Scales, Salience and Referential Safety: The Benefit of Communicating the Extreme

On an abstract level, this paper tries to show how formal semantics can contribute to theories of language evolution and vice versa. More concretely, the paper argues that signaling games fail to account plausibly for a general preference to use gradable adjectives to communicate extreme values. The reason is that these models focus too narrowly on descriptive language use. Numerical simulations show that the choice of extreme values is pragmatically beneficial in situations of referential language use under possible noise. Bringing evolutionary modeling back to formal semantics, this yields a functional explanation for otherwise rather puzzling patterns in the use of gradable adjectives.