A whole face is more than the sum of its halves: Interactive processing in face perception

Abstract

Facial information is processed interactively. Yet, such interactive processing has been examined for discrimination of face parts rather than complete faces. Here we assess interactive processing using a novel paradigm in which subjects discriminate complete faces. Face stimuli, which comprise unilateral facial information (hemifaces) or bilateral facial information from one face (consistent) or two different faces (inconsistent), are shown centrally in a face‐matching task. If each half of a complete face is processed independently, accuracy for complete faces can be predicted by the union of accuracies for right and left hemifaces. However, accuracy exceeded this independence prediction for consistent faces (facilitation) and fell below the prediction for inconsistent faces (interference). These effects were reduced or absent for inverted faces. Our findings are consistent with reports of stronger interactive processing for upright than for inverted faces and they quantify effects of interactive processing on the discrimination of complete faces.

Notes

Please address all correspondence to: Galit Yovel, Department of Brain and Cognitive Sciences, Massachusetts Institute of Technology, 77 Massachusetts Avenue, NE20‐443, Cambridge, MA 02139, USA. Email: [email protected]

We thank Satoru Suzuki and Nancy Kanwisher for their comments and Scott Grabarski for his help with data collection. This research was funded by a grant from the Brain Research Foundation at the University of Chicago to Jerre Levy and an NINDS grant NS34639 to Ken Paller.

There was a higher proportion of correct identification of the left side (23.8%) than the right side (16.9%) of inconsistent faces, t(23) = 1.91, p = .07, in Experiment 1, and for left side (28.8%) than the right side (18.8%) for upright, t(23) = 3.09, p < .01, but not inverted faces in Experiment 2.

Performance for left hemifaces was higher than for right hemifaces, t(23) = 2.81, p < .01.

Performance for left hemifaces and right hemifaces did not differ for either upright, t(23) = 1.54, p = .13, or inverted faced, t(23) < 1.