The Mode of Endurance Exercise Influences Changes in EEG Resting-State Graphs among High-Level Cross-Country Skiers : Medicine & Science in Sports & Exercise

Journal Logo


The Mode of Endurance Exercise Influences Changes in EEG Resting-State Graphs among High-Level Cross-Country Skiers


Author Information
Medicine & Science in Sports & Exercise 55(6):p 1003-1013, June 2023. | DOI: 10.1249/MSS.0000000000003122
  • Free


Participation in endurance exercise induces acute neuroplastic changes, which result in temporary modulations of brain function by means of molecular, cellular, and functional processes that eventually affect human cognition and behavior (1,2). Based on investigations of brain physiology and behavior before and after given loads of exercise, intensity has been shown to moderate the acute response of the resting brain to exercise, with the tendency toward more beneficial effects at moderate and high compared with low- and heavy-intensity exercise (3–5). These differences likely arise from differences in metabolic pathways induced by the dose of exercise according to intensity and duration (6,7). Apart from intensity and volume, the mode of endurance exercise affects the physiological responses to exercise bouts because of differences in movement constraints, muscles involved, and muscle actions (8). Despite sports-specific factors such as involved limbs/muscles, technique, and equipment that affect movement economy (9), endurance mode has been barely investigated as a factor modulating acute exercise-induced neuroplasticity of the resting brain (10,11). In contrast to exercise-intensity where metabolic processes affect brain signaling, endurance mode is likely to alter brain function because of task-specific activation of cortical areas involved into mode-specific sensorimotor control demands (12).

Among exercise modes, cross-country (XC) skiing is unique because it requires highly developed cardiovascular capacities (13) in concert with strong whole-body motor coordination abilities (8). In particular, the combination of lower and upper body and the need for a sports-specific technique to generate propulsion (13) make XC skiing a highly demanding full-body endurance sport. Considering the central nervous system as the physiological origin of motor coordination, the repeated and rhythmic coordination of limb muscles—referred to as locomotion—requires the integration of somatosensory information in brain motor areas (14). Depending on task constraints, the degree of activation of these sensorimotor brain areas may change because of factors such as limbs involved (15), upper and lower limb phasic coordination (16), and balance demands (17). For XC skiing, these factors differ from running (RUN), because XC skiing i) involves upper body muscles to actively produce force output, ii) requires antiphase temporal sequencing of upper and lower limbs, iii) takes place at increased speed, and iv) requires sports-specific equipment such as skis that affect postural control. Although these differences in motor coordination across endurance exercise modes may hypothetically change allocation of cortical resources during and after exercise, it might also be suggested that they alter brain function in nonmotor brain areas after exercise (18,19).

To investigate changes in brain function, electroencephalography (EEG) serves as a mobile and noninvasive imaging technique with a high temporal resolution (20). Apart from EEG approaches that analyze brain function during movement that is particularly limited at high intensities because of movement artifacts, the analysis of the EEG resting states after exercise became a valid approach (21). Because exercise induces acute neuromodulatory processes like the release of neurotrophins (e.g., brain-derived neurotrophic factor), which mediate neural activity on the cell level (1), brain function remains affected after exercise cessation. Thus, the neurophysiological effects could also be assessed in a resting situation, after termination of an acute bout of exercise. Next to the activity of specific brain areas, the connectedness across the cortex, defined as resting-state networks (RSN), can be calculated based on temporal coherence between signals (22). RSNs are suggested to provide information on the responsiveness of the brain to external stimuli (23) and seem to change after exercise (3,5). In our previous investigations, we applied EEG to extract RSNs and were able to describe acute exercise-induced changes in brain function (5) at an acceptable level of reliability (24). In these investigations, we calculated global graph outcomes like the clustering coefficient (CC), the characteristic path length (PL), and the small-world index (SWI) to describe changes in the efficiency of brain network communication.

To date, studies have investigated the effects of exercise on the brain, applying either RUN or cycling protocols (21). Considering the physiological differences between endurance exercise modes (8), the question arises of whether associated differences in motor coordination demands influence the acute effects of exercise on brain function. A better understanding of brain responses due to sports-specific coordination demands may allow a more precise tailoring of training programs in the future. Therefore, this study aimed to investigate the influence of exercise intensity and endurance mode on exercise-induced changes in RSNs. In a laboratory study, we compared two endurance exercise modes on the treadmill with differences in motor coordination: a RUN protocol and an XC skiing protocol with roller skis matched for intensity. To account for demands on XC skiing technique, a group of highly trained XC skiers training a similar amount of RUN and XC skiing (25,26) was tested. According to previous findings, we expected exhaustive exercise to induce detrimental changes in attention-related RSNs (5,27). With regard to differences in upper body involvement and technique between both endurance modes, we further aimed to explore the modulatory effect of exercise mode on connectivity in sensorimotor RSNs. To assess RSNs, we applied methodological approaches from previous studies (5,24). Accordingly, the findings of the present study may extend our understanding of the neuroregulatory effect of exercise with specific regard to exercise modes, including challenging upper and lower limb coordination (12). With regard to sports performance, our analyses may reveal novel insights into athlete’s acute response to exercise bouts and may forward the integration of the brain into performance monitoring (28).



We calculated a required sample size to achieve 0.8 statistical power at an alpha level of 0.05 using GPower, Version 3.1 (29), and we based assumed effect sizes on our recent publication (Büchel et al. [5]) where we observed an effect size of 0.48 of exercise intensity on brain graphs in a within-factor ANOVA design with two groups and five repetitions. This calculation led to a required participant sample size of 16. Finally, 15 healthy young male XC skiers (20.1 + 0.6 yr, 74.4 + 5.6 kg, 178.9 + 5.5 cm) participated in the present investigation. Because the menstrual cycle may cause substantial day-to-day changes in RSN organization in female participants, we preferred not to include women in this explorative study but rather examine similar aspects and the effect of menstrual cycle in follow-up studies (30). All athletes were students and competed at Norwegian national XC races and trained about 15 to 20 h·wk−1 at the time of the investigation. Physiological information on the subjects performance level can be found in Table 1. Acute or chronic states of neurological diseases as well as injuries served as exclusion criteria for study participation. Before participation, athletes signed a written informed consent form in accordance with the Declaration of Helsinki. The study procedure was approved by the ethics committee of the Norwegian University of Technology and Science.

