Abstract
The visual system can distinguish patterns over a wide variety of spatial scales. Based on physiological evidence, this robustness is thought to reflect the retinotopic scale-invariant organization of primary visual cortex, whereby the spatial area over which a cell responds is proportional to the spatial information that cell is tuned to convey. Specifically, larger receptive fields (RFs) when moving away from the foveal representation reflect the integration of information from increasingly larger regions of the visual field and a reduction in spatial sampling density. The assumption made by scale invariance is that the information sampled by a RF is constant across the visual field. Model-based fMRI methods, such as population receptive field (pRF) and population spatial frequency tuning (pSFT) analyses, have enabled the characterization of spatial tuning (eccentricity and size) and SF preferences (peak preference and tuning bandwidth) of cortical sub-populations sampled in a voxel. Using these tools, we evaluated whether scale invariance, defined as a ratio between peak pSFT and pRF size (cycles/RF), is constant across the visual field in early visual areas V1–V3 (N=8). We found a marked departure from scale invariance: there was an expansive increase in cycles/RF for voxels near the central visual field, and an eventual decrease for voxels that prefer higher eccentricities. This departure from scale invariance at innermost eccentricities was most pronounced in V1. In extrastriate cortex (V2–V3), the relationship of cycles/RF and eccentricity appeared more constant across the visual field — closer to the predictions of scale invariance. Taken together, these results reveal that SF preference does not scale linearly with RF size, a central assumption commonly used to adjust for cortical magnification factors.