Abstract
Everyday experience suggests we are able to produce and visually distinguish a large number of facial expressions of emotion. Yet, to date, most studies have focused on a small set of emotion categories: happiness, surprise, anger, sadness, fear and disgust. Combining or compounding these emotion categories generally results in rich, complex and meaningful social, affective and communicative expressions, e.g., the facial expression of happily surprised and the facial expression of angrily surprised. Here, we evaluate which facial expressions of compound emotions can be correctly labeled semantically, as well as quickly and accurately visually discriminated from other facial expressions of emotion. Stimuli included 15 facial expressions of compound emotions, the 6 typically studied emotions and a neutral face. In Experiment 1, participants made ranked-order judgments of the emotions displayed in each face. Any of combination of the six traditional emotions or neutral could be selected. For example, if Awe contains both surprise and fear, with a greater contribution of surprise, then participants should select Surprise first, then Fear and no other emotions. Data were analyzed using a single-winner election method for ordinal ranked data (Schulze method); this provided a ranked list of emotion labels for each facial expression. In Experiment 2, subjects completed a 3-alternative forced choice delayed match-to-sample task, in which a target facial expression of emotion was selected from a set of three test face emotions (1 compound emotion and 2 corresponding standard emotions). Data were analyzed using a Wilcoxon signed-rank test with Bonferroni correction. Results show that at least seven (and possibly twelve) common facial expressions of compound emotions are consistently and accurately categorized (i.e., semantically labeled and visually discriminated) by observers. These results suggest that the repertoire of facial expressions readily recognized by observers and employed by our cognitive and affective systems is larger than previously thought.
Meeting abstract presented at VSS 2014