Comparing the effect of mental fatigue-inducing models on selected cognitive and technical performance aspects in young soccer players

Abstract

Mental fatigue is a well-documented performance inhibitor in team sports, Therefore, identifying strategies to attenuate mental fatigue seems necessary. This study aimed to evaluate and compare four distinct training models—Modified Stroop, SAFT⁹⁰, T-SAFT⁹⁰, and a combined T-SAFT⁹⁰ + Stroop protocol—to identify the most effective method for inducing mental fatigue under controlled laboratory conditions, as a potential model for brain endurance training (BET) research in young soccer players. Fifteen male players (aged 16–18) participated in a randomized cross-over study. Mental fatigue was assessed via a Visual Analogue Scale (VAS; primary outcome), cognitive performance (secondary outcomes) was evaluated through response time, response accuracy, working memory capacity, visual Scanning Identification, and auditory pattern recognition. Technical performance was measured using penalty time, movement time and passing accuracy in the Loughborough Soccer Passing Test (LSPT; primary outcomes). All protocols significantly increased mental fatigue, with the largest effect observed in the combined T-SAFT⁹⁰ + Stroop model. Response accuracy declined across all models, while response time worsened in the Stroop and T-SAFT⁹⁰ conditions. Penalty time increased in the Stroop and T-SAFT⁹⁰ protocols, whereas passing accuracy decreased most significantly in the combined model. In summary, under standardized, controlled conditions, the combined cognitive-physical training model induced the highest mental fatigue and most consistently altered cognitive and technical performance. These findings provide preliminary evidence supporting its potential as a BET model for research and structured training environments, though ecological validation in real soccer contexts remains necessary.

Introduction

Soccer is a dynamic sport that places considerable cognitive, technical, and physical demands on young athletes^1–3. During competitive play, players must rapidly process complex information, make split-second decisions, and execute precise motor skills under pressure. These demands can lead not only to physical fatigue but also to mental fatigue—a psychobiological state caused by prolonged or intense cognitive activity, characterized by subjective tiredness and objective declines in cognitive and motor performance^4–6. Mental fatigue impairs core cognitive functions essential to soccer, attention, anticipation, and decision-making^7,8. Empirical studies show it reduces passing accuracy, shooting precision, and running efficiency, while increasing perceived exertion⁹. Despite this evidence, traditional youth soccer training has often centered solely on physical conditioning while neglecting cognitive fatigue management¹⁰.

Brain endurance training (BET), the integration of cognitive tasks with physical exercise, has been proposed a strategy to build cognitive resilience against mental fatigue¹¹. BET improve endurance performance and reduce the cognitive cost of effort, potentially by enhancing prefrontal oxygenation and functional brain connectivity.¹²However, BET requires an effective method to induce mental fatigue during training sessions. The choice of fatigue-inducing protocol is therefore critical, as its characteristics (e.g., ecological validity, cognitive load) may influence both fatigue magnitude and transfer to sport performance¹³.

Commonly used tasks include the Stroop test, which taxes response inhibition and selective attention¹⁴. While the classic Stroop task offers high internal validity for isolating cognitive load, the soccer-specific field protocols like SAFT⁹⁰ ¹⁵ and its technical variant T-SAFT⁹⁰ ¹⁶ which embed physical exertion with sport-specific decisions were specifically designed to mirror the intermittent, decision-intensive, and skill-demanding nature of competitive soccer, thereby enhancing ecological fidelity. Tasks that mimic decision-making under pressure or multitasking may be more relevant for soccer players and can improve both cognitive and physical performance under fatigue conditions¹⁷. Recent work even suggests that repeated technical drills (e.g., LSPT) can induce mental fatigue with higher ecological validity. A study reported that the 20-min repeated interval LSPT(Loughborough Soccer Passing Test) can induce a similar mental fatigue as the Stroop task and was associated with a better ecological validity¹⁸.

However, no study has directly compared the acute fatigue-inducing capacity of these distinct models under controlled conditions. Importantly, this study does not evaluate BET as a long-term training intervention. Instead, it aims to identify which protocol most effectively induces acute mental fatigue—a necessary first step before selecting models for future BET research. Therefore, we compared the effects of the Modified Stroop, SAFT⁹⁰, T-SAFT⁹⁰, and a combined T-SAFT⁹⁰ + Stroop protocol on cognitive (response time, accuracy, working memory, visual scanning, auditory recognition) and technical (penalty time, movement time, passing accuracy) performance in young soccer players.

Methods

Participants

Fifteen young soccer players (age: 16–18 years) were recruited from state Premier League who regularly participated in training and competition. Eligible participants had ≥ 2 years of competitive experience, with equal representation across playing positions (5 midfielders, 5 defenders, 5 forwards). An a priori power analysis was performed using G*Power 3.1 to determine the minimum sample size required for the primary within-subject analyses. The planned analysis was a repeated-measures ANOVA examining the effects of Condition (4 training models) and Time (pre, post) within participants (i.e., a within × within interaction). Based on prior studies of mental fatigue in soccer and cognitive task performance, ¹⁹ a medium effect size (Cohen’s f = 0.29) was expected. We set the type I error rate to α = 0.05 and desired statistical power to 1 − β = 0.80.

Although nonparametric statistics were ultimately used due to non-normal data distribution (confirmed by Kolmogorov–Smirnov test), a priori power analysis followed standard recommendations for crossover designs by using a repeated-measures ANOVA model in GPower¹⁹. This approach is widely accepted in sport and exercise science when estimating sample size for within-subject comparisons, even when nonparametric tests are applied post hoc²⁰. Following common practice for crossover repeated-measures designs, we conservatively assumed a moderate correlation among repeated measures (r = 0.50) and no violation of sphericity (ε = 1.0). With these inputs (number of measurements = 4 conditions × 2 time points = 8 levels for the within-subject cells, effect size f = 0.29, α = 0.05, power = 0.80, correlation among repeated measures = 0.50, ε = 1), the analysis indicated a minimum total sample size of 12 participants. To allow for possible attrition, missing data or unusable trials, we recruited 15 male soccer players.

The inclusion criteria were: (1) an experience of participating in official soccer competitions at the provincial or county level within the past two years, (2) no supplement use in the past six months, (3) health certificate approved by a physician, and (4) an age range of 16 to 18 years. The exclusion criteria included: (1) taking any supplements or medications during the study, (2) unwillingness to continue participating in the interventions, (3) an injury that affects athletic performance and research outcomes, and (4) failure to follow the research guidelines, and (5) not adherence to pre-testing protocols (e.g., avoid consuming caffeine for ≥ 12 h, 8 h of sleep, light pre-session meal).

Informed consent

was obtained from all participants prior to their inclusion in the study. For participants under the age of 18, written parental informed consent was also obtained. The study was conducted in accordance with the ethical standards of the Ethics Committee of Shiraz University (Approval No: IR.US.PSYEDU.REC.1403.033) and the principles of the World Medical Association Declaration of Helsinki. Participants were provided with written instructions outlining the studies_ procedures but were not informed of their aims. Participants did not report no adverse effect.

Outcomes

The primary outcomes of this study were subjective mental fatigue (measured via VAS) and soccer-specific technical performance (assessed by penalty time, movement time, and passing accuracy in the Loughborough Soccer Passing Test, LSPT). These were selected based on their direct relevance to the study’s aim of identifying the most effective mental fatigue-inducing model for brain endurance training in soccer.

Secondary outcomes included cognitive performance measures: response time and accuracy on the Stroop task, working memory capacity, visual scanning identification, and auditory pattern recognition (assessed via Captain’s Log software). These were included to explore the cognitive mechanisms underlying mental fatigue and its impact on technical performance.

