Cognitive traits shape the brain activity associated with mental rotation

Authors: Nadia M Bersier, Eleonora Fornari, Raffaella I Rumiati, Silvio Ionta

Abstract

Mental rotation is a spatial cognitive ability influenced by several factors, including cognitive traits. However, the relationship between mental rotation performance, cognitive traits, and brain activity is still uncertain. To fill this gap, we recorded functional magnetic resonance imaging data while 55 neurotypical participants performed mental rotation with images of geometric objects, human bodies, and real objects. Cognitive traits were evaluated through the Object–Spatial Imagery Questionnaire (visual cognitive style), a perspective-taking task, and the Cognitive Flexibility Scale. Analysis of accuracy and reaction time revealed that (i) mental rotation in spatial-visualizers was more accurate and faster than in object-visualizers, and (ii) visual cognitive style and perspective-taking positively correlated with mental rotation. Brain activity data indicated that (i) individuals with better mental rotation performance had smaller brain activation, particularly in sensorimotor regions, (ii) for the spatial–visual scale and perspective-taking, high scorers had smaller brain activity than low performers, (iii) for the object–visual scale, high scorers had greater brain activity than low scorers. Supporting a neural efficiency hypothesis, the present study highlights the influence of cognitive traits on mental rotation performance and brain efficiency, with spatial-visualizers showing more efficient neural processing. These findings contribute to our understanding of how cognitive styles shape spatial cognition.

Introduction

One of the most fundamental human cognitive abilities is visuospatial cognition (Schaie 1996). By asking participants to identify an object rotated with respect to its canonical upright position, the so-called mental rotation (MR) is an established visuospatial task that can be used to assess visuospatial cognition (Cooper and Shepard 1973; Riecanský and Jagla 2008). Many factors influence MR, such as gender (Halpern 2013; Zapf et al. 2015), sex hormones (Courvoisier et al. 2013; Scheuringer and Pletzer 2017; Bernal and Paolieri 2022; Gurvich et al. 2023), and sensorimotor experience (Terlecki et al. 2008; Jansen and Pietsch 2010; Hertanti et al. 2019). The type of stimulus used to perform MR plays an important role in MR itself (Amorim et al. 2006; Tomasino and Gremese 2016), especially in regard to stimulus-based semantics (Zeugin et al. 2020), viewpoint (Giovaola et al. 2022), visibility (Rotach et al. 2024), and sensory modality (Pamplona et al. 2022). In particular, neural dissociations have been found between MR of objects versus hands (Kosslyn et al. 1998; Tomasino et al. 2003), MR of bodily versus nonbodily images (Tomasino et al. 2004; Aso et al. 2007), MR of hands versus bodily images (Perruchoud et al. 2016), and between MR of geometrical objects, human bodies, and real objects (Bersier et al. 2024).

Furthermore, beyond demographic variables like age, gender, education, sport practice, and sociocultural background (Ginn and Stiehl 1999; Sander et al. 2010; Habacha et al. 2014), the performance in MR tasks can be influenced by psychological factors, such as field dependence/independence (Naurzalina et al. 2015), anxiety (Borst et al. 2012; Kaltner and Jansen 2014), self-confidence (Cross et al. 2011; Arrighi and Hausmann 2022), and depression (Rogers et al. 2002; Chen et al. 2013; Chen et al. 2014), as well as by cognitive traits—ways to process information and regulate individual cognitive functioning (Klein 1951) and are able to influence attitudes, strategies, preferences, and ways of perceiving/learning information (Jonassen and Grabowski 1993).

A cognitive trait that plays an important role in MR is the visual component of the verbal/visual cognitive style. The visual component refers to a preference for processing information through visual means and can be behaviorally and neurally differentiated from the verbal component (Kraemer et al. 2009). People with a visual cognitive style tend to engage in visual mental imagery even in presence of words (Bendall et al. 2016) and to activate brain regions related to visual processing (Kraemer et al. 2009). The visual cognitive style can be further divided into object- and spatial-visualization cognitive styles, for instance based on the scores in the Object–Spatial Imagery Questionnaire (OSIQ; Blajenkova et al. 2006). The OSIQ includes scales for object and spatial visualization strategies (Blajenkova et al. 2006), and has shown strong correlations between spatial visualization and MR ability (Campos and Campos-Juanatey 2014). The cognitive styles of spatial- and object-visualizers are largely uncorrelated and associated with different abilities and preferences (Haciomeroglu and LaVenia 2017). Object-visualizers excel in picture recognition, while spatial-visualizers perform better in mental spatial transformation (Kozhevnikov et al. 2013). Neuroimaging evidence supports the distinction between object- and spatial-visualizers, in that different brain activity patterns are associated with each of them (Motes et al. 2008). These 2 subclasses of the visual cognitive style are associated with measurable differences in MR tasks. In fact, spatial-visualizers are more accurate and faster than object-visualizers (Vandenberg and Kuse 1978), spatial-visualizers excel in MR while object-visualizers perform better in picture recognition (Chabris et al. 2006), and impaired visualization abilities are associated with longer reaction times in MR (Kay et al. 2024), and with a preference for spatial- than object-visualization (Pounder et al. 2018).

