Methods

Participants

Thirty-two undergraduate students in their third year studying sport sciences at the University of Poitiers participated in this study. The participants were randomly assigned into two groups: a control group and an experimental group. Their characteristics can be found in Table 1. All of the participants were native French speakers and right-handed except one in the experimental group who was left-handed. All participants had normal to corrected-to-normal vision with no reports of color blindness. None of the participants revealed a history of language difficulties, learning impairments, or severe neurological, metabolic and/or psychiatric disorders. For the sake of our handgrip endurance-dependent task, we did not recruit individuals who had been engaged in a muscle reinforcement activity for more than 6 months or anyone who had injuries or functional problems with the dominant hand. The experimental procedure was explained to the participants via a written information sheet. Then, they signed a consent form to indicate their agreement. Finally, the participants received course credit in exchange for their participation. This experiment was conducted in accordance with the Declaration of Helsinki and with the approval of the University of Tours-Poitiers ethics committee (serial number 2019-01-02).

Procedures

This experiment occurred during a single session, and the participants in both groups underwent the exact same procedures. The session started with asking the participants to complete a sociodemographic questionnaire (height, weight and age) and a hand laterality scale (Oldfield, 1971) to determine their dominant hand. The participants were also tested for color blindness (Ishihara, 1918) to determine whether they were able to perform the Stroop task. Then, the participants answered questions regarding their general state and health. Certain experimental precautions related to any parameters that could have an effect on mental strain and fatigue were considered, such as not consuming alcohol, caffeine, nicotine or psychostimulant substances; avoiding stressful events and/or intense physical activity several hours before the session; and having a sufficient amount of sleep the night before the experiment. Later, the participants underwent a familiarization period to understand the flow and instructions of the mental and physical tasks. In the familiarization phase, the participants performed 48 trials of the incongruent Stroop task and tried out the handgrip device. Afterward, they were equipped with EEG and ECG electrodes. Next, they underwent the sequential stages of the experimental protocol while their electrophysiological data were recorded on a continuous basis, as shown in Figure 1. First, the participants were asked to perform a time-to-exhaustion handgrip task to determine their physical performance baseline, while their maximal voluntary contraction (MVC) was assessed before and after this task. Second, they performed a 30-min mental task that differed in each of the two groups. The experimental group performed a 30-min modified incongruent Stroop task tapping two executive functions: inhibitory control and cognitive flexibility. The control group watched a 30-minute documentary video. After the completion of the mental task, the participants repeated the time-to-exhaustion handgrip task and the MVC measurements before and after this physical task. Every 30 seconds during each handgrip task, the participants were asked to announce the amount of effort that they invested in the task, as well as the level of pain that they experienced in their dominant forearm and hand. Several other subjective measurements were made throughout the session at four different times: T1 = Baseline, T2 = Pre-Mental Task, T3 = Post-Mental Task, T4 = Post-Handgrip Task. We used a computerized visual analog scale (VAS) with an adapted scale (0% to 100%) to rate the level of their perceived fatigue (T1 to T4), the motivation to perform the physical task (T1 and T3), the perceived difficulty of the mental task (T3) and finally the level of boredom experienced during the mental tasks (T3). The motivation to perform the dependent task and subjective mental fatigue were already assessed with a similar VAS (Brown & Bray, 2017), and this type of VAS was validated by Wewers and Lowe (1990). Finally, a computerized version of the Brief Self-Control Scale (Tangney et al., 2004) was used to assess the participants’ trait self-control due to the probable influence of this trait on the ego-depletion effect. Previously, it has been argued that individuals with a high self-control trait tend to demonstrate a lower ego-depletion effect than individuals with a low self-control trait (DeWall et al., 2007).
———————————–
Insert Figure 1 about here
———————————–