Experimental overview

This study employed a randomized, counterbalanced, crossover design with repeated measures. While participants and trainers could not be blinded to the intervention type (e.g., cognitive vs. physical tasks), outcome assessors and data analysts were unaware of session order and the specific hypotheses being tested, minimizing evaluation bias. One familiarization session preceded the four experimental sessions to control for learning effects.

Participants were randomly assigned to training sessions conducted between 17:00 and 20:00 on outdoor artificial turf, with a 4-day washout period between sessions. As indicated in Fig. 1; Table 1, the study comprised five sessions: one familiarization session and four experimental sessions which were performed in a randomized and counterbalanced order provided by a computer-generated Latin square design, ensuring that each condition appeared equally often in each position across the sample. This approach balanced potential order effects and minimized carryover bias.

Fig. 1.

Timeline of the research methodology. (a) Timeline of protocol sessions and tests (b) Timeline of research design; VAS: visual analog scale, captain, s log: cognitive performance assessment software, LSPT: Loughborough Soccer Pass Test, Stroop: Stroop task, Modified Stroop (Modified Stroop training model), T-SAFT⁹⁰ (T-SAFT ⁹⁰ training model), SAFT⁹⁰ (SAFT⁹⁰ training model), T-SAFT⁹⁰ + Stroop (Combined T-SAFT ⁹⁰ + Stroop training model). This study has used Bian, et al. protocol¹⁸with some revisions.

Table 1.

Schematic research design: LSPT (Loughborough Football Passing Test), Stroop (Cognitive Performance Assessment Test), Captain log (cognitive performance assessment software), VAS (Visual Analog Scale), Modified Stroop (Modified Stroop training model), T-SAFT90 (T-SAFT 90 training model), SAFT90 (SAFT90 training model), T-SAFT90 + Stroop (Combined T-SAFT 90 + Stroop training model)

Sequence generation and allocation concealment were performed by an independent researcher not involved in data collection using a random number generator in Microsoft Excel. The resulting sequence was concealed in sealed opaque envelopes until the morning of each session to prevent anticipation or bias. Participants and testers were blinded to the overall study hypotheses and the sequence of conditions.

To assess potential order effects, we conducted a secondary analysis using Friedman tests comparing pre-test scores (VAS, cognitive, and technical measures) across the four time points (i.e., before each of the four sessions). No significant differences were found (all p > 0.05), indicating no systematic learning or fatigue carryover at baseline. Additionally, to evaluate carryover effects, we compared post-washout pre-test values (i.e., pre-test of Session 2, 3, and 4) against the initial baseline (Session 1 pre-test) using Wilcoxon signed-rank tests. All comparisons were non-significant (p > 0.10), supporting the adequacy of the 4-day washout period in eliminating residual fatigue or practice effects.

During familiarization, participants performed all tasks to minimize learning effects. Each experimental session included warm-up, a pre-test battery tests (LSPT, Stroop Task, Captain^,s Log cognitive performance software, and VAS), a 30-minute intervention (modified Stroop, SAFT⁹⁰, T-SAFT⁹⁰, or combined T-SAFT⁹⁰ + Stroop), and an immediate post-test replicating the pre-test measures. Heart rate and rate of perceived exertion were continuously monitored. Participants rated their perceived mental fatigue on a visual analog scale (VAS). Participants were instructed to adhere to the following protocols before each subsequent visit: a minimum of 7 h of sleep, hydration of at least 35 mL of water per kilogram of body weight, abstinence from alcohol, and avoidance of vigorous physical activity on the preceding day. Additionally, they were asked to refrain from caffeine and nicotine consumption for 4 h prior to testing and to consume a similar meal during four sessions 4 h beforehand. All procedures were conducted under consistent outdoor conditions, with standardized verbal and written instructions. Participants and outcome assessors were blinded to the detailed research aims.

Measurements and tests

Mental fatigue assessment

The assessments of MF in this study can be categorized as a subjective report, cognitive performance, and soccer-specific skill performance. The changes of VAS values immediately before and after the MF-inducing tasks indicated the subjective perception of MF. The Stroop-3 in the pre-and posttest was used to assess the cognitive performance with the measures of response time and response accuracy. The Captain’s Log Cognitive Software (Captain’s Log Mind Power Builder, v2020.2, Brain Train, Inc.) was used to assess cognitive performance across multiple domains: working memory capacity, auditory pattern recognition, and visual scanning identification. This software has been utilized in previous sport science research to evaluate cognitive function under fatigue and dual-task conditions ²¹, and its tasks are based on well-validated neuropsychological paradigms (e.g., n-back for working memory, pattern recognition for auditory processing) that are sensitive to changes in cognitive load and mental fatigue²². The LSPT in the pre- and posttest was used to evaluate soccer skill performance with the measures of movement time, penalty time and passing accuracy (number of perfect passes).

Visual analog scale (VAS) test

Mental fatigue was assessed using a 100-mm VAS, where players marked their current fatigue level on a printed line in response to the question “What is your current perception of mental fatigue?” Scores were calculated in arbitrary units (AU) by measuring the distance (in millimeters) from the left end of the scale to the participant’s mark. Before each session, players received a definition clarifying that “Mental fatigue (MF) is a psychobiological state characterized by feelings of tiredness and a lack of energy and is induced by prolonged periods of demanding cognitive activity” and distinct from physical fatigue⁹. For participants needing clarification, related examples (e.g., “studying for a long exam”) were provided to illustrate mental fatigue. This test is the most frequent used instrument for mental fatigue⁴ and the most sensitive method to detect mental fatigue in athletes.²¹

Stroop cognitive performance test

The Stroop task in the pre- and posttest was obtained from the version reported by Badin et al. (2016)²³ and Gantois et al. (2020).²⁴ The Stroop test evaluates cognitive flexibility and response inhibition through three phases: (1) a practice stage (color-matching circles with unrecorded feedback), (2) a trial stage (test familiarization), and (3) the main test, featuring 48 congruent and 48 incongruent word-color trials where participants must identify ink color while suppressing word meaning. Colored words (red, blue, green, and yellow) were presented one at a time on the screen with a black background. Trials were arranged in a pseudorandom order with 50% of them being congruent (matched word and color) and 50% being incongruent, with all incongruent word–color combinations occurring with equal frequency. Players were required to respond to each trial by pressing one of the four keys on the bottom edge of the screen and then to choose the color of the word rather than its meaning. However, to increase task difficulty, if the word was displayed on screen in the color of red, the correct response was to press the key corresponding to the meaning of the word²³. The stimulus did not fade from the screen until a response was given. When the answer was correct, the stimulus disappeared, and a new one was set immediately, while any incorrect answer elicited a beep sound to prompt more accurate performance and a new stimulus subsequently appeared. In the pre- and posttest, the 3-min short version of the Stroop task (Stroop-3) was used to assess the cognitive performance, with the measures of response time and response accuracy.

In all cognitive tests, Players were instructed to sit on the bench at the corner of the court and to hold the tablet (Thinkpad, Lenovo Inc., China) to perform the cognitive task individually under the supervision of one examiner. Players were provided noise-canceling headphones to minimize distractions. Gestures were only given for indicating the start and end of the task.

Motivation assessment

Pre-session motivation was measured using a 100-mm Visual Analog Scale (VAS), consistent with Smith et al. (2016).⁹ participants rated their response to “How motivated are you to perform these tests?” by marking a plastic ruler (0 = minimum, 100 = maximum).

Auditory performance assessment

Auditory performance was measured with the Auditory Pattern Recognition Option of Captain’s Log cognitive performance software (Captain’s Log Mind Power Builder, v2020.2 @2020, Brain Train, Inc): Participants compared two rhythmic drum while viewing drum images with response buttons (red="Different”, green="Same”). If the two drum rhythmics are exactly alike, Participant should click the same button, and if two rhythmics were different, Participant click the different button. Immediate feedback was provided via visual symbols (star/stop sign) and auditory cues. Performance was graded on a 100-point scale.

