Poster #70 - Children’s Real-World Experiences of Adult Speech: Pitch, Pragmatics, Gender, and Infant-Directedness

• In Event: Poster Session 01

Abstract

Previous studies have documented higher mean pitch and pitch variability between infant-directed speech (IDS) and adult-directed speech (ADS), among other differences (e.g., Fernald & Simon, 1984; Grieser & Kuhl, 1988; Ramírez et al., 2014; Broesch & Bryant, 2018; Räsänen et al., 2018; Hilton et al., 2022). Studies have also found that acoustic properties of speech vary by communicative intent, like approval compared to prohibition (Fernald, 1989). The current project extends this prior work via two main objectives. The first is to ascertain the prevalence of various pragmatic contexts within day-long child-centered home audio recordings, and assess whether there are differences in context prevalences for IDS versus ADS registers. The second objective is to analyze how mean pitch and pitch variability differ between IDS and ADS across pragmatic contexts.

The contexts in this study include: conversational basics, inform, reading, comfort, singing, imperative, questions, and babbling/vocal play. Recordings from the IDSLabel dataset (Bergelson et al., 2019) were analyzed: the dataset consisted of audio files representing adult speech previously coded for speaker gender and IDS vs. ADS register. Trained human listeners annotated pragmatic contexts non-mutually-exclusively. Table 1 shows the number of utterances in a pilot dataset per context, adult speaker gender, and speech register.

For the pilot data, a logistic mixed effects regression found that the imperative context was significantly more frequent among IDS register utterances than among ADS register utterances (OR = 24.80; p = .024) and that the inform context was more frequent in ADS compared to IDS (OR = .34, p = .047). With regard to pitch features within the three pragmatic contexts for which there were enough instances to make comparisons, mean pitch was significantly higher for IDS than ADS utterances (conversational basics: 𝜷 = .16, p < .001; inform: 𝜷 = .16, p < .001; question: 𝜷 = .16, p = .013; Fig. 1). Similarly, across the three contexts, pitch variability was greater for IDS than ADS (conversational basics: 𝜷 = .30, p < .001; inform: 𝜷 = .30, p < .001; question: 𝜷 = .30, p = .013). No significant differences in pitch features across contexts were identified. When adult speaker gender was included as a covariate, we found significant differences between males’ and females’ mean pitch but no significant differences in pitch variability, and no significant interactions between gender and context, at least for this pilot dataset.

We are currently finalizing these pilot analyses and a pre-registered analysis plan deposited on OSF; a larger dataset is now annotated and awaiting analysis once those are finalized. At SRCD we expect to be able to present results for the complete dataset of 3,727 adult samples.

The project will provide descriptive information about the pragmatic contexts young children in a North American sample experience in their daily lives. It is also assessing whether the findings of pitch differences between IDS and ADS registers generalize to different real-world pragmatic contexts and speaker genders. Annotations from this project will be contributed to HomeBank for researchers to use in other projects.

Authors

• Anvi Brahmbhatt, University of California - Los Angeles
  Presenting Author

• Emily Neer, University of California - Los Angeles

• Anne Warlaumont, University of California