Abstract
Asymmetry in choice patterns across rewarding and punishing contexts has long been observed in behavioural economics. Within existing theories of reinforcement learning, the mechanistic account of these behavioural differences is still debated. We propose that motivational salience-the degree of bottom-up attention attracted by a stimulus with relation to motivational goals-offers a potential mechanism to modulate stimulus value updating and decision policy. In a probabilistic reversal learning task, we identified post-feedback signals from EEG and pupillometry that captured differential activity with respect to rewarding and punishing contexts. We show that the degree of between-context distinction in these signals predicts interindividual asymmetries in decision accuracy. Finally, we contextualise these effects in relation to the neural pathways that are currently centred in theories of reward and punishment learning, demonstrating how the motivational salience network could plausibly fit into a range of existing frameworks.