12th Speech in Noise Workshop, 9-10 January 2020, Toulouse, FR

How consistently do speakers apply the Lombard speech clarification effect over time?

Chen Shen(a), Esther Janse(b)
Radboud University, Nijmegen, Netherlands

(a) Presenting
(b) Attending

People with hearing impairment may experience that their interlocutors are willing to clarify their speaking style upon request, but then quickly return to their habitual speech behaviour as the conversation continues. Whereas the acoustic-phonetic changes that come with Lombard speech (i.e., speech produced in noise) and clear speech have been described extensively, it is unclear how consistent speakers’ acoustic-phonetic modifications are over a period of time. We investigated the acoustic-phonetic differences between speakers’ habitual speaking style and their speaking style in a condition where they were presented with loud noise and were also instructed to speak clearly. Our research question was whether acoustic differences in articulation rate, pitch median, pitch range, and spectral tilt between habitual and Lombard/clear speaking style would change over the course of a sentence list. Conceivably, speakers could get tired of raising their voice and of their careful speaking style, such that the differences between speaking styles would decrease over trials. Alternatively, speakers may need some practice to show the full Lombard/clear speech effect.

Seventy-eight participants read out 48 sentences in both their habitual speaking style, and in a condition where they were instructed to speak clearly while hearing loud speech-shaped noise over headphones (Lombard/clear style). The list of 48 sentences was randomised four times to create four lists assigned to different participants, such that trial effects could be isolated from sentence (i.e., item) effects.

Results from linear mixed-effects models indicate that trial main effects were present in three of the four acoustic measures (i.e., for articulation rate, pitch median, and spectral tilt). Across all four acoustic measures, sentence trial interacted with speaking style. More specifically, acoustic differences between habitual and Lombard speech increased over trials, which was sometimes due to speakers becoming ‘sloppy’ in their habitual style over trials, e.g., faster articulation rate and smaller pitch range towards the end of the list. However, speakers also enhanced some of their Lombard style modifications over Lombard trials, e.g., higher pitch median and flatter spectral tilt at later trials. Thus, despite the higher vocal effort required to produce Lombard speech, speakers in our study were able to not only maintain but even enhance their Lombard speech modifications over trials. Research with actual conversation is needed to investigate to what extent our observations generalise to more demanding speaking tasks, and to see how clear/Lombard speaking style may change over the course of a conversation.

Last modified 2020-01-06 19:23:55