Proceedings of the Annual Meeting of the Cognitive Science Society

How do we judge people wrong or right in their use of language? The words we use vary in how much their meanings de-pend on properties of the world we can all access (”wooden”), versus a speaker’s subjective construal (”pretty”). Previousstudies have obtained empirical estimates of phrases’ subjectivity by asking adults to rate how faultless a disagreementover that phrase would be (”Could both speakers be right?”). Where does this underlying dimension of subjectivity comefrom? We show that adults’ gradient judgments of faultless disagreement are systematically related to their estimates ofpopulation-level consensus (”Out of 100 people, how many would say this is a ’pretty shirt’?”) over utterance-referentpairs, but that the strength of that relation varies based on semantic class: estimated levels of consensus matter less forphrases with value adjectives, like ”pretty shirt.” Follow-ups will investigate simulating consensus as a potential develop-mental mechanism for inferring subjectivity.