Visual scanning assessment

Visual scanning performance was measured with the Visual scanning identification Option of Captain’s Log cognitive performance software (Captain’s Log Mind Power Builder, v2020.2 @2020, Brain Train, Inc): Participants identified moving shapes matching four on-screen reference boxes. Correct selections triggered a star and shape disappearance, while errors showed a stop sign. Reference shapes changed periodically, requiring memorization. Performance was graded on a 100-point scale.

Working memory capacity assessment

Working Memory performance was measured with the Working Memory Capacity Option of Captain’s Log cognitive performance software (Captain’s Log Mind Power Builder, v2020.2 @2020, Brain Train, Inc). n-back task in this option of cognitive performance software, as a commonly used task to assess working memory capacities.²⁵ In the n-back task, participants see a series of shapes (each presentation of a shapes is referred to as a trial) and have to decide whether the same shape was presented n items before. in this software, Participants tracked shapes moving right-to-left through a magnifying glass, click on a magnifying glass is only required if a so-called target trial is presented. If the shape n positions before does not match the current shape and it is therefore a non-target trial the participants should not react. In this software, the n was set to three. High amount of correct responses and faster response times, represent better working memory²⁶. This task Performance was graded on a 100-point scale.

Technical performance assessment

Technical performance was measured with The Loughborough Soccer Passing Test (LSPT). LSPT Was validated to assess the soccer-specific technical performance.²⁷ This test evaluates technical passing and control skills in a 12 × 9.5 m area featuring four benches with colored target zones (blue, white, green, red) marked by 10-cm aluminum strips for audible feedback. Participants, positioned centrally, respond to referee-called colors by completing 16 passes (8 near: red/white; 8 far: green/blue). Performance is scored via time penalties: +5 s (wrong bench), +3 s (missed zone/handball), +2 s (out of bounds/obstacle contact), +1 s (per second over 43 s), and −1 s (hitting the central strip). Outcomes include execution time, penalty time, performance time, movement time and accurate pass (correct pass) count. This test has been previously used in the topic of mental fatigue and its effects on soccer technical performance⁹.

Training models

Modified Stroop (Mental fatigue induce Model)

The computerized Stroop test assesses key cognitive functions such as response inhibition, selective attention, cognitive flexibility, and induce mental fatigue, particularly in athletic populations. With regard the Incongruent Stroop to induce mental fatigue, 30 min of incongruent Stroop task was used. In this modified version, only incongruent stimuli—color words (e.g., “red,” “blue”) displayed in mismatched hues—are presented. Participants must identify the font color while suppressing the word’s meaning, except when the word appears in red, in which case they must respond based on the word itself rather than its color (e.g., selecting “blue” for the word “blue” written in red). This task is usually used in studies focused on mental fatigue²⁸. Consistent with prior research, this study used a 30-minute computerized incongruent Stroop task to induce mental fatigue, following the same established protocol. Despite it is not a very ecological tasks, it includes response inhibition or selective attention which are also present in sport game scenarios²⁹.

SAFT⁹⁰ training model

The SAFT⁹⁰ protocol employed was developed and validated to replicate the physiological and mechanical demands made during soccer match-play³⁰. The protocol has been used in a number of research papers, and also as a late-stage rehabilitation tool for a number of professional teams in the UK¹⁶. Based on time-motion analysis from a lower-division English soccer league (2007 season), this 90-minute protocol is performed on a 20-meter shuttle course (Figure2) and incorporates five movement intensities: standing, walking, jogging, stride running, and sprinting, controlled by audio cues. Key performance metrics include: Total distance covered: 10.78 km; Frequent intensity changes: 1,269; shifts (every 4.3 s); Multidirectional demands: 1,350 directional changes. The protocol involves shuttle-based movements guided by audio cues between cones (0–20 m), divided into “Out” (0→20 m) and “Back” (20→0 m) phases. Direction changes occur only during the “Out” phase, while the return is straight-line. Audio commands begin with “up” (forward/backward running) or “side” (lateral shuffling with a cone circle). After returning to the start, participants navigate 9–11 m cones before sprinting to 20 m. Five intensity levels (standing to sprinting) are used, with jogging/stride running in the “Out” phase and walking/sprinting added in the “Back” phase. A 15 min activity profile was developed and repeated six times during the full 90 min simulated soccer match. In this study, Participations performed two 15 – minute block of SAFT⁹⁰. The protocol imposes an internal load comparable to competitive matches, making it a valid tool for training and research.

Fig. 2.

Schematic of the SAFT⁹⁰ protocol (from Lovell et al.2008)³⁰.

T-SAFT⁹⁰ training model

The T- SAFT⁹⁰ protocol employed was designed and validated by Da Silva and Ric Lovell (2020)¹⁶, which includes technical and jumping activity as an addition to SAFT⁹⁰. The task (Technical soccer-specific aerobic field test) follows the original protocol SAFT⁹⁰ (Figure 3), using audio cues to direct movements between cones (0–20 m) across two phases: “Out-bound” (0→20 m) and “Return” (20→0 m). Participants perform technical actions (passing, shooting, dribbling) and movements (standing to sprinting/jumping) on command. A soccer goal (15 m from start) includes upper (shooting) and lower (passing) circular targets (60 cm diameter), while dribbling requires weaving through agility poles at 9–11 m. All technical actions are executed with the player’s preferred foot¹⁶. A 15-minute activity profile was developed and repeated six times during the full 90 min simulated soccer match. In this study, Participations performed two 15 – minute block of T-SAFT⁹⁰.

Fig. 3.

Schematic of the T-SAFT ⁹⁰ training protocol (obtained from Da Silva and Ric Lovell, 2020,¹⁶).

Combined T-SAFT⁹⁰ + Stroop training model

This modified model (Figure 4) maintains the T-SAFT⁹⁰ framework while incorporating cognitive demands through a modified Stroop task at both the start (0 m) and turnaround (20 m) points. At each station, participants complete a visual Stroop test where they must identify either the ink color (for non-red words) or word meaning (for red-colored words). Their response determines movement direction - they proceed toward a cone matching their answer (e.g., a “yellow” response directs them to a yellow cone). After completing the Stroop task at the 20 m point with a verbal response, they continue with the standard T-SAFT⁹⁰ protocol. This dual physical-cognitive challenge repeats cyclically throughout the 30-minute session.

Fig. 4.

Schematic of the combined T-SAFT⁹⁰ + Stroop protocol training.

Statistical analysis

Data were analyzed using SPSS v27. Descriptive statistics (mean ± standard deviation) were computed for all variables. The Kolmogorov–Smirnov test confirmed non-normal data distribution, prompting non-parametric analyses: The Wilcoxon signed-rank test for paired pre/post comparisons (within/between protocols) and the Friedman test for multi-condition effects (training models on outcomes, heart rate, and motivation). Effect sizes (ES, ±95%CI) were interpreted as trivial (0.0–0.2), small (0.2–0.6), moderate (0.6–1.2), large (1.2–2.0), very large (2.0–4.0), or extremely large (>4.0)³¹.

To control for type I error inflation due to multiple endpoint testing, the Bonferroni correction was applied separately to three outcome families: (1) Mental Fatigue (VAS; adjusted α = 0.05), (2) Cognitive Performance (response accuracy/time, working memory, visual scanning, auditory recognition; adjusted α = 0.01), and (3) Technical Performance (penalty time, movement time, passing accuracy; adjusted α = 0.0167). Statistical significance was defined as p < adjusted α. All tests were two-tailed.