Another cognitive trait closely related to MR is perspective-taking (Galinsky et al. 2008). As a visuospatial ability, perspective-taking refers to the ability to reorient oneself and take the perspective of another person or object (Hegarty and Waller 2004). As a cognitive trait, perspective-taking refers to the understanding of another person’s psychological viewpoint (Davis et al. 1996; Todd and Galinsky 2014). Previous work showed the tight relationship between perspective-taking as a visuospatial ability and as a cognitive trait. For instance, the performance on visuospatial perspective-taking tasks correlates with empathic perspective-taking abilities (Erle and Topolinski 2015), the relationship between visuospatial and cognitive perspective-taking is further modulated when participants mentally adjust their spatial orientation (Gronholm et al. 2012), visuospatial perspective-taking can enhance others’ emotion recognition (Erle and Funk 2022), and the adoption of another person’s visuospatial viewpoint can lead to greater alignment with her/his psychological viewpoint (Erle and Topolinski 2017). In sum, these results support that perspective-taking as a cognitive trait is grounded in the ability to use visuospatial perspective-taking, which, in turn, has a complex relationship with MR. While MR and visuospatial perspective-taking are highly correlated (Surtees et al. 2013; Geva and Henik 2019), visuospatial perspective-taking tasks are generally more challenging than MR tasks (Geva and Henik 2019). In addition, both abilities contribute to spatial navigation performance (Kozhevnikov et al. 2006), but only visuospatial perspective-taking predicts variance in tasks requiring to update self-to-object representations (Kozhevnikov et al. 2006).

Finally, it has been suggested that MR may involve a certain degree of cognitive flexibility (Gardony et al. 2017). A common definition of cognitive flexibility concerns the individual awareness that a given situation can be solved by different options, willingness to be flexible, and self-efficacy in being flexible (Martin and Rubin 1995). The Cognitive Flexibility Scale (CFS) is a largely validated tool (Martin and Anderson 1998; Bilgin 2009; Oshiro et al. 2016) to measure such an individual self-perceived ability to consider and select between the many options possible to address a given situation (Martin and Rubin 1995). Unlike other instruments to measure cognitive flexibility that emphasize executive functions such as switching between different sets or tasks (eg Ionescu 2012), CFS focuses on self-perceived flexibility in solving a given problem by generating many options. This means that people with high scores in CFS may generate and test several options to solve an MR task, leading to better or faster overall performance with respect to people with low scores in CFS. However, beyond different possible definitions, the potential role of cognitive flexibility in MR-related performance and brain activity remains largely unexplored. To fill this gap, we examined the potential association between CFS-measured cognitive flexibility, behavioral, and neural aspects of MR. This approach was based on the reasoning that, although there is no straightforward evidence that MR and cognitive flexibility are correlated, some insights about their possible relationship exist. In particular, it has been shown that (i) the same experimental manipulation (extended sleep) affects performance in spatial rotation ability (which involves MR) and cognitive flexibility (Clark et al. 2022), (ii) virtual reality training improves both MR and cognitive flexibility (Dehn et al. 2020), (iii) the status of some regions of the cerebellum are associated with both MR and cognitive flexibility (Koppelmans et al. 2017), (iv) playing video games can enhance MR and task switching, which is closely related to cognitive flexibility (Cardoso-Leite and Bavelier 2014), and (v) performance in cognitive flexibility and MR tests is affected by epileptic seizures (Quiske et al. 2006).

Altogether, the behavioral evidence that MR is influenced by visual cognitive style and perspective-taking, and may be related to cognitive flexibility, hints to the possibility that also the brain activity associated with MR may be influenced by these cognitive traits. Furthermore, the different brain activations related to MR of different stimuli suggests that the influence of cognitive traits on neural and/or behavioral aspects of MR may be further sensitive to the type of MR stimulus. Investigating these complex interactions could provide deeper insights into how individual differences in cognitive processing influence the neural dynamics associated with MR in eventual stimulus-specific ways. While this topic may be beneficial for (i) advancing theoretical knowledge about cognitive, neural, and behavioral aspects of MR, and (ii) developing novel approaches for rehabilitation and training programs, it has not been fully investigated yet. To fill this gap, we used an event-related design during functional magnetic resonance imaging (fMRI) to investigate the relationship between MR performance with different stimuli, cognitive traits, and brain activity. We predicted that (i) MR-related brain activity would be different between high and low scorers in cognitive traits, and that (ii) these relationships would be further modulated by the type of MR stimuli.

Materials and methods

Participants

Sixty participants were recruited via local online platforms and social media. Five participants were excluded from the final sample (3 due to poor performance in MR and 2 due to bad performance in the perspective-taking task). The final sample consisted of 55 participants (27 women, age 24.33 ± 4.15 yrs, range 19 to 35; 28 men, age 25.29 ± 4.25 yrs, range 20 to 35). Prior to the experiment, participants gave their written consent and completed a safety questionnaire for the fMRI environment. Participants then completed the OSIQ (Blajenkova et al. 2006) to measure visual cognitive style and performed a perspective-taking task (as described in “Visual cognitive style”). Additionally, participants’ cognitive flexibility was assessed through the CFS (Martin and Rubin 1995). The following exclusion criteria were age outside of the predetermined range (18 to 35 yrs); uncorrected visual impairment; neurological or psychiatric disorder; incompatibility with fMRI settings; left-handedness, as defined by scores below the 60-point cut-off on the Edinburgh Handedness Inventory (Oldfield 1971). The experimental procedures were approved by the local ethics committee (CER-VD 2022-01625), and the experiment was conducted in accordance with the Declaration of Helsinki (2013).

