Figure 2. Single-unit and single-voxel tuning to viewpoint in the macaque face patches. Top, Single-unit tuning to viewpoint (based on all neurons recorded from M1, M2, and M3). In the left panel, the empty bars represent the proportions of neurons with the highest response for each of the five viewpoints (all bars sum to 100%). The filled bars represent the proportions of neurons in each class that are significantly tuned; that is, for which the mutual information (between neural activity and viewpoint) is higher than the 97.5th percentile of the null distribution. The right panel shows tuning curves for cells tuned to each of the five viewpoints within each patch. Null distributions, 95% confidence intervals (top, dashed black lines correspond to empty bars and solid black lines correspond to filled bars; bottom, shaded area) and p-values (*p < 0.05, **p < 0.01, ***p < 0.001) are all derived from a permutation test (1000 repetitions). Bottom, Single-voxel tuning to viewpoint (based on all voxels recorded from M6, M7, and M8). Middle, right, The tuning factor for each patch is compared between single units and single neurons; p-values are from a paired t test.