TABLE 1 - Comparison of physiological data derived from initial aerobic capacity assessment comparing RUN and XC skiing exercise on the treadmill.
HRpeak (bpm) 200.4 ± 13.9 196.4 ± 15.0 <0.001 0.909
V̇O2peak (mL·min−1·kg−1) 65.3 ± 4.3 63.5 ± 5.8 0.015 0.612
vV̇O2peak (km·h−1) 14.5 ± 0.9 20.4 ± 1.5 0.011 0.633
BLapeak (mmol·L−1) 11.1 ± 2.0 10.5 ± 2.7 0.060 0.496
RER (O2/CO2 ratio) 1.2 ± 0.4 1.1 ± 0.4 0.462 0.206
50%vV̇O2peak (km·h−1) 6.8 ± 0.4 10.2 ± 0.7 0.009 0.648
70%vV̇O2peak (km·h−1) 9.5 ± 0.5 14.3 ± 1.0 0.009 0.648
90%vV̇O2peak (km·h−1) 12.2 ± 0.7 18.4 ± 1.3 0.009 0.648
Data assessed include HRpeak, V̇O2peak, vV̇O2peak, BLapeak, and calculated speeds at 50%, 70%, and 90% of vV̇O2peak. P values indicate the statistical differences between both protocols, and r values indicate correlation across test variables according to Pearson.


The participants performed four experimental sessions, including i) a RUN peak oxygen uptake (V̇O2peak) test, ii) an XC skating V̇O2peak test, iii) an individual RUN protocol with EEG assessment, and iv) an individual XC skating protocol with EEG assessment. The time between the two EEG measurements did not exceed 3 wk, and all measurements took place during the competitive phase of the season (October to February), where only minor changes in physiological parameters of highly trained XC skiers can be observed (31). Because the testing sessions needed to be fitted into the training and competition schedule of the athletes, 10 athletes started with session iii and 5 athletes started with session iv.

RUN V̇O2peak test

The RUN V̇O2peak test took place on a motorized treadmill (RL 2500E; Rodby, Södertalje, Sweden). To warm up, athletes performed a submaximal lactate profile starting at 8 km·h−1 and incline of 10.5% for 5 min. After each stage, a blood sample (20 μL) was taken from the fingertip, and lactate concentration was analyzed using the stationary Biosen C-Line lactate analyzer (Biosen; EKF Industrial Electronics, Magdeburg, Germany). During the protocol, speed was increased by 1 km·h−1 per stage until lactate concentration increased absolutely by 1 mmol·L−1 compared with the previous sample and passed the 4 mmol·L−1 threshold. After cessation of the submaximal lactate profile, participants performed an 8-min recovery and cool-down period. The V̇O2peak test then started at the speed of the second last stage and was performed in an incremental manner. After every minute, the treadmill speed was progressively increased by 1 km·h−1. The test stopped at voluntary exhaustion of the participants. During the incremental test, V̇O2peak, maximum ventilation (V̇E), and respiratory exchange rate (RER) were measured breath-by-breath using a stationary gas analysis system (VYNTUS CPX, Vyaire, Germany). V̇O2peak was defined as the mean of the two highest and consecutive 30-s measurements. Accordingly, vV̇O2peak was defined as the speed at the first of these two 30-s measurements. Simultaneously, heart rate was measured using an electrocardiogram sensor connected to a chest belt (HRM3-SS; Garmin, Schaffhausen, Switzerland) to define peak heart rate (HRpeak RUN) as the maximum heart rate reached during the RUN test. Furthermore, a lactate sample was taken after cessation of the test.

XC skating V̇O2peak test

The XC skating V̇O2peak test took place on an oversize motorized treadmill (RL 3500E; Rodby, Södertalje, Sweden) in the laboratory. All athletes were instructed to perform the skating test utilizing the skating G3 subtechnique and used the same pair of skating–roller skis (IDT Sports, Lena, Norway) and poles adjusted for body size. Roller ski skating in the laboratory was shown to have large to very large correlations with XC skiing performance in the field (32). For warm-up, athletes performed a submaximal lactate profile starting at 10 km·h−1 and incline of 5% for 5 min. After each stage, a blood sample (20 μL) was taken from the fingertip, and lactate concentration was analyzed. The speed was increased by 2 km·h−1 per stage until lactate concentration increased absolutely by 1 mmol·L−1 compared with the previous sample and passed the 4 mmol·L−1 threshold. After cessation of the submaximal lactate profile, participants had 8 min to recover and cool down. The V̇O2max test then started at 10 km·h−1 and was performed in an incremental manner. After each minute, the speed of the treadmill was increased by 2 km·h−1. After reaching the speed of 20 km·h−1, the speed was increased by 1 km·h−1·min−1. The test stopped at voluntary exhaustion of the participants. During the incremental skating test, V̇O2peak, V̇E, and RER were measured breath by breath. V̇O2peak was defined as the highest average of two consecutive 30-s average values. Accordingly, vV̇O2peak was defined as the speed where athletes reached V̇O2peak. Simultaneously, heart rate was measured to define peak heart rate skating (HRpeak skate), and a blood sample was taken from the fingertip after cessation. An overview of the physiological outcomes revealed from aerobic assessment is presented in Table 1.

