Advocating a Componential Appraisal Model to Guide Emotion Recognition

Three Theoretical Perspectives On Emotion

Most research on emotion expression implicitly or explicitly used a discrete emotion perspective (Scherer, Clark-Polner, & Mortillaro, 2011) and the same is true for automatic emotion recognition systems. Discrete emotion theory has been formulated on the basis of findings concerning few intense emotions –called basic emotions – that are expected to have prototypical facial expressions and emotion-specific physiological signatures (Ekman, 1992, 1999; Ekman, Levenson, & Friesen, 1983; Ekman, Sorenson, & Friesen, 1969). This theory dominated the field for decades and it is still the most widely used. There is robust evidence about the existence of some facial configurations that are cross-culturally labeled with the same emotion terms. However, several studies show that people frequently report the experience of emotional states that are not part of this set of basic emotions (Scherer & Ceschi, 2000; Scherer, Wranik, Sangsue, Tran, & Scherer, 2004), and, more importantly, that for spontaneous and enacted emotional expression, these complete prototypical expressions rarely occur (Naab & Russell, 2007; Russell & Fernandez-Dols, 1997; Scherer & Ellgring, 2007a). Discrete emotion theorists tried to solve these problems by suggesting the concept of emotion families (Ekman, 1992). An emotion family includes several lexically marked variations of basic emotions labels; all terms within a family share a common theme (characteristics unique to the family) and variations due to individual, cultural, and contextual factors. Nevertheless, it is not clear how these variations should be modeled and whether they can be identified from vocal or facial expressions.