Materials

Time-to-exhaustion task

To measure the isometric handgrip force of the individuals’ dominant hand, we used a dynamometer (TSD121C, BIOPAC). AcqKnowledge software, version 4.2 (BIOPAC Systems Inc., Goleta, CA, USA), as well as an MP160WSW data acquisition unit, was used to record the force signal. Data were recorded online at a sampling rate of 2000 Hz and later stored and analyzed offline. The participants were asked to sit on a fixed chair with an arm support allowing for the imposition of an angle of ~90 degrees with their elbow and forearm, as shown in Figure 2. The participants were asked to remain in this anatomically neutral position throughout the handgrip task.
———————————–
Insert Figure 2 about here
———————————–
The participants started the physical task by performing an MVC to measure their true maximal force. For this measurement, they had to squeeze the dynamometer as strongly as possible for a duration of 3 seconds. This process was repeated until the time that the participants could not score higher in contraction peak force compared to their previous performances. Between each MVC measurement, there was a 30-s rest. The highest MVC measured before the first time-to-exhaustion task was used throughout the session as the reference for force feedback to calibrate the target zone for the time-to-exhaustion handgrip task (see the gauge on Figure 2). The target zone area was defined as a green arc representing 12% to 14% of the MVC. The participants were asked to maintain the contraction of their forearm in this area until exhaustion. Time to exhaustion was evaluated offline and considered the duration of the isometric contraction from the onset of the force signal to the exhaustion time (considered as staying out of the zone for more than 2 s). To assess the extent of muscle fatigue and to ensure that the participants performed the handgrip task as much as they could, we compared the difference between the MVC peak force measured before and after the time-to-exhaustion task. During the time-to-exhaustion task, every 30 s, we asked the participants to rate their perception of effort, which is referred to as the effort intensity necessary to squeeze the dynamometer while breathing and remaining in the green zone (Mangin et al., 2021) on the CR100 scale (Borg & Kaijser, 2006), as well as their perception of muscle pain, defined as the perceived pain intensity in their forearm muscles during the endurance task performance (Mangin et al., 2021) on the Cook scale (O’Connor & Cook, 2001). The ”individual isotime” method (Nicolo et al., 2019) was used to analyze the participants’ perceptions of muscle pain and effort. This method considers the shortest performance record of the participant on the handgrip task as his or her 100% individual isotime. Subsequently, the corresponding isotime points to 0%, 33%, and 66% of the individuals were calculated for their shortest performance. According to the calculated isotime points, their level of muscle pain and effort was assessed as a function of time on task.

Mental tasks

We used a computerized version of a modified Stroop task on a computer equipped with an S-R response box and E-prime software, version 2.0 (Psychological Software Tools, Pittsburgh, PA, USA). The participants sat in front of a screen and responded orally to the visual stimuli that appeared at the middle of the screen while their voices were recorded via two microphones, one to record the oral response of the participant (headset) and the other to measure his or her response time (fixed microphone). The participants underwent 888 trials of 2 seconds each. Every single trial started with a fixation point in the form of a cross that lasted 400 ms. The fixation point could be enclosed in a circle or a square for a duration of 50 ms, and then the cross remained on the screen alone. Immediately thereafter, a color name (green, red, yellow or blue) written in another ink color (e.g., “yellow” written in red) was displayed on the screen. The participants had to read the word if they saw a square (reading trials) or name the color of the ink if they saw a circle (naming ink color trials). The color word lasted on the screen until the participant’s response. If the participants did not answer (omission) or had a reaction time longer than 1250 ms, the color word lasted 1250 ms and was followed by a fixation cross lasting 300 ms. When the participant answered before 1250 ms had elapsed, the fixation cross remained in the middle of the screen for the same amount of time (1250 ms + 300 ms) to have a duration of 2 s for each trial. The presentation order of the two categories of trials was completely random. The block of trials included 50% trials in which the participants had to read the color word and 50% of trials in which they had to name the font color for the sake of increasing the task difficulty by engaging cognitive flexibility and inhibitory control (Mangin et al., 2021) and limiting the learning effect (Dulaney & Rogers, 1994). Concerning the modified Stroop task, we were interested in analyzing performance as a function of time on task. Consequently, data were divided into 4 consecutive periods of 222 trials lasting 7 minutes and 30 seconds. For each time period, we calculated different performance parameters, such as the mean reaction time for correct responses (RT) and the error rates. For the video task, we chose a documentary movie named Earth by Fothergill and Linfield (2009). Later, the participants were asked to complete a multiple-choice questionnaire related to the content of the documentary to verify whether they were actively watching the movie or not. They did not know in advance that they will have to complete this questionnaire.

EEG Acquisition and Pre-Processing Pipeline

The EEG data were collected from a 64-channel Biosemi EEG headset with the electrode distribution based on the international 10-20 system. For the sake of better eye movement artifact detection, the participants’ ocular activities were recorded via three electrooculogram (EOG) electrodes (two on the outer canthus of each eye and one on the infraorbital region of the right eye). The EEG and EOG signals were continuously recorded online throughout the whole experiment using the ActiView Biosemi version 6.05 acquisition system at a frequency of 2000 Hz referenced online to the average of the right and left mastoids. The MATLAB R2020b programming platform (MathWorks Inc., Natick, MA, USA), together with the open source EEGLAB 2021.0 toolbox (Delorme & Makeig, 2004), was used for offline data analysis. Data preprocessing steps and artifact rejection methods were adopted from a very recent reproducible workflow by Pernet and colleagues (2021). First, data were down sampled to 250 Hz, and a basic low-pass FIR filter to the higher edge of 40 Hz was applied to avoid 50-Hz line noise. Then, the clean_rawdata plugin in EEGLAB (version 2.2) was used to high-pass filter the data at 0.5 Hz (transition band [0.25 0.75]), as well as remove the bad channels (any channel with at least a 5-s flat line and/or with less than 0.8 robust estimate correlation to the other channels were considered bad channels). Next, the data were rereferenced to the average of the existing channels. Afterward, the data were decomposed into independent components using the ICA algorithm (runica algorithm with rank reduction based on the number of channels = -1, considering the average reference). An automatic algorithm was applied to label the components using ICLabel (Pion-Tonachini et al., 2019), and components labeled as muscle activities and eye movements with greater than 80% probability were omitted from the data. Then, the residual artifacts were removed using the artifact subspace reconstruction (ASR) algorithm parametrized to a 20-burst detection criteria threshold (Chang et al., 2018, 2020). Finally, we inspected the data visually and rejected any nonbrain components or artifactual portions of the data that were not detected hitherto by the automatic algorithm. No more than half of the components in each data point were rejected (the maximum number of component rejections was 32 of 64). A power spectral analysis was applied to observe the differences in the power of theta wavebands (4-7 Hz) during the Stroop task performance as opposed to the video task. The “spectopo” function in EEGLAB was used to perform the power spectral analysis and to compute each component power spectrum using the fast Fourier transform (FFT) with the following parameters: a window size of 1 second with a 50% overlap. We were also interested in the effect of time on task during the Stroop task on the theta waveband power; therefore, we also studied the data by dividing the 30 minutes corresponding to the Stroop task into 4 consecutive task periods of 7.5 minutes.