Individual XC skating and RUN protocol

To allow for comparisons between exercise modes (RUN vs XC skating), the individual exercise protocols were matched for external load based on individual vV̇O2peak. Accordingly, the protocol consisted of four exercise stages: i) RUN/XC skating at 50% vV̇O2peak for 10 min, ii) RUN/XC skating at 70% vV̇O2peak for 10 min, iii) RUN/XC skating at 90% vV̇O2peak until voluntary exhaustion, and iv) RUN/XC skating at 50% vV̇O2peak for 8 min. The cool-down exercise bout was followed by 10 min of active recovery. If the voluntary exhaustion stage lasted longer than 5 min, speed was increased by 1% vV̇O2peak per minute after the fifth minute. Athletes performed the RUN and XC skiing protocol in a randomized order. An overview of the experimental procedure is provided in Figure 1.

Overview of the experimental procedure to investigate the effect of exercise modality on brain graph modulation. For both RUN and XC skiing, an initial incremental treadmill test on separate days allowed V̇O2peak to be assessed. During the experimental session, EEG resting-state data, HR, BLa, and RPE using Borg Scale 6–20 were assessed before (PRE) and after (ACUTE 1, ACUTE 2, ACUTE 3, REC 1, and REC 2) intermittent bouts of acute exercise individualized based on vV̇O2peak.

Several physiological data considerations were assessed to estimate the individual exercise load during the different stages. Outcomes analyzed included the CR-20 Borg Scale of subjective rate of perceived exertion (RPE), the average and maximum heart rate during the exercise stages (HRmean and HRpeak, respectively), and during resting states (HRrest mean and HRrest peak, respectively), as well as blood lactate concentration based on fingertip blood samples (BLa) using the instruments introduced above.

EEG assessment

The procedures of EEG data assessment were applied according to previous EEG resting state investigations (5,24). Electrocortical activity was recorded by 65 passive wet electrodes (RNET; Brain Products, Gilching, Germany) connected to a wireless transmission system (LiveAmp; BrainProducts). Electrodes were online-referenced to FCz and attached according to the international 10–20 system. EEG recordings were obtained in a seated position with eyes open in front of a whiteboard (distance ~1.5 m) and lasted for 5 min each. Accordingly, the EEG resting states were recorded in six situations: i) before the start of the protocol (PRE), ii) after low-intensity exercise (ACUTE1), iii) after moderate-intensity exercise (ACUTE2), iv) after exhaustive exercise (ACUTE3), v) after cool-down exercise (REG1), and vi) after 10 min of recovery (REG2). An overview of the treadmill protocol is presented in Figure 1.

The fast applicability of the EEG wet electrodes allowed resting-state assessments to start within 4 ± 0.5 min after stage termination of the exercise stages. Before starting the recording, electrode impedances were kept below 25 kΩ for all electrodes. Because wearing an EEG cap while exercising at high intensities is not comfortable and may result in shifts of electrode placement, the cap was only applied to the participant’s head during each break. To consider exercise-induced changes in heart rate during the EEG resting state, timestamps were manually set using a watch (Forerunner 935; Garmin). Before and after applying the EEG cap, sweat was removed from the hair and the interelectrode space using a towel to prevent the building of electrical bridges.

EEG preprocessing

For data processing, EEG recordings were imported into the EEGLAB toolbox v14.1.2 (33) for MATLAB (Version R2019a; MathWorks Inc., Natick, MA). The detailed steps of EEG preprocessing were reported in a recent article on exercise-induced changes in EEG brain graphs (24). These automatic processing steps included the application of i) sinusoidal line noise removal (CleanLine Plugin, Version 1.04,, ii) bandpass filtering between 3 and 30 Hz, iii) average referencing and recomputing of the reference electrode FCz, v) downsampling from 500 to 256 Hz, vi) removal of bridged channels applying the eBridge tool (34), and vii) artifact-subspace reconstruction (clean_rawdata plugin, Version 2.3, After ASR, adaptive mixture independent component analysis (35) was applied to remove signal components from nonbrain sources like eye, heart rate, muscle, sweat, and channel noise. Finally, the clean data were epoched into sequences of 8 s with a phase overlap of 4 s (24). The first 40 epochs of each EEG resting-state recording were chosen for connectivity analysis.

Graph analysis

To compute brain graphs, we imported the epoched data to the BrainWave software (Version, Functional connectivity (FC) between EEG channels was calculated using the weighted Phase-Leg Index (wPLI). The wPLI is an index of the asymmetry in the distribution of phase differences calculated based on instantaneous phases of two time series. wPLI values can range from 0 to 1, whereas 1 represents perfect phase locking. By ignoring zero-phase interactions, phase-lag measures are stated to be less sensitive to signal amplitudes and volume conduction effects (36). The wPLI was computed on the scalp level for all possible connections between two electrodes for each epoch. Therefore, forty 65-by-65 matrices were computed for each participant per condition. The computation of wPLI was performed for the theta (5–8 Hz), alpha-1 (8–10.5 Hz), and alpha-2 (10.5 to 13 Hz) frequency bands according to Büchel et al. (5). Higher-frequency bands such as beta were not considered for the present study because of reasonable day-to-day differences observed in one of our previous investigations (24).

For the calculation of weighted, undirected brain graphs, the resulting connectivity matrices were imported to the MATLAB-based Brain Connectivity Toolbox (37). To obtain comparability despite the interindividual differences of absolute FC measures across participants, connectivity matrices were normalized, by bounding all wPLI values to the range from 0 to 1 based on the rank of the maximal and minimum values per participant. Because we did not apply an arbitrary threshold for the removal of weakly connected nodes, the weighted network approach allowed us to consider all possible connections across the scalp. The global graph measures CC, PL, and SWI were computed as outcomes of interest. Hereof, CC and PL were analyzed as global graph outcomes. The CC serves as a measure of network segregation and describes the tendency of a network to build local clusters within its direct neighborhood. PL serves as a measure of network integration and describes the average steps the given information needs to travel from any position in the network to another. The SWI serves as a measure of network efficiency and describes the relationship between short and long connections. For its computation, CC and PL were normalized by dividing each individual value by the mean of all other values over all frequency bands. Finally, SWI was calculated as the individual ratio between normalized CC and normalized PL that was computed.

