Visual ambiguity predicts adjective ordering preferences in referential contexts

Abstract

Recent work proposes subjectivity as a possible underlying mechanism for English’s canonical adjective order, predicting that ordering adjectives with respect to decreasing subjectivity maximizes communicative success between interlocutors. The present studies aim to contribute to our understanding of the underlying factors guiding adjective order, using two sentence judgment tasks to gather speakers’ intuitions about how visual traits might influence adjective ordering preferences. In both studies, participants saw two types of questions: for description-choice items, they chose a description from two provided options to identify a target object (a cake); for image-choice items, they used a given description to choose between two images of cakes. In the second study, context was manipulated such that participants were instructed to imagine they were working with a partner. For both description- and image-choice questions, the images contained some combination of clear and ambiguous traits: cakes were either clearly tall or short and ambiguously fancy, or clearly fancy and of medium height. If participants are gauging the subjectivity of a trait by relying on visual cues and then using this subjectivity to build descriptions that maximize communicative success, they should prefer the clearer, more objective trait in the second position, closer to the noun “cake.” In both studies, we find that participants prefer descriptions in which the clearer trait is ordered first (furthest from the noun). This suggests that while adjective order can shift systematically based on the subjectivity of visual traits, deviation from the attested order is not consistent with the subjectivity hypothesis’ predictions.