Yichen Wu^1,2 (), Sheng Li^1,2,3,4,5; ¹School of Psychological and Cognitive Sciences, Peking University, Beijing, China, ²PKU-IDG/McGovern Institute for Brain Research, Peking University, Beijing, China, ³Beijing Key Laboratory of Behavior and Mental Health, Peking University, Beijing, China, ⁴Key Laboratory of Machine Perception (Ministry of Education), Peking University, Beijing, China, ⁵National Key Laboratory of General Artificial Intelligence, Peking University, Beijing, China

Scene memory is prone to systematic distortions. A well-known distortion along the near-far axis of the three-dimensional space, called boundary transformation, demonstrates the observer’s erroneous recall of the viewing distance of scenes. The investigators argued for the role of the normalization process to the high-probability prototypical viewing distance in boundary transformation. Here, we hypothesized that the prototypical viewpoint also includes the dimension of vertical angle of view (vAOV) and may cause memory distortion along the scenes’ vertical axis. In two behavioral experiments, we found a systematic memory distortion in vAOV in both force choice (n = 79) and free adjustment (n = 30) tasks. The force choice task used the classic RSVP paradigm of boundary transformation. Subject were told to detect the vAOV change of two identical scenes. The free adjustment task adapted the RSVP paradigm and allowed subjects to adjust the scene’s pitch, yaw, and depth to relive their memory. Computational modeling suggests that the vAOV biases could be predicted by the asymmetry of the complexity information in the vertical direction of the scenes and the subjective vAOV ratings from a group of online participants. These behavioral findings are in line with the normalization process. We further conducted an fMRI and an MEG experiment with the RSVP paradigm to explore the neural mechanism of vAOV bias. The fMRI experiment (n = 24) revealed the involvement of V4, PPA, and OPA in the normalization process. The results of MEG (n = 20) experiment provided temporal evidence that memory distortion appeared approximately 140 ms after scene onset and was accompanied by the estimation of low-level visual complexity. These findings suggest that the memory distortion is embedded in the feedforward sweep of scene processing. Estimation of the complexity information could serve as the initial step of vAOV bias.