Dynamic changes in spatial representation within the posterior parietal cortex in response to visuomotor adaptation

Abstract

Recent studies used functional magnetic resonance imaging (fMRI) population receptive field (pRF) mapping to demonstrate that retinotopic organization extends from the primary visual cortex to ventral and dorsal visual pathways, by quantifying visual field maps, receptive field size, and laterality throughout multiple areas. Visuospatial representation in the posterior parietal cortex (PPC) is modulated by attentional deployment, raising the question of whether spatial representation in the PPC is dynamic and flexible, and whether this flexibility contributes to visuospatial learning. To answer this question, changes in spatial representation within the PPC and early visual cortex were recorded with pRF mapping before and after prism adaptation (PA)—a well-established visuomotor technique that modulates visuospatial attention according to the direction of the visual displacement. As predicted, results showed that adaptation to left-shifting prisms increases pRF size in left PPC, while leaving space representation in the early visual cortex unchanged. This is the first evidence that PA drives a dynamic reorganization of response profiles in the PPC. These findings show that spatial representations in the PPC not only reflect changes driven by attentional deployment but dynamically change in response to modulation of external factors such as manipulation of the visuospatial input during visuomotor adaptation.

visuospatial attention, receptive field, intraparietal sulcus, prims adaptation, pRF

Introduction

The visual cortex encodes the external environment retinotopically, such that adjacent locations in space are represented by adjacent pools of neurons (Holmes 1918; Horton and Hoyt 1991). This retinotopic organization extends to the dorsal and ventral visual pathways (Sereno et al. 2001; Silver et al. 2005; Swisher et al. 2007; Kravitz et al. 2013; Uyar et al. 2016), as well as to regions associated with visuospatial attention, such as the frontal eye fields (FEF) and the posterior parietal cortex (PPC; Bruce and Goldberg 1985; Schall et al. 1995; Silver and Kastner 2009; Szczepanski et al. 2010).

The PPC is a node in the attentional selection network that is thought to be subserving “priority” maps that drive and predict efficient attentional processing and provide top-down feedback for motor orienting and attentional allocation (Bisley and Goldberg 2010; Shomstein and Gottlieb 2016). Consistent with a role of the PPC in spatial attention, studies employing functional magnetic resonance imaging (fMRI) found that this region (along with other nodes in the attentional network including FEF and superior and inferior parietal cortices) contains topographic representations related to saccade planning and attention (Husain and Nachev 2007; Molenberghs et al. 2007; Silver and Kastner 2009; Sheremata and Silver 2015). Despite recent progress in understanding the role of topographic coding within the PPC and in attentional orienting, the question remains as to the extent to which this spatial topography is dynamic and malleable. The answer to this question has remained elusive partially due to the insufficient spatial resolution of fMRI.

Recent advances in fMRI, however, enabled fine-grained measures of spatial representation, beyond nonhuman primates, using population receptive field (pRF) mapping (Dumoulin and Wandell 2008). Specifically, pRF mapping enables measuring the fMRI response at many visual field locations to determine the set of locations that evoke a response in each voxel (Dumoulin and Wandell 2008), and thus provides estimates of each voxel’s receptive field size and preferred location. By employing this technique, recent studies have been able to provide initial evidence of PPC’s malleability. For example, it was demonstrated that orienting attention, as compared with passively maintaining fixation, increases the extent of the visual field as evidenced by increased pRF size. This finding was one of the first to demonstrate that spatial representations within the PPC differ with the extent of attentional modulation (Sheremata and Silver 2015). Interestingly, it was observed that attentional modulation of pRF size was asymmetric, such that the right PPC pRF significantly increased representation of both the contra (left) and ipsilateral (right) visual hemifield (i.e., becoming bilateral), whereas the left PPC remained confined to the contralateral (right) visual space. Although this finding provides evidence for malleability of PPC spatial representations when guided by internally initiated attentional signals, it is unclear whether such changes are exclusive to attentional modulation, or whether spatial representation can dynamically change when affected by external factors and with experience, such as manipulation of the visuospatial input during visuomotor adaptation. The question remains whether spatial representations within the PPC are malleable enough to be modulated by external factors such as visuomotor learning.

One of the well-established methods for inducing visuomotor learning is the prism adaptation (PA) technique (Helmholtz 1867; Rossetti et al. 1998). PA is a visuomotor training consisting of adaptation to displaced vision that not only modifies sensorimotor coordinates but also affects higher-order cognition in both neglect patients (McIntosh et al. 2002; Bultitude and Woods 2010; Jacquin-Courtois et al. 2010) and healthy individuals (Frassinetti et al. 2009; Bultitude and Woods 2010; Schintu et al. 2018). The degree to which PA influences behavior and the corresponding neural changes that drive it differ according to the side of displacement: right PA, by shifting visual information several visual degrees from the left to the right visual field, induces a leftward bias, whereas left PA, by shifting visual information from the right to the left visual field, induces a rightward bias. However, the effectiveness of PA also depends on the state of the system: intact versus damaged. In neglect patients, although right PA is well known to ameliorate neglect symptoms and thus reduces the pathological rightward visuospatial bias (e.g. Rossetti et al. 1998), left PA neither affects the sensorimotor nor the visuospatial domain and thus does not worsen such pathological rightward bias (Luauté et al. 2012). On the other hand, although both left and right PA produce sensorimotor aftereffects in healthy, only left PA has been extensively shown to induce a rightward visuospatial bias (Schintu et al. 2014, 2017; for a review see Michel 2016), by counteracting the inherent leftward bias called pseudoneglect (Jewell and McCourt 2000), whereas right PA has consistently failed to produce significant cognitive modulation, so much that it has become the gold standard control condition for left PA (Colent et al. 2000; Michel et al. 2003; Loftus et al. 2009; Bultitude and Woods 2010; Reed and Dassonville 2014; Schintu et al. 2014, 2017, 2018). It is worth noting, however, that several studies have been showing neural changes following right PA despite the absence of significant cognitive modulations (Crottaz-Herbette et al. 2014; Schintu et al. 2020b).

Here, we directly investigated whether spatial representations in the PPC, as measured by pRF mapping, are altered in healthy individuals in response to spatially modified visual input induced through visuomotor adaptation. Changes in pRFs were measured in healthy individuals before and after spatial input was manipulated with PA under conditions of full attention and simple fixation.