Visual cognitive style

The OSIQ is a 30-item self-assessment tool created to evaluate how individuals differ in their visual imagery preferences and experiences (Blajenkova et al. 2006) The OSIQ includes 2 distinct The object subscale measures object-visualization, which gauges the preference for processing vivid images of objects. The spatial subscale measures spatial visualization, which measures the preference for understanding schematic spatial relationships. On the basis of the OSIQ score, each participant was classified as either “object-visualizer” or “spatial-visualizer,” with an additional score (high, low) within each scale object scale, spatial scale.

Perspective-taking task

To measure perspective-taking abilities, we used the Spatial-Orientation Test (Hegarty and Waller 2004, modified by Kozhevnikov and Hegarty 2001), which is able to assess the ability to mentally simulate spatial transformations of one’s own orientation and perspective. This test is a widely recognized tool for measuring spatial orientation abilities, particularly how individuals can position themselves within a given environment. The test consists of 20 items, each printed on a separate page. For each item, participants are shown an array of images at the top of the page, along with an “arrow circle” at the bottom, which includes a specific question regarding the spatial relationship between the images. Participants are instructed to imagine themselves standing at a particular image in the array, designated in the center of the arrow circle, and facing another image, indicated at the top of the circle. From this imagined viewpoint, participants must determine the direction toward an object and draw an arrow from the center of the circle to indicate this direction. To ensure participants understand the task, a practice trial was provided prior to the test. In this practice trial, participants were asked to imagine standing at a flower (center image, first) and facing a tree (top image, second). The correct direction toward a cat (object image, third) was already marked with an arrow, demonstrating the correct procedure. Participants were encouraged to ask questions if they did not fully understand the task. Throughout the test, participants were instructed not to pick up or turn the test booklet and to refrain from making any marks on the image arrays. The focus was on accurately determining directions while avoiding excessive time spent on every single trial.

Cognitive flexibility scale

The CFS (Martin and Rubin 1995) is a 12-item questionnaire that assesses cognitive flexibility defined as the individual self-awareness of being able to generate multiple possibilities to solve a problem (eg “I have many possible ways of behaving in any given situation”). Participants are instructed to read each statement and respond by selecting how much they agree or disagree with the statements, rating their answer from 1 to 6 as 6 = strongly agree, 5 = agree, 4 = slightly agree, 3 = slightly disagree, 2 = disagree, and 1 = strongly disagree. The average score among students has been reported to be around 55 points.

MR stimuli

Participants performed MR of 3 distinct images (geometric objects, human bodies, and real objects) during the recording of fMRI data (Fig. 1). Geometric objects comprised 6 pictures that resembled the ones used by Shepard and Metzler (1971). The 6 human body images showed a complete body in 6 different postures (curtesy of Amorim et al. 2006). The 6 real objects represented a motorbike break, a wrench, a stapler, a corkscrew, a spray bottle, and a can opener. Digital manipulation was used to ensure that all images had the same overall size, symmetry, and absence of writing.

Procedure

During fMRI data recording, participants performed an MR task (Lamm et al. 2007). Sixty image pairs were presented to the participants. The first image in each pair displayed an image (geometric object, human body, real object) rotated with respect to the vertical by 1 of 10 different rotation angles (±30°, ±60°, ±90°, ±120°, and ±150°). This first image was displayed for 1500 ms, followed by a jittered blank screen that lasted between 500 and 1000 ms. A circular arrow rotating in either a clockwise or counterclockwise direction served as the cue for 500 ms, after which a screen flashed the word “Go” for 3500 ms. During this time, participants were asked to mentally rotate the first image until the vertical position. They pressed a key to indicate when they finished the task reaction time (RT). Upon receiving the participant’s response, the “Go” screen flickered on and off for 500 to 1000 ms. Then, the second image of the pair was shown for 2500 ms, representing the same or different item of the first image, but oriented vertically. Participants had to press a key to indicate whether the second image matched the result of the MR they performed on the first image. In 50% of the cases, the answer was “same,” and in 50% of the cases the answer was “different.” To avoid memorization, the “different” trials were further divided into 2 “mirrored” (as in Fig. 2) or “wrong rotation.” The 3 types of images were presented in a randomized order. To familiarize themselves with the task, participants underwent a training session of 4 trials for each type of images outside the scanner. The images used in the training session were different with respect to those used in the experiment.

MRI acquisition

We used a Siemens Magnetom Prisma Fit 3 T scanner and a 64-channel head coil (Siemens, Erlangen, Germany) to acquire both functional and anatomical brain images. A T2^*^-weighted echo-planar imaging sequence employing an isotropic voxel size of about 2 mm^3^, echo time/repetition time (TE/TR) of 30/2000 ms, 64 slices, simultaneous multislice factor of 2, field of view (FoV) of 192 mm^2^, flip angle of 80°, matrix size of 96 × 96, interleaved ascending acquisition, and 2136 Hz bandwidth per pixel was applied to get the functional images. The FoV was aligned parallel with the commissural line and included the whole cerebrum. Five dummy scans were acquired to establish steady-state magnetizations prior to each functional acquisition. To account for spatial distortion in the functional images, we acquired a pair of spin echo images with opposite phase encoding directions, matching the orientation of the functional scans. These spin echo images were obtained at the beginning of each condition, for a total of 3 sequences. All the images of the MR task were administered with PsychoPy (psychophysics software in Python; Peirce et al. 2019), projected on a screen behind the fMRI scanner, and viewed by the participants through a mirror attached to the head coil. For each item of the MR task, participants’ RT and accuracy were automatically recorded. For each of the 6 runs of the MR task, we collected 171 functional image volumes (6 min, 20 s), including 6 additional volumes for safety purposes. To synchronize code and scanner, the functional acquisition initiated the presentation code. Anatomical images were acquired using a T1-weighted (T1w) magnetization prepared rapid gradient echo sequence (Marques et al. 2010), isotropic voxel size = 1 mm^3^, FoV = 256 mm^2^, TE/TRmprage = 2.9/5000 ms, 192 contiguous sagittal slices, inversion time (TI1)/flip angle = 700 ms/4°, TI2/flip angle = 2500 ms/5°, matrix size = 240 × 240, duration = 3 min and 30 s.

