A common connectome-based neural reference space across affective, autonomic, and wakefulness arousal

Abstract

Arousal is often invoked to explain state-dependent variability in attention, emotion, memory, and decision-making. However, the term is used inconsistently, referring variously to affective experience, autonomic activation, and states of wakefulness. The extent to which different operationalizations of arousal reflect a shared neurobiological basis remains unclear, limiting efforts to unify findings across studies. We applied dynamic connectome predictive models to five fMRI datasets spanning naturalistic movie watching, story listening, wakeful rest, and sleep. We derived measures of affective, autonomic, and wakefulness arousal from subjective ratings, pupil dilation, and EEG, respectively. Models trained to predict arousal in one dataset generalized across datasets, indicating that dynamic connectivity captures shared arousal-related dynamics across measures and task contexts. The models also predicted manually scored sleep stages, suggesting that they capture arousal dynamics that extend to the graded fluctuations in arousal during sleep. Decoded arousal dynamics during movie-viewing predicted how well participants later recalled movie events, reproducing classic arousal-dependent memory enhancement effects. Model predictions were supported by overlapping functional connections, including a subset shared across all models. The largest proportion of shared connections was between the salience and somatomotor networks, suggesting that increased coordination between salience detection and action readiness may be a common feature across multiple operationalizations of arousal. Together, these results are consistent with a connectome-based neural reference space for arousal, in which different varieties share a core set of predictive connections. Our findings offer a quantitative framework for integrating arousal-related findings across tasks and modalities.