For the left PA we expect: (i) given that Sheremata and Silver (2015) pRF mapping study showed that the right PPC processes both the contralateral (left) and ipsilateral (right) visual field, whereas the left PPC solely the contralateral (right) one (Mesulam 1981), we expected left PA to produce the well-known rightward visuospatial bias in healthy individuals by increasing pRF visual field extent (i.e., size) in the left PPC and thus increasing integration of the right visual space; (ii) for the right PPC, we did not hypothesize a direction of the pRF size modulation, because given its bilateral representation of visual space (Mesulam 1981) either an increase or a decrease in pRFs size can be theoretically predicted for both visual fields canceling out the possible cognitive effect.

For right PA we expect: (i) no behavioral changes, and (ii) if any neural changes are to be observed, we predict that the effect would be opposite to the left PA one, that is, decreasing pRF visual field extent (i.e., size) in the left PPC. Furthermore, since PA is thought to modulate attentional processes (Martín-Arévalo et al. 2016) we expect maximal pRF modulation when attention is deployed toward the mapping stimulus, and given that attentional deployment enhances the neural response of the early visual areas by prioritizing regions of space (Kastner 1998) it is possible that such pRF change, albeit in smaller magnitude, would be also observed in early visual cortex.

Materials and methods

Participants

Forty adults with normal or corrected-to-normal vision and no history of neurological problems were recruited for the study. All participants were right-handed (Edinburgh Inventory; Oldfield 1971), had a right dominant eye (hole-in-card test; Miles 1930), were compensated for participation, and gave informed consent. The study was approved by the National Institutes of Health Institutional Review Board. Twenty participants were adapted to left PA (12 females; age = 26.25 ± 0.87 SEM) and the remaining 20 were adapted to right PA (13 females; age = 26.12 ± 1.05 SEM). Two participants were excluded from the right PA group, 1 for incomplete data collection and 1 because of the presence of a cyst. After pRF analysis criteria were satisfied (see pRF methods section), the final data submitted to the statistical analysis were gathered from a total of 26 participants: 16 in the left PA (10 females; age = 26.06 ± 0.9 SEM) and 10 in the right PA group (6 females; age = 24.19 ± 1.3 SEM). No difference in age was observed between the 2 groups [t(24) 1.194, P = 0.244].

Procedure

The experimental procedure consisted of two consecutive general sessions, one before and after PA, and each session included behavioral testing and neuroimaging scan. A baseline measurement of visuospatial cognition was assessed with perceptual line bisection and manual line bisection tasks (see Fig. 1, marked as PLB and MLB). Then, in the pre PA session participants underwent a neuroimaging scan that consisted of a resting state scan (results of which are reported in Schintu et al. 2020b) and a pRF scan (focus of this report), perceptual line bisection, and manual line bisection tasks, along with straight-ahead (SA) pointing task and open-loop (OL) pointing task that were used as a proxy of the PA level (Fig. 1). PA then followed, one group of participants was adapted to left-deviating prisms whereas the other group was adapted to the right-deviating prisms. In the post adaptation session, the straight-ahead and open-loop pointing tasks immediately (early-post) followed PA, participants then underwent another resting state and pRF scan, which was then followed by perceptual line bisection and manual line bisection tasks, and finally (late-post) by another repeat of the straight-ahead and open-loop pointing tasks.

Fig. 1

Experimental design. PLB = perceptual line bisection task (i.e., Landmark task); MLB = manual line bisection task; fMRI = funtional magnetic resonance imaging; SA = straight-ahead pointing task; OL = open-loop pointing task; PA = prism adaptation; pre = before prism adaptation; post = after prism adaptation. Reprinted with permission from Schintu et al. (2020b).

During behavioral assessment and PA, participants were comfortably seated in front of a horizontal wooden board with their heads supported by a chin rest. On the board, three circular targets (8 mm in diameter) were positioned at 0°, −10°, and +10° from the body midline, ~57 cm from participant’s nasion, and were used for PA, open-loop, and straight-ahead tasks.

Behavioral assessment

We employed four different tasks quantifying spatial representation at the cognitive, proprioceptive, and sensorimotor level as described previously in our studies investigating visuospatial modulation following both PA and inhibitory transcranial magnetic stimulation (TMS; Schintu et al. 2020b, 2021).

Perceptual line bisection task—the Landmark task

A modified version of the Landmark task (Milner et al. 1992) was used to quantify the visuospatial bias. The task consisted of 66 white pre-bisected lines (350 mm × ~2 mm) displayed on a black screen positioned 35 cm from the eyes. Lines were transected at the true center and 2, 4, 6, 8, and 10 mm toward the left and right of the true center. Each of the 11 different pre-bisected lines was presented 6 times in a pseudorandom order, yielding a total of 66 trials, taking ~3 min to complete. Each pre-bisected line was displayed for a maximum of 5 s or until a response was made and was then replaced by a black-and-white patterned mask, which stayed on the screen for 1 s before the next pre-bisected line was displayed. Presentation software (Neurobehavioral Systems, Inc., United States) was used to generate the stimuli, record responses, and control the timing of stimuli presentation throughout the task. Participants were instructed to fully inspect each pre-bisected line and judge whether the mark (transector) was closer to the left or right end of the line. In this two-alternative forced-choice paradigm participants answered by pressing the pedal under their left foot if the transector was perceived as being closer to the left end of the line and by pressing the pedal under their right foot if they thought it was closer to the right end of the line. Response by pedals was chosen to limit the use of the right hand, which was used to adapt to the prisms since any feedback from that hand could contribute to de-adaptation. Prior to the baseline measure, at least, 10 practice trials were given to ensure that participants properly understood the instructions and were confident answering with the pedals. For each participant, the percentage of “right” responses was plotted as a function of the position of the transector. These data were then fitted with a sigmoid function and the value on the x-axis corresponding to the point at which the participant responded “right” 50% of the time was taken as the point of subjective equality (PSE).

