Right inferior frontal contributions to visual spatial suppression and visuomotor performance

Abstract

Background

Elite racket sports demand visuomotor responses at the millisecond level, requiring rapid prioritization of task-relevant information while suppressing competing or redundant signals. Such perceptual filtering is reflected in spatial suppression, which declines with aging and in clinical cognitive disorders. This study investigated whether visual spatial suppression in expert table tennis players (TTPs) was linked to visuomotor reaction time, and whether prefrontal–visuomotor dynamics were associated with this link.

Methods

We examined 28 expert TTPs and 31 healthy controls using 7-Tesla MRI and psychophysical methods. Specifically, behavioral measures included visual spatial suppression and visuomotor reaction time (RT); an attentional cueing paradigm was used to quantify criterion shifts. Brain measures included gray matter volume (GMV) and static versus dynamic functional coupling between the right inferior frontal gyrus (rIFG) and visuomotor cortex.

Results

TTPs showed faster visuomotor reaction times and stronger spatial suppression than healthy controls (P < 0.001 and P = 0.007, respectively). Within TTPs, faster RTs were associated with stronger suppression. Structurally, GMV in rIFG tracked suppression and attentional criterion shifts (Δc: 0.371 ± 0.086 vs 0.034 ± 0.079; P = 0.006). Functionally, dynamic (but not static) coupling between rIFG and the left postcentral gyrus (lPoCG) was associated with suppression strength (large-grating thresholds, partial r = 0.595, P = 0.002) and visuomotor RT (partial r = −0.436, P = 0.033) and statistically mediated the relationship between suppression and RTs.

Conclusion

This study demonstrates the association between visuomotor reaction speed and spatial suppression in TTPs, and identifies rIFG–lPoCG dynamics as a systems-level index linking these perceptual and behavioral advantages. These cross-sectional findings suggest that perceptual filtering and rapid visuomotor behavior are correlated in expert performance.

Introduction

Detecting and isolating task-relevant objects from complex and dynamic environments is a crucial visual function. For example, when a driver encounters a car cutting abruptly into their lane, the visual system must rapidly prioritize processing the motion of that vehicle while attenuating irrelevant motion in the wider scene. Such perceptual filtering is reflected in spatial suppression, a behavioral manifestation of center–surround interactions in which broad, high-contrast motion signals in the periphery are downweighted (Tadin et al., 2003, 2019; Tadin, 2015). This enables more precise discrimination of local motion (Tadin et al., 2003, 2019; Tadin, 2015). Spatial suppression declines with age and in clinical conditions such as schizophrenia, autism, and major depressive disorder (Betts et al., 2005; Tadin et al., 2006; Schallmo et al., 2020; Song et al., 2021, 2024). Therefore, it is essential to investigate the mechanisms underlying spatial suppression.

Spatial suppression has been observed across multiple stages of the visual system, from early visual processing to the primary visual cortex (V1) and extrastriate areas such as the middle temporal visual area (MT/V5; also referred to as human MT complex, hMT+) (Tadin et al., 2003; Zenger-Landolt and Heeger, 2003; Zhang et al., 2014; Nguyen and McKendrick, 2016; Anderson et al., 2017). In particular, hMT+ contains dense direction selective neurons and is motion and velocity-sensitive. However, despite extensive characterization of the mechanisms underlying spatial suppression in occipital and temporal regions (Newsome and Pare, 1988; Born and Tootell, 1992; Tadin et al., 2011; Liu et al., 2016; Angelucci et al., 2017), the contribution of frontal control systems remains largely unexplored. The inferior frontal gyrus (IFG), a central node of the cognitive-control network, is implicated in motor inhibition, attentional reweighting, and suppression of irrelevant sensory information (Miller and Cohen, 2001; Corbetta and Shulman, 2002; Chambers et al., 2006; Hampshire et al., 2010; Aron et al., 2014; Zhuang et al., 2023). Demonstrating a role for IFG in visual surround suppression would extend inhibitory principles across perception, cognition, and action (Melnick et al., 2013; Tadin, 2015; Li et al., 2023a; Gao et al., 2024).

Expert athletes in fast ball sports offer a natural framework for studying motion discrimination and visual spatial suppression. Table tennis players (TTPs) operate under millisecond demands on visuomotor responses, requiring rapid selection of task-relevant signals while filtering distracting background motion in dynamic visual environments. Prior work has documented visual processing advantages in TTPs, including faster neural processing of visual motion. This is indexed by event-related potentials (ERPs) and superior motion discrimination against background noise (Hülsdünker et al., 2019; Goya et al., 2025). Converging evidence also suggests altered prefrontal engagement, enhanced attentional control, and faster visuomotor responses in TTPs (Wang et al., 2016, 2019; Guo et al., 2017; Zhu et al., 2022; Li et al., 2023b). Despite these findings, it remains unclear whether the noise-resilient perception observed in TTPs is related to visual spatial suppression. It is also unknown whether such suppression is primarily associated with early visual circuitry or instead emerges from distributed higher-order frontal and somatosensory control systems.

