cSCAN Rounds

Convergence of speech rate: Interactive Alignment beyond representation

It has long been known that conversational partners tend to align on common ways of talking about the world, not only in choice of syntactic structures or referring expressions, but also in manner. While theories of the alignment of representations have considered the linguistic mechanisms responsible (for example Pickering & Garrod's, 2004, Interactive Alignment account), theories of the alignment of performative aspects of conversation have largely limited their scope to motivational explanations (most notably in Howard Giles' Accommodation theory). One exception to this trend has been Wilson & Wilson's (2005) oscillator model of turn-taking. In order to explain high-coordination in turn-taking they propose that endogenous oscillators in the brains of conversational partners, representing their readiness to speak, have their frequencies determined by each others' speech rate. As these oscillators become entrained (as partners align on speech rate) turn-taking should become more seamless. We present a simulation of this model which shows that as interlocutors' speech rates converge, the amount of variance in their turn-intervals (the time between each turn) decreases. This relationship was tested in the Map Task Corpus. While interlocutors were found to align on speech rate, through a process of priming, no evidence was found that this aided turn-taking. In absence of support for this important claim of the oscillator model, but with evidence for priming of speech rate in dialogue and monologue, we suggest that the interactive alignment account may extend beyond what is said, to how it is said.