Manual line bisection task

The manual line bisection task (Schenkenberg et al. 1980; Urbanski and Bartolomeo 2008) was used to quantify the visuospatial bias. It consisted of a series of 10 black lines (350 mm × ~2 mm; identical in size to those used for the Landmark task) each drawn on A3 (297 mm × 420 mm) sheets of paper that were positioned over the computer screen, which was kept at the same distance and position as for the Landmark task (35 cm from participants’ nasion). Participants were instructed to fully inspect each line and with the pen held in their right hand, draw a vertical mark where they thought the center of the line was. Once the mark had been drawn the experimenter then turned the page to reveal the next to-be-bisected line. No time limit was imposed, and participants took on average 1 s to place the mark on each line. For each of the 10 lines, the distance between the mark placed by the participant and the true center of the line was calculated. The PSE was calculated as the average distance between the true center and the mark drawn by the participant, with marks to the right of the center coded as positive and to the left as negative.

Straight-ahead pointing task

The straight-ahead pointing task was used to quantify the proprioceptive bias (Rossetti et al. 1998). Participants performed 6 pointing movements to their perceived midline with the right index finger at a comfortable and uniform speed, whereas the left hand rested on the lap. Before each movement, participants received a verbal instruction to close their eyes and imagine a line splitting their body in half, project this line onto the board in front of them, point to the line while keeping their eyes closed, and then return their hand to the starting position. To ensure that participants had no visual feedback regarding either their movement or their landing position, vision of the arm and hand was occluded before movement onset by a cardboard baffle. The proprioceptive shift was measured as the average distance between the landing position and the true midline with an accuracy of ±0.5 cm.

Open-loop pointing task

The open-loop pointing task was used to quantify the sensorimotor bias (Rossetti et al. 1998). Participants performed 6 pointing movements to the central target (0 cm) with their right index finger while resting their left hand on the lap. Participants were verbally instructed, before each pointing movement, to look at the central target, close their eyes, point to the target at a comfortable speed while keeping their eyes closed, and then return their hand to the starting position when cued by the experimenter. Similar to the straight-ahead task described above, vision of the arm and hand was occluded. The landing position of the participant’s finger was noted with a precision of ±0.5 cm. The sensorimotor shift was measured as the average distance between the landing position and the central target.

Prism adaptation

During PA, participants were fitted with prismatic goggles with either a 15° left or right visual field deviation and performed 150 verbally cued pointing movements to the right (10°) and left (−10°) targets in a pseudorandom order as called by the experimenter. Prior to each pointing movement, participants placed their right index finger on the starting position, a 1.5cm diameter Velcro pad, placed close to the midline of their chest. Participants could not see their hand when it was in the starting position and during the first third of the pointing movement (Schintu et al. 2014). Participants were instructed to point with the index finger extended, to execute a one-shot movement at a fast but comfortable speed, and to return their hand to the starting position when prompted by the experimenter.

pRF mapping

Participants were scanned on 3 Tesla Prisma (Siemens) scanner in the National Institute of Neurological Disorders and Stroke (NINDS) functional MRI facility. Oblique slices were oriented on an AC–PC line. Whole-brain volumes were acquired using a 32-channel head coil, 46 slices, TR 2,500 ms, TE 30.0 ms, voxel size 3.0 × 3.0 ×3.0 mm, field of view = 192 × 138 × 192 mm, 64 × 46 × 64 matrix, flip angle 70°.

Fig. 2

Behavioral results. Negative and positive values represent left and right of the true center. PSE = point of subjective equality; PA = prism adaptation; pre = before PA; post = after PA. Error bars represent 1 standard error of the mean (SEM). ^* = P < 0.05.

The pRF mapping (Dumoulin and Wandell 2008) experimental design was adapted from Silson et al. (2018). Stimuli were presented through a bar aperture that moved gradually through the visual field while revealing fragments of scenes (circular aperture 8.9° diameter). During each run, the bar aperture made a total of eight sweeps through the visual field (2 orientations and 4 directions). A single sweep of the visual field took 36 s and consisted of 18 separate bar positions (each 2 s). At each bar position, five images were presented (400 ms each). PRF mapping was carried under two different conditions in which attentional deployment was manipulated. During the fixation condition, attention was directed at fixation, away from the mapping stimulus. Participants were asked to maintain fixation and indicate via button press when the white fixation dot changed to red. Color fixation changes occurred pseudo randomly, with four color changes per sweep. During the attention condition, attention was directed toward the mapping stimulus. Participants were asked to maintain fixation and indicate via button press when a specific target scene fragment (yellow sunflowers field) appeared in the bar aperture. The presence of the target image occurred pseudo randomly, with four color changes per sweep.

Each scanning session included six runs: three runs of the fixation condition and three of the attention condition. The order of the fixation or attention conditions and the hand used to answer (left or right) were counterbalanced across participants. Eye tracker recording (long-range ASL EYE-TRAC 7) was used to control fixation.

pRF mapping analysis

All data were analyzed using the Analysis of Functional NeuroImages (AFNI) software package (http://afni.nimh. nih.gov/afni; Cox 1996). Before pRF and statistical analyses, all images for each participant were motion corrected to the first image of the first run, detrended, and aligned to the anatomy, after removal of two dummy volumes to allow the magnetic field to stabilize.

The pRF mapping analysis was conducted in AFNI, using a pRF implementation for the AFNI distribution (developed by R. C. Reynolds), based broadly on previous implementations for pRF estimation (Larsson and Heeger 2006; Dumoulin and Wandell 2008). As in Silson et al. (2018), the model produces elliptical Gaussian pRF models for every possible combination of center position (x, y; 200 samples across the height and width of the screen), sigma (the size of the minor axis of the ellipse; 100 intervals over half the screen size), aspect ratio (relative length of the major and minor axes of the ellipse; 50 even intervals from 1–5), and angle (the orientation of the major axis). Sigma (minor axis) and the major length parameter (b) were then combined to calculate the area of the ellipse (A = π^*σ^*b), which takes into account not only possible changes along the minor axis (σ) but also longer axis since those two are not constrained to be the same in the pRF elliptical model (see supplementary material for related data). This procedure results in 2 × 10⁸ possible pRFs, each of which produces a unique time-series of response to the mapping stimulus. The model then uses both Simplex and Powell optimization algorithms to find the best time series/parameter sets (X, Y, sigma, aspect ratio) by minimizing the least-squares error of the predicted time series measured against the acquired time series in each voxel. The regions of interest (ROIs) in the visual (V1d and v, V2d and v, and V3d and v) and parietal (IPS 0, 1, and 2) cortex were identified via a probabilistic atlas (Wang et al. 2015). Each subarea for a given visual or parietal ROIs were combined, the summed probability threshold was set a 70%, the minimum R² was set at 0.17 (P < 0.05 for a time-series of this length) along with a minimum of 100 vertices per subject in the combined ROIs.

