Adaptation of respiratory patterns in collaborative reading

Gérard Bailly, Amélie Rochet-Capellan, Coriandre VilainInterspeech 2013

TL;DR

This paper investigates how speakers' breathing patterns adapt during collaborative reading tasks with increasing levels of synchrony. They found that breathing control switches abruptly between "vital" (automatic) and "speech production" modes, with significant coordination emerging only when speakers are actively engaged in synchronized speech.

What Problem Does This Paper Address?

The paper delves into the intricate relationship between breathing and speech production, particularly in the context of human interaction. While it's understood that breathing patterns change during speech compared to quiet breathing (e.g., faster inhalations, longer exhalations), the precise mechanisms and the degree of adaptation in collaborative settings remain less clear. Specifically, the authors question the long-held assumption of a gradual, continuous adaptation of respiratory patterns from quiet breathing to speech production. Instead, they hypothesize an abrupt switch in control mechanisms, especially when speakers need to coordinate their speech, such as in turn-taking or synchronized reading. This research aims to quantify this adaptation and understand how it's influenced by the demands of collaborative speech.

Methods

The researchers monitored the respiratory chest circumferences of eight French dyads (pairs of speakers) using BIOPAC® belts. These belts were carefully positioned to accurately capture thoracic movements. The participants read 13 short texts under four experimental conditions, each imposing a different level of coordination constraint:

  1. Reading aloud alone: No interaction.
  2. Reading aloud paragraphs in alternation: Speakers take turns reading entire paragraphs.
  3. Reading aloud sentences in alternation: Speakers take turns reading individual sentences.
  4. Reading aloud texts in synchrony/chorus: Speakers read the text simultaneously.

Additionally, six minutes of quiet breathing were recorded before the experimental conditions. Speech data was automatically forced-aligned with phonetic transcriptions, and respiratory phases (inhalation/exhalation) were detected using signal processing techniques. All segmentations were manually verified. The study analyzed three characteristics of each respiratory cycle: period, inhalation quotient, and phase relative to the interlocutor's cycle.

Datasets

The study used a corpus of 13 short French texts, each approximately 10-15 sentences long and containing 131±15 syllables. These texts were read by eight French dyads. No new datasets were introduced.

Key Insights and Ideas

  • Abrupt Switch in Breathing Control: Contrary to the idea of a gradual adaptation, the study found evidence for an abrupt switch between two distinct breathing control systems: "vital" (automatic, symmetrical breathing) and "speech production" (asymmetrical, demand-driven breathing).
  • Coordination Drives Adaptation: Significant adaptation and synchronization of respiratory patterns only emerged when speakers were actively engaged in collaborative tasks requiring mutual synchrony (alternating sentences or reading in chorus). Simple listening or reading alone did not elicit the same level of respiratory coordination.
  • Syllabic and Respiratory Cycles are Phased: The study confirmed that syllabic onsets and respiratory cycles are strongly synchronized, aligning with previous observations about the quantal nature of speech rhythm.
  • Apnea as a Coordination Tool: Some participants, particularly those with backgrounds in activities like trombone playing or apnea swimming, utilized apnea (breath-holding) as a strategy during turn-taking, further highlighting the flexible control of respiration.
  • Gender and Social Factors: While not the primary focus, the study noted that male dyads showed less adaptation in their respiratory strategy compared to their speaking rate in the synchronized condition, suggesting potential gender or social dynamics at play.

Key Results

  • Asymmetry in Speech Breathing: Respiratory profiles during speech production were highly asymmetrical (skewness close to 20), contrasting with the low skewness (3-10) observed during quiet breathing and listening.
  • Phase Locking in Synchrony: In the synchronized reading condition (condition 4), respiratory cycles of interlocutors were strongly phase-locked. This synchronization was less pronounced in alternating sentence reading (condition 3) and absent in paragraph alternation (condition 2) and reading alone.
  • Speaking Rate Adaptation: Speakers generally adapted their speaking rate to increase coordination across conditions, but this adaptation was not always mirrored in their respiratory patterns, especially for male dyads in condition 4.
  • Inhalation/Exhalation Limits: The study identified an absolute limit of approximately 7 syllables per expired second and about 1 second for inhalation duration during speech production.

Limitations and Future Work

  • Controlled Environment: Participants were in separate rooms and instructed not to move, which might not fully reflect natural conversational settings where co-presence and gesturing can influence respiratory control.
  • Familiar Dyads: The use of dyads who knew each other well, along with some training for specific participants, might have influenced the observed coordinative profiles.
  • Future Directions: The authors suggest further research using more comprehensive physiological monitoring (e.g., plethysmographs, pneumotachographs) to better understand short- and long-term control of air volume. They also propose investigating the impact of social relationships and affect on respiratory patterning during speaking and inter-speaker coordination. Exploring how speakers regulate oxygen debt during prolonged speech production is another avenue for future work.

Summary updated

BibTeX
@inproceedings{bailly2013adaptation,
  author = {Gérard Bailly and Amélie Rochet-Capellan and Coriandre Vilain},
  title = {{Adaptation of respiratory patterns in collaborative reading}},
  booktitle = {Interspeech},
  year = {2013},
}
Figure