AROUSAL SELECTIVELY MODULATES RECURRENT BUT NOT FEEDFORWARD PROCESSING

Arousal critically shapes task performance, often following an inverted U-shaped pattern in which performance is optimal at moderate arousal levels, as described by the Yerkes–Dodson law. However, the neural mechanisms through which arousal gives rise to this behavioral relationship remain unclear. Recent work suggests that moment-to-moment fluctuations in arousal modulate neural representations of external stimuli. In parallel, animal and human studies indicate that top-down factors preferentially influence late, feedback-driven stages of sensory processing, whereas early feedforward encoding is largely stimulus-driven. Yet, it remains unknown whether arousal differentially influences early feedforward encoding and later feedback-driven integration processes. Here, we investigated how pupil-linked arousal modulates distinct stages of visual processing, using EEG and pupillometry. Participants detected the Kanizsa illusion and discriminated local image contrast of masked stimuli. Multivariate pattern analysis (MVPA) assessed neural responses related to local contrast (indexing feedforward processing) and the Kanizsa illusion (indexing recurrent processing). Polynomial regression models evaluated how pupil-linked arousal relates to both task performance and distinct stages of neural processing. Behavioral performance exhibited an inverted U-shaped relationship with arousal, with optimal accuracy and reaction times at intermediate pupil sizes, corresponding to mid-levels of arousal, across task conditions (see Panel A). Crucially, decoding accuracy for recurrent processing also peaked at moderate arousal levels, whereas decoding of feedforward processing remained unaffected by arousal (see Panel B). Together, these findings suggest that the behavioral benefits of intermediate arousal may arise from selective modulation of late, feedback-driven neural processing, providing a candidate neural mechanism underlying the Yerkes–Dodson law.