Human observers can often judge emotional or affective states from bodily motion, even in the absence of facial information, but the mechanisms underlying this inference are not completely understood. Important clues come from the literature on "biological motion" using point-light displays (PLDs), which convey human action, and possibly emotion, apparently on the basis of body movements alone. However, most studies have used simplified and often exaggerated displays chosen to convey emotions as clearly as possible. In the current study we aim to study emotion interpretation using more naturalistic stimuli, which we draw from narrative films, security footage, and other sources not created for experimental purposes. We use modern algorithmic methods to extract joint positions, from which we create three display types intended to probe the nature of the cues observers use to interpret emotions: PLDs; stick figures, which convey "skeletal" information more overtly; and a control condition in which joint positions are connected in an anatomically incorrect manner. The videos depicted a range of emotions, including fear, joy, nurturing, anger, sadness, and determination. Subjects were able to estimate the depicted emotion with a high degree of reliability and accuracy, most effectively from stick figures, somewhat less so for PLDs, and least for the control condition. These results confirm that people can interpret emotion from naturalistic body movements alone, and suggest that the mechanisms underlying this interpretation rely heavily on skeletal representations of dynamic shape.
Read full abstract