Multiagent Bayesian Iterated Learning Model Revealing the Necessary Conditions for Peripheral Distribution of Dialect

Abstract

Languages evolve through diffusion processes similarly to biological evolution, forming linguistic clusters based on geographic proximity. A few mathematical modeling studies have tested the classical theory on the formation of a peripheral distribution of dialects, which originally assumed two conditions: (i) new words are innovated exclusively at a cultural center and (ii) these words spread outward due to the prestige of the cultural center. However, it is known that these special conditions are often not met. We examined whether and how the presence or absence of each condition influences the outcome using an extended Bayesian Iterated Learning Model. Our mathematical analyses revealed that peripheral distributions can emerge not only when both conditions are present but also when one of the two is absent. Furthermore, the satisfaction of one or both conditions in a population can be predicted by investigating the word age distribution there.