A First Investigation of the Timing of Turn-taking in Ruuli

Tuarik Buanzur, Margaret Zellers, Saudah Namyalo, Alena Witzlack-MakarevichInterspeech 2018

TL;DR

This paper investigates the timing of conversational turn-taking in Ruuli, an under-studied Bantu language, focusing on question-response sequences. The findings suggest that despite significant structural differences from previously studied languages, Ruuli exhibits universal patterns in turn-taking timing, particularly in how preferred vs. dispreferred responses influence conversational flow.

What Problem Does This Paper Address?

Conversational turn-taking, the intricate dance of who speaks when, is a fundamental aspect of human interaction. While it's generally understood that speakers aim for smooth transitions, minimizing both overlapping speech and long silences, the precise timing mechanisms and their cross-linguistic universality remain areas of active research. Previous studies have largely focused on Indo-European languages, leaving a significant gap in our understanding of how languages with vastly different typological structures handle these conversational cues. This paper addresses this gap by examining turn-taking timing in Ruuli, a Bantu language spoken in Uganda, specifically looking at how the timing of responses to different types of questions (polar vs. wh-questions, and preferred vs. dispreferred responses) aligns with or diverges from findings in other languages.

Methods

The researchers employed a quantitative analysis of naturally occurring conversations. They first identified question-response sequences from a corpus of Ruuli audio and video recordings. These sequences were then meticulously transcribed and translated. Key to their methodology was the annotation of these sequences using Praat, a software widely used in phonetic and linguistic analysis. They classified questions into polar and wh-questions, and for polar questions, further distinguished between type-conforming (preferred) and non-type-conforming (dispreferred) responses. Transitions between turns were then categorized as "gap" (silent pause), "overlap" (simultaneous speech), or "O-gap" (short, barely perceptible pause/overlap). The duration of these transitions was measured, providing quantitative data on conversational timing.

Datasets

The study utilized a corpus of Ruuli data collected by a project focused on creating a comprehensive bilingual dictionary and grammar for the language. Specifically, they analyzed 10 audio files totaling approximately 6 hours of conversation. These recordings were made in naturalistic settings, capturing spontaneous speech from 15 participants aged 39-83. The data represents Ruuli speakers in their everyday environments, including background noise, which the authors acknowledge as a potential source of annotation inconsistency. This dataset is novel in that it provides a linguistic resource for an under-studied language, enabling research into its specific linguistic features, including conversational dynamics.

Key Insights and Ideas

  • Universal Turn-Taking Patterns: The most significant insight is that Ruuli, despite its typological distinctiveness from previously studied languages, adheres to universal principles of turn-taking timing. This reinforces the idea that the fundamental architecture of conversational turn-taking is a human universal.
  • Timing as a Communicative Tool: The study highlights how timing isn't just about avoiding awkward silences or interruptions; it carries communicative weight. Longer pauses preceding responses, especially to polar questions, signal dispreferred actions, mirroring findings in other languages.
  • Question Type Matters: Different question types elicit different response timings. Wh-questions and dispreferred responses to polar questions tend to be associated with longer gaps, suggesting a greater cognitive load or a deliberate signaling of hesitation.
  • Preferred Responses are Quicker: Type-conforming (preferred) responses to polar questions are generally produced with shorter gaps or even overlaps, indicating a smoother, more immediate conversational flow.

Key Results

  • The mean duration of gaps between questions and responses in Ruuli is approximately 259 ms. This is consistent with findings from other languages.
  • The distribution of gaps, no-gap-no-overlaps, and overlaps between question-response pairs was found to be significantly different from a random distribution.
  • Wh-questions and polar questions with non-type-conforming responses were relatively more likely to be followed by a gap.
  • Polar questions with type-conforming responses were more likely to have their response timed with no gap or overlap.
  • Overlaps occurred in about 15.7% of transitions, which falls within the range reported for other languages.
  • While a trend was observed where gaps following wh-questions and non-type-conforming polar question responses were longer than those following type-conforming polar question responses, a linear mixed model did not achieve statistical significance, likely due to the dataset size.

Limitations and Future Work

The authors acknowledge that the dataset used for this study is relatively small, which may have limited the statistical power to detect certain trends, particularly in the linear mixed model analysis. They also note that the noisy recording environment could have introduced some inconsistencies in annotations, though they believe these are likely random rather than systematic.

Future work suggested includes expanding the analysis to a wider range of turn pairs beyond just question-response sequences. They also propose investigating the impact of linguistic structures, such as the position of wh-words within a sentence, on turn-taking timing. Furthermore, exploring the interplay of Ruuli's tone system with prosodic cues that signal turn ends is identified as a promising avenue for future research.

Summary updated

BibTeX
@inproceedings{buanzur2018first,
  author = {Tuarik Buanzur and Margaret Zellers and Saudah Namyalo and Alena Witzlack-Makarevich},
  title = {{A First Investigation of the Timing of Turn-taking in Ruuli}},
  booktitle = {Interspeech},
  year = {2018},
}
Figure