Proceedings of the Annual Meeting of the Cognitive Science Society

Language often has different ways to express the same or similar meanings. Despite this, however, people seem to have preferences for some ways over others. For example, people overwhelmingly prefer bread and butter to butter and bread. Previous research has demonstrated that these ordering preferences grow stronger with frequency (i.e., frequency-dependent preference extremity). In this paper we demonstrate that this frequency-dependent preference extremity can be accounted for by noisy-channel processing models (e.g., Gibson, Bergen, & Piantadosi, 2013; Levy, 2008). We also show that this preference extremity can only be accounted for if the listener infers more noise than the speaker produces. Finally, we show that the model can account for the language-wide distribution of binomial ordering preferences.