Conversational Turn Timing: The Effects of Prosody and Pragmatic Context in Production and Perception

Abstract

This thesis examines the production and perception of turn-transition timing in spontaneous conversation, focusing on the role of speech timing and pragmatic context. Study 1 addressed the claim that turns are timed precisely, typically after pauses of ~200ms, through a systematic review of turn-timing distributions from 25 studies. All studies operationalised turn timing as the time between the end of one turn and the start of the next, but most did not provide acoustic-phonetic definitions of turn boundaries. Turn timing distributions had high kurtosis and were generally right-skewed, with central tendency measures clustered between 0-500ms. Means of task-oriented conversations (e.g., map tasks) were distributed more widely than free conversations. Study 2 tested the perception-action entrainment hypothesis (Garrod & Pickering, 2015; M. Wilson & Wilson, 2005) that higher speech rates should correlate with faster turn timing, by analysing local and global speech rate as predictors of turn timing in question-response sequences from 17 spontaneous conversations in Dutch and English using linear regression model selection. Contrary to the hypothesis, faster questions predicted slower turn timing, while shorter final syllables predicted faster turn timing. Polar questions were answered faster than open questions, and relevant responses were initiated earlier than non-relevant responses. Moreover, speech-timing variables did not significantly explain turn-timing variance when these pragmatic variables were included. Not all turns are ‘on time’. Study 3 created a perceptual ‘timeliness’ variable from participant ratings of responses from Study 2. Speech rate predicted turn timing only for responses judged ‘timely’, which were earlier after slower questions. Timeliness judgements depended on both inter-turn pause duration and question-final syllable length, suggesting that the usual pause-based operationalisation of turn timing may not capture perceived turn timing. These findings do not support entrainment hypotheses, emphasise the importance of pragmatic factors in conversational turn timing, and suggest that perceptually-validated operationalisations of turn timing are needed.