Abstract: PURPOSE: Object recognition, an essential cognitive function in the human visual system, depends on the ventral temporal cortex (VTC). However, the functional principles and neural mechanisms of the IT cortex remain largely unexplored. Earlier studies have proposed the use of the object space model to understand the functional organization of the IT cortex in macaques (Bao, et al. 2020), but its relevance to humans is still unclear.
METHODS: To address this question, we used fMRI to measure the response to a large number (n = 500) of static object stimuli from 5 subjects.
RESULTS: The object space was defined using principal component analysis of the VTC's responses. Our results showed that the functional organization of the VTC can be represented by a low-dimensional object space, with the first two principal components accounting for 92% variance of the consistency of the representation space. These two principal components can be broadly characterized as face versus spiky objects and animal versus stubby objects. Additionally, to examine the consistency of object spaces among different participants, we used hyperalignment methods (Haxby, et al. 2020) to project responses of the VTC onto a common space and then back onto the cortex of one participant, creating a unified template. The high consistency across subjects was found not only in known category-selective areas but also in other parts of the VTC, suggesting a common space represented across different subjects. To further investigate the similarities in object representation between humans and macaques, we compared the object space between the two species. Comparisons with electrophysiological data from neurons showed that the space constructed by human VTC responses closely resembles that represented in the IT of macaques.
CONCLUSIONS: This implies that the creation of object-specific space representations is a key aspect of object recognition and that the functional organization of the IT cortex is preserved across species.

Key words: Object recognition, ventral temporal cortex, object space map, fMRI, deep neural network