We scanned alert monkeys while they passively viewed 20 s blocks

of Learned symbols, Untrained shapes (other human symbols differing in shape from the Learned symbols), and Faces, alternating with 20 s blocks of a small fixation spot (Figure 3). We first calculated maximum likelihood HSP assay maps of responsiveness to each stimulus category (Learned symbols, Untrained shapes, Faces) using general linear model methods (Boynton et al., 1996), wherein a hemodynamic impulse response function was convolved with the stimulus paradigm. We defined three category contrasts by performing t tests between responses to different pairs of stimulus categories: Learned symbols versus Faces (LvsF), Learned symbols versus Untrained shapes (LvsU), and Faces versus Untrained shapes (FvsU). Then we defined three category selectivity maps using a conjunction analysis ( Bell et al., 2009 and Price et al., 1997) on the three contrast conditions, using odd-numbered scans:

Face-selective voxels were defined as being more responsive to both F > U AND F > L, both contrasts p < 0.001 (corrected for multiple comparisons, see methods), Shape-selective regions satisfied both L > F AND U > F at p < 0.001, and Learned symbol-selective regions satisfied RO4929097 datasheet both L > U AND L > F at p < 0.001. The maps in Figure 4 and Figure 5 show these category-selective regions, projected onto semi-inflated anatomical maps for each monkey. In all six monkeys, several bilateral regions of the inferior temporal lobe were Carnitine palmitoyltransferase II more active to Faces than to either shape category (F > U AND F > L), consistent with previous reports of face selective regions in the temporal lobe (Tsao et al., 2003). These Face-selective regions showed >90% overlap between the left and right hemispheres for all six monkeys (see Table S1 available online); therefore, we averaged together the left and right Face-selective activations. We identified the three largest Face patches in each monkey as f1, f2, and f3 (posterior to anterior).

We projected the Face-selective patches from each individual monkey onto a common semi-inflated left hemisphere (Figure 6A, red patches); the patches were roughly overlapping in this projection, indicating some consistency in location from monkey to monkey, except for the most anterior Face region, which could comprise two patches or may simply be less reproducible in location from monkey to monkey. The location of the maximally selective voxels in each of the Face-selective patches in each monkey are given in Table S1. The most posterior Face patch (f1) was located in posterior area TEO, sometimes extending into anterior V4, on the ventral bank of the STS near the anterior tip of IOS, with the region of maximum overlap between monkeys at A1. The middle Face patch (f2) was mostly in area TEa with the region of maximum overlap at A8.