Ten healthy right-handed participants (three women, mean age: 26.6 ± 2.7 years) participated in the experiment. They all had normal or corrected-to-normal vision and did not report any history of neural disorders. The data of one male participant had to be discarded because of technical problems. The experiment was conducted in accordance with the Declaration of Helsinki (except for registration in a database) and was approved by the Laval University Biomedical Ethics Committee. Informed consent was obtained prior to the experiment.

The participants were seated in a dark room with their feet resting on a footstool. They were secured to the chair with a four-point belt. The chair could be manually rotated about its vertical axis by the experimenter. The rotations were recorded with an optical encoder at 1 kHz. A circular array of LEDs fixed on the floor behind the chair indicated its initial angular position and the three targets rotations (i.e., 20°, 30° and 40° in the counterclockwise direction). A light emitted by a laser diode fixed on the back of the chair provided the experimenter with visual feedback about the chair orientation along the LEDs array. The use of different rotation amplitudes together with the variability in the actual body rotation (e.g., acceleration, amplitude) for a given rotation target amplitude increased the necessity for the participants to direct their attention on information related to self-rotation to keep track of their initial orientation. Similar set-ups have frequently been used for testing vestibular-related processes (e.g., Blouin et al., 1998, 2010; Hanson and Goebel, 1998; Funabiki and Naito, 2002; Mackrous et al., 2019). Importantly, the choice of manual rotations reduced the possibility of electric noise contamination of the EEG recordings (see Nolan et al., 2009 for a discussion on this issue).

Electroencephalographic (EEG) activity was recorded using a Geodesic 64-channel EEG sensor net (1000 Hz, Electrical Geodesics Inc., Eugene, United States). This electrode density is considered appropriate for electrophysiological source imaging (Seeck et al., 2017). Data pre-processing was performed using BrainVision Analyzer 2 (Brain Products, Gilching, Germany). The recordings were first referenced to the averaged activity of the 64 scalp electrodes. Then, data recorded by all electrodes were synchronized with respect to the onset of the rotation (i.e., when chair angular velocity > 2.5°/s), with the average amplitude of the 200-ms pre-rotation epoch serving as baseline. The signals were band-passed filtered (0.5–45 Hz) and independent component analyses (ICA) were used to subtract ocular artifacts (e.g., blinks, saccades) from the EEG recordings. The recordings were visually inspected and epochs still presenting artifacts were rejected. The data were separately averaged for each participant, task, target body angular rotation amplitude (i.e., 20°, 30°, 40°) and electrode. These averages were used to estimate the sources of the cortical activities.

The cortical sources were reconstructed using Brainstorm software (Tadel et al., 2011, freely available at http://neuroimage.usc.edu/brainstorm). We employed the minimum-norm technique to resolve the inverse problem with unconstrained dipole orientations. The forward models were computed using a boundary element method (symmetric BEM, Gramfort et al., 2010) on the anatomical MRI Colin 27 brain template, a predominant volume conductor model from the Montreal Neurological Institute. We used a BEM model with three realistic layers (scalp, inner skull, and outer skull) which provides more accurate solution than a simple three concentric spheres model (Sohrabpour et al., 2015). To enhance the spatial resolution of the brain template, we used a high number of vertices (i.e., 306,716 vertices). Such source reconstruction of EEG data has proved to be suited for investigating the activity of outer and inner cortical surfaces (Ponz et al., 2014; Chand and Dhamala, 2017).

Consistent with previous studies (Schneider et al., 1996; Gale et al., 2016; Ertl et al., 2017), RotEPs were found over a large set of electrodes, but were largest over the Cz electrode (i.e., vertex). As shown in Figure 3 (left column: averaged RotEP; right column: individual RotEP), the RotEPs were composed of successive inflection points which we refer to as P₁ (overall mean 47 ± 12 ms), N₁ (136 ± 12 ms), P₂ (303 ± 35 ms), and N₂ (470 ± 55 ms). These points served as temporal landmarks for analyzing the dynamic of the cortical activities in the source space.

