Population receptive field properties change dynamically within milliseconds

Abstract

The visual system works extraordinarily fast, and perception and neuronal properties are shaped by recurrent processing on the scale of hundreds of milliseconds. Here we studied how these recurrent processes shape population receptive field (pRF) properties. To do this, we combined functional magnetic resonance imaging (fMRI), magnetoencephalography (MEG) and computational modeling to measure pRF dynamics in the human visual cortex. We first modelled the pRFs with ultra-high field fMRI using conventional pRF modeling. The participants then viewed contrast-defined bar and circle shapes and we measured stimulus-evoked responses to these stimuli with MEG. Next, we combined the pRF models and a forward model to predict the MEG sensor responses to the stimuli. We computed the goodness of fit between the predicted and measured MEG responses using cross-validated variance explained (R-squared). Last, we adjusted the pRF model parameters to find the pRF properties that optimally explained the measured MEG signal at different latencies. We measured clear stimulus-evoked MEG responses to the different contrast-defined stimuli. The MEG responses peaked at 100 and 200 ms, in opposite polarity. The pRFs measured by fMRI explained up to 81% of the MEG signal, with best fits at 100 and 200 ms in occipital sensors. Systematically adjusting the pRF parameters revealed that different pRF sizes explained the variance at different latencies. Specifically, we found that pRF sizes are smaller at 100 ms than at 200 ms. The sizes of population receptive fields vary within milliseconds. We found that pRF sizes and the polarity of the signals differ at 100 and 200 ms. We speculate that these differences reflect different contributions of feedforward and feedback processing. These results highlight the feasibility to study temporal dynamics on a millisecond timescale in the human brain.