Statistical analysis

Statistical analyses were performed in SPSS 26 (SPSS Inc., Chicago, IL). All results are given as mean ± SD. Normal distribution of the data was verified by applying the Kolmogorov–Smirnov test. ANOVA for repeated measures allowed us to analyze differences between physiological outcomes and resting-state EEG data considering two given factors: MODE (RUN vs XC) and LOAD (PRE, POST1, POST2, POST3, REC1, REC2). For analysis of physiological responses, absolute values of HR (in % HRmax), BLa (in mmol·L−1), and RPE were analyzed. For the analysis of EEG data, differences between baseline EEG before exercise at each experimental day (PRE) and POST1, POST2, POST3, REC1, and REC2 served as corresponding outcomes because we observed considerable day-to-day differences, for example, for the theta frequency band in our previous investigation (24). Because HRrun failed to match assumptions of normal distribution, Friedman tests of differences among repeated measures grouped for LOAD and MODE were conducted. Post hoc tests corrected for multiple comparison according to Bonferroni were applied in case of significant main effects, to localize effects between the different experimental conditions. The statistical level of significance was set to P < 0.05. For interpretation of effect sizes, partial eta squared (ηp2) was calculated, and 0.01, 0.06, and 0.14 were considered small, medium, and large effect sizes, respectively (38).


Physiological data

ANOVA for repeated measures revealed significant main effects of load on HRrest mean (F5,10 = 56.580, P < 0.001, ηp2 = 0.802), BLa (F5,10 = 336.355, P < 0.001, ηp2 = 0.96), and RPE (F3,12 = 240.533, P < 0.001, ηp2 = 0.945). For HRmean, Friedman test also demonstrated significant main effects of LOAD (χ2 = 119.806, P < 0.001). All outcomes demonstrated maxima after POST3. Furthermore, Friedman test also demonstrated significant main effects of MODE (χ2 = 27.74, P < 0.001) on HRmean, with higher values for XC compared with RUN. A visualization of physiological responses to exercise is provided in Figure 2.

Overview of physiological responses to different exercise loads during RUN (white) and XC skiing (black) on the treadmill. Data were assessed during different acute exercise loads: before exercise (PRE), after low-intensity exercise (ACUTE1), moderate-intensity exercise (ACUTE2), exhaustive-intensity exercise (ACUTE3), cool-down exercise (REC1), and active recovery (REC2). Data include BLa, subjective RPE, mean heart rate during RUN (HRmean), and mean heart rate during EEG resting state assessment (HRrest). For RPE, no values were assessed at PRE and REC2. For HRmean, no values were assessed at PRE. †Significantly higher than PRE. βSignificantly higher than ACUTE1. #Significantly higher than ACUTE2. *Significantly higher than ACUTE3. δSignificantly higher than REC1. §Significantly higher than REC2. Level of significance was set at P < 0.05.

Brain networks

ANOVA for repeated measures revealed significant main effects of LOAD on CC (F4,11 = 6.085, P < 0.001, ηp2 = 0.303), PL (F4,11 = 5.514, P = 0.003, ηp2 = 0.247), and SWI (F4,11 = 5.514, P < 0.001, ηp2 = 0.283) in the theta frequency. Post hoc analysis indicated significant lower CC and SWI at POST3 compared with POST1 (P = 0.049 and P = 0.028, respectively), POST2 (P < 0.001 and P = 0.002, respectively), and REC2 (P = 0.024 and P = 0.033, respectively). For PL, higher values were observed at POST3 compared with POST1 (P = 0.040) and POST2 (P = 0.004).

Furthermore, ANOVA revealed a main effect of MODE on CC (F1,14 = 7.969, P = 0.014, ηp2 = 0.363), PL (F1,14 = 5.737, P = 0.031, ηp2 = 0.291), and SWI (F1,14 = 4.720, P = 0.047, ηp2 = 0.252) in the alpha-2 frequency band. Data indicated higher CC, higher SWI, and lower PL during the XC protocol compared with RUN, respectively. An overview of extracted graph measures during different exercise conditions is presented in Figures 3–5.

Overview of modulations of EEG resting-state CC in response to different exercise loads during RUN (white) and XC skiing (black) on the treadmill. Modulations are displayed as % delta (Δ) from PRE to postconditions regarding the calculated graph outcome in the theta (θ; 4–8 Hz), alpha-1 (α-1; 8–10.5 Hz), and alpha-2 (α-2; 10–5–13 Hz) frequency bands. *Post hoc differences regarding exercise intensity. †Main effects for modality. Level of significance was set at P < 0.05.
Overview of modulations of EEG resting-state characteristic PL in response to different exercise loads during RUN (white) and XC skiing (black) on the treadmill. Modulations are displayed as % delta (Δ) from PRE to postconditions regarding the calculated graph outcome in the theta (θ; 4–8 Hz), alpha-1 (α-1; 8–10.5 Hz), and alpha-2 (α-2; 10–5–13 Hz) frequency bands. *Post hoc differences regarding exercise intensity. †Main effects for modality. Level of significance was set at P < 0.05.
Overview of modulations of EEG resting-state small-world index in response to different exercise loads during RUN (white) and XC skiing (black) on the treadmill. Modulations are displayed as % delta (Δ) from PRE to postconditions regarding the calculated graph outcome in the theta (θ; 4–8 Hz), alpha-1 (α-1; 8–10.5 Hz), and alpha-2 (α-2; 10–5–13 Hz) frequency bands. *Post hoc differences regarding exercise intensity. †Main effects for modality. Level of significance was set at P < 0.05.


