Why are face composites difficult to recognize?

Abstract

The composite task is widely used in face processing to test issues of holistic processing. In this task, composite faces are formed by taking the top half of one face and pairing it with the bottom half of another. A large body of data has demonstrated that judgments of one half are affected by the identity of the irrelevant half, particularly when the two halves are aligned into the configuration of a normal face, and this interference is thought to be due to mandatory holistic processing of a face. However, very little work has been addressed at understanding the nature of the interference observed in the composite task. We examined this issue by creating composites from faces that were identical except for the spacing of the eyes. In the experiment, Chinese participants viewed Chinese and Caucasian composites (both halves of each face were always the same race), and had to judge whether the top halves of two successive composites were identical or different. The complete design of the composite task was used, so that the composite effect is demonstrated by the top/bottom congruency x alignment interaction. Top (target) halves were always identical except (on half the trials) for the spacing of the eyes; bottom halves were either identical or from different faces. The results showed a significant three-way interaction between congruency, alignment, and orientation, which demonstrated a composite effect but only for aligned, upright faces. This effect was further modified by the significant four-way interaction between congruency, alignment, orientation, and race as our Chinese participants showed a stronger congruency effect for Chinese than Caucasian faces, but again only when they were aligned and upright. These results show that in the composite task, changes in irrelevant face information interfere with the processing of configural relationships within the target region.