The effect of structure in population and grammar space on language divergence.

摘要

This dissertation explores parallels between selection and evolution of Mendelian populations to that of language systems. The population dynamics of language learners are addressed as an instantiation of the general case of agents learning from one another. I found that population structure, language structure, and grammar network structure effect the evolutionary dynamics and influence the degree of linguistic coherence. Further, under some types of grammar networks, populations exhibit "weak cohesion" where the distribution of grammar frequencies forms a roughly Gaussian shape, similar to a quasi-species in models of molecular evolution. Weak cohesion is also characterized by a large amount of standing variation within the population which implies continued evolutionary potential. Positive frequency-dependent selection (FDS) has been of limited interest due to its obvious dynamics leading to monomorphism. However, investigations in the field of language evolution reveal that a population may exhibit standing variation even under the positive FDS if population or grammar space is structured. Recently, researchers have been considering the applicability of adaptive language to adaptive sensor networks in a heterogeneous environments. An evolutionary strategy where nodes adopt successful schemes from their neighbors with a fitness bonus for agreement is a general option with great promise. Such a system maps directly onto the systems I present.;The organization of this thesis is as follows: Chapter 1 provides the background of related research and the objectives of my thesis. The effect of population structure in the evolutionary dynamics of language is presented in chapter 2, which is published in Artificial Life and Robotics 8. Chapter 3 deals with language structure and its impact on language evolution; this material is published in Proceedings of the VIIIth European Conference on Artificial Life. Chapter 4 investigates the effects of grammar network structure in the evolutionary dynamics of language. Parts of this chapter have been submitted to Adaptive Behavior.