During natural vision, the brain must infer real-world properties from sparse and noisy sensory signals. Predictive processing accounts suggest that this is enabled by comparing incoming information to predictions based on internal models. Given that internal models are critical for perception, can inter-individual variations in internal models explain why people diverge in their perception of the world?
In this study, an inter-subject representational similarity analysis framework will be employed to test whether idiosyncrasies in objective and subjective scene perception can be predicted by similarities in internal models of the world. To characterize internal models, we will ask participants to draw prototypical versions of indoor scenes. Additionally, participants will copy scene photographs (to control for drawing ability) and provide pictures of their homes (to control for immediate visual experience). For all three conditions, we will construct inter-subject representational dissimilarity matrices (RDMs) that capture the similarity of drawings across all pairs of participants. Subsequently, these RDMs will be compared to RDMs capturing inter-subject similarity in a set of behavioral performance metrics: (i) a categorization task involving photographs of indoor scenes and (ii) five subjective ratings on the same photographs (typicality, familiarity, attractiveness, usability, and complexity).
We hypothesize that inter-subject similarities in internal models predict inter-subject variations in perception. Consequently, behavioral RDMs should correlate most strongly with the RDM constructed from typical drawings, compared to RDMs constructed on the copied drawings or scene photographs. These findings would suggest that participants with similar internal models perceive the world in similar idiosyncratic ways.