Almost everything we do that concerns other people involves talk-in-interaction. The purpose of this research is to advance understanding of how conversation works. It will investigate the ways in which speakers and listeners make use of the phonetic (sound) resources of language to shape and interpret talk in natural conversation.

Linguists have a wealth of knowledge about the way speech sounds are produced and combined together to make words and longer utterances. However, they frequently work on the basis of constructed data or speech produced in experimental settings. In consequence, there is only the most rudimentary information on the ways in which ordinary people use the phonetic resources of language in natural everyday talk. Though we know something of the way the pitch of the voice is used, for example, in asking questions, giving orders or making statements, we lack basic knowledge about the interactional work done by other phonetic features such as, tempo, loudness, or the precise way particular sounds are articulated.

As a result, we have little understanding, for instance, of how speakers signal that they have made a mistake and are now correcting it, or how they signal that they are willing to yield at turn-at-talk, or why it is that conversationalists treat some talk, which overlaps their own, as interruptive but other overlapping talk as supportive.

The proposed research develops a novel way to remedy this lack of information. Rather than working on constructed data or speech taken out of context, it will study the phonetic detail of how talk is shaped in real conversation. In particular it will investigate:

• how mistakes are signalled and corrected;
• how misunderstandings are clarified;
• how talk which overlaps or completes that of others is shaped and interpreted.

These activities are chosen because they are frequent in natural conversation and they provide rich opportunities to investigate the ways in which participants design their own talk and interpret the talk of others on a moment-to-moment basis.

The data used will consist entirely of audio and video recordings of natural conversation. The investigation will examine the full range of phonetic resources (rhythm, tempo, duration, loudness, pitch, voice quality and the articulation of sound segments) and their individual and combined roles in signalling interactional meaning.

The results of the research promise to be far-reaching. They will inform the work of linguists, sociologists, and psychologists. The development of an interactional approach to phonetics will challenge current theories of language use and language understanding. There will be practical benefits, too. The results will be a resource for researchers in computer speech recognition and for speech therapists and other practitioners concerned with the development or remediation of communication skills.