Seventy-nine (79) participants were included in this study, seventy-one (71) of whom pertained to a cohort described in a previous study (Villar-Rodríguez et al. 2023). All participants were left-handed (n = 68) or mixed-handed (n = 11) according to the Edinburgh Handedness Inventory/EHI (Oldfield 1971). EHI was scored by computing the Laterality Quotient or LQ, according to the formula (R − L/R + L)*100 (LQ lower or equal to − 50 was considered left-handed, and LQ between − 50 and + 50 was considered mixed-handed) (Szaflarski et al. 2002). Note that an alternative EHI scoring according to the method proposed by Bryden (1977) is available at the published database (see Data Availability statement). The hemispheric language lateralization of participants was calculated following the completion of an fMRI verb generation task based on which the participants were categorized as left-lateralized (n = 41), right-lateralized (n = 17), or ambilateral (n = 21) for language. We used this task for two reasons: (1) we have demonstrated its potential for determining language lateralization in the inferior frontal gyri in both presurgical patients and healthy participants (Sanjuán et al. 2010); and (2) for consistency with our previous studies on language lateralization among left-handers (Villar‐Rodríguez et al. 2020; Villar-Rodríguez et al. 2023), as well as future studies involving presurgical patients. There were no statistically significant differences between the groups in terms of age (F = 1.60; P = 0.21), sex (χ2 = 4.88; P = 0.78), or EHI-LQ (F = 1.04, P = 0.36). A slightly significant difference (F = 3.13, P = 0.049) was detected in fluid intelligence (measured via WAIS-IV matrix reasoning subtest; Wechsler 2012) between the left-lateralized and right-lateralized groups (Bonferroni’s pair-wise P = 0.053). Descriptive statistics of the lateralization groups can be found in Table 1.

Participants were recruited following a screening fMRI session in which language lateralization was roughly assessed using real-time data from the verb generation task (BrainWave software, GE HealthCare Technologies Inc.). This screening procedure was responsible for the high proportion of right-lateralized and ambilaterals (see the subsection ‘Cohort acquisition’ for more details). None of the participants reported any history of head injury resulting in the loss of consciousness, or psychiatric or neurological disorders. All participants signed an Informed Consent Form prior to participating in the study, following a protocol approved by the Universitat Jaume I Ethics Committee. All methods and procedures were carried out in accordance with the approved guidelines and current regulations.

Participation in the current study was advertised on university announcement boards and in the local media. Every healthy non-right-handed person over 16 years old was invited to participate in an fMRI session (S1). In this S1 session, general demographic variables were acquired, including in-house batteries that evaluate bilingualism history and musical experience. Standardized tests were also administered for: bilingual switching (BSWQ), sensitivity to punishment and reward (SPSRQ), and musical reward (BMRQ). Music-related variables have been the focus of a different study not yet published (note that the sample used in Villar-Rodríguez et al. [2020] is an entirely different cohort). Bilingualism-related variables are relevant for a different study we are still working on. Sensitivity to punishment/reward was acquired to complement a different line of research focused on personality traits. An audiometry test was also performed to ensure normal hearing.

During the fMRI session of S1, participants completed: (1) 3D and DTI structural acquisitions; (2) resting-state (eyes closed); (3) verb generation task (Spanish), (4) comprehension task (Spanish); (5) verb generation task (Valencian/Catalan, completed only by bilinguals); and (6) word listening task (Spanish).

We used real-time scanner software to roughly determine if the participant was typically or atypically lateralized during the verb generation task. If the participant was considered potentially atypical (crearly right-lateralized or probably ambilateral), they were invited to participate in a different fMRI session (S2) to take place in the future. Potentially typical participants were also invited to participate in S2, until matching the amount of potentially atypicals who agreed to participate in S2. We prioritized inviting potentially typical participants whose age and sex roughly matched those of the potentially atypical participants.

In total, 174 participants completed S1. However, due to time constraints and/or technical reasons, not all participants were able to complete all fMRI sequences (all 174 completed at least the verb generation task). 90 participants (45 potentially atypical, 25.86% of all participants, which is in line with published incidences of atypical lateralization among left-handers) were invited to participate in S2.

In S2, standardized tests were administered for: general intelligence (WAIS-IV matrix reasoning subtest), schizotypy traits (SPQ), autistic spectrum traits (AQ), and dyslexic traits (PROLEC-SE-R word reading subtest). Additional behavioral tests involving auditory and language processing were also administered (spoonerisms and second phoneme detection). Data regarding schizotypy traits, autistic spectrum traits, and dyslexic traits have been explored in a previous publication (Villar-Rodríguez et al. 2023). Auditory and language processing tests are currently being analyzed as part of a different study. General intelligence has been used as a control variable in all studies deriving from this cohort that have explored cognitive performance in any way (such as the accuracy during the landmark task in the current study, or the SSRT during the stop-signal task in Villar-Rodríguez et al. 2023).