Results

Descriptive statistics

Descriptive data for 15 young soccer players shows the mean age (17.33 ± 0.72 years), weight (62.53 ± 3.59 kg), height (172.73 ± 5.28 cm), and BMI (21.01 ± 1.82 kg/m²). Table 2 presents the mean heart rate values during the performance of training models, with the SAFT⁹⁰, T-SAFT⁹⁰, and combined T-SAFT⁹⁰+Stroop models eliciting similar heart rates around 177 bpm, while the Stroop-only protocol showed no significant change in heart rate due to the nature of the task. Table 2 summarizes the mean motivation scores assessed prior to each training protocol, with no significant differences observed among the groups (all p > 0.05). These findings indicate homogeneity among conditions at baseline for demographic and physiological parameters.

Table 2.

Comparison of heart rate and motivation score between training models.

Training models	Measure	Mean ± SD
Stroop	Heart rate (beat/min)	76.05±3.24
	Motivation(score)	90.93±5.33
SAFT90	Heart rate (beat/min)	177.06± 4.11
	Motivation(score)	91.46±5.38
T-SAFT90	Heart rate (beat/min)	177.46± 4.56
	Motivation(score)	91.00±5.54
T-SAFT90 + Stroop protocol	Heart rate (beat/min)	177.40±4.10
	Motivation(score)	91.13±5.37

Mental fatigue (VAS Scores)

All training models demonstrated significant post-intervention increases in mental fatigue (VAS scores: Stroop Z = -3.413, T-SAFT⁹⁰ Z = -3.413, Combined Z = -3.410, SAFT⁹⁰ Z = -3.413; all raw p < 0.001). Following Bonferroni correction for the mental fatigue outcome family (adjusted α = 0.005), all within-protocol changes remained statistically significant (all adjusted p < 0.001).

Baseline measurements showed comparable pre-test scores across groups (range: 12.06 ± 2.37 to 12.60 ± 5.66), while post-intervention values revealed pronounced between-protocol differences: combined T-SAFT⁹⁰ + Stroop showed the greatest increase (66.40 ± 13.6), followed by T-SAFT⁹⁰ (28.33 ± 3.39), Stroop (28.46 ± 4.54), and SAFT⁹⁰ (14.66 ± 2.38).

Between-protocol comparisons with Bonferroni correction (adjusted α = 0.005) revealed significant differences in post-intervention VAS scores between: Stroop vs. SAFT⁹⁰ (Z = -3.411, adjusted p = 0.001), T-SAFT⁹⁰ vs. SAFT⁹⁰ (Z = -3.413, adjusted p = 0.001), and combined vs. SAFT⁹⁰ (Z = -3.411, adjusted p = 0.001). The combined protocol produced significantly higher fatigue than both T-SAFT⁹⁰ (Z = -2.671, adjusted p = 0.024) and Stroop protocols (Z = -2.643, adjusted p = 0.025). No significant difference emerged between Stroop and T-SAFT⁹⁰ protocols (Z = -0.315, adjusted p > 0.99).

Effect sizes were very large for cognitive-loaded protocols (Stroop ES = 3.86; T-SAFT⁹⁰ ES = 3.34; combined ES = 3.77) and large for physical-only training (SAFT⁹⁰ ES = 1.50).

These findings demonstrate a clear hierarchy of mental fatigue induction: integrated cognitive-physical demands > isolated cognitive or technical-physical training > physical-only exercise. The increase in mental fatigue (ΔVAS) was approximately 20 times greater in the combined protocol than in SAFT⁹⁰ (Δ = 53.8 vs. Δ = 2.6) and approximately twice as large as in the Stroop or T-SAFT⁹⁰ protocols (Fig. 5).

Fig. 5.

Pre-post assessment of subjective mental fatigue. Circles indicate the mean of each condition, and thick lines indicate the SD value. *Significant changes between pre- posttest(p<0.05); VAS, visual analog scale; Modified Stroop (Modified Stroop training model), T-SAFT⁹⁰ (T-SAFT ⁹⁰ training model), SAFT⁹⁰ (SAFT⁹⁰ training model), T-SAFT⁹⁰ + Stroop (Combined T-SAFT ⁹⁰ + Stroop training model).

Cognitive performance

Response accuracy

Response accuracy declined across all training protocols from pre- to post-testing (all raw p ≤ 0.003). The greatest deterioration occurred in the Stroop condition, where accuracy decreased from 99.67 ± 0.47% to 96.43 ± 1.22% (Z = -3.296, raw p = 0.001, ES = 3.50). Similarly, the T-SAFT⁹⁰ protocol showed substantial impairment from 98.95 ± 1.36% to 95.64 ± 1.63% (Z = -3.274, raw p = 0.001, ES = 2.20). The SAFT⁹⁰ training produced more modest declines from 99.17 ± 1.31% to 96.63 ± 1.39% (Z = -2.956, raw p = 0.003, ES = 1.88), and the combined protocol exhibited the smallest reduction from 98.54 ± 1.20% to 95.20 ± 0.40% (Z = -3.096, raw p = 0.002, ES = 1.04) (Fig. 6a).

Following Bonferroni correction for the cognitive performance family (adjusted α = 0.001), only the Stroop condition remained statistically significant (adjusted p = 0.003). The T-SAFT⁹⁰ (adjusted p = 0.003) and combined protocols (adjusted p = 0.006) showed trends that did not survive correction. SAFT⁹⁰ demonstrated no significant change (p = 0.393, adjusted p > 0.99).

Between-protocol analyses revealed significant differences across conditions (Friedman test: χ²(3) = 29.635, p < 0.001). Post-hoc comparisons with Bonferroni adjustment (adjusted α = 0.001) showed significant differences in accuracy deterioration between: Stroop and combined conditions (Z = -3.411, adjusted p = 0.002) and T-SAFT⁹⁰ and combined protocols (Z = -3.412, adjusted p = 0.002). No significant differences emerged between Stroop and T-SAFT⁹⁰ (Z = -0.625, adjusted p > 0.99), nor between SAFT⁹⁰ and either Stroop (Z = -2.273, adjusted p = 0.092) or T-SAFT⁹⁰ protocols (Z = -2.063, adjusted p = 0.156).

These results demonstrate a hierarchy of cognitive impairment, with isolated cognitive demands producing the most pronounced effects, while combined protocols showed attenuated declines following statistical correction.

Response time

Significant post-intervention increases in response time occurred in the Stroop (918.86 ± 23.85 ms to 1000.10 ± 55.29 ms; Z = -2.842, raw p = 0.004, ES = 1.27), T-SAFT⁹⁰ (881.90 ± 58.97 ms to 960.66 ± 86.56 ms; Z = -3.181, raw p = 0.001, ES = 0.98), and combined protocols (871.50 ± 74.71 ms to 967.66 ± 44.77 ms; Z = -3.181, raw p = 0.001, ES = 1.56). In contrast, SAFT⁹⁰ training produced no significant change (911.40 ± 65.42 ms to 917.20 ± 74.63 ms; Z = -0.855, raw p = 0.393, ES = 0.08) (Fig. 6b).

Fig. 6.

Cognitive performance in pre-and posts. Circles indicate the mean of each condition, and thick lines indicate the SD value. *Significant changes between pre- posttest(p<0.05); VAS, visual analog scale; Modified Stroop (Modified Stroop training model), T-SAFT90 (T-SAFT ⁹⁰ training model), SAFT90 (SAFT⁹⁰ training model), T-SAFT90 + Stroop (Combined T-SAFT ⁹⁰ + Stroop training model).

