Abstract
Much of our understanding of human visual processing stems from measures of neural activity in the macaque visual system. However, to enable inference across species, it is critical to determine functional homologies of the different stages of processing. This is usually determined based on a conceptual understanding of functional specialization, which can be tricky for higher stages of processing where little is understood about what functionally distinguishes one stage from another. For example, the homologies of individual face-selective areas are not firmly established. The macaque middle-lateral and anterior-lateral face regions (ML/AL) have been suggested to correspond to: (a) the human occipital and fusiform face areas (OFA/FFA), respectively, (b) posterior and anterior FFA (FFA1/FFA2), or (c) FFA and the anterior temporal lobe face region (FFA/ATL). Here, we apply a data-driven approach that does not rely on specific premises of functional specialization to arbitrate between these three scenarios, by leveraging a large dataset of human fMRI responses to complex, natural images (Allen et al. 2021). We recorded neural responses in macaque visual cortex (321 multi-unit sites, 4 monkeys) to 1000 of those images and compared the neural selectivity profiles with human fMRI responses. Two monkeys had electrode arrays in ML and two in AL. For both ML monkeys, neural selectivity profiles matched responses in human FFA1 and FFA2 significantly better than in OFA or ATL. For both AL monkeys, neural selectivity profiles matched human ATL significantly better than either OFA, FFA1, or FFA2. Thus, our results confirm functional homologies between ML/AL and FFA/ATL, respectively (scenario (c)), which is most consistent with anatomical evidence. Overall, our findings suggest that evidence for functional homology can be established using a data-driven approach that evaluates the representation of a comprehensive set of natural images, thereby circumventing the need for conceptualizing functional parallels across species.