Alenka Doyle¹, Kamilla Volkova¹, Nick Crotty¹, Nicole Massa², Michael Grubb¹; ¹Trinity College, ²Mass General Brigham

Attentional control has historically been studied within a dichotomous framework based on top-down goals and bottom-up saliency. However, it has been recently established that one’s idiosyncratic history with the visual world can generate attentional biases driven by task-irrelevant and physically non-salient stimuli. In the context of learned reward associations (value-driven attentional capture, VDAC), previous work has shown that the effects of experience-driven attention hinge on reward prediction signals. Recently published work from our lab modified the VDAC training phase by employing pre-cues that provided reliable or unreliable information about the training phase target color; consequently, these pre-cues were either a reliable or an unreliable predictor of the magnitude of the upcoming reward. In the test phase, distractors rendered in colors that had been unreliably pre-cued during training slowed RTs and drew more initial fixations than distractors rendered in reliably pre-cued colors. Surprisingly, this experimental manipulation also eliminated VDAC value-dependency: the magnitude of the associated reward had no impact on RTs or eye movements. That information reliability alone modulated attentional capture in our previous study suggests that the instrumental value of information, as conceptualized in the information-seeking literature, played a critical role. Here, we sought to evaluate the hypothesis that information reliability modulates experience-driven attention more directly. We repeated our modified VDAC protocol but replaced trial-to-trial rewards with accuracy-based feedback in order to isolate the impact of information reliability on experience-driven attention. Consistent with our hypothesis, unreliably pre-cued distractors slowed RTs and drew more initial fixations than reliably pre-cued distractors in the test phase. Given that reliably and unreliably pre-cued distractors had equivalent histories as sought targets, differences between them must be due to our manipulating the reliability of information in the training phase. These results converge with recent work from multiple labs investigating the role of uncertainty in experience-driven attention.

Acknowledgements: Supported by NSF-2141860 CAREER Award to MAG