Abstract
Categorizing the age of others is a process so automatic and ubiquitous that we are rarely aware of it. On what basis does the brain make these perceptual judgments? We used the 'superstitious perception' reverse correlation paradigm (Gosselin & Schyns, 2003, Mangini & Biedermann, 2004) to address this question with faces. 12 observers with normal or corrected to normal vision (six 18-25; six 58-75 years old) participated in the experiment. On each trial, stimuli comprised an average face (averaged over 83 faces, 18 to 79 years old) with superimposed noise generated from recursive Gabor filters (Morlet wavelets of size 5, 6 orientations, 2 polarities). We instructed observers that they would be categorizing ages faces drawn from a large database, and obscured by noise. At the start of each block of 18 trials, observers choose one of 3 simultaneously presented stimuli (computed as above) as the best representative of a numerical age category (18-35, 40-55, or 60-80 years old), randomized across blocks over a total 3,240 trials. For each numerical age category, we reverse correlated the noise templates. Features associated with 'old' images were a low spatial frequency darkening of the skin and thin lips, while 'young images' showed paler skin, round cheeks and exaggerated lips. In the image domain, we validated these features with a pattern classifier, revealing the low SF pale/dark skin as a discriminatory feature of young/old face-images. In the perceptual domain, 8 independent observers validated the CIs (projected onto 12 new average faces), by classifying their age. We identified an 'other-age effect' in young observers: Young observer classification images (CIs) for 'middle aged' were not distinct from those of old, whereas old observers had distinctive CIs for the three age categories. Validation demonstrated that older observers' CIs more accurately spanned the expected age range.
Meeting abstract presented at VSS 2014