2022-04-13, 15:00–15:30 (Europe/Vienna), Room 4
We set out to develop a laboratory method to study cross-situational stylistic variation in fine phonetic detail of single speakers. It has been shown that speech varies depending on the addressee (Bell 1984). Yet, an addressee’s social role pertaining to the social constellation between the interlocutors in an interaction in combination with functional-situational aspects have not been explored so far. This study closes this gap.
To keep the interlocutor stable, we have created six pictures of the same individual varying in clothing styles (from formal jacket to T-shirt) and accessories (glasses, piercing). 86 participants rated these pictures for perceived formality. The persona judged to be the most and the one rated to be the least formal were selected. We also created short choreographed – and as much as possible identical – movies in which the person in the video listens and occasionally lightly nods or slightly changes in posture. The two films corresponding to the two levels of formality were used in the experiment.
German university students from Berlin (16 females and 4 males) and Bremen (9 females and 6 males) were asked to engage with a person on a computer screen in a presumed video-call. They were told that the person would only listen but not speak. The contexts given were e.g. this is your new neighbor [informal] or this is your boss [formal] and they were requested e.g. to explain what to do around town [nothing at-stake] or ask for a pay raise [at-stake]). To get the greatest possible contrast in formality, speakers conversing with the formal persona had something riding on the conversation (at stake situation) or just small-talked with the informal persona without having anything riding on it (nothing-at-stake).
Orthographic annotations were automatically segmented (MAUS, Kisler et al. 2017) and manually corrected at the word and phoneme level. The size of the vowel space was parametrized as the dispersion of the German vowels /iː eː ɛː aː a oː uː/ within the F1×F2 space (cf. Weirich & Simpson 2014). All vowel formants were extracted with a custom Praat script. Vowel dispersion was implemented as the median Euclidean distance of each vowel token from the center of the respective vowel space.
A linear mixed effects regression model was fit to analyze vowel dispersion with fixed effects for gender (male vs. female), setting (formal vs. informal) and speaker group (Berlin vs. Bremen), including all two-way interactions, and speaker identity as a random effect. All statistical analyses were performed in R (lme4 and lmerTest: Bates et al. 2015, Kuznetsova et al. 2017).
Results for 35 speakers indicate a significant effect of setting (p < 0.001) – vowels are more dispersed in the formal context. An effect for speaker group (p = 0.002) indicates regional differences between Bremen and Berlin: Bremen speakers show a smaller dispersion (in both formal and informal settings) compared to Berliners. Gender does not significantly contribute to the outcome. More data is currently being analyzed.
- Bates, Douglas, Martin Mächler, Ben Bolker & Steve Walker. 2015. Fitting linear mixed-effects models using lme4. Journal of Statistical Software 67(1). 1–48.
- Bell, Alan. 1984. Language style as audience design. Language in Society 13(2). 145–204.
- Kisler, Thomas, Uwe Reichel & Florian Schiel. 2017. Multilingual processing of speech via web services. Computer Speech & Language 45, 326–347.
- Kuznetsova, Alexandra, Per B. Brockhoff & Rune H. B. Christensen. 2017. LmerTest package: Tests in linear mixed effects models. Journal of Statistical Software 82(13).
- Weirich, Melanie & Adrian Simpson. 2014. Differences in acoustic vowel space and the perception of speech tempo. Journal of Phonetics 43. 1–10.