Home > Research > Publications & Outputs > Tell me something I don’t know

Links

View graph of relations

Tell me something I don’t know: Speaker salience and style affect comprehenders’ expectations for informativity

Research output: Contribution in Book/Report/Proceedings - With ISBN/ISSNChapter (peer-reviewed)peer-review

Published
  • Vilde R. S. Reksnes
  • Alice Rees
  • Chris Cummins
  • Hannah Rohde
Close
Publication date19/09/2024
Host publicationInformation structure and information theory
EditorsPhilippa Cook, Anke Holler, Catherine Fabricus-Hansen
PublisherLanguage Science Press
ISBN (electronic)9783961104819
ISBN (print)9783985541102
<mark>Original language</mark>English

Publication series

NameTopics at the Grammar-Discourse Interface
PublisherLanguage Science Press
Volume10
ISSN (Print)2567-3335

Abstract

A comprehender’s estimate of what events or situations are typical in the world is distinct from their estimate of what a speaker is likely to report on. Comprehension and production studies have shown contradicting preferences for which types of estimates are used by comprehenders and by speakers: typicality is favoured in comprehension (e.g. real-world typical content is associated with processing ease), whereas speakers’ production choices favour the inclusion of surprising or informative content (i.e., easily inferable or typical content is disfavoured). We posit that comprehenders are aware of and make use of speakers’ production preferences when anticipating upcoming content. In two studies, we elicit sentence completions as an index of comprehenders’ expectations about upcoming material and evaluate the informativity of these completions (their object typicality, presence of modification or negation, and information theoretic entropy and relative entropy scores). Experiment 1 manipulated the salience of the speaker and found that increased emphasis on the speaker led to an increase in informativity, showing that the more aware comprehenders are made of an intentionally communicating speaker, the more their expectations favour upcoming words that would yield an informative utterance. Experiment 2 further tested the malleability of this informativity bias by familiarising participants with two speakers who differ in the informativity of their utterances. When completing utterances from the two speakers, comprehenders provide more informative completions for the high-informativity
speaker, showing that comprehenders are able to adapt their expectations for informativity to individual speakers’ communicative styles. This sensitivity to speakers’ production preferences highlights a role for informativity-driven reasoning about the speaker in models of language processing.