For each participant, rotation amplitude and task, we measured the latency of each RotEP’s inflection point and then averaged the current computed in the source space over three successive time windows: P₁N₁, N₁P₂, and P₂N₂. Following this step, we collapsed (by averaging) the maps obtained for each amplitude of rotations to obtain a single map of current amplitude per participant and task. Following trial rejections, the averaged current maps were built using an average number of trials of 61.33 and 63.22 for the Updating and Control conditions, respectively. It should be noted that this method for estimating EEG sources is not impacted by potential effects of rotation amplitudes, task or inter-individual differences on RotEPs peak latencies. Computed in the source space, current amplitude is considered to reflect brain activation (Tadel et al., 2011, 2019). The last RotEP peak considered in the Updating task (i.e., N₂) occurred on average 171, 269, and 326 ms before the end of the 20°, 30° and 40° rotations. Thus, different cortical activations between the Updating and Control tasks would likely be more related to spatial updating than to the preparation of the arm motor commands in the Updating task.

To highlight those cortical regions specifically involved in spatial updating, we computed statistical maps by contrasting the current maps (i.e., each vertice) obtained in the Updating and Control tasks using t-tests (significance threshold p < 0.05, uncorrected). These analyses were performed separately for each time windows (i.e., P₁N₁: short latency, N₁P₂: mid-latency and P₁N₂: long latency) to gain insight into the spatio-temporal dynamics of the cortical activities during spatial updating. Sources were identified according to Brodmann’s areas after converting MNI to Talairach coordinates with the Non-linear Yale MNI to Talairach Conversion Algorithm (Lacadie et al., 2008a). The Brodmann’s areas definition was based on Lacadie et al. (2008b).

The amplitude of the cortical potentials increases when they are evoked by task-relevant somatosensory (Cybulska-Klosowicz et al., 2011; Saradjian et al., 2013) or visual (Lebar et al., 2015) events. In the present study, extracting the RotEPs out of the EEG recordings allowed us to determine if this effect generalized to cortical responses evoked by idiothetic information relevant for spatial updating. To this end, we compared the amplitude of the P₁N₁, N₁P₂, and P₁N₂ components between the Updating and Control tasks. We also compared between these tasks, the latency of the RotEPs, which was defined as the time elapsed between the onset of the body rotation and P₁. Variables related to the RotEP were submitted to separate 2 (Task: Updating, Control) by 3 (Amplitude: 20°, 30°, 40°) repeated-measures ANOVAs (significance threshold p < 0.05). Prior tests (Kolmogorov–Smirnov) confirmed the normality of all data.

The ANOVA revealed a significant effect of Task on P₁N₁ component amplitude (Figure 4). The amplitude of this first Rotation-Evoked Potentials (RotEP) component was greater in the Updating than in the Control task (8.83 ± 3.18 mV vs. 7.76 ± 3.14 mV, F_1,8 = 5.66, p = 0.04; ηp2 = 0.44). The analyses did not reveal significant effect of rotation Amplitude (p = 0.38) or significant interaction of Task × rotation Amplitude (p = 0.09). The experimental tasks did not have significant effect on the other RotEP components (i.e., N₁P₂: p = 0.56; P₂N₂: p = 0.09).

The latency of the RotEPs (i.e., P₁, mean 47 ± 12 ms) was not significantly different between the Updating and Control tasks (p = 0.64) or between the different rotation amplitudes (p = 0.60). The interaction Task × rotation Amplitude was also not significant (p = 0.19).

The statistical maps computed over the 3 RotEP components (i.e., P₁N₁, N₁P₂, and P₂N₂) are shown in Figures 5–7. In these figures, warm color shadings indicate that the current computed in the Updating task was significantly greater than in the Control task. Cold color shadings indicate the opposite pattern. The MNI coordinates of maximal significant current difference between the Updating and Control tasks and their corresponding Brodmann areas are presented in Tables 1–3 for P₁N₁, N₁P₂, and P₂N₂, respectively.

The mean current measured over the P₁N₁ interval, that is, between 47 and 136 ms after rotation onset, was strikingly alike between the Updating and Control tasks. The statistical map indicated that the cortical activity did not significantly differ between both tasks except for small areas of the left PPC and left LOC (see Figure 5). In these areas, the activity was smaller in the Updating than in the Control task.

Extensive differences in cortical current between both tasks emerged, however, in the second RotEP component (N₁P₂, between 130 and 303 ms after rotation onset). Overall, the cortical activity increased when participants tracked their initial position during the body rotations (see Figure 6). Significant task-related activities were mainly source-localized in the frontal and temporal cortices. Specifically, the supplementary motor complex (SMC), the dorsolateral prefrontal cortices (dlPFC), and the right anterior prefrontal cortex all showed greater activation in the Updating than in the Control tasks. Significantly greater current in the Updating task was also found in the right TC and in both medial TC, and in a small area of the right PPC. Only sparse regions of the LOC showed greater activation in the Control task than in the Updating task.

