Key points are not available for this paper at this time.
Monitoring how attention is distributed across concurrent task demands is fundamental in daily life contexts such as mobility, education and industrial control. However, existing neurophysiological measures, typically based on univariate EEG markers, are sensitive to generic cognitive load, visual complexity, or motor activity, and therefore lack specificity for attentional splitting. Here, we introduce the Attentional Split Index (ASI), a novel EEG-based metric that employ Mutual Information (MI) theory, designed to quantify the coordinated modulation of multiple neurometrics that emerges when attention is divided across tasks. Twenty-five participants completed a realistic driving protocol combining a main driving task in two different environments (Urban, Highway) with four types of attentional-split demands, i.e., Focused, Auditory Continuous Performance Test (ACPT), Matrix, Surrogate Reference Task (SURT). Traditional neurometrics (parietal alpha, inverse frontal beta, frontal theta/beta ratio) exhibited partial sensitivity to task demands but failed to selectively reflect attentional splitting. In contrast, the ASI showed a robust and systematic increase across conditions, distinguishing not only explicit multitasking segments but also subtler differences between Urban and Highway focused driving. Eye-tracking and subjective distraction rating showed a strong and significant correlation with the ASI (Urban: r ET = 0.515 and r SUB = 0.744; Highway: r ET = 0.357 and r SUB = 0.673; all p < 10 −2 ), confirming high behavioural and phenomenological coherence. Surrogate analyses demonstrated that ASI effects were absent when temporal coordination across neurometrics was artificially disrupted, and that real ASI exceeded surrogate values in the vast majority of participants during multitasking but not during eyes-open baseline. Together, these findings establish the ASI as a specific, robust, and ecologically coherent neural marker of attentional splitting, with promising implications for neuroergonomics and next-generation adaptive human–machine systems.
Ronca et al. (Tue,) studied this question.