Data analysis

Behavioral data

For MR performance, items with RTs above 3 standard deviations from the group mean for each image category (geometric objects, human bodies, real objects) were excluded from the analysis (557 out of 9,900 trials). Statistical analyses were performed with RStudio (https://rstudio.com/). A full-factorial repeated measures analysis of variance (ANOVA) was performed to examine the main effects and interactions of stimulus type (geometric objects, human bodies, real objects) and cognitive traits (ie OSIQ object-/spatial-visualizer; OSIQ high/low; perspective-taking high/low; CFS high/low) on mean RT and accuracy. As object- and spatial-visualizers groups were not balanced in terms of size (N = 41 for object-visualizers and N = 14 spatial-visualizers) another division within the scales was made, resulting in “high/low score on the object scale” and “high/low score on the spatial scale.” The high or low classification for each assessment was determined by the participant’s position relative to the group median. In particular, the median score was 51 for the OSIQ-object scale, 48 for the OSIQ-spatial scale, 340 for the perspective-taking, and 54 CFS. For each scale, participants with scores above or below these values were classified as high or low scorers, respectively. This approach resulted in 27 high and 28 low scorers in the OSIQ-object scale, 29 high and 26 low scorers in the OSIQ-spatial scale, 28 high and 27 low scorers in the perspective-taking, and 27 high scorers and 28 low scorers in the CSF scale. The score for the perspective-taking task was computed by calculating the difference in degrees between what the subject drew and what the correct answer is for the item, either in the clockwise or counterclockwise direction. Therefore, the higher the score, the worse the performance. For clarity throughout the paper, this quote has been inverted to avoid confusion. Therefore, an original high score to the perspective-taking task has been classified as low performance and vice versa. Main effects and interactions were considered significant according to a level of significance of 0.05. If a significant effect was found, post hoc analyses were performance using t-tests. P-values were adjusted for multiple comparisons according to the Bonferroni method.

Brain imaging quality assessment process

Functional and anatomical brain images were converted from the Digital Imaging and Communications in Medicine format to the Brain Imaging Data Structure (https://bids.neuroimaging.io/) using the Dcm2Bids program (https://github.com/cbedetti/Dcm2Bids). To evaluate the quality of structural and functional brain images, the MRI Quality Check (MRIQC; Esteban et al. 2017) tool was employed. Additionally, a comparative analysis was conducted using MRIQCeption (https://github.com/elizabethbeard/mriqception), which involved comparing the acquired quality metrics with a reference dataset obtained from the MRIQC online Application Programming Interface (Esteban et al. 2019).

MRI data preprocessing

Preprocessing of the fMRI data was carried out using fMRIPrep version 1.5.1rc2 (Esteban et al. 2019), built on Nipype (Gorgolewski et al. 2011). The T1w images were first corrected for intensity nonuniformity using N4BiasFieldCorrection (Tustison et al. 2010), available with ANTs version 2.3.3 (Avants et al. 2008). These corrected images served as the T1w reference for subsequent steps. Skull-stripping was performed using antsBrainExtraction.sh v2.2.0 based on the OASIS template, followed by the reconstruction of brain surfaces through the recon-all command from FreeSurfer v6.0.1 (Dale et al. 1999). The brain mask was further refined by integrating ANTs and FreeSurfer segmentations of the cortical gray matter, utilizing a customized approach informed by Mindboggle (Klein et al. 2017).

For spatial normalization, the data were aligned to the International Consortium for Brain Mapping 152 Nonlinear Asymmetric template version 2009c (Fonov et al. 2009) via the antsRegistration tool of ANTs v2.2.0 (Avants et al. 2008), utilizing brain-extracted versions of both the T1w volume and the template. Segmentation of brain tissues, including cerebrospinal fluid, white matter, and gray matter, was performed using fast (FSL v5.0.9; Zhang et al. 2001). Functional data were corrected for slice timing with 3dTshift from AFNI v16.2.07 (Cox 1996) and for motion using mcflirt (FSL v5.0.9; Jenkinson et al. 2002). Distortion was addressed with a TOPUP-based approach (Andersson et al. 2003) implemented in 3dQwarp (AFNI v16.2.07; Cox 1996). Coregistration of functional data to the T1w images was performed using boundary-based registration (Greve and Fischl 2009), with 6 degrees of freedom, via the bbregister function (FreeSurfer v6.0.1). All transformation steps, including motion correction, field distortion adjustment, BOLD-to-T1w alignment, and T1w-to-template Montreal Neurological Institute (MNI) registration, were applied in 1 step using antsApplyTransforms (ANTs v2.2.0) with Lanczos interpolation.

