Abstract
Our visual system represents visual objects according to categories. One of most basic categories is the animate vs. inanimate division. Based on results from recent studies (Livingston et al., 2015; Perrinet & Bednar, 2015) and our earlier study (Yue et al., 2014), we hypothesized that animate vs. inanimate categorization is: 1) encoded by multivoxel activity patterns; and 2) influenced to a large degree by the unique image-based features that distinguish animate from inanimate stimuli. Using a slow event-related design, we acquired fMRI scans in three fixating rhesus macaques in response to a large set of visual stimuli, including 47 sub-categories with 20 images per sub-category. We employed curved and rectilinear Gabor filters to quantify curved and rectilinear image-based features. As hypothesized, we found that multivoxel activity patterns measured with support vector machine classification encoded animate vs. inanimate categories in the monkey inferior temporal cortex; multiple dimensional scaling failed to categorize individual exemplars in the animate vs. inanimate division. This result suggests that animate vs. inanimate categorization is represented in the brain in a high dimensional space, and is not a 2-dimensional representation. Moreover, curved and rectilinear features explained a significant amount of variance in the fMRI activity patterns that encoded animate vs. inanimate categories. Our results thus support our hypothesis that animate vs. inanimate categorization in the inferior temporal cortex is influenced to a large extent by the unique image-based features (such as curved and rectilinear features) that distinguish animate vs. inanimate stimuli. The results argue against the notion that categorization stems from acquired semantic knowledge of the characteristics that distinguish object categories, and instead suggest that the unique image-based features that distinguish animate vs. inanimate stimuli give rise to the formation of categorization in the macaque inferior temporal cortex.
Meeting abstract presented at VSS 2017