Abstract
Scene construction involves automatic synthesis of fragmented visual inputs into coherent mental models. Although previous studies have focused on static landscapes, real-world scenes require the integration of dynamic agents with stable environmental anchors. How the brain represents the spatial configuration of such dynamic agents relative to a static environment remains unclear. We investigated this question using an automatic space-encoding task in a three-dimensional virtual-reality environment and functional magnetic resonance imaging representational similarity analysis. The results revealed functional segregation across the cortex, with the superior lateral occipital cortex representing the first-person perspective of dynamic character layouts, and the lingual gyrus and occipital fusiform cortex representing the environment specified by visible landmarks. The orbitofrontal cortex (OFC) integrates these streams into an allocentric spatial state that captures relational geometry beyond the current visual field. Together, these findings indicate that the OFC represents the spatial relationship between dynamic agents and the static environment, an allocentric spatial framework underlying stable scene representation, even when the scene is encoded incidentally.