Abstract
To date, the large majority of studies employing the population receptive field mapping (pRF) technique have employed a 2-D Gaussian model, whereby the pRF of a given voxel is predicted by a circlular aperture of center (x, y) and size (sigma). Although there is some evidence to suggest that the accumulation of receptive fields in a given V1 voxel roughly approximates a Gaussian, there is also reason to think there might be systematic deviations from this assumption, particularly in higher-order visual areas or close to the vertical and horizontal meridian. For example, receptive field mapping in monkey inferotemporal cortex (IT) reveals a foveal bias with the receptive fields of most neurons overlapping the foveal region. Combined with biases for the contralateral upper and lower visual fields within a given area, we tested the prediction that the most efficient way for a population of RFs to completely cover either a quadrant or hemifield and still overlap the fovea individually is for them to take an ellipitical shape with the primary axis orientated towards fovea. Here we analyze pRF mapping data in human participants with both the 2-D Gaussian and elliptical pRF implementations developed in AFNI. Our results reveal a significant correlation between eccentricity and aspect ratio within all identified retinotopic maps (V1-V4, V3A, V3B, LO1, LO2 & V7/IPS0). Further, for areas containing largely quadrant representations of the visual field (e.g. V2, V3), the average pRF shows a systematic orientation toward the fovea. Such foveally biased and elliptical pRFs suggest the possibility of greater transfer of information between fovea and an eccentric position than between two isoeccentric positions in different quadrants of the visual fieldz (e.g. V2d and V2v). The widespread assumption of circular pRFs may obscure observations of systematic pRF biases throughout visual cortex.
Meeting abstract presented at VSS 2017