Mental representations of emotional facial expressions are more complex rather than less accurate in older observers

Abstract

The brain ages as dramatically as the body, manifesting itself in the reorganization of brain networks mediating interpretation of social signals. In particular, previous research has indicated that older adults are less accurate in categorizing negative facial emotional expressions, but the specific nature of these changes is unclear. Here, using the 4D Generative Face Grammar (GFG) platform (Yu et al., 2012) we generated a set of random combinations of facial movements, which coupled with reverse correlation informed the subsets constituting socially meaningful facial expressions. Eighteen young (10 female, median age=21, 18-25) and eighteen older (10 female, median age=65, 56-88) observers each categorized 4,992 animations of young and old faces into six basic emotions (i.e. 'happy', 'surprise', 'fear', 'disgust', 'anger', and 'sad') plus 'other/don't know', and rated the intensity of each expression (Jack et al., 2012). We then reverse correlated the random facial movements and their temporal parameters with the observers' responses. This resulted in dynamic models of the 6 basic facial expressions at five levels of intensity for each observer. We then performed a group-level k-means clustering analysis on the individual models. Young observers showed high homogeneity within expression categories, with a characteristic confusion between 'anger' and 'disgust'. Older observers showed well-preserved homogeneity for the expression models of 'disgust' and 'happiness'. By contrast, models of negative emotions ('fear', 'anger', 'sadness', and 'surprise') were less homogeneous across older observers. To examine whether this arose from a higher diversity of models of negative emotions in older observers, we identified the recognition strategies of each individual observer (Delis et al., submitted) and found that older observers deploy more strategies, in particular for models of 'sadness'.

Meeting abstract presented at VSS 2014