Abstract
Rationale. Holistic processing for faces is well known to be sensitive to rotation of the face in the image plane, occuring for upright but not inverted (upside-down). This has two possible origins: an innate template coding upright faces; or greater experience with upright faces than inverted faces (either across the lifespan or during a critical period in infancy). These theories are potentially distinguishable by testing the effect of depth rotation on holistic processing. As with image-plane rotation, some depth views are more common than others: front-on is common, and profile is rare. If differential experience were the origin of the inversion effect on holistic processing, holistic processing should be weaker for profile views of upright faces than front-on. Method. Five experiments examined holistic processing with two independent techniques: the aligned-unaligned composite paradigm and the peripheral inversion method. Four depth views were tested (front, 22.5°, three-quarter, profile), view was manipulated between- and within-subjects, experiments used two face stimulus sets, with and without hair. Results. There were no effects of view on holistic processing. Even for profiles, holistic processing (e.g., aligned-unaligned difference) was found for upright but not inverted faces and was no weaker than for front-on or 3/4 faces. There was a view-frequency effect on overall performance: this arose from part-based (not holistic) processing, as evidenced by view effects being at least as strong for faces with disrupted configurations (misaligned, inverted, scrambled) as for upright intact faces. Conclusion. Results are contrary to the predictions of the differential experience hypothesis. Taken together with other findings about faces and objects, results are consistent with a theory in which an innate special face mechanism codes the structural form of an upright face in all views and supports holistic processing, while a generic part-based view-sensitive system used for objects can also be activated by faces.
Funded by Australian Research Council DP0450636