Following Bonferroni correction for the cognitive performance family (adjusted α = 0.001), the T-SAFT⁹⁰ (adjusted p = 0.005) and combined protocols (adjusted p = 0.005) remained statistically significant, while the Stroop condition showed a trend that did not survive correction (adjusted p = 0.017). The SAFT⁹⁰ protocol demonstrated no significant effect (adjusted p > 0.99).

Between-protocol analyses revealed significant differences in response time changes (Friedman test: χ²(3) = 13.470, raw p = 0.004). Post-hoc comparisons with Bonferroni adjustment (adjusted α = 0.001) showed significant differences between: Stroop and SAFT⁹⁰ (Z = -2.556, adjusted p = 0.044), T-SAFT⁹⁰ and SAFT⁹⁰ (Z = -2.897, adjusted p = 0.016), and combined protocol versus SAFT⁹⁰ (Z = -2.783, adjusted p = 0.022). No significant differences emerged among the cognitive-loaded conditions themselves (Stroop vs. T-SAFT⁹⁰: Z = -0.341, adjusted p > 0.99; Stroop vs. combined: Z = -0.511, adjusted p > 0.99; T-SAFT⁹⁰ vs. combined: Z = -1.601, adjusted p = 0.548).

The effect size progression (combined> Stroop > T-SAFT⁹⁰ > > SAFT⁹⁰) indicates that integrated cognitive-physical demands may compound temporal processing challenges beyond isolated cognitive tasks.

Visual scanning identification

Significant post-intervention declines in visual scanning performance were observed in the combined protocol (90.93 ± 6.65 to 85.46 ± 6.01; Z = -2.480, raw p = 0.013, ES = 0.86), T-SAFT⁹⁰ (90.20 ± 5.00 to 87.40 ± 4.74; Z = -2.427, raw p = 0.015, ES = 0.57), and Stroop conditions (90.80 ± 6.78 to 88.26 ± 7.81; Z = -2.949, raw p = 0.003, ES = 0.34). In contrast, the physical-only SAFT⁹⁰ protocol showed no significant change (90.53 ± 5.35 to 90.73 ± 5.56; Z = -0.504, raw p = 0.614, ES = -0.036) (Fig. 7a).

Following Bonferroni correction for the cognitive performance family (adjusted α = 0.001), only the Stroop condition showed a trend toward significance (adjusted p = 0.012), while the combined (adjusted p = 0.052) and T-SAFT⁹⁰ (adjusted p = 0.063) protocols did not survive correction. The SAFT⁹⁰ protocol demonstrated no significant change (adjusted p > 0.99).

Between-protocol analyses revealed significant heterogeneity in visual scanning changes (Friedman test: χ²(3) = 14.521, raw p = 0.002). Post-hoc comparisons with Bonferroni adjustment (adjusted α = 0.001) showed significant differences between: Stroop and SAFT⁹⁰ (Z = -2.881, adjusted p = 0.016) and combined protocol and SAFT⁹⁰ (Z = -2.530, adjusted p = 0.057). The T-SAFT⁹⁰ versus SAFT⁹⁰ comparison showed a trend (Z = -2.430, adjusted p = 0.075). No significant differences emerged among the cognitive-loaded conditions themselves (Stroop vs. T-SAFT⁹⁰: Z = -0.663, adjusted p > 0.99; Stroop vs. combined: Z = -1.382, adjusted p = 0.668; T-SAFT⁹⁰ vs. combined: Z = -1.163, adjusted p > 0.99).

These results indicate a dissociation between protocol types, with cognitive-loaded training showing trends toward visual scanning impairment while physical-only training preserved performance. The effect magnitude hierarchy (combined > T-SAFT⁹⁰ > Stroop > > SAFT⁹⁰) suggests that integrated cognitive-physical demands may compound visual processing challenges beyond isolated cognitive tasks.

Auditory pattern recognition

Analysis of auditory pattern recognition revealed protocol-specific effects on post-training performance. The combined T-SAFT⁹⁰ + Stroop protocol demonstrated a significant decline in accuracy (82.00 ± 7.02 to 75.00 ± 14.26; Z = 2.624, raw p = 0.009, ES = 0.62). In contrast, isolated protocols showed non-significant changes: T-SAFT⁹⁰ (87.00 ± 7.74 to 84.00 ± 16.16; Z = -0.898, raw p = 0.369, ES = 0.23), Stroop (84.00 ± 6.03 to 82.40 ± 9.85; Z = -0.713, raw p = 0.476, ES = 0.195), and SAFT⁹⁰ (78.46 ± 3.20 to 78.93 ± 5.32; Z = -0.774, raw p = 0.439, ES = -0.056) (Fig. 7b).

Following Bonferroni correction for the cognitive performance family (adjusted α = 0.001), no within-protocol changes reached statistical significance (combined: adjusted p = 0.036; all others: adjusted p > 0.99).

The omnibus Friedman test showed no significant between-group differences in performance changes (χ²(3) = 5.308, raw p = 0.151). Targeted pairwise comparisons with Bonferroni adjustment (adjusted α = 0.001) revealed no significant differences between protocols (combined vs. T-SAFT⁹⁰: Z = -2.762, adjusted p = 0.024; combined vs. SAFT⁹⁰: Z = -2.549, adjusted p = 0.051; all other comparisons: adjusted p > 0.99).

These findings suggest that while the combined protocol showed a trend toward impairing auditory pattern recognition, this effect did not survive strict multiple comparison correction. Isolated training modalities preserved this ability, with the selective trend in the combined condition (ES = 0.62) versus trivial effects in single-modality training (ES range: -0.056 to 0.23) indicating potential interference effects when combining cognitive and physical demands. The dissociation between auditory and visual processing outcomes highlights domain-specific fatigue effects across sensory modalities.

Working memory capacity

The analysis revealed differential effects of training protocols on working memory performance. The T-SAFT⁹⁰ protocol showed the most pronounced decline (89.20 ± 7.21 to 84.80 ± 8.15; Z = 3.309, raw p = 0.001, ES = 0.57). The combined protocol demonstrated a marginal reduction (82.40 ± 13.42 to 81.06 ± 13.49; Z = 2.272, raw p = 0.023, ES = 0.09), while SAFT⁹⁰ training showed a non-significant improvement (79.20 ± 10.58 to 81.73 ± 10.36; Z = 1.931, raw p = 0.054, ES = -0.24) and Stroop training produced negligible change (83.60 ± 9.89 to 83.40 ± 9.81; Z = -0.742, raw p = 0.458, ES = 0.02) (Fig. 7C).

Following Bonferroni correction for the cognitive performance family (adjusted α = 0.001), the T-SAFT⁹⁰ protocol remained statistically significant (adjusted p = 0.003). The combined protocol showed a trend that did not survive correction (adjusted p = 0.092), while both SAFT⁹⁰ (adjusted p = 0.216) and Stroop protocols (adjusted p > 0.99) demonstrated no significant effects.

Between-protocol analyses revealed significant heterogeneity in working memory changes (Friedman test: χ²(3) = 17.146, raw p = 0.001). Post-hoc comparisons with Bonferroni adjustment (adjusted α = 0.001) showed significant differences between: T-SAFT⁹⁰ and SAFT⁹⁰ (Z = -3.177, adjusted p = 0.005) and T-SAFT⁹⁰ and combined protocols (Z = -2.901, adjusted p = 0.016). The comparison between SAFT⁹⁰ and combined protocols showed a trend (Z = -2.731, adjusted p = 0.024), while T-SAFT⁹⁰ versus Stroop approached significance (Z = -2.258, adjusted p = 0.096). No significant differences emerged between Stroop and SAFT⁹⁰ (Z = -1.649, adjusted p = 0.396) or Stroop and combined protocols (Z = -0.917, adjusted p > 0.99).

