Emily Granger^1,2, Corey M Ziemba¹, Philipp Kreyenmeier², Hendrikje Neinborg¹, Elisha P Merriam²; ¹Labratory of Sensorimotor Research, National Eye Institute, Bethesda, MD, ²Laboratory of Brain and Cognition, National Institute of Mental Health, Bethesda, MD

Visual processing in the brain is organized hierarchically, with early areas selective for basic visual information and later areas becoming selective for complex, categorical information. Areas of the early visual cortex (V1-V3, EVC) are believed to mainly process low-level features such as position and orientation. The fusiform face area (FFA), in contrast, exhibits a strong preference for faces that relies on higher-level processing. Despite this, some studies have suggested that EVC contains category selective responses, particularly in the foveal representation, a property attributed to feedback from higher visual areas. Here we investigate whether this preference can be explained by low-level image statistics that are overrepresented in face stimuli compared with other image categories. Subjects viewed alternating blocks of rapidly presented images of isolated faces or objects while fixating on the center of the screen. Participants performed a 1-back task on the images while we measured BOLD fMRI responses with a 7T scanner. We observed strong face preference in FFA, as expected, and a robust face preference in the foveal representation of EVC. To better understand the origin of this foveal face preference, we used a texture synthesis algorithm to scramble the more complex features of the face and object images, rendering image category unrecognizable, while maintaining the low- and mid-level features. We found that FFA did not exhibit significant selectivity to face textures relative to object textures, consistent with the putative role of FFA in face perception. However, we observed robust face preference in foveal EVC that was similar for both intact and scrambled images. These results suggest that feedback from FFA is unlikely the source of category selectivity in EVC. Rather, our results demonstrate that category selective response in EVC reflect a preference for low-level natural image statistics.

Acknowledgements: We acknowledge funding from the National Eye Institute and National Institute of Mental Health Intramural Research Program at the National Institutes of Health (NIH, grant no. 1ZIAEY000570-01 to H.N. and NIH, grant no. ZIAMH002966 to E.M.)