Inferring the drivers of language change using spatial models

Abstract Discovering and quantifying the drivers of language change is a major challenge. Hypotheses about causal factors proliferate, but are difficult to rigorously test. Here we ask simple question: can 20th century changes in English be explained as consequence spatial diffusion, or have other processes created bias favour certain linguistic forms? Using two most comprehensive datasets available, which measure state at beginning end century, calibrate model so that, initialised with early state, it evolves into later. Our calibrations reveal that while some by diffusion alone, others clearly result substantial asymmetries between variants. We discuss origins these and, by-product, generate full spatio–temporal prediction for evolution features over future.