Statistical analysis

Statistical analyses were performed using SPSS (IBM, Version 27.0) with alpha set at 0.05. All data are presented as means with the within-subjects standard error of the mean (SEM). Two-tailed paired or independent t-tests were carried out for post hoc comparisons. Effect sizes are indicated for significant effects.

Results

Behavioral measurements

Perceptual line bisection—Landmark task

This task quantified the visuospatial bias by asking participants to judge a series or pre-bisected lines. The two pre-adaptation performances were collapsed because of the absence of a significant difference in the mean data [t (25) = −1.335, P = 0.194]. A mixed-model analysis of variance (ANOVA) with time (pre and post) as within-subjects variable and group (left PA and right PA) as between-subjects variable revealed a main effect of time [F(1, 24) = 7.178, P = 0.013, η²_P = 0.230], such that both groups shifted rightward of the true center after PA (from −1.14 to 0.17 mm). No other main effect or interaction reached statistical significance [Fs ≤ 0.311, Ps ≥ 0.582]. Given the consistent effect of left PA on visuospatial behavior, as opposed to the right-shifting prisms that produced no-significant effect, left PA group performance was submitted to an exploratory paired t-test, which revealed a significant rightward shift in midline judgment from (−0.94 to 0.15 mm) following adaptation [t(15) = −2.142, P = 0.049, Cohen’s d = 0.535; Fig. 2). These results suggest that the main effect of time was driven by the changes induced by left PA, which induced the well-known rightward bias in midline judgment.

Manual line bisection task

This task quantified visuospatial bias by measuring the difference between the perceived (manually marked) and actual center of the line. The two pre-adaptation performances were averaged, since there was no difference at baseline [t(25) = −1.762, P = 0.09]. A mixed-model ANOVA with time (pre and post) as within-subjects variable and group (left PA and right PA) as between-subjects variable revealed a trend toward a main effect of time [F(1, 24) = 3.607, P = 0.070, η²_P = 0.131] such that both groups tended to shift rightward of the true center after PA (from −1.19 to −0.17 mm). No other main effect or interaction reached statistical significance [Fs ≤ 0.367, Ps ≥ 0.550]. These results revealed that PA, independently of the visual displacement direction, induced a weak rightward bias in midline judgments.

Straight-ahead pointing task

Proprioceptive performance was measured by quantifying the deviation between pointing to the perceived midline and the true midline. A mixed-model ANOVA with time (pre, early-post, and late-post) as within-subjects variable and group (left PA and right PA) as between subject-variable revealed significant time × group interaction [F(2, 48) = 31.233, P < 0.001, η²_P = 0.565]. Planned comparisons indicated that for the left PA group the proprioceptive performance from baseline (−0.43 cm) shifted rightward at both early [2.50 cm; t(15) = −6.338, P < 0.001, Cohen’s d = 1.58] and late (1.24 cm) post measurements [t(15) = −4.750, P < 0.001, Cohen’s d = 1.19], and for the right PA group from baseline (1.37 cm) it shifted leftward at both early [−1.17 cm; t(9) = 4.099, P = 0.003, Cohen’s d = 1.30] and late-post adaptation measurements [−0.09 cm; t(9) = 2.907, P = 0.017, Cohen’s d = 0.92; Fig. 2]. Importantly, this measure confirmed that PA affected visuospatial processing in the direction anticipated by left- and rightward shifting prisms. The main effect of time [F(2, 48) = 0.162, P = 0.851] and group [F(1, 24) = 2.673, P = 0.115] were not significant.

To assess whether the amount of proprioceptive aftereffect differed between the two groups, the absolute value of the amount of change in proprioceptive performance (early- and late-post “minus” pre) was compared between groups. An independent t-test revealed that the amount of sensorimotor adaptation between the two groups did not differ neither at early nor at late-post (ts ≤ 0.162, Ps ≥ 0.873).

To control for the possible effect of baseline difference between the two groups [t(24) −2.488 P = 0.020 Cohen’s d = 1.793] we ran a mixed-model ANOVA with the change in pointing as dependent measure and baseline pointing as a covariate. Importantly, this analysis revealed a significant time × group interaction [F(1, 23) = 13.319, P = 0.001, η²_P = 0.367] and a no-significant time × baseline interaction [F(1, 23) = 3.205, P = 0.159], meaning that the baseline difference did not influence the changes following PA.

These results show that left PA-induced a rightward bias whereas right PA-induced a leftward bias in pointing to the perceived midline, and that this effect was comparable in magnitude across the two groups. Thus, these results show that PA successfully altered representation of space according to the prismatic deviation.

Open-loop pointing task

Sensorimotor performance before and after PA was measured by quantifying the deviation in the pointing from the landing position and the true center. A mixed-model ANOVA with time (pre, early-post, and late-post) as within-subjects variable and group (left PA and right PA) as between-subjects variable revealed a significant time × group interaction [F(2, 48) = 226.600, P < 0.001, η²_P = 0.904]. Planned comparisons revealed that for the left PA group the sensorimotor aftereffect from baseline (−0.58 cm) shifted rightward at early [4.63 cm; t(15) = −14.774, P < 0.001, Cohen’s d = 3.69] and late-post measurements [1.78 cm; t(15) = −8.336, P < 0.001, Cohen’s d = 2.08], and for the right PA group from baseline (1.10 cm) it shifted leftward at both early [−4.02 cm; t(9) = 12.310, P < 0.001, Cohen’s d = 3.89] and late-post measurements [−1.96 cm; t(9) = 7.921, P < 0.001, Cohen’s d = 2.50; Fig. 2]. Importantly, this measure confirmed that PA affected behavioral visuospatial processing in the direction anticipated by left and right PA. The main effect of group [F(1,24) = 50.041, P < 0.001, η²_P = 0.676] indicated a general rightward pointing deviation for the left PA group (1.94 cm) and leftward pointing deviation for the right PA group (−1.63 cm) whereas the main effect of time [F(2, 48) = 1.621, P = 0.208] was not significant.

