Abstract
When complementary halves of different familiar faces are combined into a new face, there is interference in the identification of either half (Young et al., Perception, 1987). This “composite face effect” has been taken as strong evidence for the recognition of faces as single and immediate wholes - “holistic processing”. Here we ask if this effect persists when the two constituent parts appear at disjoint times. We presented human subjects with familiar faces briefly flashed in a dynamic stream of visual noise. Faces were shown both upright and inverted, and on some trials a variable interval (up to 160 ms) separated one part from the other. Subjects were instructed to respond according to the identity of one designated part only. Identification of chimeric faces was slower than identification of self-consistent faces; this difference was more robust with upright than with inverted stimuli. For simultaneous composites, this replicates the classical result: inversion disrupts our natural processing of faces as unified wholes and allows the irrelevant part to be disregarded. For sequentially presented parts, this demonstrates that the composite face effect persists across temporal discontinuities. In other words, our expertise at recognizing upright faces appears to involve processes that relate different parts of a face not only across space but also across time.
Burroughs Wellcome Fund, James S. McDonnell Fund