Prior research has revealed a fundamental tripartite division of higher-level visual cortex into regions responding preferably to large objects, all animals, or small objects (Konkle & Caramazza, 2013). It has been suggested that this organization by animacy and real-world size is supported by mid-level visual curvature information (Long et al., 2018). Given the comparably small scale of the original studies, we aimed to test how these organizational principles apply to a broader, more representative set of natural images and, more generally, what other visual and semantic features could underlie the functional organization of occipitotemporal cortex. To this end, we used THINGS-fMRI (Hebart et al., 2023), featuring fMRI responses to 8,740 naturalistic images of 720 categories, and a rich behavioral dataset of object property ratings (Stoinski et al., 2023), as well as novel image-level curvature ratings. Our results confirm many facets of the characteristic animacy-size organization, but reveal a more partitioned size organization and a stronger distinction of responses to large and small animals. These findings were consistent across all subjects and robust to various modulations of our stimulus set. With respect to curvature, we observed rectilinear preferences in scene-selective regions, with interleaved curvy preferences. Nevertheless, perceived curvature alone could not explain the animacy-size division. Finally, variance partition revealed high multicollinearity of object properties, with moveability (“can move”), preciousness, and curvature explaining most unique variance. Together, our results confirm the generalizability of previous findings to representative, natural images while refining and enhancing our understanding of visual cortex organization.