Eye tracking reveals the cost of switching between self and other perspectives in a visual perspective-taking task

TLDR

People can quickly compute visual perspectives but struggle to ignore irrelevant perspectives when they differ. The study used an avatar perspective‑taking task to investigate the mechanisms behind egocentric and altercentric interference. Participants were eye‑tracked while judging disc counts from either their own or an avatar’s perspective, with trials varying in perspective consistency and preceding cue, allowing comparison of same‑vs‑different cue effects. Altercentric interference was reduced or eliminated when participants maintained their own perspective across trials, and distinct fixation patterns indicated that consistency effects arise from implicit mentalizing rather than avatar cues.

Abstract

Previous studies have shown that while people can rapidly and accurately compute their own and other people's visual perspectives, they experience difficulty ignoring the irrelevant perspective when the two perspectives differ. We used the "avatar" perspective-taking task to examine the mechanisms that underlie these egocentric (i.e., interference from their own perspective) and altercentric (i.e., interference from the other person's perspective) tendencies. Participants were eye-tracked as they verified the number of discs in a visual scene according to either their own or an on-screen avatar's perspective. Crucially in some trials the two perspectives were inconsistent (i.e., each saw a different number of discs), while in others they were consistent. To examine the effect of perspective switching, performance was compared for trials that were preceded with the same versus a different perspective cue. We found that altercentric interference can be reduced or eliminated when participants stick with their own perspective across consecutive trials. Our eye-tracking analyses revealed distinct fixation patterns for self and other perspective taking, suggesting that consistency effects in this paradigm are driven by implicit mentalizing of what others can see, and not automatic directional cues from the avatar.