Xucong Hu¹ (), Yitong Zheng¹, Qinyi Hu¹, Haokui Xu², Hui Chen¹, Mowei Shen¹, Jifan Zhou¹; ¹Zhejiang University, ²Zhejiang University of Technology

The mechanism underlying visual perspective-taking (VPT) remains contested: do people rely on perceptual simulation, reconstructing what others see, or on heuristic inference grounded in naïve optics? Although recent evidence favors heuristic-account, most support comes from explicit VPT tasks in which participants report the agent’s view. Whether implicit (spontaneous) VPT—where participants perform an unrelated task yet are influenced by the agent’s viewpoint—relies on the same mechanism remains unclear. This matters because explicit tasks permit strategic top–down correction based on high-level knowledge, potentially masking underlying visual processes. Moreover, simulation-account may seem unsupported simply because observers lack access to the agent’s visual experience—yet providing such experience in explicit tasks would reveal the answer. We therefore tested whether prior visual experience could dissociate the two accounts in spontaneous VPT. In Experiment 1, participants received plausible, no, or implausible agent-view priors but completed a viewpoint-irrelevant line-length reproduction task. The spontaneous VPT effect was defined as a systematic bias in reproduction consistent with the agent’s view. The simulation-account predicts such a bias only under plausible priors, requiring viable sensory input; the heuristic-account predicts following any prior, even implausible, because it relies on abstract, knowledge-based rules. Results showed a bias for plausible priors, no bias for no prior, and a weak bias for implausible priors. A control experiment ruled out priming. The instability of the implausible-prior effect (ratio analyses yielded only marginal effects) suggests plausibility-based integration rather than blind adherence. Experiment 2 used a Ponzo-style stimulus with five levels of prior-plausibility. A pattern emerged: participants followed the prior when it was plausible or only moderately implausible, but abandoned it when highly implausible. This pattern reflects the bounded rationality of heuristic inference: observers integrate the provided prior with visual input while it remains within a plausible range, but revert to self-visual evidence once the “plausibility” threshold is exceeded.

Acknowledgements: This research was supported by the National Natural Science Foundation of China (32371088, 62337001, and 32471093), the Science and Technology Innovation 2030 —“Brain Science and Brain-like Research” Major Project (2022ZD0210800), and the Fundamental Research Funds for the Central Universities (226-2024-00118).