Abstract
A recent study by Bao and colleagues (Nature 2020) revealed a series of three topographic maps in macaque inferotemporal cortex (IT) that respond according to the animacy and aspect ratio of viewed objects. This could suggest that face-, body- and scene-selective regions of IT are districts of a larger continuous map whose organizing principles are based on visual properties of the stimulus. Such representational spaces were also found in a convolutional neural network trained to classify objects, suggesting a similar encoding of objects across the biological and artificial visual systems. However, these maps have not yet been identified in the human brain. Using fMRI, we measured the neural response in human IT to the same stimuli used in the macaque experiment. A localizer was designed using four object conditions: animate spiky, animate stubby, inanimate spiky and inanimate stubby. In 5 out of 6 subjects, we found regions of IT that were significantly more responsive to each condition than other conditions, potentially consistent with an animacy / aspect ratio organization in human IT. Next, we asked whether these regions were clustered in a series of three repeating maps of object space arranged along the posterior-anterior axis of IT, as was found in macaque. In 5 out of 6 subjects, we identify at least one cluster of regions that could indicate a map of object space. However, no subject clearly exhibited a series of three maps along IT. Taken together, these results suggest that the topographic organization of human IT may be accounted for by one or more maps of object space based on animacy and aspect ratio.