The role of world-knowledge in setting imprecise standards for absolute gradable adjectives: Probabilistic Modeling

Team: Petra Schumacher (InfoPer2, U Cologne), Barbara Tomaszewicz (InfoPer2, U Cologne), Michael Franke (Pro3, U Tübingen), Judith Degen (Pro3, Stanford University) and Daniel Lassiter (Stanford University)

Absolute gradable adjectives, such as ‘full’, differ from relative gradable adjectives, ‘big’, in that they may have context-independent interpretations (Rotstein & Winter 2004, Kennedy & McNally 2005, Syrett 2007, Kennedy 2007, Syrett et al. 2010, Toledo & Sassoon 2011, Lassiter & Goodman 2013, 2015, Burnett 2014, Qing & Franke 2014, Lassiter 2015). Theoretical and experimental work has suggested that while both kinds of adjectives are context-sensitive, in the case of absolute adjectives only imprecise standards (i.e. deviating from a certain standard threshold, as in ‘full for a whiskey glass’) require context, whereas precise standards are always acceptable irrespective of context. While prior experiments concentrated on identifying the differences in context-sensitivity between relative and absolute adjectives, we focus on absolute adjectives and the mechanism for setting an imprecise standard for comparison. Two experiments within the InfoPer2 project, a picture-sentence verification task and an EEG study, explore how the interpretive mechanism for absolute adjective incorporates world knowledge. In collaboration with the Pro3 project and the external collaborator Daniel Lassiter, a model of probabilistic reasoning about the degree of contextually adequate imprecision will be developed to explain the judgment data. The model’s ability to capture which categorical truth-value decisions associate with the use of world knowledge and which do not will then be tested on the online processing data.