Cheryl Olman¹; ¹University of Minnesota

We collected GE fMRI data with 0.6 mm resolution with the hope of being able to characterize the orientation preference of individual voxels in primary visual cortex (V1) as a function of cortical depth (columns) and location across the cortical surface (pinwheels). Stimuli were large 2 cycle per degree gratings, briefly presented in a 2-interval forced-choice paradigm as observers performed a foveal orientation discrimination task on the portion of the grating that was in a 0.5 degree diameter circle at the center of the grating. The experiment used a block design, presenting 4 12-sec blocks of gratings at each of 8 different (average) orientations in each of 4 task scans during a scanning session. High-quality datasets were acquired from 13 of the 16 individuals scanned. Regression analysis estimated the amplitude of response for each voxel in V1 for each of the 8 different orientations. Visual responses in parafoveal regions of interest (ROIs) defined by an independent localizer were robust: 76% +/- 10% s.d. of the voxels in ROIs defined on the surface and then propagated through the cortical depth were modulated by visual stimuli at the p < 0.001 (uncorrected) level. However, orientation preference of individual voxels was not reliable (e.g., not the same in the first and second half of the scanning session in a split-halves analysis). Average voxel orientation selectivity, computed as 1 minus the circular variance of responses to the 8 orientations, was not significantly different from the null hypothesis. Interestingly, computed orientation selectivity did vary through depth, but only because the inherent contrast-to-noise ratio of fMRI data varies through depth. Also, in spite of the lack of reliable orientation preference or selectivity in individual voxels, a reliable radial bias was evident in the data.

Acknowledgements: NIH R01 NS123482, NIH R01 MH111447