The nature of the animacy organization in human ventral temporal cortex

Abstract

The principles underlying the animacy organization of the ventral temporal cortex (VTC) remain hotly debated, with recent evidence pointing to an animacy continuum rather than a dichotomy. What drives this continuum? According to the visual categorization hypothesis, the continuum reflects the degree to which animals contain animal-diagnostic features. By contrast, the agency hypothesis posits that the continuum reflects the degree to which animals are perceived as (social) agents. Here, we tested both hypotheses with a stimulus set in which visual categorizability and agency were dissociated based on representations in convolutional neural networks and behavioral experiments. Using fMRI, we found that visual categorizability and agency explained independent components of the animacy continuum in VTC. Modeled together, they fully explained the animacy continuum. Finally, clusters explained by visual categorizability were localized posterior to clusters explained by agency. These results show that multiple organizing principles, including agency, underlie the animacy continuum in VTC.

A humanness dimension to visual object coding in the brain

“…The superior performance of the human model builds on our existing understanding of the 503 representation of object categories in the brain. The model extends on the continuum idea of 504 Connolly et al (2012) and Sha et al (2015), as it represents a type of human-similarity continuum 505 (see also Thorat, et al, 2019). However, unlike the animacy continuum that is based on biological 506 classes, the human model was not limited by biology (Gobbini, et al, 2011;Tong, Nakayama, 507 Moscovitch, Weinrib, & Kanwisher, 2000).…”

“…for example, proposed that the neural representation of objects is better characterised according to 66 the object's ability to perform goal-directed actions (see also Thorat, Proklova, & Peelen, 2019). 67…”

“…This idea shifts the focus away from models based on categories, with a 535 view to a more holistic approach to object representations that considers interactions between 536 various circuits throughout the brain. In this multidimensional framework, it is essential to 537 recognize that no single feature or attribute could be able to fully explain the richness of the brain's 538 (multidimensional) representation of objects (see e.g., Thorat, et al, 2019). Recent fMRI studies 539 have sought to identify principle axes of object representations in the brain (e.g., Connolly, et al, 540 2012; Sha, et al, 2015; Thorat, et al, 2019).…”

“…In this multidimensional framework, it is essential to 537 recognize that no single feature or attribute could be able to fully explain the richness of the brain's 538 (multidimensional) representation of objects (see e.g., Thorat, et al, 2019). Recent fMRI studies 539 have sought to identify principle axes of object representations in the brain (e.g., Connolly, et al, 540 2012; Sha, et al, 2015; Thorat, et al, 2019). In the present study, we show that the human 541 similarity provided the best account of late stage processing, highlighting "humanness" as a key 542 feature in the human brain's representation of objects that shapes our experience of the world.…”

Neural representation of abstract task structure during generalization

“…In the case of LS, value, and context RDMs, the hypothesized distances were simply set so that conditions that were the same on these factors would have smaller expected distances and larger distances where these conditions were not the same. For the category RDM, hypothesized distances were based on asymmetric distances related to the perceptual-semantic distances of animate and inanimate stimuli observed in many other studies ( Chikazoe et al, 2014 ; Thorat et al, 2019 ). Namely, that animate categories (faces and animals) were expected to be more similar to each other than inanimate objects, but inanimate objects were expected to be more similar to animals than to faces.…”