The increased activation observed in the Updating task during the N₁P₂ interval persisted in several cortical regions during the last RotEP component (P₂N₂, between 303 and 470 ms after rotation onset). This was the case for the bilateral dlPFC, SMC, medial TC, and for the right PPC (see Figure 7). Other regions showed significant task-related activities exclusively in this last analyzed interval of the RotEP. These regions were the right dorsal premotor areas, the left and right cuneus, the left precuneus, and the left PPC. Increased in the Updating task during N₁P₂, the activity of the right temporal lobe was no longer altered by the spatial updating processes during P₂N₂. The left LOC continued to exhibit greater activities in the Control task.

It is of note that the motor cortex contralateral to the arm used by the participants to point toward the home position after rotation did not show significant activity in any of the three analyzed intervals. This suggests that the increased activation observed in the present Updating task was more related to spatial updating than to motor preparation per se.

As a first salient finding, the current measured in the source space remained largely similar between the Updating and Control tasks over the first component of the RotEPs, a period spanning 47–136 ms after body rotation onset. The only significant effect revealed by the statistical map during this interval was a smaller activation in the Updating task in a few vertices of the left PPC and LOC. This effect persisted at both mid- and long latencies only for the LOC. The narrowness of the regions showing this trend raises the question of the robustness of this finding. Note, however, that the LOC contributes to the allocentric coding of space (Committeri et al., 2004; Zaehle et al., 2007; Ruotolo et al., 2019). Hence, the smaller activity observed in the LOC during the Updating task may have hampered the use of this frame of reference for encoding home position during body rotations. This could have indirectly enhanced the use of an egocentric frame of reference, which appeared more relevant in the present Updating task performed in the dark.

The absence of large task-related activation during P₁N₁ suggests that the first wave of idiothetic cue processing during the body rotations was not strictly linked to spatial updating. Yet, despite their non-specific nature, brain processes during the early phase of the rotations most certainly remained critical for spatial updating. This could be the case of the processes associated to vestibular inputs, which are the main carrier of body motion information during passive displacements in the dark (Valko et al., 2012).

The scarcity of early task-related activation was somewhat unexpected given that overt and covert attention influences neural activities related to the early processing of sensory cues (<100 ms; Woldorff et al., 1987; Di Russo and Spinelli, 1999). One explanation for the lack of different cortical activations between the Updating and Control tasks may be that participants also directed their attention toward self-motion information in the latter task. This could have helped them to keep fixation on the chair-fixed LED during the rotations.

Large significant task-related activities emerged during the second RotEP component (i.e., N₁P₂, 136–303 ms after rotation onset). These activities were source-localized in a large network comprised mainly of the frontal, temporal and parietal regions. Like most task-related activities observed in the present study, the current measured in these regions were greater in the Updating than in the Control tasks.

The dlPFC and mPFC showed increased activations in the Updating task. The increased activations in these frontal areas could be associated to important executive functions for spatial processes. Their roles in maintaining spatial information in short-term working memory, and in the cognitive manipulation of information from the environment that is out of view (Wager and Smith, 2003; Gilbert and Burgess, 2008) could have been relevant in the present study for keeping track of the original orientation during entire body rotations. The persistence, in the last analyzed interval, of the strong activity of the dlPFC and mPFC in the Updating task is consistent with this suggestion.

On the other hand, the task-related activations observed in the aPFC and TC at both mid- and long latencies could be linked to the online processing of home direction during rotations. This interpretation is supported both by human studies showing that the aPFC is crucial to encoding spatial information about goals (Ekstrom et al., 2003; Ciaramelli, 2008; Spiers, 2008; see Poucet and Hok, 2017 for similar evidence in rodents) and by those reporting that the activities of the medial temporal lobe dynamically change according to the current Euclidean distance of the spatial goal during navigation (Ekstrom et al., 2003; Howard et al., 2014; Spiers and Barry, 2015; Kunz et al., 2021). Together, the sustained activation observed in the prefrontal and temporal regions while participants were passively rotated could have then contributed to the maintenance of home orientation in working memory and in the encoding of its angular distance for use in the upcoming goal-directed pointing movements.