To assess whether the amount of sensorimotor aftereffect differed between the two groups we compared the absolute value of the amount of change in sensorimotor performance (early- and late-post minus pre) between groups. The independent t-test revealed that the amount of sensorimotor adaptation between groups did not differ neither at early nor at late-post measurements [ts ≤ −1.513, Ps ≥ 0.143]. To control for the effect of the difference at baseline between the two groups [t(24) −3.196 P = 0.004 Cohen’s d = 1.288] we ran a mixed-model ANOVA with time (early-post and late-post) as within-subjects variable and group (left PA and right PA) as between-subjects variable and having the change in pointing (early- and late- post minus pre) as dependent measure and the baseline pointing as covariate. Importantly, this analysis revealed a significant time × group interaction [F(1, 23) = 96.837, P < 0.001, η²_P = 0.808] and a nonsignificant time × baseline interaction [F(1, 23) = 0.237, P = 0.631], meaning that the baseline difference did not influence the changes following PA.

These results show that left PA-induced a rightward bias, whereas right prims adaptation induced a leftward bias in pointing to a central visual target and that this effect has a comparable magnitude between groups, meaning that both groups were significantly and no-differently adapted until the end of the experiment.

pRF MAPPING

After having shown that PA successfully modulates behavioral performance, we examined the corresponding changes in spatial representation within the retinotopically organized PPC (IPS 0–1–2) and sensory visual areas (V1, V2, and V3) as a function of left- and right-deviating prisms. Given previous results we expected PA to alter pRF size and the largest alterations of visuospatial representation to be observed in the attention rather than fixation condition (Sheremata and Silver 2015).

Here, we applied the most recent elliptical model of pRF mapping (Silson et al. 2018). To analyze the changes in pRF size we used the sigma parameter as it relates most closely to the size estimates of the well-established 2-D Gaussian model (e.g. Dumoulin and Wandell 2008; Silson et al. 2015). See supplementary material for calculated size and angle parameter analysis.

pRF sigma

To examine changes in PPC and early visual areas, pRF sigma parameter was submitted to an omnibus mixed-model ANOVA with ROI (IPS, V1, V2, and V3), time (pre and post), condition (attention and fixation), and hemisphere (left and right) as within-subjects variables, and group (left PA and right PA) as between-subjects variable. This analysis revealed a significant 5-way interaction ROI × time × condition × hemisphere × group [F (3, 72) = 3.506, P = 0.020, η²_P = 0.127], along with an ROI × condition × hemisphere × group [F(3, 72) = 3.722, P = 0.015, η²_P = 0.134], ROI × time × condition × group [F(3, 72) = 3.011, P = 0.036, η²_P = 0.111], ROI × time × group [F(3, 72) = 2.986, P = 0.037, η²_P = 0.111], and ROI × condition [F(3, 72) =3.472 P = 0.020, η²_P = 0.126] interaction. There were also a significant main effect of ROI [F(3, 72) = 11.142, P < 0.001, η²_P = 0.317] and condition [F(1, 24) =15.141, P = 0.001, η²_P = 0.387]. Other main effects and interactions did not reach significance (Fs ≤ 2495 Ps ≥ 0.067).

To unpack the 5-way interaction, the corresponding hubs in the visuospatial attention network (Szczepanski et al. 2010) were separated for further statistical analyses. Namely, parietal and early visual cortex ROIs were analyzed independently.

Parietal cortex

pRF sigma parameter was submitted to a mixed-model ANOVA with time (pre and post), condition (attention and fixation), and hemisphere (left and right) as within-subjects variables, and group (left PA and right PA) as between-subjects variable. The analysis revealed a significant time × condition × hemisphere × group interaction [F(1, 24) = 4.240, P = 0.05, η²_P = 0.150];  follow-up planned comparisons showed a significant increase (from 1.07° to 1.25°) for the attention condition in the left hemisphere following left PA [t(15) −2.241 P = 0.041 Cohen’s d = 0.560], and a marginally significant decrease (from 1.25° to 0.98°) after right PA [t(15) 2.135 P = 0.062 Cohen’s d = 0.675] (Fig. 3). This analysis also revealed a main effect of condition [F(1, 24) = 9.228, P = 0.006, η²_P = 0.278] such that overall the pRF sigma parameter was larger in the attention (1.16°) as compared with fixation (1.02°) condition, replicating earlier findings (Sheremata and Silver 2015). Other main effects and interactions did not reach significance (Fs ≤ 4.075 Ps ≥ 0.055). All other comparisons controlling for possible baseline condition differences both between groups (ts ≤ −1.233 Ps ≥ 0.229) and within groups (ts ≤ 1.723 Ps ≥ 0.105) did not reach statistical significance. Overall, this analysis revealed that left PA increased pRF size in the left hemisphere when attention was deployed. These results show that spatial representation in PPC dynamically changes in response to alteration of the spatial input.

Fig. 3

Parietal cortex. pRF sigma expressed in degrees of visual angle. PA = prism adaptation; pre = before PA; post = after PA; LH = left hemisphere; RH = right hemisphere. Error bars represent 1 SEM. ^* = P < 0.05; § = P < 0.06.

Early visual cortex

pRF size data were submitted to a mixed-model ANOVA with ROI (V1, V2, and V3), time (pre and post), condition (attention and fixation), hemisphere (left and right) as within-subjects variables, and group (left PA and right PA) as a between-subjects variable. The analysis showed an ROI × time × group interaction [F(2, 48) = 5.800, P = 0.006, η²_P = 0.195] such that pRF size in V3 at baseline was smaller in the left PA (1.02°) than right PA (1.11°) group [t(24) = −2.167, P = 0.040, Cohen’s d = 0.874] and in V1 at post pRF size was smaller in the left PA (0.80°) than right PA (1.02°) group [t(24) = −2.468, P = 0.021, Cohen’s d = 0.995; others ts ≤ −1.629 Ps ≥ 0.103]. There was also an ROI × condition × hemisphere interaction [F(2, 48) = 4598, P = 0.015, η²_P = 0.161] such that, independently of the visuospatial manipulation, in V2 in the left hemisphere pRF size for the attention condition (1.02°) was larger than the fixation condition (0.98°) [t(25) = 2.179, P = 0.039, Cohen’s d = 0.427] and similarly pRF size in V3 in the right hemisphere was larger for the attention (1.11°) than fixation (1.07°) condition hemisphere [t(24) = 2.749, P = 0.011, Cohen’s d = 0.539; others ts ≤ 1.728 Ps ≥ 0.096].