The present study aimed to investigate the influence of endurance exercise mode on exercise-induced changes in RSNs by applying EEG graph analysis after intensity-matched RUN and XC in highly trained XC skiers. In accordance with our hypothesis, data indicated increased modulation of efficiency in alpha-2 networks after XC skiing compared with RUN. For both endurance exercise modes, similar modulations of brain network efficiency appeared after exhaustive exercise load.

The main finding of the present investigation was an increased modulation of efficiency in alpha-2 networks after XC skating compared with RUN. Because we observed highly comparable physiological responses and subjective perception of exertion for RUN and XC skiing, we assume that the higher increases in alpha-2 network efficiency may result from the sports-specific demands of XC skating compared with the more general mode of RUN. Accordingly, increased sensorimotor demands and full-body coordination (39) in XC skating may involve more cortical resources during exercise and may therefore alter communication within the cortex after exercise.

In general, oscillations in the alpha-2 frequency band are associated with task-related information processing, as well as corticocortical and thalamocortical information flow (40). Functionally, alpha oscillations are linked to active inhibition of nonessential neuronal processing and therefore gait information flow within the brain (41). Accordingly, the degree of desynchronization in the alpha-2 band was previously associated with the complexity of task-relevant information processing (42,43) and increasing task demands during sensorimotor tasks (44). According to the assumption that cortical reallocation relies on exercise demands (18), we assume that higher demands on movement coordination during XC skating induced a greater change in the communication of task-relevant neural circuits after exercise because of concurring resources. To discriminate movement coordination between RUN and XC skating, two aspects of movement coordination might be considered: i) the role of the upper body during XC skating and ii) the complexity of movement coordination. Previous findings indicated that the simultaneous control of both arms and legs induces stronger desynchronization of alpha oscillations in brain motor areas compared with leg activity only (15). Therefore, the active involvement of the upper body to produce propulsion during XC skiing (39) may require additional cortical resources to coordinate this movement. Furthermore, increased information flow in the alpha-2 frequency was reported in finger movement reproductions as a function of task complexity, because more complex movements evoked stronger desynchronization in motor-related brain areas (45). Because XC skating requires a precise temporal and spatial sequencing of movement between the upper and lower body (46), complexity of motor coordination during XC skating may exceed coordinative demands during RUN and require increased allocation of resources, resulting in increased corticocortical information exchange after exercise cessation (18). Consequently, the modulation of the alpha-2 network after XC skating might be interpreted as a function of coordinative demands, as more cortical resources are involved to coordinate movement during exercise.

In the current body of literature, increased small-world characteristics in the alpha-2 network were considered a long-term maladaptation of brain function after chronic diseases such as Alzheimer’s disease (47) or obsessive-compulsive disorder (48). In this regard, increased small-world characteristics in the alpha-2 frequency are interpreted as a state of reduced specificity of information exchange in the resting brain (48). Because we investigated short-term responses of the brain immediately after exercise, increased alpha-2 network modulations after XC skating compared with RUN may demonstrate a temporary deterioration from “normal network efficiency,” possibly arising from more complex coordination demands during preceding exercise, namely, upper body involvement and sports-specific technique.

Another observation of our study was that the efficiency in the theta network decreased significantly after the most intense exercise bout for both XC skiing and RUN. Considering that all individuals demonstrated maximal physiological responses after this exercise stage, it might be stated that exhaustive exercise load induces short-term distortions of attention-related brain networks.

Theta frequency oscillations are associated with activity in the structures of the prefrontal cortex and processes of attentional control and executive function (49). Therefore, theta frequencies are suggested to play a major role in attentional processes associated with motor coordination (50). Thus, a modulation of corticocortical communication within the theta network may be associated with an exhaustion-induced disturbance of executive brain function (5,27). In endurance modes with challenging motor coordination like XC skiing or swimming, this state of reduced network efficiency might be associated with exhaustion-induced decrements in technique (51). Because the present data revealed reduced network efficiency in the theta band for both endurance protocols, this specific pattern might be associated with exercise intensity rather than exercise mode. Therefore, the metabolic responses induced by exhaustive exercise may contribute to changes in brain network communication. Accordingly, the accumulation of metabolites was shown to interact with receptors in the brain and consequently changes brain function (52). Depending on metabolite concentration, the excitability of neural ensembles can be affected and therefore temporarily modulate brain function (52). In agreement with our observations, Tamburro et al. (53) extracted brain graphs from participants during graded cycling exercise and observed that facilitations of brain network efficiency appearing at moderate exercise intensity disappear when approaching heavy exercise intensity. Hence, our data support previous findings, which demonstrated that exercise intensity may act as a strong moderator of acute brain network responses to exercise (3,5,53).

Significance for the field

The present findings revealed that the acute neuroregulatory effect of exercise is modulated by two training variables: exercise intensity and mode. Although intensity and duration are frequently discussed as mediators in the literature (21), only few studies have investigated exercise mode as a potential moderator of the brain’s response to exercise. Our findings demonstrated that exercise with more complex demands on motor coordination might induce stronger responses in sensorimotor, task-related brain networks. During exercise modes with challenging coordination demands like XC skiing, the brain is continuously processing information, because multiple parts of the body act in concert and/or sports-specific technique becomes more complex. On an neural level, these demands may induce an additional stimulus to the brain compared with less complex exercise modes like RUN. This observation is supported by previous evidence reporting mode-specific modulations of RSN applying functional magnetic resonance imaging in the context of balance and coordination training (11,54). Therefore, exercise mode may be considered as a factor affecting exercise-induced modulations of brain networks. Considering the standardized treadmill setting the athletes were exposed to—fixed slope, speed, and surface throughout the stages—one may hypothesize that the observation of mode-specific responses might become even more evoked moving laboratory to field conditions and increasing the variability in technique and physiological demands (55). Based on the neural efficiency hypothesis (56), it might be hypothesized that acute, exercise mode-specific modulations may result in long-term reorganization of athletes´ brains when repeated consistently over time.

