A pragmatic theory of generic language
share
Generalizations about categories are central to human understanding, and generic language (e.g. "Dogs bark.") provides a simple and ubiquitous way to communicate these generalizations. Yet the meaning of generic language is philosophically puzzling and has resisted precise formalization. We explore the idea that the core meaning of a generic sentence is simple but underspecified, and that general principles of pragmatic reasoning are responsible for establishing the precise meaning in context. Building on recent probabilistic models of language understanding, we provide a formal model for the evaluation and comprehension of generic sentences. This model explains the puzzling flexibility in usage of generics in terms of diverse prior beliefs about properties. We elicit these priors experimentally and show that the resulting model predictions explain almost all of the variance in human judgments for both common and novel generics. We probe the theory in more detail, and find that generic language depends in a fundamental way on subjective beliefs, not mere frequency. This theory provides the mathematical bridge between the words we use and the concepts they describe.
READ FULL TEXT