5. Category-selective regions show graded activation profiles for images of their preferred category. A, If some images consistently activated a region more strongly than other images of the same category (i.e., graded within-category activation profile), the within-category ranking order should replicate across sessions. We computed the replicability of within-category ranking by selecting the samecategorysubsetofimagesinbothsessionsandcorrelatingtheiractivationestimatesusingSpearman’sr.order AMG9810 Thisprocedureisillustrated forthewithin-faceactivationprofileinrightFFAdefinedat128voxelsinonespecificsubject.ColorcodingisthesameasinFigure1.B,Group analysisofreplicabilityofwithin-categoryactivationprofilesforcategory-selectiveregionsFFAandPPAandforcontrolregionshITandEVC. Analysis was performed for the image subsets of faces (top) and places (bottom), either using the concatenation approach (left) or the averaging approach (right) for combining single-subject data. Analysis of GLPG0187 chemical information concatenated single-subject activation profiles is sensitive to replicablerankingregardlessofdifferencesinparticularrankingorderamongsubjects,whileanalysisofsubject-averageactivationprofiles is sensitive to replicable ranking that is consistent among subjects. We performed a standard one-sided test on Spearman’s r to determine whether replicability of within-category activation profiles was significantly higher than expected by chance (H0: r 0). p values were corrected for multiple comparisons as described in Figure 1. Black boxes highlight the ROI sizes that results were displayed at in Figures 1 (FFA and PPA) and 2 (hIT and EVC).Mur et al. ?Single-Image Activation of Category RegionsJ. Neurosci., June 20, 2012 ?32(25):8649 ?8662 ?tive regions in IT cortex are categorical, yet graded.DiscussionFFA and PPA might respond more strongly to every single member of their preferred category than to any nonmember We measured single-image activation of human category-selective regions to 96 object images from a wide range of categories, and investigated whether category selectivity holds in general or is violated by particular single images. We found good discrimination of preferred from nonpreferred stimuli based on single-image activation of category-selective regions FFA and PPA across a wide range of ROI sizes. Furthermore, we did not find evidence for violations of category-consistent ranking by particular single images, except in left FFA. Together, these findings suggest the possibility that right FFA and left and right PPA respond more strongly to every single member of their preferred category than to any nonmember. This conclusion is consistent with several single-image studies in monkeys that showed strong face-selectivity in the macaque middle and anterior superior temporal sulcus (STS) (Foldiak et al., 2004; ??Tsao et al., 2006). These studies reported cells that responded almost exclusively to faces. It should be noted that many of the recorded cells in the middle macaque face patch, a suggested homolog of FFA located in the STS (Tsao et al., 2003, 2006), also responded significantly to several nonface images (Tsao et al., 2006). These nonface images shared lower-level visual properties with face images (e.g., round shape). However, at the population level (i.e., when responses were averaged across the population of visually responsive cells in the middle face patch), the responses elicited by these nonface images were4 Figure 6. Category steps and graded activation profil.5. Category-selective regions show graded activation profiles for images of their preferred category. A, If some images consistently activated a region more strongly than other images of the same category (i.e., graded within-category activation profile), the within-category ranking order should replicate across sessions. We computed the replicability of within-category ranking by selecting the samecategorysubsetofimagesinbothsessionsandcorrelatingtheiractivationestimatesusingSpearman’sr.Thisprocedureisillustrated forthewithin-faceactivationprofileinrightFFAdefinedat128voxelsinonespecificsubject.ColorcodingisthesameasinFigure1.B,Group analysisofreplicabilityofwithin-categoryactivationprofilesforcategory-selectiveregionsFFAandPPAandforcontrolregionshITandEVC. Analysis was performed for the image subsets of faces (top) and places (bottom), either using the concatenation approach (left) or the averaging approach (right) for combining single-subject data. Analysis of concatenated single-subject activation profiles is sensitive to replicablerankingregardlessofdifferencesinparticularrankingorderamongsubjects,whileanalysisofsubject-averageactivationprofiles is sensitive to replicable ranking that is consistent among subjects. We performed a standard one-sided test on Spearman’s r to determine whether replicability of within-category activation profiles was significantly higher than expected by chance (H0: r 0). p values were corrected for multiple comparisons as described in Figure 1. Black boxes highlight the ROI sizes that results were displayed at in Figures 1 (FFA and PPA) and 2 (hIT and EVC).Mur et al. ?Single-Image Activation of Category RegionsJ. Neurosci., June 20, 2012 ?32(25):8649 ?8662 ?tive regions in IT cortex are categorical, yet graded.DiscussionFFA and PPA might respond more strongly to every single member of their preferred category than to any nonmember We measured single-image activation of human category-selective regions to 96 object images from a wide range of categories, and investigated whether category selectivity holds in general or is violated by particular single images. We found good discrimination of preferred from nonpreferred stimuli based on single-image activation of category-selective regions FFA and PPA across a wide range of ROI sizes. Furthermore, we did not find evidence for violations of category-consistent ranking by particular single images, except in left FFA. Together, these findings suggest the possibility that right FFA and left and right PPA respond more strongly to every single member of their preferred category than to any nonmember. This conclusion is consistent with several single-image studies in monkeys that showed strong face-selectivity in the macaque middle and anterior superior temporal sulcus (STS) (Foldiak et al., 2004; ??Tsao et al., 2006). These studies reported cells that responded almost exclusively to faces. It should be noted that many of the recorded cells in the middle macaque face patch, a suggested homolog of FFA located in the STS (Tsao et al., 2003, 2006), also responded significantly to several nonface images (Tsao et al., 2006). These nonface images shared lower-level visual properties with face images (e.g., round shape). However, at the population level (i.e., when responses were averaged across the population of visually responsive cells in the middle face patch), the responses elicited by these nonface images were4 Figure 6. Category steps and graded activation profil.