Abstract
A multitude of studies have shown that high-level regions of the ventral visual pathway differentially code for the animacy of objects. However, less clear are the organizing principles that structure the neural responses to images of “animate” exemplars. On the one hand, some studies have used images of human and animal faces and bodies (Kiani et al. 2007; Kriegeskorte et al. 2008, Cichy et al. 2014), which suggest that body partonomics may play a role. Such a possibility is reinforced by the well-known selectivity for body and face stimuli in circumscribed regions of lateral occipital and ventral temporal cortex. On the other, some studies suggest that intuitive hierarchies for biological class membership are represented in these more general regions (Connolly et al. 2012, 2016; Sha et al. 2015). We investigated the relative contribution of these two factors in accounting for the neural responses in multiple regions of the ventral pathway. Animate stimuli consisted of a single close-up face and full-body image of 24 animals from intuitive groupings of primates, mammals, birds, reptiles/amphibians, fish, and insects/invertebrates (48 images total). In behavioral and neural testing these were contrasted with (subsets) of 48 images of natural objects (flowers, fruits/vegetables, fungi). In total we collected data for seven behavioral tasks involving either similarity judgments (face, body, humanness, animacy) or categorization decisions (animate vs inanimate, face vs non-face, face vs body). The responses from these tasks were used to construct dissimilarity matrices (DM) to carry out representational similarity analysis, and compared with DMs constructed from neural responses from multiple ventral pathway regions selective for objects, faces, and bodies measured with human fMRI, as well as layers of a deep convolutional neural network. We found that both factors, body partonomics and biological class, are complementary predictors of neural responses across regions of the ventral pathway.