The analysis also showed a main effect of ROI [F(2, 48) =28.646, P < 0.001, η²_P = 0.544] such that overall pRFs size increased from V1 (0.89°) to V2 (1°) to V3 (1.07°) (all ts ≥ −3.169 Ps ≤ 0.004), a main effect condition [F(1, 24) = 4.357, P = 0.048, η²_P = 0.154] such that overall pRFs were larger in the attention (1°) than fixation (0.97°) condition. Finally a main effect of group was also found [F(1, 24) = 4.781, P = 0.039, η²_P = 0.166] such that pRFs were overall larger in the right PA (1.07°) than left PA (0.93°) group. Other main effects and interactions did not reach significance (Fs ≤ 2,842 Ps ≥ 0.105; Fig. 4).

Fig. 4

Early visual cortex. pRF sigma expressed in degrees of visual angle. PA = prism adaptation; pre = before PA; post = after PA. Error bars represent 1 SEM. ^* = P < 0.05.

Overall, results from the early visual cortex show that, unlike the parietal cortex, early sensory V1–V3 regions do not show changes in pRF size exclusively driven by altering spatial input by PA. However, these findings show that, independently of the condition (attention or fixation) and hemisphere, pRF size in V1 was smaller in the left PA as compared with the right PA group and that the same difference was present for V3 at baseline.

pRF x-center position

To examine possible changes in pRFs preferred location along the x axis, the x-center position parameter was submitted to an omnibus mixed-model ANOVA with ROI (IPS, V1, V2, and V3), time (pre and post), condition (attention and fixation), and hemisphere (left and right) as within-subjects variables, and group (left PA and right PA) as between-subjects variable. The analysis revealed a significant ROI × time × hemisphere interaction [F(3, 72) = 3.045, P = 0.034, η²_P = 0.113] such that, independently of whether it was the fixation or attention condition or whether participants were adapted to left or right prisms, solely in V2 the x-center position shifted further leftward from pre (−1.51°) to post (−1.69°) in the right hemisphere [t(25) 3.263, P = 0.003, Cohen’s d = 0.640] and further rightward in the left hemisphere [from 1.78° to 1.91°; t(25) −2.329, P = 0.028, Cohen’s d = 0.457], all others (ts ≤ 1.676, Ps ≥ 0.106). The ROI × time × condition interaction was also significant [F(3, 72) = 2.725, P = 0.050, η²_P = 0.102] such that in IPS at post, the x-center position marginally differed between the attention (−0.31°) and fixation (−0.04°) condition [t(25) −1.983, P = 0.058, Cohen’s d = 0.389] independently of the direction of the visuospatial modulation. There was also a significant ROI × hemisphere interaction [F(3, 72) = 92.683, P < 0.001, η²_P = 0.794] such that in the left hemisphere the x-position differed across all ROIs (ts ≥ −2.168, Ps ≤ 0.040) except for V1 vs. V2 [t (25) = −0.707, P = 0.486], and in the right hemisphere the x-position differed across all ROIs (ts ≥ 2.173, Ps ≤ 0.049) except for V1 versus V2 [t (25) = −0.101, P = 0.920], and a condition × hemisphere interaction [F(1, 24) = 14.793, P = 0.001, η²_P = 0.381] such that the x-position differed between and within condition across all ROIs (ts ≥ 2.476, Ps ≤ 0.020). The analysis revealed also a main effect of hemisphere [F(1, 24) = 3455.77, P < 0.001, η²_P = 0.993] such that the x-position in the right hemisphere was negative (−1.47°), thus coding the left side of the visual field, and positive in the left hemisphere (1.6°), thus coding for the right side of the visual field, and a main effect of ROI [F(3, 72) = 5.800, P = 0.001, η²_P = 0.195] such that the x-position of each V areas differed from IPS (ts ≥ 2.537, Ps ≤ 0.018) but not difference was found within V1-V3 (ts ≤ −0.782, Ps ≥ 0.442). Other main effects and interactions did not reach significance (ts ≤ 1.914, Ps ≥ 0.135).

Overall, the x-center position results show no changes in such parameters due to the direction of the visuomotor adaptation, but rather a change over time restricted to V2 area that was also independent of task condition.

pRF y-center position

To examine possible changes in pRFs preferred location along the y axis, the y-center position parameter was submitted to an omnibus mixed-model ANOVA with ROI (IPS, V1, V2, and V3) time (pre and post), condition (attention and fixation) and hemisphere (left and right) as within-subjects variables, and group (left PA and right PA) as between-subjects variable. The analysis revealed a significant ROI × time × condition × group interaction [F(3, 72) = 4.393, P = 0.007, η²_P = 0.155], planned comparison revealed that solely for V1 and following left PA the y-position increased/moved further up in the visual field in both the attention [from 0.24° to 0.40°; t(15) = −2.270, P = 0.038, Cohen’s d = 0.568] and fixation [from 0.09° to 0.22°; t(15) = −2.789, P = 0.014, Cohen’s d = 0.697] condition (all others ts ≤ 1.513 Ps ≥ 0.151). The analysis also showed an ROI × hemisphere interaction [F(3, 72) = 3.133, P = 0.031, η²_P = 0.115] driven by an overall left (−0.53°) versus right (−0.27°) hemisphere difference solely for V2 [t(25) = −3.690, P < 0.001, Cohen’s d = 0.724], and a main effect of ROI [F(3, 72) = 62.843, P < 0.001, η²_P = 0.724] such that the y-position differed across all ROIs (ts ≥ 4.437 and Ps ≤ 0.001) except between IPS and V1 [t(25)–1.639, P = 0.114]. Other main effects and interactions did not reach significance (Fs ≤ 2.752 and Ps ≥ 0.110).

Overall, this analysis showed that solely in V1 following adaptation to left-shifting prims the y-center position bilaterally increased (moved further up) in the visual field over time, independently of whether pRFs were quantified during the attention or fixation task. We do not have a further interpretation of this result as we had no a-priory hypothesis regarding a shift in y-center position. Further investigations are warranted to examine the putative shift in y-center in V1.

Correlation between neural and behavioral changes

To investigate the possible relationship between neural and behavioral changes following PA we computed a Pearson correlation between the change in pRF sigma parameter and the shift in line bisection judgment as measured by the Landmark task. The analysis revealed that, across both groups, pRF sigma changes in the left hemisphere for the attention condition predicted the shift in perceptual midline judgment (r = 0.44, P = 0.025), meaning that the increase in pRF size in the left PPC following PA predicts a rightward shift as measured by the Landmark task (Fig. 5).

