Obstacle coding in scene-selective cortices.

Navigating through complex environments requires detecting navigational constraints that arise from the interaction between paths and scene elements-so-called obstacles. Yet, the neural representation of such path-object interactions within scenes remains largely unknown. Across three fMRI experiments, we examined whether scene-selective regions-particularly the occipital place area (OPA)-encode the alignment between a path and an object that determines obstacle presence and its navigational significance. In Experiment 1, multivoxel pattern classification revealed that the OPA distinguished scenes in which a path was obstructed by an object versus unobstructed, suggesting that obstacle presence, defined by path-object alignment, is a key feature organizing its activity patterns. Experiment 2 ruled out an explanation based on low-level visual features: after removing high-level scene information through texturization, the OPA no longer showed significant obstacle decoding, unlike the early visual cortex. Finally, in Experiment 3, representational similarity analysis tested competing models of how obstacles are represented. Neural patterns in the OPA were best explained by the availability of passage, indicating that this region encodes the functional navigability of a scene rather than the mere presence of a blocking obstacle. Together, these findings demonstrate that the OPA integrates path and object information to represent action-relevant spatial structure, extending its established role in boundary representation to flexible, interaction-defined constraints on navigation.