Rhythm in vocal emotional expressions: the normalized pairwise variability index differentiates emotions across languages

The voice is an important channel for emotional expression. Emotions are of- ten characterized by differences in pitch, loudness, the duration of segments, and spectral characteristics (Scherer, 2003). The rhythmic aspect of emotional speech has been largely neglected, most studies limit themselves to segment duration and speech rate. Since languages are often characterized by their rhythmic class (as either “stress timed” or “syllable timed”), we wanted to know whether the rhythmic structure plays a role in vocal emotional expres- sions. To characterize the rhythmic structure, we used the normalized pairwise variability index. The normalized pairwise variability index (nPVI) characterizes the rhythm of a language in a more continuous way (Grabe & Low, 2002). Quin- to, Thompson, and Keating (2013) found that the nPVI differentiated emotio- nal expressions from non-emotional ones. However, their study was limited to English (a stress timed language) and their nonsensical carrier sentence contai- ned real words, possibly influencing the role of speech rhythm. This contribu- tion investigates whether the nPVI can be used to characterize the possible rhythmic differences between emotions in a stress timed and a syllable timed language (Dutch and Korean). We do so by using an existing corpus (Goudbeek & Broersma, 2010) of eight posed emotional expressions (balanced for valance and arousal) by speakers of Dutch and Korean. The findings show, as expected, that Dutch and Korean differ in their nPVI, but, importantly, that the different emotions in the corpus also differ in their nPVI. Further analysis shows that emotional valence is an important contributor to these differences. Finally, the effects are different for Dutch and Korean, indicating the importance of studying different languages when investigating vocal emotional expression