Arousal Optimizes Neural Evidence Representation for Human Decision-Making

Arousal-related fluctuations of cortical activity are ubiquitous in the mammalian brain and vary spontaneously with neuromodulatory catecholamine levels. How such endogenous changes in brain state impact on behaviour is incompletely understood. Here, we investigate the neural mechanisms by which arousal-related brain state fluctuations, indexed by pre-stimulus pupil dilation, impact on perceptual and value-based choices in humans. We find that during elevated arousal, both types of choices are faster and more accurate. Computational modelling indicates that these arousal-dependent behavioural enhancements reflect increased precision in the representation of task-relevant evidence. Functional imaging demonstrates that changes in evidence representation predicted by behavioural modelling correspond to neural gain increases in regions that represent the choice-relevant decision variables. Moreover, the neural gain increases in these two regions correlated with participants’ reward rates on the corresponding tasks. Thus, our data show that arousal optimizes decision making by enhancing choice-relevant evidence representation in the human brain