Task-related theta activity

According to the methodology used by Arnau et al. (2021), we conducted an analysis of stimulus-locked theta power during the Stroop task. Throughout the Stroop task, when the color word was displayed on the screen, a marker was sent to ActiView Biosemi with the help of the Eprime program indicating that the ongoing trial is either a ‘naming ink color’ trial or a ‘reading’ trial. Later, during the EEG offline analysis, the data were analyzed in the following time window: from 100 ms before the stimulus onset to 2000 ms after this onset. After removing the baseline recorded during the time window from 100 ms before the stimulus to the onset of the stimulus, we separated the 444 trials belonging to the ‘naming ink color’ condition from the 444 trials belonging to the ‘reading’ condition for further analyses. Then, the power spectral analysis was applied to observe the changes in theta wave band by comparing these two types of trial (reading vs. naming ink color) as the function of TOT.

Source localization technique

The DIPFIT plugin (Oostenveld & Oostendorp, 2002) in the EEGLAB 2021.0 toolbox (Delorme & Makeig, 2004) was used to calculate an equivalent current dipole model for each independent component through a four-shell spherical head model. A bilaterally symmetric dual dipole model was fitted for the components with bilaterally distributed scalp maps. Only components located inside the model of brain volume, for which their best-fitting single or dual equivalent dipole showed less than 15% residual variance from the spherical forward-model scalp projection, were contemplated for further analysis.

ECG

Heart rate variability (HRV) was continuously recorded using the Electrocardiograph (BIOPAC Systems Inc., Goleta, USA) and Acqknowledge 4.2 software (BIOPAC Systems Inc., Goleta, CA, USA) at a frequency of 2000 Hz by placing three EL503 electrodes on the participant’s thorax as recommended by the American Heart Association (Kligfield et al., 2007). The Kubios HRV Premium software, version 3.5.0 (Tarvainen et al., 2014), was used to analyze the data. After applying the medium automatic artifact correction algorithm available in Kubios, the residual artifacts were rejected via manual inspection. One subject in the control group was excluded from further analysis due to the low quality of his or her ECG data. We studied the changes in HRV on the temporal basis of four equal time windows of 7 minutes and 30 seconds. We analyzed the data by examining the time and frequency domain parameters of HRV. For the time domain analysis, we chose the SDNN (standard deviation of NN intervals), and for the frequency domain analysis, the LF power (log power of the low-frequency band from 0.04 to 0.15 Hz), as well as HF (log power of the high-frequency band from 0.15 to 0.4 Hz), was assessed (Shaffer & Ginsberg, 2017). The SDNN demonstrates the components that are responsible for the variability in the recording period (Malik, 1996). High frequency can be considered the cardiac parasympathetic tone index (Reyes del Paso et al., 2013) since it reflects vagal tone (Laborde et al., 2017; Malik, 1996). Low frequencies are assumed to be markers of cardiac outflow, which is under the influence of both the sympathetic and parasympathetic autonomic nervous systems (Laborde et al., 2017; Malik, 1996).

Statistical analysis

The statistical analyses were performed with Jamovi software, version 2.2.5, and Jasp software, version 0.16.3.0. The statistical analysis of EEG data was performed in EEGLAB using EEGLAB Study parametric statistics. We set the alpha level for statistical significance to α = .05. When the results were significant and marginal, the effect sizes were calculated: Cohen’s d for the t test (only Student’s t test was used in this study), rank biserial correlation for its equivalent nonparametric Mann-Whitney U test and partial eta square (ηp2) for analysis of variance (ANOVA). When testing an effect involving a repeated-measures factor with more than two levels (e.g., time on task), we applied a Greenhouse-Geisser correction to consider any violation of the sphericity assumption. For EEG data, the FDR multiple comparison correction available in EEGLAB statistics was considered.