Population receptive field (pRF) modeling is a popular method to map the retinotopic organization of the human brain with fMRI. While BOLD-based pRF-maps are qualitatively similar to invasively recorded single-cell receptive fields in animals, it remains unclear what neuronal signal they truly represent. We address this question with whole-brain fMRI and large-scale neurophysiological recordings in awake nonhuman primates. Several pRF-models were independently fit to the BOLD signal, multi-unit spiking activity (MUA) and local field potential (LFP) power in distinct frequency bands. Our results provide a retinotopic characterization of cortical and subcortical areas, suggest brain-wide compressive (ie, sublinear) spatial summation, and demonstrate a visually tuned deactivation of default mode network nodes. Cross-signal analysis of pRF-map structure (eccentricity-size relation) indicates that the neural underpinnings of BOLD-pRFs are area-specific. In V1, BOLD-pRFs mirror MUA, while in V4 they are more similar to the tuning of the gamma LFP-power.