|dc.description.abstract||Compound contributions (CCs) – dialogue contributions that continue or complete an
earlier contribution – are an important and common device conversational participants
use to extend their own and each other’s turns. The organisation of these cross-turn
structures is one of the defining characteristics of natural dialogue, and cross-person CCs
provide the paradigm case of coordination in dialogue.
This thesis combines corpus analysis, experiments and theoretical modelling to explore
how CCs are used, their effects on coordination and implications for dialogue models.
The syntactic and pragmatic distribution of CCs is mapped using corpora of ordinary
and task-oriented dialogues. This indicates that the principal factors conditioning the
distribution of CCs are pragmatic and that same- and cross-person CCs tend to occur in
In order to test the impact of CCs on other conversational participants, two experiments
are presented. These systematically manipulate, for the first time, the occurrence
of CCs in live dialogue using text-based communication. The results suggest that syntax
does not directly constrain the interpretation of CCs, and the primary effect of a
cross-person CC on third parties is to suggest to them a strong form of coordination or
coalition has formed between the people producing the two parts of the CC.
A third experiment explores the conditions under which people will produce a completion
for a truncated turn. Manipulations of the structural and contextual predictability
of the truncated turn show that while syntax provides a resource for the construction of
a CC it does not place significant constraints on where the split point may occur. It also
shows that people are more likely to produce continuations when they share common
ground. An analysis using the Dynamic Syntax framework is proposed, which extends
previous work to account for these findings, and limitations and further research possibilities