In this study, we used psychophysics and 7-Tesla magnetic resonance imaging (7T MRI) to examine whether the IFG contributes to visual spatial suppression in expert TTPs. We hypothesized that elite TTPs would show enhanced spatial suppression accompanied by IFG-related neural adaptations linking perceptual filtering with visuomotor performance. We report four main findings. First, TTPs exhibit enhanced spatial suppression, reflected by elevated thresholds for large, surround-sensitive gratings. Second, stronger suppression correlates with faster visuomotor responses in athletes. Third, right IFG (rIFG) morphology is associated with both suppression strength and attentional criterion shifts. Finally, dynamic coupling, but not static connectivity, between rIFG and left postcentral gyrus (lPoCG) tracks suppression and mediates its relationship with reaction time. Together, these findings position rIFG–lPoCG dynamics as a key mechanism linking inhibitory control across perceptual and visuomotor domains.

Methods

Participants

Based on previous studies), 59 volunteers were recruited from Zhejiang University, China, including 28 student TTPs (13 females; mean ± SD age = 21.66 ± 2.73 years; body mass index (BMI) ± SD = 22.11 ± 4.24; training duration ± SD = 12.00 ± 2.60 years; weekly practice ± SD = 5.37 ± 3.80 h) and 31 healthy controls (HCs) (12 females; mean ± SD age = 22.42 ± 2.21 years; BMI ± SD = 21.90 ± 3.08). Detailed subject recruitment information and criteria can be found in the references (Li et al., 2023b; Liu et al., 2024; Zheng et al., 2024). Groups were matched for age, sex, and BMI, and they had no significant demographic differences (all P > 0.1; Table 1). All TTPs were nationally certified first-class athletes according to the Athlete Technical Grade Standard issued by the General Administration of Sport of China; they had received more than 6 years of systematic training and were actively practicing at the time of participation. HCs had no history of regular athletic training or sports specialization. A subset of participants had been included in our previous publication; the present study includes additional participants and new analyses not reported previously. Participants completed structural and resting-state functional MRI (rs-fMRI) scans, as well as three psychophysical tasks assessing spatial suppression, visuomotor reaction time, and attentional cueing. MRI data were successfully obtained from all participants. All participants were right-handed, had normal or corrected-to-normal vision, and reported no neurological or psychiatric medical history that could affect cognitive or perceptual function. All procedures were approved by the Research Ethics Review Committee of Wuyunshan Hospital ([2021] Clinical Research Medical Ethics Review No. 004) and conducted in accordance with the Declaration of Helsinki. Written informed consent was obtained from all participants prior to participation, and each participant received 200 RMB as compensation upon completion.

Table 1.

Demographic characteristics of table tennis players (TTPs) and healthy controls (HCs).

	TTP (n = 28)	HC (n = 31)	χ2 or t	P value
Age (years)	21.66 ± 2.73	22.42 ± 2.21	1.177	0.244a
Sex (female/male)	13/15	12/19	0.359	0.549b
BMI	22.11 ± 4.24	21.90 ± 3.08	0.213	0.832a
Train time (h/week)	5.37 ± 3.80	NA	NA	NA
Duration (years)	12.00 ± 2.60	NA	NA	NA
Mean FD (mm)	0.07 ± 0.03	0.06 ± 0.02	1.721	0.091a

Data were expressed as mean values ± standard deviation (SD). BMI, body mass index; FD, frame displacement; NA, not applicable. Sample size varies by measure; see the Methods section for exclusions and task-specific sample sizes.

^aTwo-sample t-test (two-tailed).

^bχ² test (two-tailed).

Predefined exclusion criteria were applied: participants were excluded if their maximum head translation or rotation exceeded 3 mm or degrees, respectively, or if they failed to complete task calibration. Behavioral data were partially missing due to incomplete responses (grating orientation: 28/28 TTP, 29/31 HC; visuomotor reaction: 28/28 TTP, 31/31 HC; attentional cueing: 25/28 TTP, 26/31 HC).

Psychophysical tasks

Three psychophysical tasks were used to quantify spatial suppression, visuomotor speed, and attentional bias. Detailed stimulus parameters are provided in the Supplementary Information. For the grating motion direction discrimination task, participants sat 60 cm from the display and judged whether a vertically drifting sinusoidal grating moved left or right. Gratings appeared on a gray background of 56 cd m⁻² at 50% contrast and had either a 10° (large) or 2° (small) diameter; the spatial frequency was 1 cycle/° and the drift speed was 4° s⁻¹. Each block comprised 160 trials, with stimulus duration adjusted using four interleaved 3-down/1-up staircases (two per stimulus size). For each stimulus size, trials from the two interleaved staircases were pooled to compute accuracy as a function of duration. A cumulative Gaussian was then fit to these data, and the 75% correct point on the psychometric function was taken as the duration threshold (Song et al., 2021):

Participants also completed a visuomotor reaction task by standing 1 m from the display and pressing a wall-mounted button at shoulder height as soon as a red target appeared; RT was defined as the interval between stimulus onset and button press across 10 trials. Finally, in an attentional cueing paradigm with moving-dot fields oriented at 45° or 135°, a pre-stimulus arrow either indicated the probable acceleration direction (Cue) or provided no directional information (No-cue); following signal detection theory, we derived Δc, Δd′, and ΔRT by contrasting Cue versus No-cue trials.

