Aging, pattern separation, and categorical perception of faces

Categorical perception (CP) is the phenomenon by which observers view linear changes that occur across a continuum as distinct categories. Although categorical perception is a perceptual phenomenon, it may be subserved by mnemonic processes such as pattern separation. To examine this hypothesis, following standard CP tasks, we assessed younger and older participants' abilities to identify and discriminate between members of pairs of famous or non-famous faces. We hypothesized that if CP is dependent upon neural pattern separation, which declines with aging, discrimination ability as indexed by CP would be compromised in older adults, as was found in our study. Since familiarity promotes pattern separation, CP should be enhanced for famous, as compared to non-famous faces, even in older adults. We found that all participants benefited from familiarity, but younger adults outperformed older adults overall. We next examined the effects of face inversion on CP for both famous and non-famous faces. If pattern separation, and CP, is determined solely by the similarity across physical features, then CP should be similar for upright and inverted faces since these features are perceptually invariant across orientation. If, however, pattern separation, and CP, depends on how stimuli are represented, then orientation may matter as upright and inverted faces are represented holistically or part-based, respectively. We found that inversion disrupted CP in younger adults whereas older adults performed similarly across both conditions, suggesting that face-representation is more part-based in older adults.

Automated summary

The study found that younger adults performed better than older adults in categorical perception tasks involving famous and non-famous faces. Additionally, face inversion disrupted categorical perception in younger adults, while older adults performed similarly across both conditions, indicating that the representation of faces in older adults is more part-based.