Syntactic priming, message formation, and successful communication in unscripted dialogue

Thesis (Ph. D.)--University of Rochester. Dept. of Brain & Cognitive Sciences, 2011.This dissertation describes a series of interactive dialogue experiments addressing the question: how does syntactic repetition function in an unscripted task-based dialogue? In Experiments 1-3, pairs of participants took turns producing picture descriptions, which could optionally contain prenominal or postnominal adjective-containing structures (e.g., the striped cat, the cat with stripes). Experiment 1 tested the hypothesis that syntactic priming is stronger in dialogue than in response to a pre-recorded voice. When an adjective was used, the syntactic form in which it appeared - prenominal or postnominal - was influenced by the preceding structure for dialogue prime trials, but not pre-recorded prime trials. Experiments 2 and 3 tested the hypothesis that syntactic priming affects language production at the message formulation stage, as opposed to the grammatical encoding stage. This was done by comparing the effects of syntactic priming with adjective-containing structures on two properties of the subsequent utterance: syntactic structure and information content. Priming had no effect on the information content of the subsequent message. But priming with an infrequent structure did have a small effect on the subsequently produced structure, suggesting that syntactic priming influenced grammatical encoding. Finally, Experiment 4 made use of a more complex cooperative dialogue task, as a step toward assessing the differential role that local syntactic priming and conceptual pacts play in determining referential form choice in dialogue. Pairs of participants took turns instructing each other about how to move 18 objects around until all the objects were in the same locations for both of the participants. The results showed that local syntactic priming only occurred if a referent was part of the situation model throughout the task, in which case it was quite strong. In addition, there was significantly more syntactic repetition within contrast sets than there was between two adjacent, but conceptually unrelated descriptions. This was true even when lexical overlap was ruled out as an explanation. Finally, when local syntactic priming did occur, it predicted successful communication on three different measures of communicative success: completion time, fluency, and accuracy. This supports previous claims that syntactic priming promotes success in task-based dialogues