To reduce physiological noise, a Component Based Noise Correction Method (CompCor; Behzadi et al. 2007) was employed. A subcortical mask was created by eroding the original brain mask, and 6 aCompCor components were derived from the intersection of this mask with the union of cerebrospinal fluid and white matter masks generated from T1w images. These components were projected onto the native space of each functional run. Additionally, frame-wise displacement and the derivative of RMS variance over voxels (DVARS) (Power et al. 2014) were computed for each run using Nipype.

Subsequent processing steps involved masking the functional data based on the brain mask provided by fMRIPrep. Fourteen confounding variables from fMRIPrep (including 6 motion parameters, frame-wise displacement, standardized DVARS, and 6 aCompCor components) were regressed out at the voxel level using the Denoiser tool (Tustison et al. 2010). The final step involved spatial smoothing of the functional data using a Gaussian kernel with a 6-mm full-width at half-maximum.

First-level GLM analysis

For the first-level general linear model (GLM) analysis, the FSL FEAT software was used (www.fmrib.ox.ac.uk/fsl). A separate GLM model was constructed for each participant and each run, where the 3 images (geometric objects, human bodies, real objects) served as the regressors of interest, and their temporal derivatives were included as regressors of no interest. The regressors were convolved with a double-gamma hemodynamic response function and timed with the beginning and end of the video stimulus. We used FMRIB’s Improved Linear Model prewhitening to adjust for autocorrelation, and a high-pass filter with a 100-s cutoff was used to remove low-frequency drifts.

Univariate analysis

The purpose of the univariate analysis was to identify brain regions that showed significant activation during MR tasks, particularly in relation to cognitive traits and performance levels (high, low). By isolating the 4-s-long period following the cue for MR, we aimed to focus on the neural processes directly associated with mental transformation. This analysis allowed us to explore how individual differences and stimulus type modulate neural activity, contributing to our understanding of neural efficiency and compensatory mechanisms in MR.

Using 2-sample unpaired t-tests, 6 separate contrasts were created to examine the distinctive neural activations corresponding to each condition, with group added as a regressor to determine whether any observed effects were influenced by the OSIQ style (object- or spatial-visualizer), the accuracy of MR performance, or by obtaining a “high” or “low” in the concerned cognitive trait (OSIQ subscales, perspective-taking task, CFS). To investigate potential links with accuracy, this covariate was added in a separate model. Accuracy was mean centered by subtracting the overall mean accuracy from each individual score and added as a third explanatory variable in the model, for both positive and negative effect.

Results

The ANOVA on MR accuracy revealed no significant main effects nor interactions related to CFS on either behavioral (accuracy, RTs) or neural (brain activity) responses to MR of any type of stimulus.

Mental rotation

During MR of human bodies, with respect to participants with high MR accuracy, participants with low accuracy in MR had stronger activity in 4 clusters located in the left hemisphere and comprising, respectively, (i) the middle occipital lobe and angular gyrus, (ii) the middle frontal gyrus, (iii) the precentral gyrus and opercular part of the inferior frontal lobe, and (iv) the superior frontal gyrus and supplementary motor area (yellow clusters in Fig. 2A and B). During MR of geometric objects, with respect to participants with high MR scores, those with low MR scores showed stronger activity in a cluster covering the left parietal and occipital lobe, angular gyrus (blue cluster in Fig. 2A and B). The 2 contrasts overlapped in the angular gyrus (green cluster in Fig. 2A and B). All the details of the clusters resulting from these contrasts are reported in Table 1. The ANOVA on MR accuracy revealed the significant main effect of stimulus [F(2,157) = 11.38, P < 0.001]. Post hoc t-tests Bonferroni-corrected for multiple comparisons showed a significant difference between the geometric objects and human bodies [t(1, 107) = 2.8, P = 0.005] and between the geometric objects and real object images [t(1, 107) = 4.65, P < 0.001] (Fig. 2C). The ANOVA on MR RTs revealed the significant main effects of stimulus [F(2,157) = 50.10, P < 0.001]. Post hoc t-test Bonferroni-corrected for multiple comparisons revealed that RT for MR of human bodies were significantly longer compared to MR of both geometric objects [t(1, 99) = 7.03, P < 0.001] and real objects [t(1, 99) = 8.77, P < 0.001] (Fig. 2D).

Visual cognitive style

The brain activity associated with MR of geometric objects and human bodies was different between spatial- and object-visualizers, as classified by the OSIQ. During MR of geometric objects, object-visualizers showed stronger brain activity in a cluster comprising the primary motor cortex and several somatosensory regions (orange clusters in Fig. 3A and B and Table 1). In addition, during MR of geometric objects, the object-visualizers had stronger brain activity in (i) the left superior frontal gyrus and the left anterior and middle cingulum, (ii) the middle cingulum bilaterally, and (iii) the left postcentral, precentral, and supramarginal gyri (purple clusters in Fig. 3A and B). All the details of the clusters resulting from these contrasts are reported in Table 1. The ANOVA on MR accuracy revealed the significant main effect of visual cognitive style [F(1,157) = 9.81, P = 0.002], with the spatial-visualizers being more accurate than object-visualizers (Fig. 3C). The ANOVA on MR RTs revealed the significant main effect of OSIQ style [F(1,157) = 7.3, P = 0.007], with object-visualizers being slower than spatial-visualizers (Fig. 3D). During MR of real objects, the comparison between object- and spatial-visualizers did not result in significantly different brain activations.

