Abstract
Many researchers assume that face recognition is driven by configural processing. This assumption, however, is rarely tested in a direct way. Here we examine whether an observer's level of configural processing can predict his or her face recognition performance.
Last year we reported data demonstrating a lack of correlation between the magnitude of the composite face effect (CFE) and accuracy on a face identification (FI) task (VSS Abstract #601, 2007). The CFE task required subjects to make same-different judgments about top halves of faces while bottom halves were either aligned or misaligned (blocked). The FI task was to determine if the target was present in the lineup and, if so, to select the face that matched the target. The idea behind the study was that if CFE taps into a process that helps with face recognition in naturalistic contexts, then there should be a correlation between the magnitudes of CFEs and accuracies on face identification tasks.
Here we report two follow-up experiments that compared CFEs and accuracies on two different FI tasks. The CFE task was administered three times per subject. One FI task was a 10-AFC task with unlimited viewing time of 1 target and 9 distractors; the second FI task was a 4-AFC task that limited the viewing time of the target face to 200 msec, which matched the stimulus duration used in the CFE task.
There were large but stable individual differences on the CFE task and on both FI tasks. The correlations between the magnitude of the CFE (accuracy or RT) and accuracies in the two FI tasks were not significant (r≤0.25, p≥0.5). These results indicate that the mechanisms that produce the composite face effect may not facilitate accurate face identification, thus the common assumption that face recognition is driven by configural processing should be reassessed.
This research was supported by NSERC grants 42133 and 105494, and the Canada Research Chair programme.