We also found significant increased activities in the Updating task in the right superior PPC during N₁P₂ which became bilateral during P₂N₂. These results were to be expected given the well-recognized importance of the superior PPC for spatial processes, including those specifically linked to the updating of spatial representations during movements (Duhamel et al., 1992; Medendorp et al., 2003; Merriam et al., 2003; Pisella and Mattingley, 2004; Ventre-Dominey and Vallee, 2007; Gutteling et al., 2015; Gutteling and Medendorp, 2016). The activity of the PPC could have contributed to estimating the angle of home position during and after body rotation (Spiers and Maguire, 2007; Howard et al., 2014; see Ekstrom et al., 2017 for a review), which was an important egocentric process for planning the pointing movement.

The increased activation observed in the superior PPC could also be related to the maintenance of spatial attention (see Ikkai and Curtis, 2011, for a review), which is a critical cognitive process for spatial updating. It should be noted that our observation that the increased activity in the right PPC preceded the increased activity over the left PPC is consistent with studies suggesting a right-hemispheric dominance for visual processes and remapping (Corbetta et al., 2000; Marshall and Fink, 2001; Pisella et al., 2011).

The SMC was the last region where significant task-dependent activities were observed at mid-latencies. These increased activities also lasted until the final analysis time window (i.e., P₂N₂). The SMC comprises the supplementary motor area (SMA), the supplementary eye fields (SEF) and the pre-supplementary motor area (preSMA) (Nachev et al., 2008). The SEF might have played a crucial role in the Updating task. This is suggested by studies showing that lesions affecting the SEF impair the accuracy of saccades toward a memorized visual target only if the patients are rotated before triggering the saccade (Pierrot-Deseilligny et al., 1993, 1995). The SEF was first considered as an oculomotor area (Schlag and Schlag-Rey, 1987). Studies in Monkey, however, have identified a large population of SEF neurons that increase their activities during arm movements (40% of 337 neurons in Fujii et al., 2001; 42% of 106 neurons in Mushiake et al., 1996). These arm-related cells provide grounds for a plausible contribution of the SEF to providing relevant spatial information in the present study for pointing, after the rotations, toward the original body orientation.

Bilateral increased activations were observed at long-latency (P₂N₂) in the cuneus, which is part of the medial visual cortex. In the present study, the only available visual input was a chair-fixed LED which the participants fixated throughout the trials. Because this visual input was present in both tasks, the contrast of Updating and Control tasks most likely canceled out the neural activity evoked by the LED fixation. This possibility is well supported by the fact that the latency of the task-related activity observed in the cuneus (>300 ms) was much longer than the latency of visual-evoked potentials recorded in occipital lobe either by EEG or magnetoencephalography (<80 ms, Vianni et al., 2001; Ellemberg et al., 2003; Lebar et al., 2015). More likely, the late activity of the visual cortex may have resulted from non-visual top-down signals (mediated for instance by parietal or frontal regions, Michelli et al., 2004). This could have enabled the use of a visual-like representation to encode initial direction during the rotation, perhaps through visual mental imagery (see Kosslyn et al., 1999; Strokes et al., 2009). The fact that visual mental imagery activates the earliest visual cortex (BA 17 and 18) (Slotnick et al., 2005) affords this possibility.

It is interesting to note that the task-related activities observed in the medial surface of the left PPC (i.e., precuneus) at long latencies are consistent with the use of an egocentric frame of reference for encoding home position (Byrne et al., 2007; Wolbers et al., 2008; Chadwick et al., 2015). Based on idiothetic and gaze direction cues (Paillard, 1987; Jeannerod, 1991), this frame of reference appears most relevant in the present spatial task for updating body orientation during rotations in the dark. The egocentric spatial information contained in the precuneus would serve in contexts with body displacement but not for the mere egocentric judgments of objects location in steady body conditions (Chadwick et al., 2015). We noted that the increased activation of the precuneus observed in the Updating task was left lateralized. This lateralization was also found by Chadwick et al. (2015) when using searchlight analysis to characterize neural activity in an fMRI navigation study, but not in their follow-up analyses using a more liberal threshold. Bilateral activation of the precuneus was also observed in the Wolbers et al.’s (2008) fMRI study when participants indicated the position of memorized objects following self-displacements in virtual environments. Note, however, that the low temporal resolution of fMRI scanners does not allow to determine whether activations in both hemispheres occur simultaneously. The results of the present EEG study may suggest that processes related to egocentric representations may have shorter latencies in the left than in the right precuneus, especially when encoding spatial positions from the contralateral visual hemifield.