In the context of high-performance sports, the present findings reveal novel insights into the brain’s role in exercise physiology. Future studies should indeed consider long-term adaptations of the brain to exercise. Prioritizing technique-specific endurance modalities in specific phases of the season (25) may result in chronic adaptations of the nervous system. To understand whether changes in brain communication contribute to or interfere with sports-specific performance, longitudinal investigations are required that consider both behavioral (e.g., technique) and neurophysiological (e.g., network efficiency) outcomes. Especially for exercise modes with even higher focus on sensorimotor control, brain network data may improve the understanding of differences in skill level and training adaptation (28).

Study limitations

Despite providing novel insights into the role of exercise mode on brain function, the experimental design and methodological aspects limit the generalizability of the observed findings. Notably, the investigated sample was small and consisted of highly trained male athletes only, because fitness level, technical execution, and menstrual cycle may interfere in with the results (56). However, this inclusion of highly trained individuals only limited the general availability of participants, leading to a relatively small sample size of n = 15. According to the neural efficiency hypothesis (57) and explorative insights from Ludyga et al. (58), the brain’s responses to exercise may vary depending on the levels of expertise. Therefore, it should be considered that the high training status of the investigated sample may have resulted in sample-specific modulations of brain function. To increase transferability, future investigations should consider including subgroups with different performance levels to investigate the effect of exercise mode on brain function (58), and female participants should be examined.

A further limitation was that the protocol order was not fully counterbalanced, because the experimental sessions were arranged according to the training schedule of the athletes. Therefore, more athletes started with RUN and less athletes (5 of 15) started with XC. Although the included XC athletes were familiar with the two investigated endurances modes and were regularly tested at the laboratory environment at the study site, potential limitations due this should be considered. Despite the fact that XC skiers train high amounts in both investigated modes throughout the entire year, they prioritized skiing modes in the competitive phase of the season (13). Therefore, variations in the distribution of the exercise modes across the season may have affected the familiarization and confounded mode-specific modulations of brain function. Taken together, experimental aspects like fitness level, sex, mode counterbalancing, and exercise familiarity should be considered as potential confounders of mode-specific modulations of brain function in athletes.

Furthermore, EEG-specific limitations affect the interpretation of the obtained findings. One important limitation comes with the choice of analyzing RSN organization in the channel space because of the lower reliability of reconstructing cortical sources in repeated-measures designs (59). Thus, possible effects of volume conduction on the computed connectomes cannot be ruled out by certainty, even if the volume conduction robust wPLI was chosen. However, Miljevic et al. (60) suggested that particularly for intraindividual analysis of global graph metrics, the volume conduction of underlying sources will remain stable and should not affect the data too much, resulting in more reliable results compared with source space approaches. Another methodological issue is the contribution of exercise-induced sweating to the EEG data. Sweating on the scalp surface increases the risk of electrical bridging and further results in slow drift artifacts contaminating the measured electrocortical potential changes. During processing, the impact of slow-drift sweating artifacts was minimized because of a high-pass filter (61), and bridged channels (0.2 ± 1.1 channels per recording) were identified and interpolated (34). In addition, the use of the wPLI as a FC metric insensitive for volume conduction and zero-lag connectivity particularly reduced the impact of electrical bridges on the FC estimation (36). However, systemic physiological responses such as sweating on noninvasive neurophysiological methods like the EEG remain complex sources of artifact and need to be considered in exercise neuroscience investigations (62).


The present investigation revealed insights into the role of exercise intensity and exercise mode on resting-state brain network responses to acute bouts of exercise. We observed that an acute bout of coordinatively challenging endurance exercise affects brain networks differently from an acute bout of RUN for athletes who are similarly trained in both modes. Furthermore, we observed a modulation of brain network efficiency after maximal intensity exercise bouts independent of the endurance mode performed. Therefore, we propose that future studies extend perspectives on brain–exercise interactions and consider endurance mode a moderator of brain network responses. This may allow training programs to be tailored in the context of prevention, rehabilitation, and high performance by stimulating specific physiological adaptations to exercise, including those of the brain.

The authors declare no conflict of interest. The results of the study are presented clearly, honestly, and without fabrication, falsification, or inappropriate data manipulation. The results of the present study do constitute endorsement by the American College of Sports Medicine.

Substantial contributions to the conception and design of the work were done by D. B., P. Ø. T., T. L., Ø. S., and J. B. The work draft and revision, as well as the approval of the final version, were done by D. B., P. Ø. T., T. L., Ø. S., and J. B. In addition, all aspects of this work related to the accuracy or integrity were ensured by D. B., P. Ø. T., T. L., Ø. S., and J. B.


