Heterogeneous single-unit selectivity in an fMRI-defined body-selective patch.

Although the visual representation of bodies is essential for reproduction, survival, and social communication, little is known about the mechanisms of body recognition at the single neuron level. Imaging studies showed body-category selective regions in the primate occipitotemporal cortex, but it is difficult to infer the stimulus selectivities of the neurons from the population activity measured in these fMRI studies. To overcome this, we recorded single unit activity and local field potentials (LFPs) in the middle superior temporal sulcus body patch, defined by fMRI in the same rhesus monkeys. Both the spiking activity, averaged across single neurons, and LFP gamma power in this body patch was greater for bodies (including monkey bodies, human bodies, mammals, and birds) compared with other objects, which fits the fMRI activation. Single neurons responded to a small proportion of body images. Thus, the category selectivity at the population level resulted from averaging responses of a heterogeneous population of single units. Despite such strong within-category selectivity at the single unit level, two distinct clusters, bodies and nonbodies, were present when analyzing the responses at the population level, and a classifier that was trained using the responses to a subset of images was able to classify novel images of bodies with high accuracy. The body-patch neurons showed strong selectivity for individual body parts at different orientations. Overall, these data suggest that single units in the fMRI-defined body patch are biased to prefer bodies over nonbody objects, including faces, with a strong selectivity for individual body images.