These results indicate that technical-physical training (T-SAFT⁹⁰) induced the most substantial working memory impairment following statistical correction, while physical-only training (SAFT⁹⁰) showed a non-significant tendency toward improvement. The combined protocol’s marginal effect (ES = 0.09) suggests potential interference between cognitive and physical demands, though less pronounced than in isolated T-SAFT⁹⁰ training. The dissociation between protocol effects highlights the sensitivity of working memory to specific training characteristics.

Penalty time

Significant increases in penalty time were detected in the Stroop (9.80 ± 4.29 to 13.06 ± 4.06 s; Z = 2.823, raw p = 0.005, ES = 0.78), T-SAFT⁹⁰ (7.40 ± 2.32 to 10.60 ± 1.95 s; Z = 3.432, raw p = 0.001, ES = 1.49), and combined protocols (7.40 ± 2.09 to 11.66 ± 1.87 s; Z = 3.200, raw p = 0.001, ES = 2.14). The SAFT⁹⁰ protocol showed no significant change (15.13 ± 5.55 to 15.53 ± 4.92 s; Z = -0.243, raw p = 0.808, ES = -0.076) (Fig. 7a).

Fig. 7.

Assessment of soccer skill performance. Circles indicate the mean of each condition, and thick lines indicate the SD value. *Significant changes between pre- posttest(p<0.05); VAS, visual analog scale; Modified Stroop (Modified Stroop training model), T-SAFT⁹⁰ (T-SAFT ⁹⁰ training model), SAFT⁹⁰ (SAFT⁹⁰ training model), T-SAFT⁹⁰ + Stroop (Combined T-SAFT ⁹⁰ + Stroop training model).

Following Bonferroni correction for the technical performance family (adjusted α = 0.00167), the T-SAFT⁹⁰ (adjusted p = 0.002) and combined protocols (adjusted p = 0.005) remained statistically significant, while the Stroop condition showed a trend that did not survive correction (adjusted p = 0.020). The SAFT⁹⁰ protocol demonstrated no significant effect (adjusted p > 0.99).

Between-protocol analyses revealed significant differences in penalty time changes (Friedman test: χ²(3) = 14.790, raw p = 0.002). Post-hoc comparisons with Bonferroni adjustment (adjusted α = 0.00167) showed significant differences between: SAFT⁹⁰ and combined protocols (Z = -3.305, adjusted p = 0.004) and SAFT⁹⁰ and Stroop protocols (Z = -2.682, adjusted p = 0.036). The comparison between SAFT⁹⁰ and T-SAFT⁹⁰ showed a trend (Z = -2.342, adjusted p = 0.095).

Effect sizes ranged from negligible in SAFT⁹⁰ (ES = -0.076) to moderate for Stroop (ES = 0.78) and large to very large for T-SAFT⁹⁰ (ES = 1.49) and combined protocol (ES = 2.14).

Movement time

Significant post-intervention increases in movement time were observed in the Stroop (37.18 ± 4.74 to 41.26 ± 2.51 s; Z = 3.076, raw p = 0.002, ES = 1.07), T-SAFT⁹⁰ (39.64 ± 2.98 to 42.40 ± 1.78 s; Z = 2.443, raw p = 0.015, ES = 1.12), and combined protocols (38.64 ± 3.47 to 42.02 ± 3.79 s; Z = 3.415, raw p = 0.001, ES = 0.93). The SAFT⁹⁰ protocol showed no significant change (39.65 ± 4.46 to 40.82 ± 4.21 s; Z = 1.025, raw p = 0.305, ES = 0.26) (Fig. 7b).

Following Bonferroni correction for the technical performance family (adjusted α = 0.00167), the combined protocol remained statistically significant (adjusted p = 0.003), while the Stroop condition showed a trend (adjusted p = 0.010) and the T-SAFT⁹⁰ protocol did not survive correction (adjusted p = 0.070). The SAFT⁹⁰ protocol demonstrated no significant effect (adjusted p > 0.99).

Between-protocol analyses for movement time changes approached but did not reach statistical significance (Friedman test: χ²(3) = 7.824, raw p = 0.050). Post-hoc comparisons with Bonferroni adjustment (adjusted α = 0.00167) revealed a significant difference specifically between the SAFT⁹⁰ and combined protocols (Z = 2.342, adjusted p = 0.095). No other between-protocol comparisons reached significance.

These findings demonstrate that cognitive-loaded protocols impaired movement speed, with the combined protocol showing the most consistent effect following statistical correction. Physical-only training preserved movement time, suggesting that cognitive load rather than physical exertion primarily drives movement timing impairments under fatigue conditions.

Pass accuracy (perfect pass)

Passing accuracy outcomes showed declines across training protocols, with significant reductions observed in the Stroop (pre: 7.80 ± 1.97, post: 6.06 ± 1.83; Z = 3.415, raw p = 0.001, ES = 0.64), T-SAFT⁹⁰ (pre: 7.60 ± 1.76, post: 6.00 ± 1.60; Z = 3.181, raw p = 0.001, ES = 0.60), and combined protocols (pre: 7.40 ± 1.88, post: 5.60 ± 1.35; Z = 3.296, raw p = 0.001, ES = 0.56). The SAFT⁹⁰ protocol showed borderline significance (pre: 7.46 ± 1.55, post: 6.53 ± 1.36; Z = 2.941, raw p = 0.003, ES = 0.50) (Fig. 7C).

Following Bonferroni correction for the technical performance family (adjusted α = 0.00167), the Stroop (adjusted p = 0.003), T-SAFT⁹⁰ (adjusted p = 0.005), and combined protocols (adjusted p = 0.003) remained statistically significant, while the SAFT⁹⁰ protocol did not survive correction (adjusted p = 0.012).

Between-protocol analyses revealed no significant differences in passing accuracy changes (Friedman test: χ²(3) = 2.910, raw p = 0.406). Post-hoc comparisons with Bonferroni adjustment (adjusted α = 0.00167) showed no significant differences between any protocol pairs (all adjusted p > 0.99).

These findings demonstrate that cognitive-loaded protocols consistently impaired passing accuracy following statistical correction, while physical-only training showed only a trend toward decline. The similar magnitude of change across protocols despite variations in baseline performance levels suggests a generalized effect of training load on passing precision, with cognitive demands exacerbating rather than qualitatively altering this pattern.

Fig. 8.

Assessment of soccer skill performance. Circles indicate the mean of each condition, and thick lines indicate the SD value. *Significant changes between pre- posttest(p<0.05); VAS, visual analog scale; Modified Stroop (Modified Stroop training model), T-SAFT⁹⁰ (T-SAFT ⁹⁰ training model), SAFT⁹⁰ (SAFT⁹⁰ training model), T-SAFT⁹⁰ + Stroop (Combined T-SAFT ⁹⁰ + Stroop training model).

Discussion

The study examined the effects of different training protocols—Stroop (cognitive), T-SAFT⁹⁰ (technical-physical), SAFT⁹⁰ (physical-only), and a combined T-SAFT⁹⁰+Stroop—on mental fatigue and technical performance. All protocols increased mental fatigue, with the combined protocol inducing the greatest fatigue and largest performance impairments across domains. Cognitive-loaded protocols (Stroop, T-SAFT⁹⁰, combined) led to significant declines in response accuracy, processing speed, visual scanning, working memory, and technical performance (e.g., increased movement and penalty times), while physical-only training (SAFT⁹⁰) preserved most cognitive and motor functions. The combined protocol, despite producing the highest mental fatigue, sometimes showed smaller cognitive accuracy deficits. This pattern may reflect task prioritization, speed–accuracy trade-offs, or attentional reallocation under dual-task demands, rather than a true mitigation of cognitive fatigue. However, it consistently caused the largest penalties in motor performance and timing, indicating a compounded fatigue effect. Overall, results reveal that integrated cognitive-physical training has the greatest impact on both mental fatigue and performance, while isolated cognitive or physical training has more domain-specific effects which may be justified by various mechanisms³².

