Abstract
Many studies have shown that humans can discriminate faces on the basis of changes in the spatial arrangement of features. However, it is not known whether sensitivity to differences in distances among features is sufficient to support real-world face identification. The current study examined if inter-pupillary distance (IPD) or nose-to-mouth distance (NMD) could serve as ecologically-valid cues for face discrimination. Observers discriminated faces that differed in IPD or NMD and then performed a realistic face identification task (Bruce et al., 1999). The primary question was whether IPD and/or NMD discrimination thresholds were correlated with face identification accuracy. Observers also performed a control task requiring them to discriminate face contrast. This control task is unlikely to rely on processes that encode spatial relations, but (like the other tasks) requires vigilance and focused attention. IPD and NMD discrimination thresholds followed Weber's law, and were not significantly affected by variation in facial identity or size. Interestingly, a composite measure of IPD and NMD thresholds was significantly correlated with face identification accuracy, even after controlling for the relation between face identification and contrast discrimination. This result suggests that face identification is supported in part by the encoding of spatial relations. However, anthropometric data from a large population of adult faces reveals that variation in IPD and NMD is small relative to the thresholds measured in most of our observers. Therefore, neither IPD nor NMD cues on their own would be particularly useful for discriminating or identifying faces in natural contexts.
NSERC grants 42133 and 105494, and the Canada Research Chair programme.