CATEGORICAL PERCEPTION OF FACES AND FACIAL EXPRESSIONS: THE AGE FACTOR

ABSTRACT

Effects of normal aging on categorical perception (CP) of facial identities and facial emotional expressions were investigated. The stimuli were morphs of three identities and of three expressions. In Experiment 1, 38 healthy participants (22 to 79 years old; three age groups) had to identify morphed identities (three continua) and expressions (three continua) varying with respect to their distance from the original photographs. This experiment allowed the computation of the boundary between categories (expressions and identities). The results showed that the locus of this boundary was not affected by age (except for the continuum disgust-happiness), but that the rate of intrusions and latencies of responses increased with age. In addition, CP was suggested, as latencies increased as a function of the distance between the displayed morph and the original, unmorphed photograph. CP predicts an easier discrimination of two morphs located on both sides of the boundary than discrimination of two morphs located on the same side. This was investigated in Experiment 2 (48 new participants; three age groups). For expressions, the prediction was verified in all groups for both accuracy and correct latency; in addition, an effect of age was observed only for morphs located on both sides of the boundary. For identities, the same pattern of results emerged, with one exception: for accuracy, the predicted advantage of pairs located on both sides of the boundary was not verified in the oldest group. Thus, CP of facial identities seems to vanish with normal aging, whereas CP of facial expressions appears to be more robust.

Notes

S. Campanella is founded by the Belgian National Fund for Scientific Research (FNRS).

¹Following the recommendations of Snodgrass and Corwin (1988), many studies used discrimination and bias measures derived from the signal detection theory. The score d′ evaluates the subject's ability to discriminate between old and new items. The score β is an index of bias known as the likelihood ratio (the ratio of the likelihood of obtaining an observation equal to a criterion given an old item, to the likelihood of obtaining this observation given a new item). C score is the z score for the false alarm rate minus d′/2, and this index is preferable to β because the latter is not independent of. d′ Negative C scores (most lax) indicate a bias favoring the judgment in question, positive scores (most conservative) indicate a bias against this judgment, and 0 indicates no bias. Generally, nonparametric indices are used because the empirical data are rarely normally distributed: A′ is computed instead of d′, and B′′ instead of C. The maximal A′ = 1, and B′′ lies between − 1 and + 1, with B′′ = 0 indicates a neutral criterion, B′′ > 0 indicates a liberal criterion, and B′′ < 0 indicates a conservative criterion.

²As the recognition of facial expressions is acquired as soon as the first years of the life, no preliminary learning was required, so that five different expressions were used. However, for facial identities, a preliminary learning was required, as the set of used identities had first to become familiar. For that reason, and because elderly participants were involved, we limited this set to three faces. One could argue that famous faces were available, but we were in awe of important individual differences between the participants with respect to face familiarity.