1. El-Sayes J, Harasym D, Turco CV, Locke MB, Nelson AJ. Exercise-induced neuroplasticity: a mechanistic model and prospects for promoting plasticity. Neuroscientist. 2019;25(1):65–85.
2. De Sousa Fernandes MS, Ordônio TF, Santos GCJ, et al. Effects of physical exercise on neuroplasticity and brain function: a systematic review in human and animal studies. Neural Plast. 2020;2020:8856621.
3. Schmitt A, Upadhyay N, Martin JA, Rojas S, Strüder HK, Boecker H. Modulation of distinct intrinsic resting state brain networks by acute exercise bouts of differing intensity. Brain Plast. 2019;5(1):39–55.
4. Woo M, Kim S, Kim J, Petruzzello SJ, Hatfield BD. Examining the exercise-affect dose–response relationship: does duration influence frontal EEG asymmetry? Int J Psychophysiol. 2009;72(2):166–72.
5. Büchel D, Sandbakk Ø, Baumeister J. Exploring intensity-dependent modulations in EEG resting-state network efficiency induced by exercise. Eur J Appl Physiol. 2021;121(9):2423–35.
6. Hashimoto T, Tsukamoto H, Takenaka S, et al. Maintained exercise-enhanced brain executive function related to cerebral lactate metabolism in men. FASEB J. 2018;32(3):1417–27.
7. Herold F, Müller P, Gronwald T, Müller NG. Dose–response matters!—a perspective on the exercise prescription in exercise–cognition research. Front Psychol. 2019;10:2338.
8. Sandbakk Ø, Haugen T, Ettema G. The influence of exercise modality on training load management. Int J Sports Physiol Perform. 2021;16(4):605–8.
9. Losnegard T, Schäfer D, Hallén J. Exercise economy in skiing and running. Front Physiol. 2014;5:5.
10. Brümmer V, Schneider S, Abel T, Vogt T, Strüder HK. Brain cortical activity is influenced by exercise mode and intensity. Med Sci Sports Exerc. 2011;43(10):1863–72.
11. Ueta K, Mizuguchi N, Sugiyama T, Isaka T, Otomo S. The motor engram of functional connectivity generated by acute whole-body dynamic balance training. Med Sci Sports Exerc. 2022;54(4):598–608.
12. Netz Y. Is there a preferred mode of exercise for cognition enhancement in older age?—a narrative review. Front Med (Lausanne). 2019;6:57.
13. Sandbakk Ø, Holmberg HC. Physiological capacity and training routines of elite cross-country skiers: approaching the upper limits of human endurance. Int J Sports Physiol Perform. 2017;12(8):1003–11.
14. Takakusaki K. Neurophysiology of gait: from the spinal cord to the frontal lobe. Mov Disord. 2013;28(11):1483–91.
15. Kline JE, Huang HJ, Snyder KL, Ferris DP. Cortical spectral activity and connectivity during active and viewed arm and leg movement. Front Neurosci. 2016;10:91.
16. Weersink JB, Maurits NM, de Jong BM. Amble gait EEG points at complementary cortical networks underlying stereotypic multi-limb co-ordination. Front Hum Neurosci. 2021;15:691482.
17. Sipp AR, Gwin JT, Makeig S, Ferris DP. Loss of balance during balance beam walking elicits a multifocal theta band electrocortical response. J Neurophysiol. 2013;110(9):2050–60.
18. Dietrich A, Audiffren M. The reticular-activating hypofrontality (RAH) model of acute exercise. Neurosci Biobehav Rev. 2011;35(6):1305–25.
19. Kao S, Cadenas-Sanchez C, Shigeta T, et al. A systematic review of physical activity and cardiorespiratory fitness on P3b. Psychophysiology. 2020;57(7):e13425.
20. Mehta RK, Parasuraman R. Neuroergonomics: a review of applications to physical and cognitive work. Front Hum Neurosci. 2013;7:889.
21. Gramkow MH, Hasselbalch SG, Waldemar G, Frederiksen KS. Resting state EEG in exercise intervention studies: a systematic review of effects and methods. Front Hum Neurosci. 2020;14:155.
22. Imperatori LS, Betta M, Cecchetti L, et al. EEG functional connectivity metrics wPLI and wSMI account for distinct types of brain functional interactions. Sci Rep. 2019;9(1):8894.
23. Fox MD, Snyder AZ, Vincent JL, Corbetta M, Van Essen DC, Raichle ME. The human brain is intrinsically organized into dynamic, anticorrelated functional networks. Proc Natl Acad Sci U S A. 2005;102(27):9673–8.
24. Büchel D, Lehmann T, Sandbakk Ø, Baumeister J. EEG-derived brain graphs are reliable measures for exploring exercise-induced changes in brain networks. Sci Rep. 2021;11(1):20803.
25. Solli GS, Tønnessen E, Sandbakk Ø. The training characteristics of the world’s most successful female cross-country skier. Front Physiol. 2017;8:1069.
26. Torvik PØ, Solli GS, Sandbakk Ø. The training characteristics of world-class male long-distance cross-country skiers. Front Sports Act Living. 2021;3:641389.
27. Porter S, Silverberg ND, Virji-Babul N. Cortical activity and network organization underlying physical and cognitive exertion in active young adult athletes: implications for concussion. J Sci Med Sport. 2019;22(4):397–402.
28. Perrey S. Training monitoring in sports: it is time to embrace cognitive demand. Sports (Basel). 2022;10(4):56.
29. Faul F, Erdfelder E, Lang AG, Buchner A. G*Power 3: a flexible statistical power analysis program for the social, behavioral, and biomedical science. Behav Res Methods. 2007;39(2):175–191.
30. Haraguchi R, Hoshi H, Ichikawa S, et al. The menstrual cycle alters resting-state cortical activity: a magnetoencephalography study. Front Hum Neurosci. 2021;15:652789.
31. Losnegard T, Myklebust H, Spencer M, Hallén J. Seasonal variations in VO2max, O2-cost, O2-deficit, and performance in elite cross-country skiers. J Strength Cond Res. 2013;27(7):1780–90.
32. Talsnes RK, Solli GS, Kocbach J, Torvik PØ, Sandbakk Ø. Laboratory- and field-based performance-predictions in cross-country skiing and roller-skiing. PLoS One. 2021;16(8):e0256662.
