Abstract
In three previously collected neuroimaging datasets (HCP 7T Retinotopy, n = 181; Temporal Decomposition through Manifold Fitting, n = 5; Natural Scenes Dataset, n = 8), we find large asymmetries in the magnitude of the BOLD response in primary visual cortex (V1): stimulus-evoked BOLD responses, expressed as percent change above baseline, are up to 50% stronger along the representation of the horizontal meridian compared to the vertical meridian. While this effect is quite robust, it remains unclear whether magnitude differences in the BOLD signal between locations can be interpreted as differences in local neural activity. To investigate this issue, we systematically evaluated the potential contribution of a number of non-neural factors to the observed effect, including cortical thickness, cortical folding, cortical depth, and vasculature. We find that only two of these factors have substantial relationship to the size of BOLD responses: cortical thickness and presence of macrovasculature (as indexed by T2*-weighted signal intensity). The relationship generalizes across subjects and brain regions, indicating that variation in BOLD response magnitudes across brain locations partly reflects differences in cortical thickness and vascularization. To compensate for these confounding factors, we then implement several regression-based correction methods and show that BOLD responses, after correction, become more spatially homogeneous. However, even after correction, BOLD response asymmetries in V1 still persist. This suggests a genuine neural substrate for previously reported asymmetries in behavioral performance, wherein performance is best along the horizontal meridian (Carrasco et al., Spatial Vision, 2001). More broadly, these results enhance the ability of neuroscientists to make correct inferences about local neural activity from fMRI data.