Abstract
Orientation selectivity in primate visual cortex is organized in cortical columns that form pinwheel motifs along the cortical surface. While cortical columns are at a finer spatial scale than the sampling resolution of standard BOLD fMRI measurements, numerous fMRI analysis approaches have been proposed to peer past these spatial resolution limitations of the technique. It was recently reported that these methods are predominantly sensitive to an interaction of the oriented stimulus with the aperture edge—an effect called vignetting (Roth et al., 2018). Beyond vignetting, it is not clear whether, and to what degree, orientation-selective neural responses contribute to BOLD measurements. As a result, the field is at an impasse. Here, we leverage a large dataset of visual cortical responses, measured using high-field 7T fMRI, and preprocessed using state-of-the-art methods (Allen et al., 2021). We fit these responses using two image-computable models based on the steerable pyramid (Simoncelli et al., 1992). The constrained model, which includes both location tuning and spatial frequency tuning, but pools across orientation-selective filters, is sensitive to the effects of stimulus vignetting. The full model, which includes orientation tuning, is sensitive to any additional orientation-selective responses beyond the effects of stimulus vignetting. Using these two models, we compensate for vignetting but nonetheless find clear evidence for reliable tuning for orientation at the macroscopic scale probed by fMRI. Our results further reveal a striking widespread map of orientation preference. This map possesses signatures of both a radial bias and a cardinal bias, and may constitute the neural basis for perceptual orientation anisotropies. Taken together, our findings settle a long-standing debate in human neuroimaging, and lay the groundwork for a deeper understanding of stimulus feature encoding and organizational principles in human visual cortex.