Abstract
Purpose: Multivariate analyses are widely employed in human fMRI studies. But there is considerable controversy over what decoding analyses reveal about underlying neural architecture. We previously demonstrated that a coarse-scale bias for radial orientations in human primary visual cortex (V1) is both necessary and sufficient for orientation decoding. It has been hypothesized that the radial bias is related to the edge of the stimulus aperture, rather than a neural preference for radial orientation. We tested this possibility by measuring fMRI activity to oriented gratings while systematically manipulating the shape of a modulator grating. Method: Stimuli consisted of a carrier grating multiplied by a modulator. The carrier was a Cartesian sinusoidal grating (0.5 cycles/deg) that filled an annulus (0.5 to 10 deg). Orientation of the stimulus cycled through sixteen evenly-spaced angles (0-180 deg) in 24 s. The modulator was a second sinusoidal grating that was constant throughout each fMRI run. On half the runs, the modulator had an angular orientation, producing a series of 'spokes' centered on the fixation cross. On the other half of runs, the modulator had a radial orientation, producing 'rings' emanating from fixation. The fMRI data analysis characterized the preferred carrier grating orientation for each voxel. Results: The majority of voxels in V1 exhibited robust and reliable orientation preferences to the carrier grating, confirming that fMRI responses in human V1 are selective for orientation. However, the orientation preferences of most voxels were determined by the orientation of the modulator. For the radial modulator, voxels exhibited a radial bias. For the angular modulator, voxels exhibited a tangential bias (i.e., rotated 90 deg from radial). We observed no evidence for orientation tuning that was unaffected by the orientation of the modulator grating. Conclusions: Ostensible orientation tuning in fMRI activity arises from interactions with the edge of the stimulus aperture.
Meeting abstract presented at VSS 2018