MRI data acquisition

MRI scanning experiments, including rs-fMRI and structural MRI, were performed using a Nova Medical 32-channel head coil in a 7T whole-body magnetic resonance system (Siemens Healthcare, Erlangen, Germany). High-resolution rs-fMRI data were acquired using a multiband echo-planar imaging (EPI) sequence (University of Minnesota; cmrr_mbep2d_bold) with an isotropic voxel size of 1.5 mm³ (TR/TE = 2000/20.6 ms, flip angle = 70°, FOV = 228 × 228 mm², 90 slices, multiband factor = 5; 160 volumes). Structural MRI data were acquired using an MP2RAGE sequence with a finer isotropic voxel size of 0.7 mm³ (TR/TE = 2300/3.09 ms, flip angle = 7°, FOV = 225 × 225 mm²). During scanning, participants were instructed to close their eyes, remain awake, and minimize head motion to ensure data quality. Foam padding was used to restrict head motion, and earplugs were used to attenuate scanner noise.

Structural MRI preprocessing and analysis

Structural MRI images were analyzed using the Computational Anatomy Toolbox (CAT12) running within the Statistical Parametric Mapping (SPM12) framework in MATLAB R2019a (http://dbm.neuro.uni-jena.de/cat12/; http://www.fil.ion.ucl.ac.uk/spm/software/spm12/). Preprocessing included denoising, skull stripping, tissue segmentation (into gray matter, white matter, and cerebrospinal fluid), and normalization to standard Montreal Neurological Institute (MNI) space using the International Consortium for Brain Mapping 152 (ICBM152) template. The initial template for DARTEL registration was constructed from the average image, after which participants’ gray matter images were registered to this template; the template was iteratively refined by averaging across images. A Gaussian kernel with 3-mm full width at half maximum (FWHM) was applied to smooth the images. No participants were excluded from the final structural dataset. Voxel-based morphometry (VBM) was used to assess group differences in gray matter volume (GMV). GMV estimates from three subregions of the rIFG (rIFGoperc, rIFGtriang, and rIFGorb), defined in the Automated Anatomical Labeling 3 (AAL3) atlas, were extracted from each participant’s preprocessed structural images for subsequent statistical analyses (Rolls et al., 2020).

Rs-fMRI data preprocessing and analysis

Resting-state functional images were preprocessed using RESTplus v1.25 (Jia et al., 2019) (http://restfmri.net/forum/restplus) implemented in MATLAB R2019a (Mathworks, Natick, MA, USA). The preprocessing steps included: removal of the first five time points, realignment, coregistration of structural and functional images, segmentation of structural images using the NewSegment tool in RESTplus, normalization to standard MNI space with a voxel size of 1.5 mm³, spatial smoothing with a Gaussian kernel (FWHM = 3 mm), linear detrending, nuisance regression (including 24 head-motion parameters, white-matter signal, and cerebrospinal-fluid signal as covariates) (Friston et al., 1996; Chen et al., 2026), and band-pass filtering at 0.01–0.08 Hz. Participants who exhibited head motion exceeding 3 mm translation or 3° rotation were excluded (Tang et al., 2023; Song et al., 2024). Consequently, rs-fMRI data from 27 of 28 TTPs and all 31 HCs were retained for functional connectivity (FC) analyses.

Before time-series extraction, regions of interest (ROIs) were defined. Bilateral hMT+ were identified using a cytoarchitectonic probabilistic map (Malikovic et al., 2007) and represented by 5-mm-radius spheres centered at (−43, −73, 10) and (49, −70, 11) in MNI space. After preprocessing, the left and right hMT+ served as seed regions for computing static and dynamic functional connectivity (sFC and dFC) across the whole brain using RESTplus v1.25. For sFC, Pearson’s correlation coefficients (r) were calculated between the mean time series of each seed and every voxel in the brain (Bandettini et al., 1993). The resulting correlation maps were converted to Fisher z-transformed values to yield individual sFC maps for group-level analysis. For dFC, we followed previous studies (Li et al., 2023b; Liu et al., 2024; Zheng et al., 2024) and used a sliding-window approach (Leonardi and Van De Ville, 2015) with a window length of 50 TRs and a 1-TR step size, generating 106 overlapping windows. Within each window, voxelwise FC maps were computed as in the sFC analysis. The temporal variability of FC strength across all windows was then quantified using the PerAF method to obtain individual whole-brain dFC maps (Li et al., 2023b; Liu et al., 2024).

After group-level sFC and dFC analyses, we further examined the rIFGtriang region. Using whole-brain analyses seeded in both left and right hMT+, we defined the rIFGtriang ROI as the union of voxels showing differences between the TTP and HC groups in sFC or dFC. In our data, this union coincided with the sFC cluster. We then used this rIFGtriang ROI to compute whole-brain sFC and dFC following the same procedures described above.

