Marks, Tim
Coauthors(s): Debra L. Mills (UCSD), Marissa Westerfield (UCSD), Scott Makeig (Salk Institute), Tzyy-Ping Jung (Salk Institute), Ursula Bellugi (Salk Institute), Terrence J. Sejnowski (Salk Institute)
UCSD
Cognitive Science
Mail Code 0515 9500 Gilman Drive La Jolla, CA 92093

Face Processing in Williams Syndrome: Using ICA to Discriminate Functionally Distinct Independent Components of ERPs in Face Recognition

Event-related potentials (ERPs) are electrical potentials on the scalp time-locked to particular events. Traditional ERP analysis is based on the amplitude and latency of peaks in averaged response waveforms at individual scalp electrodes. If a response is composed of two or more spatially fixed components that overlap in time, peak-based methods may be unable to resolve the response into its component parts. Independent Component Analysis (ICA), a new approach to linear decomposition (Bell and Sejnowski, 1995), can overcome this limitation (Makeig et al., 1996, 1999). In an ERP study (Mills, 1998; Mills et al., in press), subjects were shown sequentially-presented photographic pairs of upright or inverted faces and were asked to indicate whether the second (target) face did or did not match the first face. Matching faces were non-identical photographs of the same face. ERPs were recorded while the task was performed by normal adults and adults with Williams Syndrome (WMS), a genetically-based disorder involving a deletion on chromosome 7. Face processing in WMS adults is of particular interest because despite their impaired performance in many cognitive domains including spatial cognition, WMS adults perform in the normal range on face processing tasks (Bellugi et al., 1999). Previously, Mills et al. found characteristic differences between the ERPs of the WMS group and those of normals and other groups. Peak-based analyses of these ERP data identified a single late positive component, called P500, encompassing the brainwave activity in both subject groups from 400 ms through 800 ms after stimulus onset. In normal adults but not in WMS adults, the P500 amplitude was larger in response to mismatched targets than to matched targets. By applying ICA to the same ERP data, we show that the P500 in this task is not unitary, but rather is composed of at least three spatially fixed, functionally distinct independent components. In both the control and WMS groups, a sustained positive component, SP, had relatively constant amplitude throughout the late wave. A second component, P7um, was larger during presentations of upright matched target faces. A third component, P5t, was active in all target conditions in the control group and was largest in response to targets in mismatched pairs. In the WMS group, a P5t analog was active only in response to mismatched targets. Comparing the identified late components from the WMS group with those from the normal control group may shed new light on the similarities and differences between the brain mechanisms underlying face processing in normal adults and in WMS adults. Components P7um and P5t may account for the P500 match/mismatch effect found in normal adults in previous studies, and appear to extend the effect to WMS adults.