Using the eyes to encode and recognize social scenes

Abstract

In a previous study, we found that observers look mostly at the eyes when viewing natural scenes containing one or more people (Birmingham et al. submitted). This prioritization of eye regions occurred regardless of the type of scene being viewed (e.g. scenes with one person vs. scenes with several people, see Figure 1). The finding that observers attend preferentially to the eyes when freely viewing scenes suggests that they are the most informative regions of the scene. As a consequence, one might also expect that observers encode and/or recognize scenes through information from the eyes. This prediction is in line with the finding that when viewing object scenes in preparation for a later memory test, observers tend to fixate more informative objects more frequently than less informative objects (Henderson et al. 1999).