Statistical analysis

SPSS 26 (IBM, USA) was used to examine differences in demographic information between the TTP and HC groups. A χ² test was performed to compare sex, and a two-sample t-test was used to compare age, BMI, and mean framewise displacement (FD). Differences were considered not significant when P > 0.05.

For behavioral data, we used SPSS 26 (IBM, USA) to test for between-group differences and perform correlation analysis. For duration thresholds of small and large stimuli, a two-way ANOVA was used to examine the effects of stimulus size and group. When the effect was significant, post hoc tests were then conducted. For other behavioral indicators, two-sample t-tests were performed to examine the differences between the TTP and HC groups. Controlling for age, sex, and BMI, partial correlation was used to analyze the relationship between behavioral indicators. In this part specifically, the partial correlation only included the behavioral data of participants who had complete data for all three psychophysical tasks.

For GMV, two-sample t-tests were used to test for between-group difference analysis of GMV estimates in three subregions of the rIFG (rIFGoperc, rIFGtriang, and rIFGorb) in the AAL3 atlas using SPSS 26 (IBM, USA). Partial correlation analysis was used to analyze the associations between GMV in three subregions of the rIFG and spatial suppression indicators (SI, small grating thresholds, and large grating thresholds) as well as visuomotor reaction times, controlling for age, sex, BMI, and total intracranial volume (TIV), with multiple comparisons corrected using the false discovery rate (FDR) method. Subsequently, after determining the correlation between duration thresholds for large stimuli and GMV of the rIFGtriang, partial correlation analysis was used to calculate the associations between GMV of the rIFGtriang and behavioral indicators of the attentional cuing task (Δc, Δd′, and ΔRT) with age, sex, BMI, and TIV as covariates.

For sFC and dFC, the two-sample t-test was used for between-group comparisons in RESTplus v1.25, with covariates (age, sex, and BMI) regressed out. Multiple comparisons were controlled with Gaussian random field (GRF) correction for voxel-level P < 0.05 and cluster-level P < 0.05. To assess the relationship between behavioral indicators and FC measures, we first extracted the sFC and dFC values of each individual in both groups using RESTplus v1.25. Then, we calculated their partial correlation coefficients with all behavioral indicators using SPSS 26, controlling for age, sex, and BMI. Differences or correlations were considered statistically significant if P < 0.05.

PROCESS v3.4, a toolbox in SPSS 26, was used to conduct the mediation model. Based on the results obtained through partial correlations, we examined whether the association between visuomotor response ability and spatial suppression was mediated by dFC between the rIFGtriang and the lPoCG, controlling for age, sex, and BMI. A bias-corrected bootstrap method was used to evaluate the significance of the direct and indirect effects, in which a 95% confidence interval (CI) was calculated with 5000 bootstrap resamples. If the bootstrap 95% CI of the indirect effect did not contain zero, the indirect effect was considered significant in the model. All statistical tests were two-tailed unless otherwise noted.

Results

Spatial suppression is enhanced in table tennis players

We first quantified spatial suppression using a motion direction discrimination task in which large (10°) and small (2°) drifting gratings were interleaved across 160 trials (Fig. 1A). Thresholds for each stimulus size and the resulting spatial suppression index (SI) are shown in Fig. 1B and C. Thresholds were estimated using interleaved 3-down/1-up staircases. After excluding two control participants with non-converging staircases, the final sample for the motion discrimination task consisted of 28 TTPs and 29 HCs (Table 1; see the Methods sectionfor sample sizes by measure). A two-way mixed-effects ANOVA revealed significant main effects of group (F₁,₅₅ = 7.134, P = 0.010, partial η² = 0.115) and stimulus size (F₁,₅₅ = 125.602, P < 0.0001, partial η² = 0.695), as well as a significant group × size interaction (F₁,₅₅ = 6.876, P = 0.011, partial η² = 0.111). Post hoc analyses showed that TTPs exhibited substantially higher motion-discrimination thresholds for large, surround-sensitive gratings (TTPs: 6.579 ± 0.382 ms; HCs: 5.317 ± 0.240 ms; P = 0.007, Cohen’s d = 0.747; Fig. 1B). Meanwhile, thresholds for small central stimuli did not differ between groups (TTPs: 3.832 ± 0.130 ms; HCs: 3.612 ± 0.125 ms; P = 0.226, Cohen’s d = 0.325; Fig. 1B). Under high-contrast conditions, increasing the size of a moving stimulus impairs, rather than improves, directional motion perception (Tadin et al., 2003). As stimulus size increases, it extends into the surroundings and increases broader suppressive signals, leading to stronger spatial suppression (Schallmo et al., 2018)). In this paradigm, stronger spatial suppression selectively impairs motion discrimination for large, surround-sensitive gratings, making motion direction more difficult to perceive. As a result, participants with stronger spatial suppression require longer stimulus durations to process accurate motion discrimination. This leads to higher duration thresholds for large gratings. Consequently, SI was significantly higher in TTPs (TTPs: 0.223 ± 0.023; HCs: 0.163 ± 0.015; t_46.23 = 2.175, P = 0.035, with Welch’s correction, Cohen’s d = 0.580; Fig. 1C).

