. This exploratory analysis can provide further information on functional similarities between regions, and, more specifically, on the extent to which activation profiles of category-selective Biotin-VAD-FMK price regions are inherited from EVC. As for the replicability of within-category ranking (Fig. 5), we combined data PP58 msds across subjects either by concatenating or averaging the activation profiles across subjects. The concatenation approach is sensitive to inter-region correlations of activation profiles even if the particular activation profiles differ across subjects. The averaging approach is sensitive to inter-region correlations of activation profiles that are consistent across subjects. We investigated the inter-region correlations for (1) the full activation profile, (2) the within-face activation profile, and (3) the withinplace activation profile. Statistical inference was performed by a standard one-sided test on Spearman’s r. p values were corrected8658 ?J. Neurosci., June 20, 2012 ?32(25):8649 ?Mur et al. ?Single-Image Activation of Category Regionsfor multiple testing using Bonferroni correction based on the total number of tests performed. Figure 7 shows the inter-region correlation results. The main pattern that emerges is that activation profiles are correlated between hemispheres for corresponding regions, and between hIT and EVC (red blocks on diagonals). We subsequently inspected the within-face correlation between FFA and EVC, and the within-place correlation between PPA and EVC. The within-face activation profile was correlated between left but not right FFA and EVC, and the within-place activation profile was not significantly correlated between PPA and EVC. Results were similar across ROI sizes. These results suggest that EVC is not a major contributor to the within-category activation profiles of PPA and right FFA. We then inspected the correlation between category-selective regions (FFA/PPA) and EVC for the full activation profile (top row). The full activation profile was correlated between EVC and both categoryselective regions, especially PPA. One interpretation of this finding would be that some degree of category selectivity is already present at the level of EVC, implying that low-level feature differences contribute to some extent to categoryselective responses. For places, this seems a plausible interpretation, consistent with our finding that single-image activation of EVC can discriminate places from nonplaces at an above-chance level (Fig. 2). For faces, this interpretation seems less likely: the correlation between EVC and FFA is not significant for the subjectaverage activation profile, suggesting that the correlation is driven by subjectspecific effects (e.g., idiosyncratic arousal effects) and not by face-selectivity of responses (shared across subjects in FFA). Categorical, yet graded Figure 8 summarizes our results. Singleimage activation profiles of categoryselective regions (1) show near-perfect discrimination of preferred from nonpreferred images and no preference inversions for particular object images, (2) show a step-like drop-off at the category boundary, and (3) are graded within and outside the preferred category. It can further be noted that single-image category selectivity is stronger in right than left FFA. In addition, gradedness seems to be more pronounced in FFA; the category step seems to be more pronounced in PPA. In sum, our findings indicate that the activation profiles of category-selec-Figure.. This exploratory analysis can provide further information on functional similarities between regions, and, more specifically, on the extent to which activation profiles of category-selective regions are inherited from EVC. As for the replicability of within-category ranking (Fig. 5), we combined data across subjects either by concatenating or averaging the activation profiles across subjects. The concatenation approach is sensitive to inter-region correlations of activation profiles even if the particular activation profiles differ across subjects. The averaging approach is sensitive to inter-region correlations of activation profiles that are consistent across subjects. We investigated the inter-region correlations for (1) the full activation profile, (2) the within-face activation profile, and (3) the withinplace activation profile. Statistical inference was performed by a standard one-sided test on Spearman’s r. p values were corrected8658 ?J. Neurosci., June 20, 2012 ?32(25):8649 ?Mur et al. ?Single-Image Activation of Category Regionsfor multiple testing using Bonferroni correction based on the total number of tests performed. Figure 7 shows the inter-region correlation results. The main pattern that emerges is that activation profiles are correlated between hemispheres for corresponding regions, and between hIT and EVC (red blocks on diagonals). We subsequently inspected the within-face correlation between FFA and EVC, and the within-place correlation between PPA and EVC. The within-face activation profile was correlated between left but not right FFA and EVC, and the within-place activation profile was not significantly correlated between PPA and EVC. Results were similar across ROI sizes. These results suggest that EVC is not a major contributor to the within-category activation profiles of PPA and right FFA. We then inspected the correlation between category-selective regions (FFA/PPA) and EVC for the full activation profile (top row). The full activation profile was correlated between EVC and both categoryselective regions, especially PPA. One interpretation of this finding would be that some degree of category selectivity is already present at the level of EVC, implying that low-level feature differences contribute to some extent to categoryselective responses. For places, this seems a plausible interpretation, consistent with our finding that single-image activation of EVC can discriminate places from nonplaces at an above-chance level (Fig. 2). For faces, this interpretation seems less likely: the correlation between EVC and FFA is not significant for the subjectaverage activation profile, suggesting that the correlation is driven by subjectspecific effects (e.g., idiosyncratic arousal effects) and not by face-selectivity of responses (shared across subjects in FFA). Categorical, yet graded Figure 8 summarizes our results. Singleimage activation profiles of categoryselective regions (1) show near-perfect discrimination of preferred from nonpreferred images and no preference inversions for particular object images, (2) show a step-like drop-off at the category boundary, and (3) are graded within and outside the preferred category. It can further be noted that single-image category selectivity is stronger in right than left FFA. In addition, gradedness seems to be more pronounced in FFA; the category step seems to be more pronounced in PPA. In sum, our findings indicate that the activation profiles of category-selec-Figure.