Fig. 5

Brain behavior correlation. Pearson correlation between the change in pRF size in the left hemisphere for the attention condition and the shift in midline judgment at the Landmark task.

pRF control measures

pRF stimuli detection

During the pRF mapping experimental run, participants performed a detection task. Performance was analyzed to ensure that the attention and fixation conditions did not differ in difficulty and that PA did not affect basic visual abilities such as visual detection. The percentage of correct detection was submitted to a mixed-model ANOVA with time (pre and post) and condition (attention and fixation) as within-subjects variables, and group (left PA and right PA) as a between-subjects variable. The analysis revealed solely a main effect of time [F(1, 24) = 7.448, P = 0.012, η²_P = 0.237], such that both the left PA and right PA group showed a common practice effect as performance improved after adaptation (from 95 to 97%). Other main effects and interactions did not reach significance (Fs ≤ 0.155 Ps ≥ 0.697). This analysis confirmed that the attention and fixation condition did not differ in difficulty level, that PA did not alter visual detection abilities, and that both groups improved detection with practice.

Eye tracking analysis

Given that attending to the stimulus could result in a larger spread of gaze position (which would increase pRF size) and since PA might affect eye position according to the direction of the lateral displacement, the standard deviation of the eye × position was submitted to a mixed-model ANOVA with time (pre and post) and condition (attention and fixation) as within-subjects variables and group (left PA and right PA) as between-subjects variable. The results for those participants for whom eye tracker data was not lost during pRF mapping (left PA = 8 and right PA = 4) did not reveal any significant main effects or interactions (Fs ≤ 1.414 all Ps ≥ 0.262), meaning that there was no difference in accuracy in fixating the central cross across conditions and that neither left nor right PA altered it.

Discussion

The aim of the current investigation was to test whether spatial representation within the PPC is modulated in response to manipulation of the visual input induced by visuomotor adaptation. After having confirmed that PA altered behavioral performance, changes in the PPC and early visual cortex were quantified by the elliptical pRF model before and after left or right PA. Based on behavioral modulation produced by left PA in healthy individuals (Colent et al. 2000; Schintu et al. 2014) and attentional modulation of pRF size in the PPC (Sheremata and Silver 2015) we anticipated left PA to produce a rightward visuospatial bias by increasing the representation of space (i.e., pRF size) in the left PPC, since any change in the right PPC would impact both visual fields (Mesulam 1981). In agreement with the prediction, the results showed that left PA increases pRF size in the left PPCs when attention is deployed, therefore increasing the integration of visual space in the left hemisphere, and thus possibly allocating more attention to the right side of space.

Behavioral changes

The effect on visuospatial behavior suggested that PA-induced a significant rightward bias as measured by the Landmark task. At baseline participants exhibited the typical leftward bias in the subjective midline judgment (i.e., pseudoneglect Jewell and McCourt 2000), which then significantly shifted rightward after PA, and exploratory paired comparison suggested that such modulation was driven by the left PA group (Fig. 2). It is possible that the group by time interaction failed to reach significance because of the fluctuating nature of the visuospatial bias induced by left PA (Schintu et al. 2014), or because the cognitive effect decreased its power by the time it was tested at the very end of the experiment, which was over 40-min long (Schintu et al. 2014). It could also be possible that the cognitive effect worn off earlier as compared with other studies, due to the fact that subjects may have received a considerable amount of sensory and proprioceptive feedback, which could have speeded up the de-adaptation procedure since they were constantly engaged in a visual task and some of them answered by using the hand that was also used for pointing during PA (instead of feet like in Schintu et al. 2014, 2017, 2020a, 2020b). The absence of a visuospatial bias modulation following right PA is not surprising given the established observation that it fails to produce significant cognitive changes (Schintu et al. 2017) while producing neural changes (Crottaz-Herbette et al. 2014; Schintu et al. 2020b). Findings in the literature reporting the absence of cognitive modulation in healthy individuals following adaptation to right-shifting prisms has not been discussed but rather accepted as a fact, so much so that right PA has become the control condition for the left one. The absence of a significant modulation at the manual line bisection task in both groups was not surprising since this is an ideal screening tool for neglect’s symptoms but its sensibility to detect visuospatial alteration in healthy is lower than the Landmark task (McIntosh et al. 2019; Schintu et al. 2020a). When considering the open-loop and straight-ahead pointing tasks, the left PA group showed the expected rightward bias, and the right PA group showed the expected leftward bias. Both groups were significantly adapted until the end of the experiment as evidenced by the two pointing tasks’ performance (Fig. 2), ensuring that while the neural data were acquired participants were in a state of sensorimotor adaptation. The groups’ sensorimotor and proprioceptive aftereffect did not differ (Fig. 2), ensuring that any between groups difference, at both behavioral and neural level, could not be ascribed to a difference in the adaptation strength produced by the visual shift direction.

Neural changes

Changes in spatial representation were measured within the PPC, given this region’s retinotopically organized topography (Silver and Kastner 2009) and its sensitivity to changes in spatial representation following attentional modulation (Sheremata and Silver 2015). Importantly, here, we further restricted our analysis to the three visuospatial subregions, IPS 0–1–2, which have been shown to affect visuospatial performance when perturbed by TMS (Szczepanski and Kastner 2013).

At the center of our investigation was the question of whether spatial representation in the PPC, as measured by pRF mapping, would show dynamic updating to reflect visuospatial adaptation. Consistent with the prediction, findings reveal changes in pRF size in the left hemisphere following left PA. We also observed, for the same hemisphere, that pRF size changes following right PA were in the opposite direction and close to reaching significance (Fig. 3), possibly due to a smaller sample size. Moreover, pRF changes in the PPC correlated with the PA-induced visuospatial bias (Fig. 5). Confirming the hypothesis that PA affects attentional allocation, changes in pRF sizes were observed exclusively in the attention, not fixation, condition. Namely, pRF size for the attention condition increased after left PA, meaning that, in agreement with previous findings (Sheremata and Silver 2015), pRF size was modulated solely when attention was deployed toward the mapping stimulus. Importantly, the pRF modulation in the attention condition cannot be attributed to a difference in task difficulty across conditions, as both groups improved their performance after adaptation independently of condition and direction of the visual displacement. Such pRF modulation cannot be explained either by eye movement since the variability of the x-position, measured via eye tracker during the experimental sessions, did not differ across time, conditions, and groups. Furthermore, the absence of a significant shift on the pRF x-center position suggests that PA does not affect the pRF preferred visual field location and could be interpreted as further support for PA action over the visual field extent (i.e., pRF size).

