Distinct mechanisms of contextual modulation for dynamic natural scenes in motion- and scene-selective cortex

Neural responses in visual cortex to a central stimulus are modulated by its surrounding context. Although this contextual modulation is well established for simple stimuli, how it manifests during dynamic naturalistic scene perception remains unclear. Using functional magnetic resonance imaging (fMRI), we examined how motion congruence and categorical similarity between central and surrounding scenes shape contextual modulation across the visual hierarchy. Central and surrounding scenes systematically varied in categorical similarity (identical exemplar, different exemplar of the same basic-level category, different basic-level category, or different superordinate category) and in motion direction (same or opposite direction). Neural responses were measured in primary visual cortex (V1), motion-selective cortex (hMT+), and scene-selective occipital and parahippocampal place areas (OPA and PPA). hMT+ showed robust motion-dependent contextual modulation, with stronger suppression for same-direction motion. V1 responses were sensitive to low-level similarity between center and surround, with reduced responses when center and surround were physically identical and moved in the same direction. In contrast, scene-selective regions (OPA and PPA) did not show univariate modulation by motion congruence or categorical similarity, but multivoxel activity patterns reflected the categorical relationship between center and surround. Multivariate analyses further showed that decoding accuracy in OPA and PPA increased as categorical similarity decreased. Together, these results demonstrate that contextual modulation in natural vision differs across visual regions, reflecting sensitivity to motion relationships in hMT+, low-level similarity in V1, and higher-level categorical structure in scene-selective cortex. These findings highlight how contextual influences operate at multiple levels of visual representation to support the processing of dynamic natural scenes.NEW & NOTEWORTHY Visual perception depends on contextual interactions across the visual field. Using functional magnetic resonance imaging (fMRI) and dynamic natural scenes, we show that contextual modulation differs across the visual system: it is motion-dependent in motion-selective cortex (hMT+), reflects low-level similarity in primary visual cortex (V1), and is shaped by categorical relationships between center and surround in scene-selective cortex. Together, these findings reveal differences across functionally distinct visual regions in how contextual information is represented during natural vision.