Breathing behaviour and non-lexical vocalisations in talk-in-interaction

Context Speaking relies on breathing; and breathing behaviour underpins many underexamined sounds in conversation: audible in- and out-breaths, clicks (‘tsk’, ‘tut’ sounds); laughter, sighs and gasps. These non-lexical vocalisations are common in conversation, and have crucial functions like regulating turn-taking, commenting on talk, displaying shared pleasure or expressing a feeling; yet such sounds are rarely replicated in experimental data. A growing body of work in the psychological and cognitive sciences argues that humans are designed for dialogue rather than monologue, implying that sounds like these must be central to how we understand conversation and social relationships. The Challenges the Project Addresses Our knowledge of the phonetic form and positioning of non-lexical vocalisations and the link to breathing behaviour in spoken talk-in-interaction is very limited: We understand rather little about their variability, or what role this variability plays in spoken interaction. We lack conversational data with high-quality audio, video and respiratory kinematic data that allows us to track breathing patterns. We need an interactionally informed understanding of how breathing and speaking in conversation are intertwined. By combining the methods of Conversation Analysis and Phonetics, we will achieve a more nuanced understanding of the variability we see in speech, using categories grounded in tasks like turn-taking, turn-construction and the organisation of social actions, and seeing conversation as a joint, social achievement. Aims and Objectives The project has four main aims: To investigate the relationship between breathing and affiliated vocalisations in natural spoken interaction in English (including as a second language), German and French. To identify and explain the variable forms and functions of breathing and affiliated sounds, such as audible breath noises, laughter, and tongue clicks. To investigate how such sounds contribute to the fine-grained timing, turn construction, and social organisation of conversation. To explore the emergent nature of turns at talk, marked by e.g. pauses, turn-taking cues, unfolding sentence structures, and signs of speech planning and ultimately visualise and model the core elements of this linguistic organisation. To achieve this, we will collect and analyse a novel dataset of high-quality conversational data, complete with synchronised audio, video, and non-invasive respiratory kinematic data using respiratory belts. Potential Applications and Benefits The fundamental knowledge produced by this project is expected to yield significant benefits to the academic and wider community. Advancing our understanding of conversation: The research will model speech as a dialogic process, enriching and challenging current theoretical and methodological paradigms in linguistics. It will provide the necessary foundational detail on the form, position, and function of breathing-centred sounds to allow for more informed future quantitative analyses. New resources and tools: A primary output will be a richly annotated, multilingual corpus of multimodal data (audio, video, respiratory signals). The project will also develop new visualisations to help conceptualize and test new theoretical ideas about how breathing and talk-in-interaction are intertwined. Wider applications: The resulting knowledge about the social norms that govern how and when people produce audible in- and exhalations, laugh, or click can be used to model natural human behaviour. This has practical applications in fields such as: Speech synthesis and conversational agents: Making virtual assistants and synthesized speech more human-like and interactionally competent. Forensic phonetics: Providing new metrics for speaker comparison by studying individual vocal behaviour.