Co-speech laughter in a conversational speech task

Laughter plays a role in the natural flow of conversation. This study examines four dyads participating in a semi-structured collaborative conversational speech task to assess the effects of participant role—listener versus speaker—and the location of laughter within a speech turn on laughter distribution and duration. We hypothesize that speakers’ laughter will be more constrained than listeners’ due to its limitation to speech turn boundaries and its accommodation of a speaker’s ongoing speech planning requirements. Additionally, based on prior findings on acoustic phrase-final lengthening in speech prosody, we expect that laughter at the end of a speech turn will be longer than turn-initial or -medial laughter. Our results demonstrate substantial individual variability, with a majority of participants having longer laughter duration as listeners than as speakers, in part, we propose, due to laughter’s use as a form of backchanneling. Participants also demonstrate longer durations of turn-initial laughter than turn-final or turn-medial laughter. This unexpected result may be due to overlapping speech and laughter that is often seen at the end of speech turns. Our results suggest that the patterning of laughter with respect to speech turns may help structure interactive speech discourse. [Work supported by NIH.]Laughter plays a role in the natural flow of conversation. This study examines four dyads participating in a semi-structured collaborative conversational speech task to assess the effects of participant role—listener versus speaker—and the location of laughter within a speech turn on laughter distribution and duration. We hypothesize that speakers’ laughter will be more constrained than listeners’ due to its limitation to speech turn boundaries and its accommodation of a speaker’s ongoing speech planning requirements. Additionally, based on prior findings on acoustic phrase-final lengthening in speech prosody, we expect that laughter at the end of a speech turn will be longer than turn-initial or -medial laughter. Our results demonstrate substantial individual variability, with a majority of participants having longer laughter duration as listeners than as speakers, in part, we propose, due to laughter’s use as a form of backchanneling. Participants also demonstrate longer durations of turn-initial laught...