Abstract
Selective gaze-contingent masking of the fixated feature in face recognition increases the processing difficulty for inverted, compared to upright faces. Preventing the use of peripheral information by revealing only the fixated feature through a gaze-contingent window, however, results in a decreased inversion effect (Van Belle et al., JOV, 2010). This suggests that the inversion effect is caused by problems in simultaneously perceiving multiple facial features present outside of the fixated feature, for inverted faces. However, gaze contingent masking or windowing explicitly forces human observers to change their processing strategy. Here we aimed at directly observing the differential use of fixated (central) and peripheral information in upright/inverted faces, without manipulating the processing strategy. Therefore, we conducted a matching task in which each trial consisted of three simultaneously presented faces in full view (Figure 1). Participants’ task was to indicate which of two faces presented side by side on the lower half of the screen resembled most the reference face in the top part of the screen. The reference face, however, consisted of a combination of both bottom faces in a gaze contingent way. The part of the reference face in the center of gaze equaled the corresponding part of one of both alternative faces. The remaining part of the reference face equaled the remaining part of the other answering alternative. The reference face was therefore updated upon each gaze position shift. Although the overall preference for choosing the central or peripheral face differed considerably between participants, the proportion of choices for the answering alternative corresponding to the centrally presented part of the reference face was significantly and consistently higher for inverted than for upright faces. These observations confirm the narrower perceptual field for inverted than for upright faces, supporting a more holistic processing strategy for upright than for inverted faces.
Meeting abstract presented at VSS 2012