Improving the reliability and accuracy of population receptive field measures using a ‘log-bar' stimulus

The population receptive field (pRF) method is a powerful tool for investigating the functional organization of human early visual cortex using fMRI (Dumoulin and Wandell, 2008). The fMRI response to a time-varying stimulus is used to estimate the size and location of each voxel’s ‘receptive field’ by fitting a voxel’s time-course response to a forward model that windows the stimulus with a Gaussian in space followed by convolution in time by a hemodynamic response function. We ran simulations of the forward model with realistic pRF parameters and additive noise. Using the traditional moving bar stimulus, which has a fixed bar width, estimated pRF locations are biased away from the fovea and estimated pRF sizes are biased toward larger Gaussians, especially for pRFs with centers near the fovea. These biases seem to be due to the cortical magnification of the fovea. When a fixed width bar stimulus in visual space is projected onto the cortical surface, the cortical projection of the bar size is larger and travels more slowly near the fovea. Here, we introduce a ‘log-bar’ stimulus which is a moving bar that is logarithmically warped along its eccentricity dimension. This log-bar stimulus has a relatively constant width and speed when projected on the cortical surface. Simulations show that the ‘log-bar’ stimulus reduces the eccentricity and size biases found with a traditional fixed-bar stimulus. In the scanner, we estimated pRFs in 12 subjects using either a fixed width bar stimulus or the log-bar stimulus. In foveal and parafoveal early retinotopic visual areas, the log-bar stimulus produced more reliable pRF size and location estimates and smaller pRF size estimates. These results are consistent with our simulations and indicate that a log-bar stimulus generates more reliable and accurate estimates of pRF location and size, especially near the fovea.