Melinda Sabo^1,2 (), Edmund Wascher², Manuel Varlet³, Tijl Grootswagers³; ¹Wu Tsai Institute | Yale University, ²Leibniz Research Centre for Working Environment and Human Factors, ³MARCS Institute, Western Sydney University

Attentional control (also known as cognitive control) is a fundamental function that regulates information processing to align with goal-directed behavior (von Bastian et al., 2020). Traditionally, its role has been demonstrated through classical tasks like the Stroop, Simon, and Go/No-Go, which have been argued to involve conflict resolution. However, recent behavioral evidence suggests that these tasks do not reflect a single, domain-general construct of attentional/cognitive control but rather task-specific processes. This raises the question: do these tasks rely on a shared neural mechanism orchestrated by a unified attentional/cognitive control network, as traditionally proposed (e.g., Gratton et al., 2017)? To address this, we analyzed data from the ongoing Dortmund Vital study, involving ~500 participants aged 20–70 years. Participants completed tasks traditionally associated with attentional/cognitive control—Simon, Stroop, Go/No-Go, and a perceptual discrimination task—while their brain activity was recorded using electroencephalogram (EEG). A linear classification model was trained on EEG data to distinguish conflict from non-conflict trials within each task. While the classifier successfully learnt this distinction within tasks, cross-task decoding revealed below-chance performance. Specifically, a leave-one-out cross-task decoding procedure (i.e., training on three tasks and testing on the fourth) provided strong evidence against generalization across tasks. Notably, these results remained consistent even when decoding parameters (e.g., input features, different preprocessing steps etc.) were varied. These findings suggest that the neural mechanisms underlying conflict processing are task-specific, likely governed by highly specialized and potentially non-overlapping attentional/cognitive control sub-networks. This aligns with prior behavioral evidence and challenges the traditional view of a shared neural substrate for cognitive control across tasks.

Acknowledgements: This work was supported by the International Visiting Scholar Program initiated by the MARCS Institute (Western Sydney University)