Behavioral session results, face-matching task. (A) Accuracy. (B) RT. Results revealed an effect of group membership for both dependent variables [d′: F(2,42) = 17.39, *P < 0.001; RT: F(2,42) = 20.64, *P < 0.001]. Follow-up Tukey tests revealed the same pattern of results for both variables: both the WS and DA groups were outperformed by the CA group, with respect to both accuracy and RT (P < 0.001 for all t-test comparisons). The performances of the WS participants and DA controls, as expected, were not different (P > 0.2 for both d′ and RT).

Activation in ventral occipito-temporal regions and the amygdala. (A) Mean z-score images depicting active voxels (P < 0.05, corrected) in CA-matched controls (top), developmental controls (middle) and WS (bottom). Data are displayed in radiological convention (right is on left). Green circles denote significant activation in both control groups in the right amygdalar region, absent in the WS group. Green arrows denote activation in all three groups in the fusiform gyrus, a region in ventral occipito-temporal cortex, i.e. highly sensitive to faces. Except for one focus in a more superior aspect of the fusiform gyrus, where the WS group showed more robust activation than both control groups [Talairach coordinates (38, –69, –9); % signal change = 0.21 for WS, 0.079 for DA controls and 0.13 for CA controls), most ventral occipito-temporal regions showed comparable levels of activation between the WS and DA groups. (B) Mean activation within a spherical region-of-interest (radius = 4.5 mm) placed within the right amygdalar region.

Task. (A) Each block of six face identity-matching trials was preceded by a warning cue (green ‘smiley face’). For each trial, two face stimuli were presented in series, followed by a delay of either 500 or 1750 ms. Participants were then given 3500 ms to respond to the third face stimulus, indicating whether the identity matched either of those seen in the first two stimuli. A match was presented in half of the trials. Participants were encouraged to respond as quickly and accurately as possible. The first 18 trials used a 500 ms delay between the second reference stimulus and the test stimulus, in accordance with our previous face-processing study of WS (Paul et al., 2002). The second 18 trials used a 1750 ms delay, in line with the fMRI paradigm that has been successfully employed in the UCSD laboratory for several years (Passarotti et al., 2003; Stiles et al., 2006). Comparison of accuracy data from the two trial types across the three participant groups did not reveal a significant group × trial type interaction effect (P > 0.1). These data, therefore were collapsed over the two trial types in the present analyses. (B) Each 4 min, 50 s scanning run consisted of six 47.5 s blocks of face identity matching and control trials, in an interleaved fashion. Control trials followed the same presentation sequence as identity-matching trials. Control trial blocks were preceded by a distinct cue (an image of a black handprint, not shown), in order to warn participants as to which trial type would be shown in the upcoming task block.