Abstract
The pulvinar is richly interconnected with visual and association cortices that support a variety of visual functions. Studies in non-human primates have shown that inferior and lateral regions of the pulvinar are connected primarily to early visual cortex, have a clear retinotopic organization, and are sensitive to low-level visual features. However, the pulvinar’s potential role in high-level visual functions such as scene perception and object recognition has been little studied. To address this gap, we investigated subcortical activity in the Natural Scenes Dataset (NSD). In NSD, 8 humans completed a visual fMRI experiment during which they were exposed to 9,000–10,000 unique natural scenes. Whole-brain fMRI data were collected at 7T with 1.8-mm resolution, ensuring high-quality measurements from subcortical regions. We fit population receptive field (pRF) models to individual voxels, systematically evaluating different stimulus features (contrast, saliency, faces, bodies, foreground, background) that might be encoded in voxel responses. The pRF analysis confirmed that the LGN and inferior-lateral portions of the pulvinar are selective for contrast presented in the contralateral visual hemifield and have clear retinotopic organization, consistent with prior work. Interestingly, the analysis revealed a subregion of the pulvinar, located medial and posterior to the contrast-selective region, that is selective for bodies and faces presented in the contralateral visual hemifield. In-between these two subregions, we also identified a region of the pulvinar sensitive to foreground objects in general. These results indicate a role of the pulvinar in object recognition, specifically in the processing of high-level visual features such as faces and bodies. The functional organization of the pulvinar appears to parallel the hierarchical organization of ventral visual cortex, and suggests that the classic two-stream model of the visual system extends to subcortex.