Izes for PPA. ROI size did not have a noticeROI sizes able effect on largest-gap inverted-pairs results for either FFA We performed our analyses of category discriminability and prefor PPA (Fig. 4 B). erence inversions for five different ROI sizes, ranging from 10 to In sum, these findings indicate that the strong single-image 300 BMS-5 web voxels for unilateral FFA and PPA and from 20 to 600 voxels preference for faces over nonfaces in FFA and places over nonfor bilateral hIT and EVC. Testing across multiple ROI sizes enplaces in PPA can be found at all ROI sizes. Nevertheless, ROI size ables assessment of the robustness of our effects against changes does affect measured category selectivity. Strongest category sein ROI size. Figure 1 shows discrimination performance (AUC) lectivity is found at smaller ROI sizes for FFA and at larger ROI for an intermediate ROI size (128 voxels) chosen to approxisizes for left PPA. The clear decrease in discrimination performately match the previously reported volume of right FFA (Kanmance for left FFA with increasing ROI size might simply reflect wisher et al., 1997). Discrimination performance for the other the previously reported finding that left FFA contains fewer ROI sizes can be found in Table 1. strongly face-selective voxels than right FFA (Kanwisher et al., The top panel of Table 1 shows very good discrimination of 1997). faces from nonfaces based on single-image activation of both left and right FFA at all ROI sizes. Discrimination performance was best for the smallest ROI size (10 most face-selective voxels) and Within-category activation profiles are graded decreased with increasing ROI size. The effect of ROI size was Figure 1 suggests that, within the preferred category, some images more pronounced for left than right FFA, resulting in a considactivated category-selective regions more strongly than others. erable difference in performance between left and right FFA for We tested this hypothesis by examining the replicability ofMur et al. ?Single-Image Activation of Category RegionsJ. Neurosci., June 20, 2012 ?32(25):8649 ?8662 ?consistent across subjects. In addition, effects in EVC were small and were significant at one ROI size only. Right, but not left, PPA showed replicable ranking for places at most ROI sizes (Fig. 5B, bottom). This indicates that some places consistently activated right PPA more strongly than others. The significant results for the group-average activation profiles suggest that within-place activation profiles were similar across the four subjects. This Pedalitin permethyl ether site conclusion was supported by visual inspection of singlesubject within-place activation profiles and by intersubject correlation analyses. Right FFA and control regions hIT and EVC showed replicable within-place ranking as well. Right FFA showed effects at smaller ROI sizes and hIT at larger ROI sizes. Effects in hIT and EVC were present for the group-average activation profiles, and effects in right FFA were only present for concatenated single-subject activation profiles. These findings suggest that withinplace activation profiles were similar across the four subjects for hIT and EVC, but not for right FFA. This conclusion was supported by visual inspection of single-subject within-place activation profiles and by intersubject correlation analyses. Figure 3. Most inverted pairs do not replicate. How prevalent are true inversions of category preference in FFA and PPA? We These findings confirm that categoryinvestigated this.Izes for PPA. ROI size did not have a noticeROI sizes able effect on largest-gap inverted-pairs results for either FFA We performed our analyses of category discriminability and prefor PPA (Fig. 4 B). erence inversions for five different ROI sizes, ranging from 10 to In sum, these findings indicate that the strong single-image 300 voxels for unilateral FFA and PPA and from 20 to 600 voxels preference for faces over nonfaces in FFA and places over nonfor bilateral hIT and EVC. Testing across multiple ROI sizes enplaces in PPA can be found at all ROI sizes. Nevertheless, ROI size ables assessment of the robustness of our effects against changes does affect measured category selectivity. Strongest category sein ROI size. Figure 1 shows discrimination performance (AUC) lectivity is found at smaller ROI sizes for FFA and at larger ROI for an intermediate ROI size (128 voxels) chosen to approxisizes for left PPA. The clear decrease in discrimination performately match the previously reported volume of right FFA (Kanmance for left FFA with increasing ROI size might simply reflect wisher et al., 1997). Discrimination performance for the other the previously reported finding that left FFA contains fewer ROI sizes can be found in Table 1. strongly face-selective voxels than right FFA (Kanwisher et al., The top panel of Table 1 shows very good discrimination of 1997). faces from nonfaces based on single-image activation of both left and right FFA at all ROI sizes. Discrimination performance was best for the smallest ROI size (10 most face-selective voxels) and Within-category activation profiles are graded decreased with increasing ROI size. The effect of ROI size was Figure 1 suggests that, within the preferred category, some images more pronounced for left than right FFA, resulting in a considactivated category-selective regions more strongly than others. erable difference in performance between left and right FFA for We tested this hypothesis by examining the replicability ofMur et al. ?Single-Image Activation of Category RegionsJ. Neurosci., June 20, 2012 ?32(25):8649 ?8662 ?consistent across subjects. In addition, effects in EVC were small and were significant at one ROI size only. Right, but not left, PPA showed replicable ranking for places at most ROI sizes (Fig. 5B, bottom). This indicates that some places consistently activated right PPA more strongly than others. The significant results for the group-average activation profiles suggest that within-place activation profiles were similar across the four subjects. This conclusion was supported by visual inspection of singlesubject within-place activation profiles and by intersubject correlation analyses. Right FFA and control regions hIT and EVC showed replicable within-place ranking as well. Right FFA showed effects at smaller ROI sizes and hIT at larger ROI sizes. Effects in hIT and EVC were present for the group-average activation profiles, and effects in right FFA were only present for concatenated single-subject activation profiles. These findings suggest that withinplace activation profiles were similar across the four subjects for hIT and EVC, but not for right FFA. This conclusion was supported by visual inspection of single-subject within-place activation profiles and by intersubject correlation analyses. Figure 3. Most inverted pairs do not replicate. How prevalent are true inversions of category preference in FFA and PPA? We These findings confirm that categoryinvestigated this.