Shaozhi Nie¹ (), Daniel Kersten¹, Stephen A. Engel¹; ¹University of Minnesota, Twin Cities

Segmented objects show enhanced neural response relative to background in early visual cortex. However, it is unclear whether such response enhancement simply reflects salience, and whether it is visible in natural scenes and for multiple objects. Using fMRI, we tested the condition under which V1/V2 locations whose population receptive fields (pRFs) overlapped with objects showed higher response than when they overlapped the background. We analyzed the Natural Scenes Dataset (NSD), which consists of 7T-fMRI data from eight participants, each viewing 10,000 images from the Common Objects in Context dataset (COCO). NSD also includes pRFs of the image pixels driving each cortical location. The COCO images have rich annotations of segmented pixels. We operationalized objecthood using the “thing” annotation, which included most objects, and contrasted it with "stuff" including common backgrounds, such as sky and grass. In addition, we used a state-of-the-art deep network (FOCUS, developed to identify foreground, You et al., 2025) to calculate the salience of COCO segments. For each image, we computed the objecthood and salience at each pixel and projected these onto cortical locations by dot product with pRFs. Average V1/V2 responses were higher for pRFs on an object than background (3.4% signal change for object vs 2.9% change for background, p < 0.01 by permutation test). To control saliency, we used FOCUS scores to predict each vertex's response. Residuals were higher of vertices on an object, confirming an effect beyond salience. A similar analysis accounted for low-level features by fitting a V1-like Gabor filter-bank model, and again found effects of objecthood in the residuals. Effects of objecthood were also found when analysing separately the first, second, and third most salient objects. Our results demonstrated that object segmentation increases response in the human visual cortex for multiple objects in complex naturalistic scenes.