Assessing common ground in conversation: The effect of linguistic and physical co-presence on early planning.

摘要

Speakers routinely adjust their behavior upon assessing the information they share in common with their conversational partners, but there remains controversy over when and how these adjustments happen. In this dissertation I address two debates regarding partner-specific adjustments: (a) whether they recruit the language processing system in a way that is so automatic as to be inflexible, affecting more inferential processes (e.g., utterance planning) but not fast-acting ones (e.g. articulation), and (b) what aspects of an experience with a partner become indexed in episodic traces for shared information. Specifically, I investigate whether co-presence conditions (i.e., whether information is shared linguistically, physically, or both) become indexed in episodic traces, consequently affecting both utterance planning and articulation.;Experiments 1 and 2 involved referential communication tasks in which Directors instructed two Matchers, separately, on how to arrange cards. In Experiment 1 materials were items that were difficult to describe, whereas in Experiment 2 materials had common labels. In the first two rounds (Phase 1) cards were distributed as follows: with each Matcher, some cards were shared linguistically and physically, others only linguistically, others only physically, and others were completely absent. In the subsequent two rounds (Phase 2) Directors matched all cards from Phase 1 with each Matcher. I examined whether the Directors' descriptions in Phase 2 reflected sensitivity to the co-presence conditions in Phase 1.;Indeed, Directors' initial descriptions in Experiment 1 showed sensitivity to how information had been shared. Directors used fewer definite expressions for items that had been mentioned in Phase 1 compared to items that had not been mentioned. At the same time, adjustments in the amount of detail and provisionality of their initial descriptions showed sensitivity to the specific conditions of co-presence, suggesting that episodic traces did not merely encode a binary (mentioned vs. unmentioned) distinction: Directors included more words, idea units, reconceptualizations and hedges for items they had shared only physically with their Matchers compared to items they had shared only linguistically, and in turn included more words, idea units, reconceptualizations and hedges for items they had shared only linguistically with their Matchers compared to items they had shared both linguistically and physically. When taken together, these adjustments reflect appropriate strategies in initial audience design, driven by speakers' memory for how information had been previously shared: Referents that had been shared both linguistically and physically involved attenuated initial descriptions (fewer words, idea units), fewer markers of provisionality and more markers of definiteness. Referents that had been shared only linguistically were described with just as many markers of definiteness, signaling to the conversational partner that these referents had been previously mentioned. But at the same time they were described with more detail, with just as many words and idea units as completely new referents, reflecting the degree of grounding that the conversational partners had previously achieved.;Directors' explicit reports in a source monitoring questionnaire on how they had shared items with their Matchers in Phase 1 of Experiments 1 and 2 provide corroborating evidence that to some extent people can actually recall the conditions of co-presence.;The intelligibility of lexically identical expressions culled from Experiment 2 was assessed by a new group of listeners in Experiment 3. Listeners' judgments revealed that Directors also distinguished the intelligibility of their expressions according to how information had been shared: expressions for items that had been shared previously only physically were rated as clearer than those shared only linguistically or both linguistically and physically. In other words, although items in Experiment 1 lacked conventional labels and were negotiated more than those in Experiment 2, Directors kept track of co-presence and adjusted their utterance planning and articulation accordingly in both experiments.;Together, these findings suggest that episodic traces do index the conditions of co-presence of shared information: speakers' adjustments in utterance planning reflect grounding techniques appropriate to how information had been previously shared. Moreover, the effects of co-presence extend to the relatively automatic process of articulation, suggesting that partner-specific adjustments are deployed flexibly. When the informational needs of the conversational partner are represented easily and are cued rapidly, as a small set or relevant constraints, speakers adjust their early planning at multiple grains of linguistic processing.