Figure 1.

Behavioral analysis of motion spatial suppression and visuomotor speed. (A) Task schematic illustrating the motion spatial suppression paradigm. Participants including table tennis players (TTPs) and healthy controls (HCs) determined the direction of briefly presented drifting gratings of either large (10°) or small (2°) size. (B) Motion discrimination thresholds for large and small gratings in TTPs and HCs (Bonferroni-corrected post hoc test). (C) The resulting spatial suppression index (SI), defined as the log difference between thresholds for large and small gratings (two-sample t-tests). (D) SI correlated positively with the large-grating threshold in both groups. (E and F) Visuomotor reaction times (RTs) and their relationship with suppression strength. Correlations were calculated while controlling for age, sex, and body mass index (BMI). Error bars represent the standard error of the mean. Lines represent linear fits and shaded ribbons represent the 95% confidence interval (CI) of the regression lines. *P < 0.05, **P < 0.01, ***P < 0.001, ^****P < 0.0001, ^nsP > 0.05.

To further examine the source of this enhancement, we conducted partial correlations while controlling for age, sex, and BMI. SI scaled tightly with large-grating thresholds (Fig. 1D; TTPs: partial r = 0.761, 95% CI [0.523, 0.889], P < 0.0001; HCs: partial r = 0.617, 95% CI [0.302, 0.811], P = 0.003), but not with small-grating thresholds (Fig. S1). This indicated that expertise selectively enhanced surround-dependent processing. This pattern is consistent with strengthened center–surround interactions in the visual system of trained athletes.

Stronger spatial suppression is associated with faster visuomotor responses

Visuomotor reaction times (RTs) and their relationship with suppression strength are shown in Fig. 1E and F. In a cued visuomotor task, participants responded to lateralized visual targets by pressing a wall-mounted button positioned at arm’s length. TTPs responded significantly faster than HCs (TTPs: 0.533 ± 0.010 s; HCs: 0.604 ± 0.014 s; t_51.5 = −4.130, P < 0.001, with Welch’s correction, Cohen’s d = −1.054; Fig. 1E). Within the TTP group, SI values were inversely associated with RTs (partial r = −0.428, 95% CI [−0.704, −0.040], P = 0.047; Fig. 1F). Critically, this relationship indicated a correlation between athletes with stronger spatial suppression and shorter RTs. A similar, though nonsignificant, negative trend was observed between RT and large-grating thresholds (partial r = −0.295, 95% CI [−0.618, 0.113], P = 0.182), while small-grating thresholds were not associated with RT (partial r = 0.190, 95% CI [−0.222, 0.544], P = 0.397; Fig. S2a and b). All correlations were estimated while controlling for age, sex, and BMI. This association was absent in HCPs: SI and RT were not significantly related (partial r = −0.093, 95% CI [−0.464, 0.305], P = 0.689; Fig. 1F), nor were RTs associated with large- or small-grating thresholds (all P > 0.1; Fig. S2a and b). Together, these findings suggest that TTPs may have a stronger linkage between perceptual filtering changes and action execution.

Right IFG morphology supports perceptual and attentional inhibition

We examined whether structural features of the right IFG (Fig. 2A), a region associated with inhibitory and attentional processing (Aron et al., 2004; Chambers et al., 2006; Hampshire et al., 2010; Zanto et al., 2011), were associated with spatial suppression. GMV was extracted from three IFG subdivisions: right pars opercularis (rIFGoperc), right pars triangularis (rIFGtriang), and right pars orbitalis (rIFGorb). Among TTPs, rIFGtriang GMV significantly predicted large-grating thresholds (partial r = 0.514, 95% CI [0.140, 0.760], P = 0.030; Fig. 2B). This relationship was specific to rIFGtriang; no associations were observed for rIFGoperc (TTPs: partial r = 0.150, 95% CI [−0.270, 0.522], P = 0.484; Fig. 2C) or rIFGorb (TTPs: partial r = 0.199, 95% CI [−0.222, 0.558], P = 0.525; Fig. 2C). In HCs, none of the IFG subdivisions showed associations with suppression (rIFGtriang: partial r = −0.052, 95% CI [−0.438, 0.350], P = 0.929; rIFGoperc: partial r = 0.031, 95% CI [−0.369, 0.421], P = 0.929; rIFGorb: partial r = −0.019, 95% CI [−0.411, 0.379], P = 0.929; Fig. 2B and C). The P values in the above analysis were corrected by an FDR method.

Figure 2.