It is crucial to emphasize that all fatigue induction and performance assessments in this study occurred in a highly controlled, simulated environment. Although protocols like T-SAFT⁹⁰ and its combined variant incorporate soccer-specific movements and decision-making, they remain pre-scripted, non-interactive, and devoid of the tactical unpredictability, emotional pressure, and opponent-driven dynamics inherent to real match-play. Consequently, the mental fatigue documented here reflects a laboratory-elicited cognitive strain rather than the complex, context-dependent fatigue experienced during competitive soccer. While our findings inform BET protocol selection under controlled conditions, their transfer to authentic performance settings requires cautious interpretation and empirical validation.

Induction of perceived mental fatigue across training models

A central finding of this study is the significant increase in perceived mental fatigue across all training models, as measured by the Visual Analog Scale (VAS). The results confirm that both purely cognitive (Stroop) and combined cognitive-physical (T-SAFT⁹⁰, SAFT⁹⁰, and T-SAFT⁹⁰ + Stroop) protocols effectively induced subjective mental fatigue. These findings are consistent with prior research demonstrating that the Stroop task, particularly the incongruent version, is a robust method for eliciting mental fatigue due to its high demands on selective attention, inhibitory control, and conflict resolution^9,21, also with prior research demonstrating that the combination of mental and physical fatigue led to even greater impairments in tasks requiring both cognitive and physical effort^28,33.

Notably, between-group analyses revealed significant differences in the magnitude of mental fatigue induction. Specifically, the combined T-SAFT⁹⁰ + Stroop model elicited significantly higher levels of perceived fatigue compared to the SAFT⁹⁰ model, as did the Stroop and T-SAFT⁹⁰ models when compared to SAFT⁹⁰. However, no significant difference was found between the Stroop and T-SAFT⁹⁰ models, suggesting that both impose a comparable cognitive load despite their different modalities. This equivalence—supported by very large effect sizes (ES = 3.768 for Stroop, ES = 3.864 for T-SAFT⁹⁰, ES = 3.345 for combined)—indicates that T-SAFT⁹⁰, traditionally viewed as a physical conditioning drill, carries a substantial cognitive component, likely due to rapid decision-making, spatial awareness, and response inhibition under physical duress.

While the T-SAFT⁹⁰ protocol and its combined variant incorporate soccer-specific movements, decisions, and physical demands—and thus offer greater ecological fidelity than traditional lab-based cognitive tasks (e.g., seated Stroop)—their ecological validity remains relative and partial. These protocols are still pre-scripted, non-interactive, and devoid of opponent behavior, tactical unpredictability, and emotional pressure inherent to actual matches. Therefore, their representativeness should be viewed as a step toward ecological relevance, not equivalence with real-game conditions. The combined T-SAFT⁹⁰ + Stroop model, which integrates physical effort with sustained cognitive challenge, demonstrated the highest overall impact, reinforcing the idea that simultaneous physical and cognitive loads can amplify perceived mental fatigue²⁸. The use of the VAS for measuring mental fatigue in soccer has several limitations: first, The VAS relies on self-reported data, which can be influenced by individual perceptions and biases. Players may underreport or over report their fatigue levels based on their mood or motivation at the time of assessment³⁴. Second, Participants often show reluctance to use the extreme ends of the scale (i.e., the highest and lowest points), which can lead to a clustering of responses around the middle of the scale. This can reduce the sensitivity of the VAS in capturing the full range of mental fatigue.³⁴Third, the context in which the VAS is administered (e.g., before or after a game) can affect the results. Factors such as physical fatigue, stress, or environmental conditions may confound the results, making it difficult to isolate the effects of mental fatigue alone³⁴. These limitations suggest that while the VAS can provide valuable insights, it should be used in conjunction with other measures for a more comprehensive assessment of mental fatigue in soccer players.

Cognitive performance: response Accuracy, response Time, visual scanning identification, working memory Capacity, auditory pattern recognition

It has been suggested that performance decrement during a cognitive task is the gold standard measure of mental fatigue³⁵. All training models except for SAFT⁹⁰ resulted in a significant decline in response accuracy on the Stroop test, with the largest effects observed in the combined T-SAFT⁹⁰ + Stroop, T-SAFT⁹⁰, and Stroop models. The relatively smaller decline in response accuracy observed in the combined T-SAFT⁹⁰ + Stroop condition (ES = 1.04) compared to the isolated Stroop protocol (ES = 3.50) should be interpreted cautiously. This pattern does not necessarily reflect a ‘protective’ or ‘mitigating’ effect of physical activity on cognitive fatigue. Instead, it may result from task prioritization, where participants allocate limited attentional resources to physical execution at the expense of cognitive task precision, or from a speed–accuracy trade-off, wherein maintaining movement tempo under dual-task demands leads to strategic sacrifice in response accuracy. Alternatively, attentional reallocation toward motor control during physical exertion could reduce cognitive monitoring capacity. These alternatives are consistent with resource-competition models of dual-task performance and highlight the complexity of interpreting accuracy changes in integrated protocols. These results align with previous studies showing that sustained cognitive effort leads to reduced inhibitory control and impaired decision-making^9,24,36. The physical-only training (SAFT⁹⁰) preserved most cognitive and motor functions, likely because it did not impose the additional cognitive load that leads to mental fatigue.

The deterioration in accuracy reflects a depletion of cognitive resources, particularly in brain regions such as the anterior cingulate cortex (ACC), which is involved in conflict monitoring and error detection³⁷. The lack of significant change in the SAFT⁹⁰ model suggests that its cognitive demands are insufficient to induce measurable cognitive fatigue, positioning it as a lower cognitive load protocol compared to the others.

Similarly, response time on the Stroop task significantly increased in the Stroop, T-SAFT⁹⁰, and combined T-SAFT⁹⁰ + Stroop models, but not in SAFT⁹⁰ (P = 0.393). This indicates that tasks requiring sustained attention and conflict resolution are particularly vulnerable to mental fatigue. The effect sizes further support this: while SAFT⁹⁰ showed a trivial effect (ES = 0.082), Stroop (ES = 1.27), T-SAFT⁹⁰ (ES = 0.98), and the combined model (ES = 1.56) demonstrated large to very large effects, reinforcing the notion that cognitive load is a key determinant of performance decline.

Interestingly, the auditory pattern recognition task showed significant impairments only in the combined T-SAFT⁹⁰ + Stroop model, with moderate effect size, while the other models—including the purely cognitive Stroop—showed no significant changes. This suggests that the combined model imposes a uniquely high cognitive load, potentially disrupting auditory processing pathways. One possible explanation is that the integration of physical exertion with cognitive demand increases overall mental strain, reducing the brain’s capacity to process non-visual sensory information²⁸. The neuroendocrine response to simultaneous cognitive and physical demands can exacerbate fatigue, as both systems compete for limited resources. This interaction can lead to greater impairments in motor performance and timing, as seen in our results³⁸. This finding highlights the sensitivity of dual-task paradigms in detecting subtle cognitive impairments.

Impact on technical performance: passing Accuracy, penalty time and movement time

The most practically relevant finding of this study is the significant negative impact of mental fatigue on technical performance, particularly in the Loughborough Soccer Passing Test (LSPT). All models except SAFT⁹⁰ led to a significant increase in penalty time—a marker of reduced passing accuracy and efficiency. Between-group comparisons revealed significant differences between SAFT⁹⁰ and all other models, with the combined model showing the largest effect size (ES = 2.14, very large).

