggests a very interesting link between structural and functional lateralization: In women, visuospatial cognition was predominantly correlated with left parietal gray matter volume differences reflecting a relative increase in the somatodendritic tissue necessary for information processing. In men, however, visuospatial cognition was predominantly correlated with right parietal white matter volume differences reflecting a relative increase in connecting axons necessary for transportation of information (Gur et al., 1999). For a detailed review of the evidence in favor of a parietal sexual dimorphism hypothesis with respect to relative gray versus white matter volume differences, see Hänggi et al. (2008).
Even if one accepts the causal role of sex differences in functional hemispheric lateralization based upon sex differences in structural asymmetry, however, the question emerges whether these sex differences are caused genetically or rather hormonally or as a result of experience and usage-dependent neural plasticity. Genetic and hormonal factors may have considerable effects on, among others, neural networks relevant for spatial cognition. For example, women with Turner syndrome (X-chromosome deficiency) are impaired in spatial cognition and exhibit structural anomalies in the parietal lobes (Brown et al., 2004; Molko et al., 2004). Besides genetic factors, sex hormones exert considerable effects on spatial cognition, at least in women (Bayer et al., 2008; Sherwin, 2003), and modulate gray and white matter volumes even in the parietal lobes (Goldstein et al., 2001).
However, experimental evidence also suggests that learning- and training-induced cortical plasticity is also accompanied at a structural level by an increase in gray matter volume (e.g., Draganski et al., 2004, 2006), reflecting an increase in somatodendritic tissue. Because boys move more frequently than girls (Campbell & Eaton, 1999), boys encounter more differentiated environments that place greater demands on neural networks responsible for spatial cognition. Additional studies revealed that sex-stereotyped space-related activities predict visuospatial performance and that women improve in spatial cognition when engaged in male-associated space-related activities (Ginn & Pickens, 2005; Devlin, 2004; Signorella, Jamison, & Hansen Krupa, 1989; but see Heil, Rösler, Link, & Bajric, 1998). Thus, whether sex differences in functional and structural hemispheric lateralization are caused by biological (genetic, hormonal) or environmental (spatial activities, socialization) factors remains to be answered (e.g., Casey, Nuttall, & Pezaris, 1999).
All of the theoretical questions, problems, and even pitfalls mentioned above, however, are in no way exclusive to mental rotation, but in fact apply to cognitive neuroscience as a whole field. We look with optimism to the future of mental rotation research because we truly believe that this experimental paradigm might act as a prime example for solving these fundamental questions of cognitive neuroscience, thanks to the many fields within both psychology, biology, and neurosciences dealing with this aspect of spatial cognition for some time.
In summary, the present study provides evidence for a sex-related hemispheric lateralization during mental rotation in children as young as 5 years. Thus, a fundamental sex difference in hemispheric lateralization of spatial cognition exists long before puberty. Menstrual cycle effects in postpuberty samples modulate sex differences in hemispheric lateralization that already existed before puberty-related sex hormone activation.
We thank B. Rolke, K. Lange, and two anonymous reviewers for their helpful comments. This work was supported by grants from the German Research Foundation (DFG) and from the Federal Ministry of Education and Research (BMBF) to P. J. and M. H.
Reprint requests should be sent to Prof. Dr. Petra Jansen, Department of Sport Science, University of Regensburg, 93040 Regensburg, Germany, or via e-mail: Petra.Jansen@psk.uni-regensburg.de.