Attentional capture by Pavlovian signals of reward readily generalises across contexts

Reward-signalling stimuli capture attention and gaze—a phenomenon known as value-modulated attentional capture (VMAC). Prior work in which there is initially an instrumental relationship between attending to the reward-associated stimulus and receiving reward has shown that VMAC is context-specific. However, it is unclear whether VMAC remains context-specific when attending to the reward-associated stimulus is never a requirement of the task (i.e., stimuli have only a Pavlovian relationship with reward). Across three experiments, participants made saccades to a target shape to earn reward. Coloured distractors signalled different reward values depending on background context images, but making a saccade to the reward-signalling distractors caused the omission of reward. Unlike the context-specific biases observed previously, we found that VMAC trained under these conditions readily generalised across contexts: distractors that signalled high-value reward in one context captured attention in the high-value context, as well as in contexts where they signalled low reward, and in novel contexts. This pattern of context-independent capture occurred despite participants reporting explicit awareness of the context-stimulus-reward contingencies. These findings suggest fundamental differences in the way that Pavlovian and instrumental reward learning mechanisms interact with attentional control. We propose a dual-pathway account, where Pavlovian stimulus-reward relationships result in context-independent increases in attentional priority, while instrumental stimulus-reward relationships trigger context-dependent activation of target templates. Understanding these distinct pathways has important implications for theories of attention and for addressing maladaptive reward-related attentional biases in clinical populations.