Associations between right inferior frontal gyrus triangular subregion (rIFGtriang) gray matter volume (GMV) and behavioral measures. (A) Subregions of the right inferior frontal gyrus (rIFG), including opercular (rIFGoperc), triangular (rIFGtriang), and orbital (rIFGorb). (B and C) Relationships between rIFG subregions GMV and duration thresholds for large motion gratings in TTPs and HCs. (D) Attentional cueing paradigm. Participants judged whether a brief acceleration of random-dot motion stimuli occurred in a specified direction (ISI, inter-stimulus interval; ITI, inter-trial interval). (E) Resulting group differences in criterion shifts (Δc) (two-sample t-test, two-tailed). (F) Correlation between rIFGtriang GMV and Δc. Correlations were calculated while controlling for age, sex, body mass index (BMI), and total intracranial volume (TIV). Error bars represent the standard error of the mean. Lines represent linear fits and shaded ribbons represent the 95% confidence interval (CI) of the regression lines. *P < 0.05, **P < 0.01.

To determine whether rIFGtriang also contributed to attentional processing, we examined performance in a speed-up cueing paradigm (Fig. 2D). Here, a directional cue either predicted an upcoming acceleration or provided no useful information. Signal-detection analyses revealed that TTPs demonstrated larger cue-induced changes in decision criterion shifts (Δc) compared with HCs (TTPs: 0.371 ± 0.086; HCs: 0.034 ± 0.079; t₄₉ = 2.891, P = 0.006, Cohen’s d = 0.810; Fig. 2E), indicating greater flexibility in altering perceptual decision boundaries, such as top-down adjustment. Within TTPs, but not HCs, rIFGtriang GMV predicted the magnitude of this criterion shift (partial r = 0.483, 95% CI [0.065, 0.757], P = 0.026; Fig. 2F). No significant correlations were observed for sensitivity change (Δd′) or reaction-time change (ΔRT) in either group (all P > 0.4; Fig. S3). Together, these findings suggest that rIFGtriang morphology anchors individual differences in both perceptual and attentional inhibition among trained athletes.

Dynamic IFG–lPoCG coupling predicts suppression strength

To identify functional pathways linking IFG to spatial suppression, we computed dFC using sliding-window correlations. Seed regions included bilateral human middle temporal complex (hMT+) and the rIFGtriang cluster identified from earlier analyses. Whole-brain comparisons indicated widespread group differences: TTPs exhibited increased dFC between rIFGtriang and frontal, temporal, parietal, and occipital association cortices (voxel-level P < 0.05, cluster-level P < 0.05; Figs S4a and S5a and Tables S1 and S2), but decreased dFC with bilateral postcentral gyri (lPoCG: TTPs, 0.875 ± 0.089, HCs, 2.937 ± 0.397, t_32.96 = −5.066, P < 0.0001, with Welch’s correction, Cohen’s d = −1.250; rPoCG: TTPs, 1.086 ± 0.110, HCs, 3.690 ± 0.560, t_32.18 = −4.565, P < 0.0001, with Welch’s correction, Cohen’s d = −1.125; Fig. 3B and C; Fig. S5a and Table S4)—which are primary somatosensory and sensorimotor areas.

Figure 3.

Dynamic (dFC) and static functional connectivity (sFC) between rIFGtriang and the postcentral gyrus (PoCG) and its behavioral relevance. (A) Schematic of the sliding-window analysis used to calculate dFC and sFC. (B) Cortical map of group differences in dFC between rIFGtriang and left and right PoCG (lPoCG, rPoCG). Shading indicates t-statistics for table tennis players (TTPs) versus healthy controls (HCs). (C) Group differences in dFC between rIFGtriang and bilateral PoCG (two-sample t-tests, two-tailed). (D–F), Behavioral relevance of rIFGtriang–PoCG dFC, showing associations with large-grating thresholds (D), spatial suppression index (SI; E), and visuomotor reaction times (RTs; F) (main plots show lPoCG; insets show rPoCG). (G), Group differences in sFC between rIFGtriang and bilateral PoCG. (H–J), Behavioral relevance of rIFGtriang–PoCG sFC, showing associations with large-grating thresholds (H), SI (I), and RTs (J). Correlations were calculated while controlling for age, sex, and body mass index (BMI). Error bars represent the standard error of the mean. Lines represent linear fits and shaded ribbons represent the 95% confidence interval (CI) of the regression lines. *P < 0.05, **P < 0.01, ^****P < 0.0001, ^nsP > 0.05.

Because sensorimotor regions are a known site for convergence between perceptual and motor processes, we focused on rIFGtriang–PoCG interactions. Within TTPs, stronger rIFGtriang–left PoCG (lPoCG) dFC was associated with higher large-grating thresholds (partial r = 0.595, 95% CI [0.252, 0.805], P = 0.002; Fig. 3D), greater SI values (partial r = 0.511, 95% CI [0.136, 0.758], P = 0.011; Fig. 3E), and faster visuomotor RTs (partial r = −0.436, 95% CI [−0.714, −0.040], P = 0.033; Fig. 3F). These relationships were not observed for rIFGtriang–right PoCG (rPoCG) coupling and did not emerge in HCs (Fig. 3D–F). The specificity to the left PoCG aligns with the predominantly right-handed motor behavior of participants and suggests a role for cross-hemispheric cortical coordination. Notably, static (time-averaged) connectivity showed no relationship with suppression or RT (Fig. 3G–J; Figs S4b and S5b and Tables S3 and S4), suggesting that temporal fluctuations are more associated with behavior than average coupling strength.

