Abstract
Although much is known about the orientation-selective properties of visual cortex in animals, such selectivity has proven difficult to measure non-invasively in humans. Previously, we have shown that ensemble fMRI activity in the human visual cortex allows for accurate neural decoding of seen and attended orientations (Kamitani & Tong, VSS, 2004, Nature Neuroscience, 2005). Our method capitalizes on the fact that random variations in the spatial distribution of orientation-selective neurons or columns could lead to weak local biases, which remain detectable at coarse spatial resolutions. By pooling the information from many coarse-scaled voxels, we can obtain robust measures of ensemble orientation selectivity. Here, we investigated whether visual areas differ in ensemble orientation selectivity across variations in stimulus contrast and spatial frequency. In the first experiment, bilateral sine-wave gratings of ±45 degrees were presented at 4%, 20%, or 100% contrast. Whereas orientation selectivity was somewhat greater in V1 than in higher extrastriate areas at high stimulus contrasts, this pattern reversed at low contrasts. Thus, orientation-selective responses became more contrast invariant at higher levels of the visual hierarchy. In the second experiment, oriented gratings were presented at varying spatial frequencies of 0.25, 1.0 and 4.0 cycles/degree. Although areas V1–V3 showed robust orientation selectivity at low and moderate spatial frequencies, only V1 continued to show strong orientation selectivity at high spatial frequencies. Taken together, it appears that human V1 shows greater spatial sensitivity but also greater contrast dependence than higher areas, consistent with neurophysiological studies in animals. Results reveal that our measures of ensemble orientation selectivity are sufficiently sensitive to detect key functional differences between visual areas. Since all stimuli were well above threshold, variations in orientation selectivity cannot be explained by whether the observer perceived a global oriented pattern. Instead, these ensemble responses seem to depend primarily on local orientation processing.
Research support: NEI R01-EY14202 to FT and NSERC graduate fellowship to SH