During the fMRI session of S2, participants completed: (1) 3D and FLAIR structural acquisitions; (2) stop-signal task (Villar-Rodríguez et al. 2023); (3) landmark task (current study), and (4) reading task.

In total, 90 participants completed S2. However, due to time constraints and/or technical reasons, not all participants were able to complete all fMRI sequences (hence the sample differences between this study and Villar-Rodríguez et al. 2023). Also, in both this study and Villar-Rodríguez et al. (2023), different participants had to be excluded due to low engagement during the relevant tasks.

Expressive language function was measured by way of fMRI verb generation task (Sanjuán et al. 2010) that consisted of a block design paradigm with activation and control conditions. During the activation condition, participants were presented with a series of nouns and were requested to say the first verb that came to mind when seeing each word. During the control condition, participants were asked to read aloud visually presented pairs of letters. The task was administered using E-prime 2.0 (https://pstnet.com/products/e-prime) and included 6 activation and 6 control blocks. Each block lasted 30 s with each stimulus duration of 1500 ms and with a blank inter-stimulus interval of 1500 ms. Prior to performing the task in the scanner, each participant received detailed instructions on performing the task and completed a practice trial that lasted 2 min. Stimuli were presented using MRI-compatible googles (VisuaStim Digital, Resonance Technology Inc.) and verbal responses were recorded with a noise-cancelling microphone (FOMRI III + , Optoacoustics Ltd.) to ensure task compliance.

Visuospatial processing was examined by fMRI landmark task (Ciçek et al. 2009). During this task, participants were presented with a series of horizontal lines pre-bisected with a short vertical line and were required to respond by pressing the index button on the left response grip if the line was bisected correctly (task condition) and the thumb button if not. In this condition, the lines were bisected correctly in 40% of the trials or deviated to the right or left of the midline by 2.5% (hardest difficulty), 5% (medium difficulty) or 7.5% (easiest difficulty) of the line’s length, each deviation presented in 10% of the trials (see Fig. 1). During the control condition, the participants were required to respond whether the presented horizontal line and the bisection mark touched (index button) or not (thumb button). In this condition, the lines were touching in 40% of the trials, and not touching in 60% of the trials. These task parameters are identical to those described in Ciçek et al. (2009), which were also used by Cai et al. (2013) and Gerrits et al. (2020b), differing only slightly from Badzakova-Trajkov et al. (2010) and Zago et al. (2016). The task was administered using E-prime 2.0 (https://pstnet.com/products/e-prime) and included 7 activation and 7 control blocks. Each block lasted 22 s and started with a 4-s instruction, followed by a 215 ms blank inter stimulus interval and a 1.6-s presentation of a total of 12 line images. To avoid the use of the screen center as reference during the activation trials, and to ensure that participants correctly engaged in visuospatial processing, line images were not centered on the screen but slightly tilted to the left or right, alternating between trials. Prior to performing the task in the scanner, each participant received detailed instructions and completed a practice trial that consisted of 1 activation and 1 control block. Stimuli were presented using MRI-compatible goggles (VisuaStim Digital, Resonance Technology Inc.), and goggles-adapted corrective lens were available to ensure perfect visual perception for all participants. During the task, data on accuracy (% of correct responses) across all types of trials was recorded to measure task performance. Responses were registered with an MRI-compatible response grip (ResponseGrips, NordicNeuroLab). All participants responded using their left hand.

Showcase of the different trial types presented during the ‘activation’ condition of the landmark task. Participants had 1.815 s to respond whether the presented line was perfectly bisected or not

Images were acquired on a 3 T General Electric Signa Architect magnetic resonance imaging (MRI) scanner using a 32-channel head coil. All slices were acquired in the sagittal plane. A 3D structural MRI was acquired for each subject using a T1-weighted magnetization-prepared rapid gradient-echo sequence (TR/TE = 8.5/3.3 ms; flip angle = 12; matrix = 512 × 512 × 384; voxel size = 0.47 × 0.47 × 0.5). For the functional images, a gradient-echo T2*-weighted echo-planar imaging sequence was used to acquire 150 functional volumes for the verb generation task (TR/TE = 2500/30 ms; flip angle = 70; matrix = 64 × 64 × 30; voxel size = 3.75 × 3.75 × 4). A different gradient-echo T2*-weighted echo-planar imaging sequence was used to acquire 185 functional volumes for the landmark task (TR/TE = 2000/30 ms; flip angle = 70; matrix = 64 × 64 × 27; voxel size = 3.75 × 3.75 × 4.5).

