LLF – Bât. ODG – 5e étage – Salle du conseil (533)
Seung Jin Hong and Jean-Pierre Koenig (Université de Paris)
Scalar predicates, categorization, and contraries
It is well-known that the lower bound of scalar properties such as tallness is both context-dependent and vague (Kennedy 2007). What is less well established is how that threshold is determined in language production or comprehension. It is often assumed it is determined on the basis of the height of members of the comparison class (Kennedy 2007) and some experimental work seems to confirm that hypothesis (Barner and Snedeker, 2008; Syrett et al., 2010). In this paper, we suggest on the basis of five experiments that, in fact, when confronted with sets of new objects that vary in other ways than mean or maximum height, speakers use percentiles of numbers of objects to set up the threshold of tallness and choose a threshold that minimizes categorization errors.