Distraction is a universal feature of human cognition, yet the reasons why it occurs remain poorly understood. Theories of sustained attention often point to failures of cognitive control in maintaining the task-set, while data-driven approaches suggest that distraction may instead reflect a breakdown in the selection of task-relevant information. This study aimed to better characterize the neural mechanisms of distraction and to test whether its EEG-based signature reflects a unique pattern or merely overlaps with failures in selective attention or task-set maintenance. Twenty adults completed a sustained attention go/no-go task (3,200 trials) while focusing on either numbers or letters, with EEG recorded simultaneously. Distraction was examined at two complementary levels: (i) trial-level lapses, defined as no-go errors, and (ii) attentional states, derived from reaction-time variability and categorized as 'in-the-zone' versus 'out-of-the-zone'. Across both levels, distraction was associated with attenuated event-related potentials, most notably a reduced P3 amplitude over parietal regions. Whole-scalp inter-electrode correlation analyses revealed weaker large-scale neural coordination during distraction. To isolate a unique EEG marker, we trained a machine-learning classifier to decode attentional state from EEG activity. Cross-validated decoding accuracy reached ~80% and remained robust even after controlling for reaction time, demonstrating that the signal captures information beyond overt performance. Finally, representational similarity analysis and additional classifiers confirmed that this neural signature is unique, as it was dissociable from other forms of attention, including the selection of the relevant stimulus side (left vs. right) and the control required for task-set maintenance (letter vs. number). Together, these findings reveal convergent neural markers of distraction and demonstrate the existence of a single EEG-derived signature that reliably predicts distraction, independently of behavior and other cognitive processes.
Chidharom et al. (Thu,) studied this question.