Abstract
A previous fMRI study (Curby & Gauthier, 2002) found that RD, a patient with an Autistic spectrum disorder (ASD), lacks a fusiform face area (FFA) for unfiltered images, consistent with previous ASD research (Schultz et al., 2000), but surprisingly has face-selective activity in the fusiform gyrus for HSF (high spatial frequency) images (Curby & Gauthier, 2002). Using Gosselin and Schyns' (2001) Bubbles technique, we measured RD's relative use of information from different spatial scales in a face identification task. Bubbles randomly reveals limited information from different locations and SF bandwidths, allowing for the computation of an effective face stimulus, based on RD's performance, depicting the information used for identification. Compared to 15 typical control participants from Schyns et al. (2002), RD not only uses much less information to identify the faces but also relies on a unique distribution of information from the different scales. Controls used relatively more information from mid-to-coarse scales (between 11 to 22 cycles per face) whereas RD appears to rely mostly on the fine scales (45 to 90 cycles per face). Using controls as a benchmark of information use, RD appeared to be more efficient with fine scale information (45 to 90 cycles per face). Efficiency progressively decreased with lower resolutions. RD's effective face stimulus appeared to be atypical with a dominance of the right eye compared to controls in which both eyes were well defined. The nose was also poorly defined suggesting this feature plays little role in identification for RD. The results combined with RD's fMRI activity pattern are consistent with recent work suggesting that the FFA operates on inputs that are segregated for different spatial scales (Gauthier et al., 2002; Eger et al., 2002). Further studies are required to establish if this bias for HSF information is characteristic of ASD as this may provide a new perspective on visual processing in this population.