Perceptual and acoustic differences between authentic and acted nonverbal emotional vocalizations

Most research on nonverbal emotional vocalizations is based on actor portrayals, but how similar are they to the vocalizations produced spontaneously in everyday life? Perceptual and acoustic differences have been discovered between spontaneous and volitional laughs, but little is known about other emotions. We compared 362 acted vocalizations from seven corpora with 427 authentic vocalizations using acoustic analysis, and 278 vocalizations (139 authentic and 139 acted) were also tested in a forced-choice authenticity detection task (N = 154 listeners). Target emotions were: achievement, amusement, anger, disgust, fear, pain, pleasure, and sadness. Listeners distinguished between authentic and acted vocalizations with accuracy levels above chance across all emotions (overall accuracy 65%). Accuracy was highest for vocalizations of achievement, anger, fear, and pleasure, which also displayed the largest differences in acoustic characteristics. In contrast, both perceptual and acoustic differences between authentic and acted vocalizations of amusement, disgust, and sadness were relatively small. Acoustic predictors of authenticity included higher and more variable pitch, lower harmonicity, and less regular temporal structure. The existence of perceptual and acoustic differences between authentic and acted vocalizations for all analysed emotions suggests that it may be useful to include spontaneous expressions in datasets for psychological research and affective computing