In line with previous studies employing similar paradigms (32), we specified an expected effect size of f = 0.30 and a desired statistical power of 0.80 to determine the required sample size. Twenty-nine right-handed participants were recruited in the study, with fifteen assigned to the swimming group (SG) and fourteen to the control group (CG), all without a prior history of upper extremity or severe musculoskeletal injury. After exclusion of participants with incomplete testing, 24 participants were included in the final analysis (12 in SG, 12 in CG). The general information for each group is shown in Table 1. The training background information for the SG is shown in Table 2. Participants were recruited from Shanghai Sport University, with swimmers being members of the university's competitive swimming team and having competed at national levels. Mean sport-specific training age is 11.6 years. The performance level of the swimmers was assessed using FINA points, calculated based on their fastest records (http://www.fina.org/content/fina-points). The control group comprised college students with no formal swimming training. The study was approved by the Ethics Committee of Shanghai Sport University. All experiments were conducted in strict accordance to the ethical guidelines outlined in the Declaration of Helsinki, with comprehensive informed consent obtained from all participants.

The participants underwent two experimental sessions, wherein they were randomly allocated to execute elbow flexion and extension movements in aquatic and terrestrial environments. EEG signals were continuously recorded during these sessions for subsequent neurophysiological analysis (Figure 1B).

An offline band-pass filter of 1–40 Hz was applied. Independent component analysis (ICA) was performed using MATLAB-based EEGLAB toolbox with default parameters to remove electrooculogram (EOG) and electromyogram (EMG) components. EEG Data were re-referenced using the reference electrode standardization technique (REST) (42). Multitaper spectral estimation was used for power spectra of theta (4–8 Hz), alpha (8–12 Hz), beta (13–30 Hz), and SMR (12–15 Hz) waves with the FieldTrip toolbox (43). The data were segmented into 2-second epochs, with all available trials included in the analysis. The Fast Fourier Transform (FFT) was applied using the ft_freqanalysis function in Fieldtrip toolbox with the following settings: a 3 Hz smoothing frequency, DPSS tapers, and zero-padding to the next power of two for optimal frequency resolution. The power in the specified frequency band was averaged across trials. The power ratio was defined as the power in the task condition divided by that in the rest condition. The resulting frequency data were then stored for further statistical analysis. For MF, the formula is as follows (44), where P(f) was defined as power spectrum. I and J were the lower and upper limits of the corresponding frequency band.f¯=∑f=ij(P(f)×f)∑f=ij(P(f))The percentage deviation between task and rest states was calculated using the formula Δ% = [(T value—R value)/R value] × 100, where T value represents the EEG eigenvalues during the task condition, and R value denotes the EEG eigenvalues during the rest state. The mean frequency for each cortical region was calculated as the average of all channels within that region. To characterize the overall trend, EEG data from the left and right elbow tasks were averaged prior to statistical analysis. The Schematic overview of the EEG data-processing and the result of power spectrum are provided in the Supplementary Materials (Supplementary Figures S1–S3).

The statistical analysis was conducted using SPSS 27.0 software, and descriptive statistics were performed on the raw data. Outliers in the raw data were defined as data points exceeding three times the interquartile range, with approximately 2.2% of the total data being replaced. The Shapiro–Wilk test was employed to assess the normality of the data. In this study, the testing time was the within-group factor (rest, terrestrial, aquatic), and the training experience was the between-group variable (SG, CG). For normally distributed data, analysis of variance (ANOVA) was used for comparisons of factors within groups, and independent t-tests were used for comparisons of factors between groups in the same condition. For non-normally distributed factors within groups, non-parametric statistical analysis was performed using nonparametric factorial ANOVA with aligned rank-transformed (ART) values. ART is a modification of the rank transform procedure by aligning data to strip the interaction effect from the main effects, as well as the main effects from each other and the interaction, and then ranking, factorial ANOVA is possible (45, 46). Mauchly's test of sphericity was performed to test the assumption of equal variances of the differences between conditions. When the assumption was violated, degrees of freedom were adjusted using the Greenhouse–Geisser correction. The Mann–Whitney U Test was employed for non-normally distributed factors between groups. For Δ%(MF), Δ% (SMR power), and Δ% (Mu power), a t-test was implemented for those that exhibited a normal distribution, and the Wilcoxon Signed-Rank Test or Mann–Whitney U Test was employed for those that did not conform to a normal distribution. Multiple comparisons using a paired t-test with Bonferroni correction were used as a post-hoc test.

