EEG Synchronization Signatures for Decoding Attentional States During Continuous Force Control

Yilei Zheng¹Qiaoxiu Wang^2*Qianqian Tong³Bohao Tian⁴Peng Su¹Yonghong Xu¹Dangxiao Wang⁴

• ¹College of Mechanical and Electrical Engineering, Beijing Information Science and Technology University, Beijing, China
• ²Qiyuan Laboratory, Beijing, China
• ³Pengcheng Laboratory, Shenzhen, China
• ⁴Beihang University, State Key Laboratory of Virtual Reality Technology and Systems, Beijing, China

Mind wandering, the shift of attention from an ongoing task to task-unrelated thoughts, is a pervasive cognitive phenomenon often accompanied by detrimental consequences for task performance. While extensively studied in visual and auditory paradigms, attentional fluctuations during visuo-haptic tasks, such as force control, remain underexplored despite their high relevance to real-world skilled activities such as surgical operations or robotic-assisted manipulation. There exists a critical deficiency in exploring signatures of mind wandering from the perspective of neural synchronization. This study investigated EEG-based synchronization features to decode attentional states during a novel continuous force control task using the thought-probe method. Nine healthy male participants tracked a dynamically varying target force while scalp EEG and high-frequency force data were recorded synchronously. EEG epochs preceding self-reported attentional probes were labeled as on-task or mind-wandering states. Spectral power and three synchronization features – cross-frequency coupling, functional connectivity, and neural-behavioral synchronization – were extracted and compared between on-task and mind-wandering states. Results revealed that the mind-wandering state was characterized by increased alpha power (8-10 Hz) over frontal-posterior regions and reduced occurrence of high alpha-theta harmonic ratios. It also exhibited increased functional connectivity within sensorimotor networks and decreased mutual information between frontal EEG activity and force errors. Support vector machine classifiers for the binary attentional-state classification, utilizing combined spectral power and synchronization features, achieved 75.53% within-participant and 71.57% cross-participant accuracy, outperforming single-feature models. These findings highlight EEG synchronization signatures of mind wandering and demonstrate their feasibility for decoding attentional states during the force control task. This work may provide a foundation for future exploration of haptic-based neurofeedback systems, which could potentially complement existing visual and auditory modalities in applications such as neurocognitive rehabilitation or skilled motor training.