Participants viewed unique instances of 768 gray-scale photographic images in four blocks for each of the categories of faces (males and females of various ages, races, facial expressions and views), objects (abstract sculptures), places (indoor and outdoor), and textures (created by randomly scrambling object pictures into 225, 8 × 8-pixel squares).

(a) The FFA was defined in each participant as a cluster of contiguous face selective voxels with activation peaking in the mid-fusiform gyrus (faces > objects, P < 10⁻⁶, uncorrected). Blue lines point to the right FFA in coronal, sagittal and horizontal views from a representative TD participant. Color bar indicates T values.
(b) Same as (a) but from a representative WS participant.
(c) Bars show the volume of the FFA, defined as a contiguous cluster of activation peaking in the mFUG plotted against the minus logarithm (base 10) at three different thresholds (faces > objects, P < 10⁻³, P < 10⁻⁴, P < 10⁻⁶), using non-spatially smoothed data, or at P < 10⁻³, using spatially smoothed data (3s). At each threshold, FFA volume was averaged across 13 TDs (light gray), and 13 WS participants (black). Error bars show group SEM. Red bars, WS participants matched to the TD group for %Res in FUS. The volume of the rFFA was significantly larger among WS than TDs across all four comparisons (WS > TD, * P < 0.05, 2-tailed F-test, repeated measures ANOVA).
(d) For each participant, the percentage of mid-fusiform activation was calculated as the total volume of face-selective voxels (faces > objects), regardless of contiguity, within the anatomically defined left or right mid-fusiform gyrus, divided by the volume of that anatomical ROI. The percentage of mid-fusiform activation is plotted against the minus logarithm (base 10) at six statistical thresholds between 10⁻⁴ and 10⁻¹² (uncorrected) for all participants (TD in gray and WS in black), and for a subset of WS participants who were matched to TD for %Res in FUS (in red). Diamonds: TD, n = 13; black squares: WS (n = 13); red squares: WS matched to TD participants for %Res in FUS (n = 8). Error bars show group SEM. The proportional volume of face-selective activations relative to the volume of the FUS was significantly higher in WS than in TD across the various thresholds tested in both (WS > TD, * P < 0.05, 2-tailed F test, repeated measures ANOVA).
(e) FFA’s (faces > objects, P < 10⁻⁴, uncorrected) responses to faces, objects, places and textures based on beta coefficients derived from the GLM for each image category relative to baseline. Bars represent each group’s data as in (c). Amplitude of responses to visual stimuli were not significantly different among groups and there were no significant (NS) group by category interactions. Note that the response magnitudes are not independent of the data used to select the FFA. Error bars show group SEM.

(a) Anatomical boundaries of the mid-fusiform gyrus (FUS) were drawn for each participant in the right and left hemisphere to include gray and white matter. In red are the anatomical FUS ROIs in coronal, sagittal and horizontal views from a representative TD participant
(b) Same as (a) but from a representative WS participant.
(c) The volume of the gray matter within the anatomically defined FUS. Gray: TD adults n = 13; black: WS adults n = 13. Error bars show standard error of mean for each group (SEM, TD > WS; † P < 0.001, †† P < 0.0001, t-test).
(d) From the anatomically defined FUS ROIs, we calculated response amplitudes to faces, objects, places and textures, based on beta coefficients derived from the GLM for each image category relative to baseline. Gray and black gars represent each group’s data as in c. Red bars show data from a subset of 8 WS participants, matched to the TD group for %Res (which reflects BOLD related noise and goodness of GLM fit) in FUS. The overall response amplitudes in the WS and TD groups were not statistically different. However, in the right FUS there was a significant group by category interaction (P < 0.002, F test, ANOVA), as response to faces was higher in WS than in TDs († P < 0.05 one-tailed t-test). Note that the response magnitudes are independent of the anatomical data used for delineating the ROI.
(e) Face selectivity was calculated based on the differential responses to faces versus one of the other stimulus categories (i.e object, place or texture) as indicated along the x-axis (e.g. (face – object)/%Res, see methods). Positive values along the y-axis indicate preference for faces, and negative values indicate preference for objects or places or textures, as indicated by pictures. Bars represent TD and WS groups as in (d). Error bars show SEM. Face selectivity was significantly higher among WS compared to TDs in the right hemisphere (* P < 0.015, F-test repeated measures ANOVA).

