Abstract
A variety of well-established phenomena in face perception are often explained based on the 'face diet'—the contents of daily exposure to faces. For example, the face inversion effect and the other-race effect are believed to stem from minimal exposure to inverted and other-race faces, respectively. Face perception involves a protracted development period and remains malleable well into adulthood (Germine, Duchaine, Nakayama, 2011; Rhodes et al. 2011). Previous studies focused on the face diet in infancy (Jayaraman, Fausey, Smith, 2015; Sugden, Mohamed-Ali, Moulson, 2013). We present the first report on the face diet of adults (N=29) via a head-mounted camera. Our aim was to 1) provide empirical evidence for widely-believed assumptions such as the predominance of own-race and upright faces; 2) examine unknown aspects of the face diet regarding gender, familiarity, and pose; 3) test whether visual experience with faces occurs most often at close social distances, a hypothesis that was borne out of previous work (Oruc, Barton, 2010; Yang, Shafai, Oruc, 2014). Out of the 25,691 recorded frames (>214 hours of footage), 7475 faces were detected and annotated for gender, ethnicity, size, familiarity and pose. No inverted faces and no differences between exposure to faces of either gender were observed. Majority of faces were own-race (>85%). Faces were most frequently seen in the three-quarters pose, followed by frontal, and then by profile. Over 75% of faces were familiar, reflecting the longer periods of interaction with familiar individuals in contrast to brief encounters with strangers. Seventy-three percent of all faces were viewed from 1.5 m or closer, peaking at 89 cm (median), corresponding to a face size of about 8-9 degrees visual angle. These results confirm widely-accepted assumptions regarding the contents of the face diet and furthermore provide a potential explanation for effects of size in face recognition performance.
Meeting abstract presented at VSS 2017