IFG–lPoCG dynamics mediate the relationship between suppression and reaction time

Given that rIFGtriang–lPoCG dFC was associated with both suppression and reaction time in TTPs, we performed mediation analyses to test whether connectivity statistically accounted for the association between perceptual inhibition and visuomotor performance. The mediation model showed that rIFGtriang–lPoCG dFC significantly mediated the relationship between reaction time and large-grating thresholds (Fig. 4). Faster reaction times predicted stronger dynamic coupling, which in turn predicted enhanced spatial suppression. When the mediation direction was reversed, the indirect effect was no longer significant (Fig. S6), suggesting directional specificity. Although mediation models do not establish causal relationships, these findings support the idea that dynamic rIFGtriang–lPoCG interactions are associated with the integration of perceptual and motor processes in TTPs.

Figure 4.

Mediation of the spatial suppression–visuomotor association by rIFGtriang–lPoCG dynamic coupling in table tennis players (TTPs). Statistical mediation model indicating that rIFGtriang–lPoCG dynamic functional connectivity (dFC) is statistically consistent with the relationship between visuomotor reaction time (RT) and spatial suppression. Path a represents the association between visuomotor RT (predictor) and rIFGtriang–lPoCG dFC (mediator), and path b represents the association between rIFGtriang–lPoCG dFC and large-grating threshold (outcome). Paths c and c′ denote the total and direct effects of RT on large-grating thresholds before and after accounting for the mediator. The analysis showed a significant indirect pathway (a × b), whereas the direct effect (c′) was not significant. Coefficients a, b, c, and c′ are unstandardized regression path coefficients. CI, 95% confidence intervals; se, standard errors. The effect ratio denotes the proportion of the total effect carried by the indirect pathway. The model controlled for age, sex, and body mass index (BMI). *P < 0.05, **P < 0.01.

Discussion

This study provides new evidence that visual spatial suppression is not solely a property of early visual cortex, but is also associated with rIFG–lPoCG dynamic coupling. We report four main findings. First, TTPs exhibited enhanced spatial suppression, which was associated with faster visuomotor responses. Second, rIFG morphology was associated with both suppression strength and attentional criterion shifts. Third, at the neural systems level, rIFG-related structural and functional markers indicated expertise-related differences were associated with suppression strength and visuomotor performance. Finally, dynamic coupling between rIFGtriang and lPoCG emerged as a key feature linking visuomotor speed and spatial suppression. Together, these findings support the view that IFG-centered dynamics provide a systems-level mechanism underlying spatial suppression within a broader inhibitory control architecture.

Although the role of the rIFG in inhibitory and attentional control has been extensively documented (Corbetta and Shulman, 2002; Aron et al., 2004; Hampshire et al., 2010), considerably less is known about its contribution to visual spatial suppression during rapid visuomotor performance. Elite TTPs provide a particularly suitable model for addressing this question because they undergo prolonged training in an open-skill fast-ball sport that requires continuous visual processing and rapid visuomotor transformations. Our findings extend the established role of the rIFG by demonstrating that rIFG-related structural and dynamic functional adaptations are associated with visual spatial suppression and superior visuomotor performance in this expert population. The rIFG is commonly divided into three anatomically distinct subregions: the pars opercularis, pars triangularis, and pars orbitalis (Boen et al., 2022). Previous functional neuroimaging studies suggest a posterior-to-anterior functional gradient within the rIFG, with posterior regions being associated with action execution and response inhibition, and more anterior regions being involved in increasingly abstract cognitive processes, including reasoning and cognitive control (Hartwigsen et al., 2019). In the present study, all three rIFG subregions were examined using the same analytical framework. However, only the rIFGtriang showed significant structural and dynamic functional associations with visual spatial suppression. Variations in rIFGtriang morphology were associated with individual differences in both suppression strength and attentional criterion shifts, which are key components of inhibitory control (Hampshire et al., 2010; Lennert and Martinez-Trujillo, 2011; Aron et al., 2014; Brown et al., 2023). While traditional models attribute visual spatial suppression primarily to computations in MT/hMT+ and surrounding extrastriate regions (Tadin et al., 2003, 2011; Liu et al., 2016; Angelucci et al., 2017), spatial suppression is understood not as a nonspecific gain change, but as an efficient center–surround regulation mechanism. At the visual level, hMT+ functions as a motion-selective hub that integrates local motion signals and contributes to surround suppression. Our findings extend this framework by demonstrating that prefrontal control mechanisms, particularly within the rIFG, contribute to spatial suppression. Importantly, spatial suppression should not be viewed as an isolated sensory phenomenon, but rather as a behaviorally relevant component of visuomotor optimization. The convergence of perceptual, attentional, and structural findings implicates rIFGtriang as an important contributor to modulating center–surround computations.

