Cortical Selectivity to Real-World Object Size In Natural Images

Real-world size is a fundamental property of objects that supports adaptive interactions with the physical environment. Prior fMRI studies have identified large-scale organization in higher visual cortex associated with real-world object size but have relied primarily on decontextualized objects presented on simple backgrounds. Here, we ask whether this neural organization generalizes to objects embedded within the context of complex natural scene images. To address this question, we used voxelwise analyses and encoding models to assess neural responses to natural images from two independent fMRI datasets. These analyses identified reliable organization encompassing alternating regions of sensitivity for small and large objects across the ventral and lateral surfaces of occipitotemporal cortex, which largely aligns with prior findings using isolated objects. At the same time, we identify additional, previously unreported size-related organization on the ventral surface at a finer spatial scale. We also find real-world size sensitivity for both animate and inanimate objects, with broadly corresponding cortical patterns across the two domains. Importantly, controlling for low- and mid-level visual features, object composition, and scene clutter attenuated the magnitude of size sensitivity in some regions, but closely preserved the spatial organization and directionality of neural size preferences. Finally, we characterized the shape of voxelwise tuning functions for real-world object size, and found that a logarithmic-like relationship predominated, although profile shape varied across regions and between small- and large-preferring voxels. Overall, our findings support real-world object size as a behaviorally-relevant dimension of cortical organization during natural vision.