Impact of task structure on mental fatigue induction and work performance

Mental fatigue poses significant risks in safety-critical environments, yet how task structure affects fatigue progression and its physiological expression remains unclear. This study recruited 40 participants to perform a Stroop-based mental fatigue task under two task structures (NORMAL vs. HALF) with identical durations but different block segmentation. Multimodal physiological signals-electrocardiography (ECG), electrodermal activity (EDA), electroencephalography (EEG), electrooculography (EOG), and respiration (RSP)-were recorded together with behavioural performance. Behavioural results indicated a shift from compensatory control to increased automaticity as fatigue accumulated, reflected in reduced reaction times and declining accuracy. EDA and ECG emerged as the most reliable single-signal indicators, while five-signal fusion with LightGBM achieved the highest classification performance (F1 = 0.9186). Compared with the NORMAL condition, the HALF structure showed attenuated physiological differentiation at higher fatigue levels. These findings advance understanding of task-structure-dependent fatigue dynamics and highlight the value of multimodal fusion for continuous fatigue monitoring in high-risk settings.