The effect of noise and second language on turn-taking in task-oriented dialogue

Anna Josefine Sørensen, Ewen N MacDonald, Michal Fereczkowski

Research output: Contribution to conferenceConference abstract for conferenceResearchpeer-review


Previous studies of floor-transfer offsets (FTO), the interval between one talker
stopping and the other starting, suggest that normal conversation requires interlocutors to predict when each other will finish their turn. We hypothesized that noise and/or speaking in a second language (L2) would result in longer FTOs due to increased processing demands. Conversations from 20 pairs of normal hearing, native-Danish talkers were elicited using the Diapix task in four conditions consisting of combinations of language (Danish vs. English) and noise background (quiet vs. ICRA 7 noise presented at 70 dBA). Overall, participants took longer to complete the task in both noise and in L2 indicating that both factors reduced communication efficiency. However, L2 had very little effect beyond completion time, likely because the participants were very good in English. In contrast to our predictions, in the presence of noise, the median of the FTO distribution decreased by approximately 30 ms and the standard deviation decreased by approximately 10 %. However, the average duration of inter-pausal units (i.e., utterances of continuous speech) increased by 40 % in noise. These findings are consistent with talkers holding their turn for longer, allowing more time for speech planning.
Original languageEnglish
Publication date2017
Publication statusPublished - 2017
EventInternational Conference on Cognitive Hearing Science for Communication, - Sweden , Linkoping, Sweden
Duration: 18 Jun 201721 Jun 2017
Conference number: 4


ConferenceInternational Conference on Cognitive Hearing Science for Communication,

Bibliographical note

Abstract Book p.30


Dive into the research topics of 'The effect of noise and second language on turn-taking in task-oriented dialogue'. Together they form a unique fingerprint.

Cite this