Object and spatial subscales of OSIQ

The comparison of high and low scores separately for object- and spatial-visualizer subscales of OSIQ showed 2 main significant differences. First, with respect to low-scoring object-visualizers, high-scoring object-visualizers showed stronger activity in a cluster comprising the left superior and inferior parietal lobes and angular gyrus during MR of geometric objects compared to real objects (red clusters in Fig. 3E and F). Second, during MR of geometric objects compared to human bodies, with respect to high-scoring spatial-visualizers, low-scoring spatial-visualizers had stronger activity in a cluster covering the right angular gyrus, middle temporal lobe, and inferior occipital lobe (blue clusters in Fig. 3E and F and Table 1). All the details of the clusters resulting from these contrasts are reported in Table 1. The ANOVA on MR accuracy revealed the significant main effects of OSIQ score (high, low) in the object scale [F(1,157) = 4.03, P = 0.04] and of OSIQ score (high, low) in the spatial scale [F(1,157) = 4.75, P = 0.03]. Object-visualizers with high scores in the object scale of OSIQ were generally less accurate in MR task compared to object-visualizers with low scores in the object scale of OSIQ (Fig. 3G). Spatial-visualizers with high scores in the spatial scale were more accurate in the MR task compared to spatial-visualizers with low scores in the spatial scale (Fig. 3H).

Perspective-taking

During MR of human bodies with respect to geometric objects, low-scoring participants in perspective-taking had significantly stronger activity in 2 clusters, compared to participants with high scores in perspective-taking. One cluster comprised the left the postcentral and precentral gyri, as well as the inferior parietal lobe. The second cluster included the supplementary motor area, the paracentral lobule, and the right middle cingulum (Fig. 3I and J and Table 1). All the details of the clusters resulting from this contrast are reported in Table 1. The ANOVA on MR accuracy revealed the significant main effects of score at the perspective-taking task [F(1,157) = 4.52, P = 0.03], according to which participants that scored higher in the perspective-taking task had higher/lower accuracy with respect to those who scored lower (Fig. 3K).

Discussion

The present study investigated whether cognitive traits have a stimulus-specific influence on the neural and behavioral correlates of MR. To this aim, we analyzed the association between brain activity associated with MR of different stimuli and cognitive traits. The results showed that the brain activity associated with MR was stimulus-specifically influenced by visual cognitive style and perspective-taking, as well as by the performance level in the MR task.

Visual cognitive style

Our results showed that MR-related bran activity and performance were differentially influenced by object versus spatial-visual cognitive style. Our finding that participants identified as spatial-visualizers were faster and more accurate than object-visualizers is consistent with previous evidence that spatial-visualizers are better equipped to handle the spatial manipulations required in MR tasks (Kozhevnikov et al. 2005; Blajenkova et al. 2006). This finding was corroborated by neuroimaging results, in that the stronger activity in sensorimotor and frontal regions shown by object-visualizers during MR of geometric objects, compared to spatial-visualizers, was associated with lower overall accuracy. The higher brain activity could reflect a compensatory mechanism in which object-visualizers use additional neural resources to complete the task, such as additional top-down control (Barbey et al. 2011) or the use of a motor strategy (Tomasino and Gremese 2016). This interpretation is in line with previous evidence that object- and spatial-visualizer may differentially recruit neural resources to accomplish the same tasks (Motes et al. 2008).

During MR of human bodies, object-visualizers showed a larger activation of the somatosensory cortex compared to spatial-visualizers, suggesting that they may rely more on body-based processes to solve MR. This finding is congruent with previous evidence that larger sensorimotor activity during MR can be associated with lower accuracy (Bersier et al. 2024). Thus, the poorer accuracy of object-visualizers may be due to the involvement of sensorimotor regions. It is indeed well known that these regions are crucial for generating a kinesthetic image of the body in space and simulating body movements during MR tasks (Parsons 1994; Maravita and Iriki 2004; de Lange et al. 2006; Perruchoud et al. 2016). On this basis, we propose that object-visualizers’ preference for vivid, concrete, manipulable images, may result in them recruiting more concrete sensorimotor processes, which would not be as precise as visual ones. Thus, the activation of sensorimotor regions may interfere with the use of a purely spatial process, resulting in lower performance.

The lack of significant reverse contrast suggests that spatial-visualizers showed more efficient neural processing, with less activation in the same regions, especially during the (more complex) MR of human bodies and geometric objects images. This finding is in line with previous evidence that the preference for object-visualization style is linked to a more efficient utilization of cognitive resources, leading to reduced neural activity to perform object-related processing (Motes et al. 2008). These results suggest that spatial-visualizers may have more optimized neural pathways for processing MR tasks, which may explain their superior performance at the behavioral level.

The influence of high and low scoring in visual cognitive style.