Dorsal motor areas of the frontal lobe were strongly activated in the last analyzed rotation interval of the Updating task. These areas have direct connections with spinal motor neurons (Chouinard and Paus, 2009). Importantly, however, we found that the increased activations in the dorsal motor areas were circumscribed to the right hemisphere, which was ipsilateral to the pointing arm. Although a fraction of ipsilateral connections reaches the spinal level (Kuypers, 1981), the absence of significant task-related activities in the contralateral dorsal motor areas suggests that the functions linked to their activations were more cognitive than motoric in nature. The marked activities found in the right dPMC in the Updating task is consistent with studies showing evidence of functional hemispheric differences between the left and right dPMC, with a salience for the right dPMC for spatial working memory processes (Jonides et al., 1993; Smith and Jonides, 1999). The dPMC could have beneficiated from inputs sent by prefrontal areas which showed task-related activities with shorter latencies (136–303 ms vs. 303–470 ms after rotation onset). Indeed, the dPMC, and particularly the areas with only sparse projections to motoneurons, have dense connections with the PFC (Lu et al., 1994; Genon et al., 2017).

The P₁N₁ component of the RotEP, as measured here over the vertex, had a significantly greater amplitude when the participants had to update their orientation during body rotations. This finding indicates that the amplitude of the cortical potential recorded at scalp level is a reliable marker of cognitive process enhancement related to spatial updating (including processes related to attention). This task-related effect on the P₁N₁ amplitude has found little echo, however, in the cortical current measured in the source space during the same P₁N₁ interval. This could be explained by the fact that the latter variable was obtained by averaging the cortical current over the P₁N₁ interval and that N₁ marked the upper bound of this interval. The differential effect of the updating task on these two EEG measurements (i.e., P₁N₁ amplitude and current in the P₁N₁ interval) might suggest that the processes specifically dedicated to space updating essentially started with a latency close to that of N₁, measured here as 136 ms after body rotation. Computing this latency with respect to the first cortical arrival of idiothetic cues (i.e., P₁: 47 ms after rotation onset), probably provides a better estimate of the latency of the updating processes. Viewed from this perspective, our EEG recordings suggest that the cortical processes specifically involved in spatial updating have a latency of ∼89 ms.

Ertl et al. (2017) reported that the amplitude of the first RotEP component increases by increasing linear acceleration of the participants. In the present study, despite the angular acceleration increased with the rotation amplitude (see Figure 2), the amplitude of P₁N₁ did not significantly differ between the different body accelerations. Although it is difficult to compare linear and angular accelerations, one can notice that the maximal linear accelerations differed more between the different translation conditions in Ertl et al.’s (2017) study (from 1 to 7.67 times greater) than the maximal angular accelerations did between the different rotation conditions of the current study (from 1 to 1.36 times greater). Hence, the absence of effect of angular acceleration on the amplitude of the RotEP could be due to the small difference of acceleration, in the present study, between the 20°, 30°, and 40° rotation amplitudes.

Some limitations in the present study should be considered. Having the objective that task-related cortical activities observed in the Updating task could not merely reflect motor-related processes linked to the preparation of the pointing movements produced after body rotations, we restrained our analyses to the early spatial updating processes [i.e., <470 ms (N₂) with respect to the onset of body rotation]. Cortical activities involved in spatial updating that had greater latencies could therefore not be identified. Also, because the present vestibular memory-contingent task involved goal-directed arm movements, task-related activities observed here could be, in part, specific to spatial updating in contexts of motor actions, and perhaps more specifically of goal-directed arm movements.

Another limitation concerns the statistical analyses of the cortical sources. The present experiment was designed to investigate the dynamics of the cortical activation (including from inner surfaces) during spatial updating. To enhance the spatial resolution of the brain template, we used a high number of vertices (i.e., 306,716 vertices), without correcting for multiple comparisons. Further studies could build on the present results to focus the analyses on more circumscribed regions of the cortical network shown here as being involved in spatial updating, allowing for multiple comparisons corrections.