It May or May Not Be Relevant: Testing the Relationship Between Relevance and Informativeness
Abstract
The dominant view in cognitive science and formal pragmatics is that relevance is a matter of informativeness: a sentence is relevant if it is informative, and irrelevant otherwise. We provide experimental evidence against this thesis. For polar questions (e.g. Is Mary coming to the party?), we constructed two kinds of uninformative replies: replies that clearly violate relevance (uninformative-off-topic condition; e.g. Mary lives in an old house) and replies that are equally uninformative yet do not intuitively incur a relevance violation, such as uncertainty or disagreement reports (target condition; e.g. Her parents disagree about that). In a within-subject study (__ = 100; 12 scenarios), participants judged both kinds of replies uninformative, but a forced-choice continuation diagnostic revealed a sharp dissociation in perceived relevance: target replies were judged relevant 95%, versus 3% for the uninformative-off-topic condition. These results show that informativeness is not a necessary condition for relevance.