Further results emerge when considering participants based on the basis of their individual scores on the object and spatial subscales of OSIQ. Participants with high scores on the object scale showed significantly stronger activity in the angular gyrus in the MR of geometric objects with respect to human bodies. The angular gyrus has been implicated in a variety of cognitive processes, including spatial reasoning (for a review, see Seghier 2013) and MR. For instance, a damage to the angular gyrus is associated with impaired MR of letters (Doganci et al. 2024), the length of mental spatial transformations during MR of hands and chairs is associated with activity in the angular gyrus (Menéndez Granda et al. 2022), and the involvement of the angular gyrus in MR of real objects has been shown both in childhood and adolescence (Hirai et al. 2022). Furthermore, the angular gyrus is involved in detecting discrepancies between predicted and actual consequences of actions across different sensory modalities (van Kemenade et al. 2017), playing a critical role in integrating different sensory inputs for further cognitive processes (Bonner et al. 2013; Thakral et al. 2017; Tibon et al. 2019). Altogether, we propose that the stronger activity of the angular gyrus in high-scoring object-visualizers, with respect to low-scoring object-visualizers, may reflect a relatively larger effort to activate sensory simulation strategies to perform MR of geometric objects than human bodies. Considering that a reasonable affordance of geometric objects could be that they are perceived as more “manipulable” than human bodies, we argue that participants with a higher tendency to focus on objects’ features (high-scoring object-visualizers) would need stronger mental efforts to perform the mental spatial transformations required by MR of items that, additionally, induce the mental simulation of the sensory inputs that would be generated by their physical manipulation. This mental effort would be reflected in higher brain activity of regions processing sensory inputs for further cognitive processing, such as the angular gyrus.

People who scored low on the spatial subscale also showed an additional activation in the right middle temporal gyrus when the geometric objects was contrasted with the human bodies condition. The middle temporal gyrus and inferior temporal gyrus are involved in memory processing (Tranel et al. 1997; Chao et al. 1999; Cabeza and Nyberg 2000), visual perception (Ishai et al. 1999; Herath et al. 2001), and multimodal sensory integration (Mesulam 1998). The fact that people who scored lower on the spatial subscale also showed additional brain activation in these regions could mean that they rely more heavily on memory processes, visual perception, and multimodal sensory integration to compensate for their less efficient spatial processing abilities. This compensatory activation suggests that these individuals may be attempting to use alternative cognitive strategies, such as activating stored visual memories or integrating sensory information from multiple sources, to perform MR. However, this reliance on additional neural resources may not be as effective as the more streamlined and spatially focused strategies used by those with higher spatial abilities, leading to their overall lower performance.

Perspective-taking

With respect to participants with lower scores in perspective-taking, those with higher perspective-taking scores showed better accuracy on the MR task. This supports that the ability to mentally adopt another perspective and manipulate spatial relationships is closely related to MR performance. In addition, the fMRI results showed that individuals with lower scores in perspective-taking showed increased neural activity in motor brain regions, with respect to participants with high scores in perspective-taking. These motor activations may reflect low-scoring participants’ greater effort to perform MR of items that are easier to embody, such as human bodies, with respect to geometric objects. The recruitment of motor areas in participants with lower perspective-taking ability suggests that motor strategies may be used to achieve similar outcomes when cognitive reorientation is insufficient.

Cognitive flexibility

While some insights may suggest interactions between MR and cognitive flexibility (Quiske et al. 2006; Cardoso-Leite and Bavelier 2014; Koppelmans et al. 2017; Clark et al. 2022), our findings indicate no significant interaction between the CFS-measured cognitive flexibility and MR-related performance and brain activity. The absence of significant interactions between CFS scores and MR-related performance and brain activation suggests that the specific aspects of cognitive flexibility measures by CFS do not relate to MR. While MR requires spatial transformation and mental manipulation of objects, the flexibility assessed by the CFS may be more related to how individuals approach problem-solving, in contrast to other definitions of cognitive flexibility more focused on executive-level cognitive control mechanisms like task-switching or working memory updating (Ionescu 2012). Thus, while our finding indicates that the type of cognitive flexibility captured by the CFS (awareness that problems can be solved in many ways and self-perceived efficacy in flexibility) may not directly relate to MR; nevertheless, this does not mean that other aspects of cognitive flexibility, possible measured through other tools, cannot relate to MR.

The role of the left angular gyrus

The left angular gyrus was associated with lower performance in several of our contrasts and may be indicative of several underlying neural mechanisms. This region is known to play a critical role in spatial cognition, attention, and the integration of sensory information, making it essential for tasks requiring complex spatial processing (for a review, see Seghier 2013). Increased activation in the left angular gyrus among lower performers in terms of MR accuracy may reflect a compensatory effort to cope with the demands of the MR tasks. This increased activity may indicate less efficient processing, with the brain struggling to perform the necessary spatial transformations, resulting in greater cognitive load. In addition, the angular gyrus is involved in working memory and the allocation of attention, suggesting that individuals with lower performance in terms of accuracy may have difficulty maintaining and manipulating spatial information over time (Corbetta and Shulman 2002; Wager and Smith 2003). The overactivation observed in this region could also indicate a processing bottleneck, where the brain’s attempts to integrate visual and spatial inputs become less effective, further contributing to the observed deficits in task performance (Hubbard et al. 2005). Taken together, our findings suggest that the angular gyrus, while critical for spatial tasks, may become a site of inefficiency when overactivated, particularly in individuals with lower spatial abilities.