The processing of the functional images was carried out using the Statistical Parametric Mapping software package (SPM12; Wellcome Trust Centre for Neuroimaging, London, UK) and MATLAB (version R2018b, MathWorks, Natick, MA). The default pipeline was followed during preprocessing steps that included: (a) aligning the functional data to the AC‐PC plane by using the anatomical image; (b) head motion correction, realigning and reslicing the functional images to the mean functional image; (c) coregistration of the anatomical image to the mean functional image; (d) re‐segmentation of the anatomical image; (e) spatial normalization of the functional images to the MNI (Montreal Neurological Institute, Montreal, Canada) space with a 3 mm3 resolution; followed by (f) spatial smoothing with a 4-mm full-width-at-half-maximum (FWHM) Gaussian kernel. The general linear models (GLM) for both the verb generation task and the landmark tasks were defined for each participant by contrasting activation > control blocks. For both tasks, the BOLD (Blood‐Oxygen‐Level‐Dependent) signal was estimated by convolving each task’s block/trial onsets with the canonical hemodynamic response function (HRF). Six motion realignment parameters were included as nuisance regressors, and a high‐pass filter (128 s) was applied to the contrast images to account for low-frequency drifts.

Functional lateralization for each task was assessed by obtaining the Laterality index (LI) using the bootstrap method implemented in the LI-toolbox for SPM12 (Wilke and Lidzba 2007). The LI is computed by calculating the proportion of activation differences between the two hemispheres for each individual subject. For the verb generation task, we explored the LI of the areas of the inferior frontal gyrus responsible for language production, specifically, pars opercularis and pars triangularis (Price 2012). For the landmark task, the LI calculation centered on the posterior areas involved in visuospatial attention during this task (Fink et al. 2001; Ciçek et al. 2009; Cavézian et al. 2012; Cai et al. 2013; Zago et al. 2016), specifically: supramarginal gyrus, angular gyrus, and the superior division of the lateral occipital cortex (Harvard–Oxford atlas). Masks were defined using the maximum probability Harvard–Oxford atlas (Frazier et al. 2005; Makris et al. 2006; Desikan et al. 2006; Goldstein et al. 2007), and were fitted to our functional images via the mask pre-processing step in LI-toolbox. The LI ranges from + 100 (total left functional lateralization) to -100 (total right functional lateralization), thus providing information about the direction and degree of hemispheric lateralization during a given task. The participants were thus categorized as left-lateralized if their LI was > 40, right-lateralized if their LI was <  − 40, and ambilateral if their LI was in between − 40 and 40. We used ± 40 as a cut-off point based on previous findings that emphasized the importance of lateralization strength when grouping individuals (Mazoyer et al. 2014; Labache et al. 2020). Considering that one of our objectives was to disentangle the differences between strongly lateralized and weakly lateralized, we opted for a cut-off that maximized that contrast by ensuring the strong lateralization of both the left-lateralized and right-lateralized groups.

A series of analyses were performed to test the hypothesis of crossed dominance of the parietal network involved in visuospatial processing in individuals with atypical language lateralization. First, a Kruskal–Wallis test was computed to check if significant differences existed in landmark LI between the left, right and ambilateral groups according to language. Post hoc pair-wise comparisons were calculated using the Dunn test. Next, Spearman’s correlation was used to study the linear relationship between the LIs for the verb generation and the landmark tasks.

Voxel-wise whole-brain activations during the landmark and verb generation tasks were also explored in relation to language and visuospatial lateralization groups, respectively. One-sample t-tests were computed to describe the activation pattern during the ‘activation > control’ condition across the whole sample (voxel-wise P < 0.001; FWE cluster-corrected at P < 0.05). Then, voxel-wise two-sample t-tests were used to examine activation differences between the groups with left, right, and ambilateral language lateralization according to the verb generation task and the landmark task (voxel-wise P < 0.001; FWE cluster-corrected at P < 0.05).

We also studied behavioral performance during the landmark task in relation to hemispheric lateralization. First, accuracy (%) when correctly detecting the bisected lines was compared across all groups using two separate ANOVA designs (one for language-based groups, and one for visuospatial-based groups), including age and fluid intelligence (WAIS-IV score) as covariates of control. Then, two separate repeated-measures ANOVAs (one for language-based groups, and one for visuospatial-based groups) were also computed exploring the trials requiring the detection of incorrectly bisected lines. These models included difficulty (easy, medium, hard) as within-subject factor, lateralization group (left, right, or ambilateral) as between-subject factor, and the accuracy rate as a dependent variable. Age, fluid intelligence (WAIS-IV score), and accuracy rate during correctly bisected trials were used as covariates of control.