Unlike the PPC, early sensory V1–V3 regions did not show a consistent pRF size change following visuomotor adaptation. As shown in Fig. 4, independent of condition (attention vs. fixation) and hemisphere, pRF size in the post PA session was larger for the right PA as compared with the left PA group, but such difference was present also in V3 at baseline. On one hand, if the changes induced by the visuomotor manipulation are attentional in nature, pRF changes following left PA would be expected in early visual areas given that attention enhances the neural response of the early visual areas by prioritizing regions of space (Kastner 1998). On the other hand, if attentional and visuomotor adaptation had been observed in early sensory areas our results would be open for an interpretation that visual changes, not attentional ones, driven by sensorimotor adaption are the ones responsible for reorganization in the PPC. Strong conclusions based on the data derived from V1 to V3 would be premature, thus we elected to report the data, and rather than making conclusions, offer possible interpretations. The presence of pRF modulation following left PA in the PPC and its absence in early visual areas provides insight into the possible mechanism responsible for such spatial representation changes following manipulation of the visual input. The fact that changes in the PPC do not affect visual areas as previously shown (Kastner 1998) allows us to speculate that PA-induced changes occur at a higher level of the neural processing hierarchy. It also suggests that visuomotor adaption following PA not only induces attentional changes but also locally modulates spatial selectivity (i.e. size of the receptive field). One has to be careful, however, as not to overinterpret null results. Left PA was also found to modulate the y-center position but solely in V1 and independently of whether pRFs were quantified during the attention or fixation task. This result is unexpected and further investigations are warranted to examine the putative y-center shift in V1.

Our findings show that spatial modulation of visual input significantly altered space representation in the PPC and is consistent with Crottaz-Herbette et al. (2014) conclusion. However, their findings show an increase of blood oxygen level-dependent (BOLD) signal after right prism in the left inferior parietal lobule (IPL) while subjects were performing a visual search task. It is possible that different (i) techniques employed (pRF vs. event-related fMRI) and (ii) level of attentional deployment required, along with (iii) brain locations analyzed (IPS 0–1–2 vs. IPL) may account for this discrepancy. The same reasons could explain the difference between the local pRF modulation reported here and the more widespread change in connectivity measured at rest (Schintu et al. 2020b), which is, nonetheless calling further investigation to link local and global changes within the PPC in the context of spatial representation.

It is also worth mentioning that our predictions were based on the only other study (Sheremata and Silver 2015) that investigated pRFs change following attentional modulation with findings supporting the model of attention proposed by Mesulam (1981), and thus did not take into consideration the possible role of the asymmetry in PPC-to-PPC communication that could have affected our results (Koch et al. 2011; Gigliotta et al. 2017). Koch and collaborators (2011) showed that the PPCs exert asymmetric inhibitory control over each other with a stronger right-to-left than left-to-right inhibition. Grounding on the Rivalry theory (Kinsbourne 1977) and based on evidence in neglect patients, right PA could ameliorate neglect symptoms similarly to inhibitory TMS over the left PPC (Koch et al. 2008), that is to say by decreasing left PPC “activation.” Such decrease of the left PPC activation would also decrease the amount of inhibition that the left PPC exerts over the right one. Since in neglect patient the left PPC has been found hyperactive (Koch et al. 2008), a decrease in its activation level would restore the interhemispheric balance. Such change in the amount of left-to-right inhibition driven by right-PA would produce cognitive changes in neglect patients since their right PPC is lesioned and thus the right-to-left inhibition is compromised. In healthy individuals, instead, such decrease in left-to-right interhemispheric inhibition would have a null effect given the already strong (maybe at ceiling) right-to-left inhibition. The fact that only one study has been able to successfully induce a leftward bias in healthy individuals by inhibiting the left PPC with TMS (Szczepanski and Kastner 2013), supports this speculation that, however, as such deserves further investigation.

Finally, when pRF changes in the PPC were also investigated by analyzing the calculated size parameter (see supplementary material) instead of sigma alone, significant changes following left PA were present in both the left and right PPC, no trend toward significance was observed for the right PA group (Supplementary Fig. 1) along with no significant correlation between neural and cognitive modulation. The sigma-based results are mostly compatible with the calculated size measure (Fig. 3 and Supplementary Fig. 1), as we can visually appreciate that in both cases there is an increase of pRF size following left-PA. However, the sigma parameter appears to show more sensitivity to the visuomotor adaptation induced modulation than the calculated size one, possibly meaning that most of the pRF modulation happens along the ellipse shorter axis. Further experiments utilizing more in-depth pRF methods (i.e.,  symmetric vs. asymmetric pRF estimates) are warranted to adjudicate between different pRF parameters and models (Dumoulin and Wandell 2008; Silson et al. 2018).

In conclusion, here we present the first evidence of the dynamic nature of spatial representation in the PPC following visuomotor adaptation as measured with the pRF method. We show that using pRF mapping of the PPC is not only sensitive enough to detect attentionally induced modulations (Sheremata and Silver 2015) but also allow us to link fMRI signal in the visual pathway to neuronal receptive field at the single voxel level (Dumoulin and Wandell 2008). Our results have important implications for understanding the neural instantiation of space representation in the PPC and offer valuable insights into the PA’s mechanism of action and its usefulness as a rehabilitation tool for neurological disorders exhibiting deficits in spatial allocation, such as neglect syndrome.

Funding

This work was supported by the National Institutes of Health Ruth L. Kirschstein National Research Service Award (to Schintu), Biotechnology and Biological Sciences Research Council (BB-V003917-1 to Silson), the Clinical Neurosciences Program of the National Institute of Neurological Disorders and Stroke (1ZIANS002977-20 to Wassermann), and the National Science Foundation (BCS-1921415 to Shomstein and BCS-2022572 to Shomstein & Kravitz).

Conflict of interest statement

The authors declare no competing financial interests.