In the theta band, both groups exhibited a general trend of decreased MF during tasks in terrestrial and aquatic conditions compared to the resting state, as illustrated in Figure 2A, with partial statistical significance. Specifically, in the SG, MF during the task condition significantly decreased relative to baseline in the central-parietal region (p = .001) and parietal region (p = .020). In the CG, significant MF reductions were observed in the central (p = .002), central-parietal (p = .007), and parietal regions (p = .009). No significant group differences were found in MF change rates within the theta band (Figures 2D,G).

In the alpha band (Figure 2B), MF(SG) significantly decreased from baseline across all five regions: frontal (p = .009), frontal-central (p = .001), central (p = .001), central-parietal (p = .001), and parietal (p = .001). MF(CG) significantly decreased in the frontal-central (p = .016), central (p = .001), and central-parietal (p = .008) regions. About MF change rate, SG shows a significant decrease in an aquatic task at the frontal, frontal-central, and parietal regions compared with a terrestrial task. Compared with CG, the SG exhibited a significantly bigger MF change at the central region in terrestrial task, and also a significant bigger MF change at the parietal region in aquatic task (Figures 2E,H).

In the beta band (Figure 2C), MF(SG) showed significantly increased MF in all five regions during both task conditions compared to resting baseline: frontal (p = .003), frontal-central (p = .002), central (p = .001), central-parietal (p = .001), and parietal (p = .001). MF(CG) significant increases were found only in selective regions: frontal-central (p = .018), central (p = .004), central-parietal (p = .002), and parietal (p = .007) in Figure 2C. Notably, the SG demonstrated significantly greater MF change rates than the CG during the land condition at the frontal-central (p = .012), central (p = .030), and central-parietal (p = .034) regions. Interestingly, MF(SG) in the aquatic task remained stable or even decreased relative to the terrestrial task, while MF(CG) in the aquatic task keeps slight increasing compared to a terrestrial task (Figures 2F,I).

In the theta band, power ratios were comparable between the aquatic and terrestrial conditions (Figures 3A,D). In the alpha band, task-related power was consistently lower than at rest. In the control group, power ratios during the aquatic task were lower than during the terrestrial task across all cortical regions (Figure 3B), whereas in the swimmer group lower aquatic than terrestrial power ratios were mainly observed over frontocentral, central, and central-parietal areas (Figure 3E). In the beta band, task-related power was consistently higher than at rest the control group showed higher power ratios in the aquatic than the terrestrial task across regions (Figure 3C), while in the swimmer group power ratios were largely comparable between tasks (Figure 3F). None of these trends reached statistical significance for either group or task condition in the theta, alpha, or beta bands.

At the Cz electrode, both groups exhibited significantly reduced SMR power during task conditions compared to baseline (SG: p = .003; CG: p = .002), as showed in Figure 4A. post-hoc tests revealed that in the CG, SMR power significantly decreased in both terrestrial (p = .027) and aquatic (p = .013) environments. The SG exhibited a significant SMR power decrease exclusively during the aquatic environments (p = .019). Additionally, the SG displayed significantly lower SMR change rates when comparing aquatic to terrestrial conditions (p = .037) in Figure 4B.

The above sections summarize changes in mean frequency and power in the theta, alpha, beta, and SMR bands. Detailed statistics are provided in the Supplementary Materials (Supplementary Tables S1–S5).

MF provides a quantitative measure of cortical firing intensity proximal to the recording electrode (37, 47). Theta-band oscillations are closely associated with the coordinating functions of cognitive control and motor adaptation, serving as the foundational mechanism underlying various forms of action regulation (20). Such oscillations are also suited for integrating information across spatial domains (39). In this study, the control group exhibited significantly lower theta MF in the central, central-parietal, and parietal regions during task performance, while the swimmer group showed significant MF reductions in the central-parietal and parietal regions. These findings suggest coordinated engagement of motor-related regions (central and central-parietal) (48, 49) and somatosensory-related regions (parietal) (50, 51) in the execution of sensorimotor tasks (repetitive elbow flexion-extension). Theta oscillations are also implicated in working memory encoding processes (52). In our experiment, participants were first asked to memorize a target elbow angle, and then reproduce it as accurately as possible. The theta activity observed may reflect the encoding and retrieval processes of joint angle in task execution (53, 54).

