Abstract
Functional localization of visual category-selective brain regions (e.g. face-selective FFA, scene-selective PPA, object-selective LOC) with fMRI has suggested the existence of discrete, functionally distinct regions of cortex, easily identifiable with single five-minute scans. These regions are implicitly assumed to have sharp boundaries and increasing resolution could potentially increase the precision of these boundaries. We investigated the distribution and number of scene-, object-, and face-selective voxels throughout cortex within individual subjects across a range of resolutions (1.2, 1.6, 2 and 3 mm isotropic voxels) at both 3T and 7T. To maximize power we used multiple five-minute localizer runs (up to 16 per subject per resolution). If such regions are discrete, then the number of selective voxels identified should asymptote with increasing power. However, we found that the number of selective voxels increased almost linearly as a function of number of localizer runs, regardless of the categories used, for both 3T and 7T and for both smoothed and unsmoothed data. With high power, the face- and scene-selective voxels appear to form two parallel streams emerging from central and peripheral early visual cortex, respectively. At high resolution at 7T, where medial-temporal lobe susceptibility artifacts are minimized, these streams extend into the anterior temporal lobe where scene selectivity terminates in the rhinal cortices and face selectivity in more lateral aspects of the anterior temporal lobe. Within these streams there are peaks of selectivity that may co-localize with large draining vessels. Thus, the relative strength and topography of category selectivity across the streams may be driven in part by vascular anatomy and not just cortical specialization. These findings highlight limitations of using contrasts to investigate the topography of selectivity, emphasize the difficulty of interpreting null results in fMRI, and argue for richer descriptions of the type of processing across cortical regions.
Meeting abstract presented at VSS 2014