These results are in line with prior research showing that mental fatigue affects technical performance in small-sided games and passing accuracy^16,23. The combined model’s pronounced effect suggests that the interaction between mental and physical fatigue exacerbates performance decrements. This supports the shared-resources theory, where mental fatigue increases perceived effort and reduces motivation, leading to suboptimal technical execution^4,30.

The lack of significant effect in the SAFT⁹⁰ model may be attributed to its lower cognitive demand, despite high physical load. This implies that physical fatigue alone, without concurrent cognitive strain, may not be sufficient to impair technical precision in skilled athletes. This could be due to the engagement of different neural pathways during physical activity that help maintain cognitive function¹². This distinction is crucial for coaches and sports scientists, as it suggests that technical drills should be scheduled carefully to avoid periods of high cognitive load.

Visual scanning and attentional control

Significant impairments in visual scanning performance were observed after the Stroop, T-SAFT⁹⁰, and combined T-SAFT⁹⁰ + Stroop models, but not after SAFT⁹⁰). This indicates that tasks requiring top-down attentional control—such as scanning the field for passing options—are particularly vulnerable to mental fatigue. Theoretical explanations include a shift from goal-directed (top-down) to stimulus-driven (bottom-up) attention, making players more reactive and less strategic^39,40.

This shift may result in reduced situational awareness and slower decision-making, which are critical in dynamic team sports. The findings are consistent with study showing that mentally fatigued athletes commit more unforced errors and exhibit poorer spatial awareness⁹. The lack of impairment in the SAFT⁹⁰ model further supports the idea that cognitive load, rather than physical exertion alone, drives attentional deficits.

Mechanisms underlying mental fatigue and performance decline

Several mechanisms may explain the observed performance decrements. First, impaired inhibitory control—a core function of the prefrontal cortex and ACC—is disrupted by mental fatigue, leading to slower reaction times and reduced accuracy^9,24. Second, increased cognitive load from sustained attention tasks depletes working memory resources, impairing executive functions such as decision-making and task switching⁴¹.

Third, altered neural activity in key brain regions, particularly reduced dopaminergic transmission in the ACC, may underlie the decline in cognitive performance⁴². Fourth, increased perceived effort makes tasks feel more challenging, leading to early disengagement or reduced motivation⁴³. Finally, attentional resource depletion forces athletes to prioritize task completion over quality, resulting in more errors and longer compl etion times⁹. The combined T-SAFT⁹⁰ + Stroop model likely amplifies these mechanisms by simultaneously taxing physical and cognitive systems, creating a synergistic effect on fatigue and performance.

Practical implications for coaches and sports scientists

The findings should be interpreted within the context of a controlled, laboratory-based investigation. While the T-SAFT⁹⁰ protocol—and particularly the combined T-SAFT⁹⁰ + Stroop model—demonstrated robust effects on mental fatigue and performance under experimental conditions, their direct transfer to real match or training scenarios cannot be assumed. Nonetheless, these protocols may serve as standardized tools for BET research or as templates for designing structured dual-task drills in controlled training settings (e.g., small-sided games incorporating cognitive challenges like color-word conflicts or decision-based passing cues).

Coaches are encouraged to cautiously explore the integration of cognitive load into physical drills—but only after validating such adaptations for their specific context, player age, and competitive demands. Scheduling highly cognitively demanding tasks immediately before technical sessions or matches should be avoided, as even lab-induced mental fatigue can impair passing accuracy and decision speed. Future work should focus on translating these controlled protocols into ecologically valid formats that retain experimental integrity while mirroring the tactical, emotional, and interactive nature of real soccer.

Ecological validity and Real-World applicability

While our protocols were designed to simulate soccer-specific demands, they remain fundamentally constrained by the artificial, predictable nature of laboratory-based testing. Real soccer matches involve continuous adaptation to unpredictable opponent behaviors, dynamic tactical shifts, emotional arousal, and social pressure—factors absent in our controlled setting. In contrast, tasks like the Stroop or even T-SAFT⁹⁰ are pre-scripted, non-interactive, and lack the emergent decision-making required when responding to a defender’s movement or a teammate’s off-ball run.

Consequently, the mental fatigue observed here reflects cognitive strain under standardized dual-task load, not the holistic, context-dependent fatigue experienced during competition. This distinction is critical: performance decrements in LSPT or Captain’s Log may not directly translate to on-field passing errors or poor scanning during a match, where motor automatization, team coordination, and motivational factors may buffer or amplify fatigue effects.

Thus, our findings should be interpreted as preliminary evidence of protocol efficacy under controlled conditions, not as proof of in-game impact. Future studies must validate these models in ecologically richer environments, such as small-sided games with embedded cognitive distractors (e.g., color-coded passing targets requiring real-time inhibition), or during competitive matches using ecological momentary assessment (EMA) and neurocognitive wearables.

Limitations and recommendations

The findings are based on laboratory-based protocols and cannot be generalized to real competitive soccer contexts without further validation.

The most fundamental limitation of this study is its restricted ecological validity. All interventions and assessments were conducted in a controlled, simulated environment that lacks the tactical complexity, opponent interaction, and emotional intensity of real soccer matches. Although protocols like T-SAFT⁹⁰ and the combined T-SAFT⁹⁰ + Stroop incorporate sport-specific movements and decision points, they remain pre-determined, non-adaptive, and isolated from the unpredictable flow of actual play. As a result, the mental fatigue and performance decrements documented here represent responses to standardized experimental load, not the multifactorial fatigue of competition. Therefore, these findings provide preliminary, laboratory-based evidence that cannot be directly extrapolated to real-world performance without further validation. Additionally, emotional states such as stress, anxiety, or mood were not assessed or controlled for in this study. Given that emotional arousal can influence both subjective fatigue ratings and cognitive-motor performance, future research should incorporate measures of affect or emotional load to better isolate the effects of mental fatigue from emotional factors.

The use of VAS to measure perceived mental fatigue, while common, is subjective and may be influenced by physical fatigue, stress, or environmental factors⁴⁴. Future studies should incorporate objective measures such as EEG, pupillometry, or neurocognitive biomarkers to complement subjective reports.

Also, the cross-sectional design limits causal inference, and the lack of a control group makes it difficult to rule out practice or fatigue effects unrelated to the interventions. Third, emotional status, motivation, and psychological pressure were not fully controlled, which may have influenced performance outcomes⁴⁵.

Future research should explore long-term adaptations to dual-task training, the role of individual differences (e.g., fitness level, age, position), and the effectiveness of recovery strategies in mitigating the effects of mental fatigue. Additionally, longitudinal studies are needed to assess whether repeated exposure to mental fatigue can lead to cognitive resilience.

Conclusion

These findings underscore the importance of mental fatigue as a performance-limiting factor under controlled conditions and suggest that combined cognitive-physical protocols like T-SAFT⁹⁰ + Stroop may be valuable tools for BET research. These findings offer a foundation for BET protocol development, but their transfer to real soccer contexts remains speculative. Future research must test these models in ecologically valid settings—such as small-sided games under cognitive load or during competitive matches—to determine whether observed impairments translate to authentic performance scenarios.

Author contributions

A.S. contributed to the study conception and design, conducted the data collection and experiments, performed the statistical analysis, and was the primary author of the manuscript. D.M. contributed significantly to the theoretical framework and study design, and provided critical revision of the manuscript. R.R. assisted with the methodology, and data analysis. G.N. contributed to methodology and analysis. M.K.J. supervised the project, acquired funding, provided resources, and contributed to the conception of the study and writing and editing of the manuscript.

Funding

This study was supported by postgraduate students grant from Shiraz University (no: 0INB3M1899).

Data availability

Data will be available on reasonable request by corresponding author.

Declarations

Competing interests

The authors declare no competing interests.