cSCAN Rounds

Syntactic alignment and participant role in dialogue

There is substantial evidence that speakers in a dialogue implicitly affect each other's behaviour. One aspect of this is that they start to converge, or align, their speech at levels of structure as diverse as speech rate, choice of vocabulary and rhetorical structure (e.g., Giles & Smith, 1979; Clark & Wilkes-Gibbs, 1986; Schenkein, 1980). Recent research has shown that speakers also align with respect to their grammar; in other words, they tend to use the same syntactic structure in a 'target' utterance as their interlocutor has just used in a 'source' utterance (Branigan, Pickering & Cleland, 2000). In this talk, I will present a series of experiments that examine whether syntactic alignment occurs in dialogues that involve more than two participants, and in particular whether it is influenced by each participant's role within the dialogue: whether a participant acts as speaker, as addressee, or as side-participant with respect to a given utterance. In our experiments, we manipulated the participant role of naive participants in a picture-description and -matching game, and examined how this influenced the extent to which they syntactically aligned with confederate participants' prior utterances. I will present evidence that implicit effects of syntactic alignment occur in multi-party dialogues just as in two-party dialogues, and are not restricted to speaker-addressee dyads. However, alignment is stronger between speakers and addressees than between speakers and side-participants in ways that cannot be explained in terms of simple reciprocity effects. Rather, it seems to be the case that speakers' behaviour during production of a target utterance is influenced by their participant role at an earlier point where they initially processed a source utterance. These results suggest that participant role may play an important part in determining linguistic behaviour in dialogue. I will show how these results can be explained within theories of dialogue that assume that speakers address their contributions directly to their addressees, but indirectly to side-participants (e.g. Clark & Carlson, 1982), and will argue that the effects reflect differences in the depth of processing that addressees and side-participants undertake in relation to an utterance.