Beyond the contribution of rIFGtriang itself, our results reveal that dynamic interactions between rIFGtriang and lPoCG provide an additional mechanism linking visual spatial suppression with visuomotor performance. Furthermore, only the dFC between the rIFGtriang and left PoCG significantly mediated the association between visuomotor RT and visual spatial suppression. This perceptual–behavior relationship was supported by moment-to-moment coordination between prefrontal control regions and posterior visuomotor-related cortex, suggesting a dynamic link between perceptual inhibition and visuomotor behavior. Notably, static connectivity measures were not related to suppression or performance, highlighting the importance of temporal flexibility to prefrontal–visuomotor communication. Sliding-window dFC has been increasingly applied to characterize temporal variations in FC (Liegeois et al., 2017). In the present study, the observed rIFGtriang–lPoCG dFC was anatomically specific, associated with higher large-grating thresholds, spatial suppression, and visuomotor RT, whereas sFC did not show comparable behavioral associations. Together with the convergent structural MRI findings, these results support the biological relevance of the identified dFC pattern. Together, these findings suggest that visual surround suppression emerges from coordinated interactions across multiple cortical systems rather than from a single sensory locus, reflecting a systems-level process shaped by the interaction between bottom-up visual organization and top-down control dynamics.

Another key insight from our findings is that perceptual suppression and motor control share a common neural infrastructure. Behaviorally, TTPs with stronger spatial suppression exhibited faster visuomotor responses, indicating an association between perceptual filtering and action speed. Importantly, this effect was selective; enhanced suppression was observed for large, surround-sensitive gratings, while sensitivity to small central stimuli remained preserved (Tadin et al., 2003). This pattern suggests that spatial suppression reflects an optimization of center–surround processing rather than a generalized change in sensory gain, enabling more efficient selection of task-relevant signals in dynamic environments (Tadin et al., 2003, 2019; Li et al., 2023a). This framework provides a potential explanation for why expertise in fast ball sports is associated with a more efficient integration of perceptual filtering and action readiness. Rather than improving basic sensory acuity or simple reaction speed, table tennis may involve a more flexible capacity to regulate the gain of relevant versus irrelevant information during dynamic visuomotor behavior. Together, these results support a brain activity change in which spatial suppression and visuomotor performance are statistically linked through dynamic processes that regulate the gain of relevant versus irrelevant information across perceptual and motor systems. Because our design is cross-sectional, this training-related interpretation is speculative and requires longitudinal or interventional confirmation. We also examined whether the observed visuomotor RT effect could be associated with general attentional enhancement by analyzing the relationship between performance in the attentional cueing task and visuomotor RT. Although ΔRT was significantly correlated with visuomotor RT, no significant association was observed between ΔRT and rIFG–left PoCG dFC, suggesting that general attentional enhancement alone is unlikely to account for the observed brain–behavior relationship. Nevertheless, the present study was not designed to dissociate the respective contributions of perceptual suppression, attentional processing, and motor execution to visuomotor RT. Future studies incorporating more comprehensive behavioral measures will help further clarify the mechanisms underlying visuomotor RT and other factors contributing to expertise-related performance.

There are some strengths and limitations we would like to mention. Strengths include a multimodal design (psychophysics, 7T structural MRI, resting-state dFC), predefined ROIs (IFG, hMT+), and convergent behavioral–structural–functional links in athletes. Limitations include the cross-sectional comparison between athletes and controls, which precludes causal inference, and a modest sample size for both structural and functional MRI analyses, which may increase susceptibility to individual variability. Future longitudinal follow-up and interventional studies in larger cohorts will be essential to better characterize the causal mechanisms. Future studies should also incorporate behavioral specificity controls to help clarify which processes contribute to expertise-related performance differences. Beyond the athletic brain, our findings raise implications for aging and clinical populations characterized by impaired inhibitory control. Visual surround suppression is known to weaken with age (Betts et al., 2005; Tadin et al., 2019) and in disorders such as depression, schizophrenia, and autism (Tadin et al., 2006; Tadin, 2015; Schallmo et al., 2020; Song et al., 2021), contributing to perceptual crowding and slowed visuomotor responses. By identifying IFG-related network dynamics as a correlate of spatial suppression, our findings raise the possibility that cognitively demanding visuomotor training may represent a potential avenue for enhancing perceptual regulation.

Author contributions

Wenming Liu (Conceptualization, Funding acquisition, Investigation, Methodology, Project administration, Resources, Writing—original draft, Writing—review & editing), Shuqin Ye (Data curation, Formal analysis, Investigation, Methodology, Project administration, Visualization, Writing—original draft, Writing—review & editing), Yuyang Li (Data curation, Formal analysis, Investigation, Methodology, Project administration, Writing—original draft, Writing—review & editing), Sophia Wang (Formal analysis, Writing—review & editing), Bing Yun (Data curation, Investigation, Methodology), Le Luo (Methodology, Supervision), Mengyuan Gong (Conceptualization, Methodology, Software, Writing—review & editing), Hui Lu (Conceptualization, Methodology, Visualization, Writing—review & editing), and Chanying Zheng (Conceptualization, Funding acquisition, Investigation, Methodology, Project administration, Resources, Supervision, Writing—review & editing)

Conflicts of interests

The authors declare no conflicts of interest.