In the alpha band, both groups showed lower MF and power spectrum during task conditions compared to the resting baseline. Significant MF decreases were observed in the swimmer group across all five regions (frontal, frontal-central, central, central-parietal, parietal), while the control group showed significant reductions in the frontal-central, central, and central-parietal regions. Reduced alpha-band MF is generally associated with cortical activation and the suppression of task-irrelevant activity during motor performance (21). These MF and power reductions likely reflect increased sensorimotor engagement during task execution. Notably, the swimmer group showed significantly bigger MF changes in an aquatic environment, especially in the frontal (FR), frontal-central (FCR), and parietal (PR) regions. Alpha-band event-related desynchronization (ERD) has been linked to attentional control, particularly in sharpening attentional focus and suppressing irrelevant information processing (55). These results suggest that swimmers may reduce interference from buoyancy and drag force in water, thereby facilitating improved motor performance in aquatic environments.

Physical exercise is commonly accompanied by beta-band cortical desynchronization (ERD) (23). However, a clear increase in beta-band mean frequency (MF) and power during movement was recorded in this research. We speculate that the task paradigm (periodic elbow flexion–extension) imposed an alternating start–stop structure, requiring participants to repeatedly and rapidly terminate ongoing actions, thereby amplifying beta oscillations. In recent research, transient post-movement increases in beta power—post-movement beta rebound (PMBR) or beta event-related synchronization (β-ERS)—can occur after each movement during repetitive actions (56). PMBR is typically observed after voluntary movement completion or following somatosensory stimulation (57, 58), and successful movement inhibition is likewise associated with increased beta activity (59, 60). PMBR generally peaks ∼0.5–1.0 s after movement offset and returns to baseline within ∼1 s, and has been reported in the sensorimotor cortices, premotor cortex, SMA, and medial prefrontal cortex (61–63). Accordingly, PMBR is widely considered a physiological marker of movement inhibition, reflecting active suppression within motor cortical networks following movement termination and/or sensory feedback (64).

In previous research, when participants were required to terminate more rapidly, PMBR was stronger immediately after movement termination (64). This likely reflects greater recruitment of executive control networks to rapidly terminate the movement (64). In other words, PMBR enhancement reflects increased demands on movement termination and inhibitory control. In our study, the control group showed a slight increase in beta-band MF and power during aquatic vs. terrestrial tasks, whereas swimmers exhibited stable or slightly reduced beta-band MF and power under aquatic conditions. This pattern may suggest that controls experience greater perturbations when maintaining arm position in water, increasing the demands on movement termination. However, swimmers may show greater resilience to aquatic perturbations.

Notably, interpreting enhanced beta activity during continuous elbow flexion–extension solely in terms of motor inhibition should not be the only explanation. Such complex movements likely involve mixed processes, potentially including working memory and attentional demands in addition to inhibitory control (59, 65). Future work should further examine alternative mechanisms that may contribute to the observed beta enhancement.

In the swimmer group, SMR power at Cz during terrestrial tasks were significantly diminished compared to baseline, and the ΔSMR% during aquatic tasks were significantly greater than during terrestrial tasks. These observations indicate that the sensorimotor cortex of swimmers exhibited reduced activation on land tasks but demonstrated enhanced neural engagement in aquatic environments. Given the inherent buoyancy and drag force in water, the observed enhanced activation likely reflects increased sensorimotor processing demands in response to environmental perturbations. Notably, this pattern was absent in the control group, which demonstrated comparable levels of cortical activation across different environmental conditions. Pianists exhibited significantly lower alpha-band power during complex vs. simple keypress tasks, whereas non-pianists showed no significant task-condition differentiation (12). This research indicates that both task complexity and motor expertise shape cortical activation patterns during task execution. In our study, SMR power modulation (12–15 Hz) across environmental conditions appears to reflect the interactive effects of environmental interference and motor expertise on sensorimotor control. Previous studies have consistently linked elevated SMR power with reduced somatosensory interference (66), improved attentional performance, and optimized motor output (67). Our result suggests that long-term swim training may facilitate SMR-related cortical plasticity, consequently enhancing motor performance in aquatic environments.

Overall, controlling limb movements in water entails greater external interference than in a terrestrial environment. During aquatic movement, both swimmers and controls likely require additional cognitive control, sustained attentional focus, and suppression of task-irrelevant processing, accompanied by heightened task-related activation in the relevant cortical regions. Notably, the two groups also exhibited different levels of environmental adaptation. In swimmers, enhanced engagement of water-specific sensorimotor processing was reflected by a further reduction in SMR power. In contrast, controls may have recruited executive control networks to a greater extent to implement active inhibition in water, consistent with the broadly increased beta activity observed across regions during movement.