(a) The FUS_obj was defined in each participant as a cluster of contiguous object selective voxels with activation peaking in the mid-fusiform gyrus (objects > textures, P < 10⁻⁶, uncorrected). Blue lines point to the right FUS_obj in coronal, sagittal and horizontal views from a representative TD participant. Color bar indicates T values.
(b) Same as (a) but from a representative WS participant.
(c) Bars show the volume of the FUS_obj, defined as a contiguous cluster of object-selective activation (objects > textures) peaking in the mFUG plotted as in Fig 2c. Error bars show group SEM. There were no significant (NS) between-group differences.
(d) For each participant, the percentage of object-selective activation (objects > textures) in the mid-fusiform was calculated and plotted as in Fig 2d. Error bars show group SEM.
(e) Responses from the FUS_obj (objects > Textures, P < 10⁻⁴, uncorrected) to visual stimuli are plotted as in Fig 2e. In the right hemisphere, there was a significant group by category interaction (P < 0.04, F-test ANOVA), due to significantly higher response magnitudes to faces in WS than in TD (†P < 0.04 one-tailed t-test). Note that the response magnitudes are not independent of the data used to select FUS_obj. Error bars show group SEM.

(a) The pSTS_face was defined in each participant as a cluster of contiguous face selective voxels with activation peaking in the pSTS (faces > objects, P < 10⁻⁶, uncorrected). Blue lines point to the right pSTS_face in coronal, sagittal and horizontal views from a representative TD participant. Color bar shows T values.
(b) Same as (a) but from a representative WS participant.
(c) Bars show the volume of the pSTS_face defined as a contiguous cluster of activation (faces > objects) peaking in the pSTS plotted as in Fig 2c. Error bars show group SEM. The volume of pSTS_face was significantly smaller in WS than in TD controls only for the spatially smoothed data defined at the threshold of P < 10⁻³ noted as 3s (†P < 0.05 one-tailed t-test, not corrected for repeated comparisons).
(d) For each participant, the percentage of face-selective activation (faces > objects) for the anatomically defined left or right pSTS was calculated and plotted as in Fig 2d. There were no significant (NS) between-group differences.
(e) Responses from the pSTS_face (faces > objects, P < 10⁻⁴, uncorrected) to visual stimuli are plotted as in Fig 2e. Red bars, WS matched to TD participants for %Res in pSTS. Response amplitudes to visual stimuli were not significantly different among groups and there were no significant (NS) group by category interactions. Note that the response magnitudes are not independent of the data used to select the pSTS_face. Error bars show group SEM.

(a) The AMG_face was defined in each participant as a cluster of contiguous face selective voxels within the anatomical boundaries of the amygdala (faces > objects, P < 10⁻⁶, uncorrected). Blue lines point to the right AMG_face in coronal, sagittal and horizontal views from a representative TD participant. Color bar shows T values.
(b) Same as (a) but from a representative WS participant.
(c) Bars show the volume of the AMG_face defined as a contiguous cluster of activation (faces > objects) peaking in the amygdala plotted as in Fig 2c. There were no significant (NS) between-group differences. Error bars show group SEM.
(d) For each participant, the percentage of face-selective activation (faces > objects) for the anatomically defined left or right amygdala was calculated and plotted as in Fig 2d.
(e) Responses from the AMG_face (faces > objects, P < 10⁻⁴, uncorrected) to visual stimuli are plotted as in Fig 2e. Red bars, WS matched to TD participants for %Res in amygdala. Error bars show group SEM. In the right hemisphere, response amplitudes to faces were significantly lower among WS than TD controls († P < 0.03 one tailed t-test). Note that the response magnitudes are not independent of the data used to select the AMG_face.

(a) Performance on the Benton recognition of upright faces is plotted as percent correct responses against the right FFA volume (faces > objects, P < 10⁻⁴, uncorrected) for the subset of TD participants who took the test (n = 9). The correlation between performance on the Benton test and right FFA volume did not reach statistical significance.
(b) Performance of WS on the Benton test is plotted against the volume of the rFFA (faces > objects, P < 10⁻⁴, uncorrected) for all participants with Benton scores (squares, double solid line), the subset of WS that were statistically not different from TDs on %Res (after removing the participants with the highest %Res in right FUS (black squares) and the subset of WS who were matched on %Res in right FUS with TD (open red squares, dashed line). There was a significant correlation between performance on the Benton test and right FFA volume among the later two subsets of WS participants.