Neural efficiency

A final general observation is that, in our study, greater brain activation was always associated with less efficient performance in MR. In line with the neural efficiency hypothesis, this finding suggests that individuals who perform better on cognitive tasks tend to show less brain activation, particularly in regions associated with task-relevant processing (Haier et al. 1992). This implies that their brains are working more efficiently, requiring less effort to achieve the same or better performance with respect to those who performs less well and has higher brain activation. Several studies have used fMRI to investigate differences in brain activity between experts and novices, with experts showing lower or more localized brain activation during task performance compared to novices. For example, radiologists show reduced activation in visual and nonvisual brain regions during diagnostic tasks (Ouellette et al. 2020), while elite archers show more localized neural activity in the dorsal pathway and cerebellum while aiming (Kim et al. 2014). In this sense, a parallel could be drawn between spatial-visualizers and spatial expertise to explain our imaging results. Our results suggest that similar to experts in other domains, spatial-visualizers may possess more efficient neural networks that allow them to perform MR tasks with greater efficiency, thereby requiring less widespread brain activation. This reduced activation may indicate that their cognitive processes are more streamlined, allowing them to process spatial information with minimal effort. Conversely, object-visualizers and those with lower MR performance may need to recruit additional neural resources, reflecting a less efficient cognitive strategy that requires more effort and results in increased brain activation. Thus, our results not only confirm the neural efficiency hypothesis in the context of MR, but they also extend this hypothesis by highlighting the role of cognitive traits in determining the efficiency of brain function during visuospatial tasks.

These findings have important implications for understanding the neural and cognitive mechanisms underlying MR performance. The differential brain activation patterns observed between spatial- and object-visualizers provide evidence for the existence of distinct cognitive and neural strategies in spatial tasks. This could inform future interventions aimed at improving MR performance, particularly in individuals who struggle with spatial visualization.

Limitations

First, a possible limitation is that 2 cognitive traits (visual cognitive style and cognitive flexibility) were evaluated on the basis of self-reports (OSIQ, CFS). While these tools are widely used and validated, they are subject to biases related to subjective perception, including over- or under-estimation of abilities due to factors like self-awareness, social desirability, and item interpretation. This reliance on self-reports can lead to discrepancies between reported cognitive traits and the actual strategies or behaviors used during tasks. Future research may consider the use of objective measures, such as behavioral task, to complement self-report data and provide a more accurate and comprehensive assessment.

Second, the measurement of cognitive flexibility relied on the CFS, which captures the individual awareness that many options are available to address a situation and the self-perceived efficacy in flexibility. Although our results suggest that the type of cognitive flexibility assessed by the CFS (which includes the awareness that problems can be approached in multiple ways and a self-perceived ability to be flexible) may not have a direct connection to MR, this does not imply that other facets of cognitive flexibility, potentially evaluated by different instruments, cannot be related to MR. Future studies aiming to explore the relationship between MR and cognitive flexibility may consider employing additional measures of cognitive flexibility (Liu et al. 2016; Schmitz and Krämer 2023; Parekh 2024).

Third, beyond cognitive traits and stimulus type, other factors may influence the neural and/or behavioral dynamics of MR. These factors may include, for instance, gender, age, field dependency/independency, or depression. While these factors could have an important role, their inclusion in the experimental design of the present study would have added other levels of complexity to an already quite intricate experimental approach. Aiming to find the best compromise between investigating the role of cognitive traits in MR and keeping the design as simple as possible, we did not examine other factors. Nevertheless, this could be the focus of future studies.

Finally, it might be argued that the fact that participants were asked to rotate an image to its upright position may have compelled them to use an allocentric strategy, even if they might have preferred an egocentric or other strategies, leading to greater neural effort. While we do not exclude that the task could be achieved through an allocentric strategy, previous research has shown that the same task can be performed in both allocentric and egocentric perspective. For instance, Tomasino and Rumiati (2004) used this task to investigate differences between allocentric and egocentric strategies in MR, and Zacks and colleagues used this task to study first-person perspective (1999), third-person perspective (2000), and mental spatial transformations of objects and bodies (2001, 2002). This previous evidence supports that the task does not necessarily require participants to use an allocentric perspective, further depending on individual preferences or stimulus properties.

Conclusions

The present study investigated the relationship between cognitive traits such as object–spatial visual style, perspective-taking, and cognitive flexibility and MR-related brain activity and performance. Our findings show that participants with a spatial visualization style outperformed those with an object-visualization style. This performance difference was further corroborated by neuroimaging data, in that object-visualizers exhibited activations possibly related to mental effort, particularly when dealing with complex stimuli like geometric objects and bodies. These findings suggest that object-visualizers may rely more on compensatory neural mechanisms, possibly due to their preference for concrete, manipulable images, which may interfere with pure spatial processing and reduce overall efficiency of the MR task. Lower perspective-taking scores were associated with lower accuracy in MR and higher activity in motor regions, suggesting the employment of motor-based simulation strategies to mentally rotate items with a stronger manipulability affordance. Cognitive flexibility did not significantly influence MR-related performance nor brain activity. This finding suggests that only some specific cognitive traits affect MR, while cognitive flexibility may play a role in other cognitive domains.

At the neural level, the left angular gyrus was consistently associated with poorer performance across several analyses, suggesting that increased activation in this region may reflect compensatory efforts to manage the demands of MR tasks. This increased activity could indicate less efficient processing, as the brain struggles to integrate visual and spatial inputs, leading to greater cognitive load and reduced task performance.

Overall, our findings highlight the importance of some specific cognitive traits in shaping spatial cognition and support the neural efficiency hypothesis in the context of MR. These findings may inform future interventions aimed at improving MR performance, particularly for individuals who struggle with spatial visualization.