Object category warps the topography of object space for behavior

Abstract

Recent neuroimaging work suggests that the primate ventral visual stream is organized according to a 2D object space, defined by two axes, namely animate-inanimate and stubby-spiky. While this 2D geometry explains substantial neural variance, does it serve as a static, category-independent scaffold for behavior? We addressed this question using a large-scale visual foraging dataset in humans (n = 511). By employing a rank order analysis and a novel reverse-correlation-inspired "landscape analysis" to map behavioral preferences in visual foraging "back to" the object space, we revealed a critical dissociation. We found that the behavioral preference topography in this space is highly stable within object categories, yet it fails to generalize across categories, frequently exhibiting reversed polarities. These findings challenge a "universal topography" hypothesis where the 2D object space guides behavior regardless of object category. Instead, our results demonstrate that object categories warp the topography of the 2D object space, supporting a "category-dependent topography" view. Taken together, our current results raise the question of how the object space is flexibly remapped in support of human behavior.