Manipulating uncertainty in value-driven attentional capture
Undergraduate Just-In-Time Abstract
Poster Presentation 43.359: Monday, May 22, 2023, 8:30 am – 12:30 pm, Banyan Breezeway
Session: Undergraduate Just-In-Time 2
Schedule of Events | Search Abstracts | Symposia | Talk Sessions | Poster Sessions | Abstract Numbering
Nicole Massa1, Nick Crotty1, Ifat Levy2, Michael Grubb1,2; 1Trinity College, 2Yale School of Medicine
Previously rewarded stimuli slow RTs during visual search, despite being physically non-salient and no longer task-relevant or rewarding. Such value-driven attentional capture (VDAC) is typically measured in a training-test paradigm. In the training phase, the search target is rendered in one of two colors (one reliably predicting high and the other low reward). Target color is unknown at the trial start. Here we asked, how does resolving target-color uncertainty in advance of the search array impact attentional capture at test? Our modified training phase used four differently-colored targets. Two target colors were always preceded by a "valid" precue (two squares at fixation rendered in the same color), signaling the next target color with 100% certainty. The other two target colors were always preceded by a “distributed" precue (fixation rendered in both potential target colors), signaling either possible target color with 50% certainty. Within each color pair, one color reliably predicted high and the other low reward. We used the traditional VDAC test phase; on half the trials, one of the distractors was rendered in one of the former target colors. Using a preregistered data collection and analysis plan, we obtained a highly-powered sample of 79 participants. We confirmed robust capture with the most widely used metric of VDAC: previously-high-reward colors from the distributed precue condition slowed RTs relative to the distractor-absent condition. Of interest to reward prediction accounts of VDAC is that we found a main effect of precue type: when precues, rather than targets themselves, signaled the upcoming reward during training (valid condition), significantly reduced capture was observed at test. Complicating reward prediction accounts, we found no main effect of reward magnitude and no precue x reward interaction (both showing strong evidence for the null hypothesis with Bayesian statistics). A potential role for information prediction errors will be discussed.
Acknowledgements: Supported by NSF-2141860 CAREER Award to MAG and NIH/NIMH grant R01MH118215 to IL