| dc.description.abstract | This thesis is concerned with two different accounts of how speakers coordinate
conversation. In both accounts it is suggested that aspects of the manner in
which speech is performed (its disfluency and its rate) are integral to the smooth
performance of conversation.
In the first strand, we address Clark's (1996) suggestion that speakers design
hesitations, such as filled pauses (e.g. uh and um), repetitions and prolongations,
to signal to their audience that they are experiencing difficulties during
language production. Such signals allow speakers to account for their use of time,
particularly when they experience disruptions during production. The account
is tested against three criteria, proposed by Kraljic and Brennan (2005), for evaluating
whether a feature of speech is being designed: That it be produced with
regularity, that it be interpretable by listeners, and that its production varies according
to the speaker's communicative intention. While existing literature offers
support for the first two criteria, neither an experiment with dyads nor analyses
of dialogue in the Map Task Corpus (MTC; Anderson et al., 1991) found
support for the third criterion. We conclude that, rather than being signals of
difficulty, hesitations are merely symptoms which listeners may exploit to aid
comprehension.
In the second strand, we tested Wilson and Wilson's (2005) oscillator theory of
the timing of turn-taking. This suggests that entrainment between conversational
partners' rates of speech allow them to make precise predictions about when each
others' turns are going to end, and, subsequently, when they can begin a turn of
their own. As a critical test of the theory, we predicted that speakers who were
more tightly entrained would produce more seamless turn-taking. Again using
the MTC, we found no evidence of a relationship between how closely entrained
speakers were and how precisely they timed the beginning of their turns relative
to the ends of each others' turns. | |
