Validation of population receptive field estimates in human visual cortex
Population receptive field (pRF) analysis has become a popular method for non-invasively inferring the spatial tuning properties of the human visual system and for reconstructing brain activity in visual space. Yet few studies have sought to validate pRF parameter estimates or systematically compared them between different methods. Here we used pRF models to reconstruct the visual cortex response to pseudo-randomly placed ‘constellation’ stimuli. We present different methods for visualizing brain activity in visual space. Compared to typically used back-projection of pRF profiles, we show that a searchlight approach using only pRF location greatly improves the spatial precision of reconstructions. We further quantify the precision with which different pRF estimates distinguish between stimulated and unstimulated parts of the visual field. Both combined wedge-and-ring and more conventional sweeping-bar stimuli afford excellent localization of the presented test stimuli. Even a probabilistic pRF model based on cortical anatomy without any retinotopic mapping data performs well albeit with lower precision than empirical data. Our findings demonstrate that pRF analysis is an accurate and robust method for mapping the position preference of voxels in human visual cortex.