33. Delorme A, Makeig S. EEGLAB: an open source toolbox for analysis of single-trial EEG dynamics including independent component analysis. J Neurosci Methods. 2004;134(1):9–21.
34. Alschuler DM, Tenke CE, Bruder GE, Kayser J. Identifying electrode bridging from electrical distance distributions: a survey of publicly-available EEG data using a new method. Clin Neurophysiol. 2014;125(3):484–90.
35. Palmer J, Kreutz-Delgado K, Makeig S. AMICA: An Adaptive Mixture of Independent Component Analyzers with Shared Components. San Diego: Swartz Center of Computational Neuroscience; 2011. pp. 1–15.
36. Vinck M, Oostenveld R, Van Wingerden M, Battaglia F, Pennartz CMA. An improved index of phase-synchronization for electrophysiological data in the presence of volume-conduction, noise and sample-size bias. Neuroimage. 2011;55(4):1548–65.
37. Rubinov M, Sporns O. Complex network measures of brain connectivity: uses and interpretations. Neuroimage. 2010;52(3):1059–69.
38. Lakens D. Calculating and reporting effect sizes to facilitate cumulative science: a practical primer for t-tests and ANOVAs. Front Psychol. 2013;4:863.
39. Holmberg HC, Lindinger S, Stöggl T, Eitzlmair E, Müller E. Biomechanical analysis of double poling in elite cross-country skiers. Med Sci Sports Exerc. 2005;37(5):807–18.
40. Halgren M, Ulbert I, Bastuji H, et al. The generation and propagation of the human alpha rhythm. Proc Natl Acad Sci U S A. 2019;116(47):23772–82.
41. Klimesch W. Alpha-band oscillations, attention, and controlled access to stored information. Trends Cogn Sci. 2012;16(12):606–17.
42. Krause CM, Sillanmäki L, Koivisto M, et al. The effects of memory load on event-related EEG desynchronization and synchronization. Clin Neurophysiol. 2000;111(11):2071–8.
43. Zhu Y, Wang Q, Zhang L. Study of EEG characteristics while solving scientific problems with different mental effort. Sci Rep. 2021;11(1):23783.
44. Büchel D, Lehmann T, Ullrich S, Cockcroft J, Louw Q, Baumeister J. Stance leg and surface stability modulate cortical activity during human single leg stance. Exp Brain Res. 2021;239(4):1193–202.
45. Boenstrup M, Feldheim J, Heise K, Gerloff C, Hummel FC. The control of complex finger movements by directional information flow between mesial frontocentral areas and the primary motor cortex. Eur J Neurosci. 2014;40(6):2888–97.
46. Meyer F, Kocbach J, Tjønnås J, et al. Temporal and kinematic patterns distinguishing the G2 from the G4 skating sub-technique. Sports Biomech. 2021;1–18.
47. Vecchio F, Miraglia F, Alù F, et al. Human brain networks in physiological and pathological aging: reproducibility of electroencephalogram graph theoretical analysis in cortical connectivity. Brain Connect. 2022;12(1):41–51.
48. Tan B, Liu Q, Wan C, Jin Z, Yang Y, Li L. Altered functional connectivity of alpha rhythm in obsessive-compulsive disorder during rest. Clin EEG Neurosci. 2019;50(2):88–99.
49. Sauseng P, Klimesch W, Schabus M, Doppelmayr M. Fronto-parietal EEG coherence in theta and upper alpha reflect central executive functions of working memory. Int J Psychophysiol. 2005;57(2):97–103.
50. Cheron G, Petit G, Cheron J, et al. Brain oscillations in sport: toward EEG biomarkers of performance. Front Psychol. 2016;7:246.
51. Bucher E, Sandbakk Ø, Donath L, Roth R, Zahner L, Faude O. Exercise-induced trunk fatigue decreases double poling performance in well-trained cross-country skiers. Eur J Appl Physiol. 2018;118(10):2077–87.
52. Magistretti PJ, Allaman I. Lactate in the brain: from metabolic end-product to signalling molecule. Nat Rev Neurosci. 2018;19(4):235–49.
53. Tamburro G, di Fronso S, Robazza C, Bertollo M, Comani S. Modulation of brain functional connectivity and efficiency during an endurance cycling task: a source-level EEG and graph theory approach. Front Hum Neurosci. 2020;14:243.
54. Voelcker-Rehage C, Godde B, Staudinger UM. Cardiovascular and coordination training differentially improve cognitive performance and neural processing in older adults. Front Hum Neurosci. 2011;5:26.
55. Haugnes P, Kocbach J, Luchsinger H, Ettema G, Sandbakk Ø. The interval-based physiological and mechanical demands of cross-country ski training. Int J Sports Physiol Perform. 2019;14(10):1371–7.
56. Del Percio C, Babiloni C, Marzano N, et al. “Neural efficiency” of athletes’ brain for upright standing: a high-resolution EEG study. Brain Res Bull. 2009;79(3–4):193–200.
57. Li L, Smith DM. Neural efficiency in athletes: a systematic review. Front Behav Neurosci. 2021;15:698555.
58. Ludyga S, Gronwald T, Hottenrott K. The athlete’s brain: cross-sectional evidence for neural efficiency during cycling exercise. Neural Plast. 2016;2016:4583674.
59. Fraschini M, La Cava SM, Didaci L, Barberini L. On the variability of functional connectivity and network measures in source-reconstructed eeg time-series. Entropy (Basel). 2021;23(1):5.
60. Miljevic A, Bailey NW, Vila-Rodriguez F, Herring SE, Fitzgerald PB. Electroencephalographic connectivity: a fundamental guide and checklist for optimal study design and evaluation. Biol Psychiatry Cogn Neurosci Neuroimaging. 2022;7(6):546–54.
61. Beniczky S, Schomer DL. Electroencephalography: basic biophysical and technological aspects important for clinical applications. Epileptic Disord. 2020;22(6):697–715.
62. Stone DB, Tamburro G, Fiedler P, Haueisen J, Comani S. Automatic removal of physiological artifacts in EEG: the optimized fingerprint method for sports science applications. Front Hum Neurosci. 2018;12:96.


Copyright © 2023 by the American College of Sports Medicine