Comparing categorical and dimensional ratings of emotional speech

The purpose of this study was to examine how the judgments of several emotional categories map onto the Arousal-Valence emotional space, and how this mapping is influenced by the intensity of these emotions. 20 judges were asked to rate excerpts of emotional speech taken from corpora of acted and elicited emotional speech, using two methods: 1) categorical labeling; and 2) 5-point scales representing Arousal and Valence. Results showed the commonly used negative emotions (Anger Sadness and Fear) to have large degree of overlap, though their centroids were distinct. When the centroids were plotted as a function of emotional intensity, they formed relatively straight lines pointing roughly to Neutral. However, in contrast to assumptions in the literature, Neutral was found to center around a negative value of arousal. Thus the entire emotional space was somewhat biased in the direction of negative arousal. Overall, the variability in judgment of Arousal was found to be large in comparison to the variability in Valence.