Accessibility, Duration, and Modeling the Listener in Spoken Dialogue

Referring expressions are thought to be tailored to the needs of the listener, even when those needs might be very costly to assess, but tests of this claim seldom manipulate listener’s and speaker’s knowledge independently. The Map Task enables us to do so. We examine two ‘tailoring’ changes in repeated mentions of landmark names: falling clarity of word articulation and rising accessibility of referring expression. Clarity results replicate Bard et al. (2000). Standardized word duration fell for speaker-Given listener-New items (Expt 1). Hence it was unimportant whether the listener heard an earlier mention. Reduction between mentions was no greater when it could be inferred that the listener could see the named item (Expt 2), and no less when the listener explicitly declared that they could not (Expt 3). Hence it was unimportant whether the listener could see the landmark. Reduction was unaffected by whether the repeater could see the mentioned landmark (Expt 4). Articulation thus depends only on what the speaker has heard previously. In contrast, accessibility was more sensitive both to listener (Expt 1) and speaker knowledge (Expt 4). The results conform most closely to a Dual Process model: fast, automatic, word-by-word processes let the speaker’s own experience prime articulation, while computationally costly assessments of listener knowledge control influence referring expression design where competing tasks permit.