Classification images characterize age-related deficits in face discrimination

Face perception is impaired in older adults, but the cause of this decline is not well understood. We examined this issue by measuring Classification Images (CIs) in a face discrimination task in younger and older adults. Faces were presented in static, white visual noise, and face contrast was varied with a staircase to maintain an accuracy rate of ≈71%. The noise fields were used to construct a CI using the method described by Nagai et al. (2013) and each observer's CI was cross-correlated with the visual template of a linear ideal discriminator to obtain an estimate of the absolute efficiency of visual processing. Face discrimination thresholds were lower in younger than older adults. Like Sekuler, Gaspar, Gold, and Bennett (2004), we found that CIs from younger adults contained structure near the eyes and brows, suggesting that those observers consistently relied on information conveyed by pixels in those regions of the stimulus. CIs obtained from older adults were noticeably different: CIs from only two older adults exhibited structure near the eye/brow regions, and CIs from the remaining older observers showed no obvious structure. Nevertheless, face discrimination thresholds in both groups were strongly and similarly correlated with the cross-correlation between the CI and the ideal template, suggesting that despite older observers' lack of consistent structure, the CI method is sensitive to between-subject differences in older observers' perceptual strategy.