From species to languages : a phylogenetic approach to human prehistory

Abstract

Languages, like species, evolve. Just like biologists, historical linguists infer relationships between the lineages they study by analysing heritable features. For linguists, these features can be words, grammar and phonemes. This linguistic evidence of descent with modification plays an important role in our understanding of human prehistory. However, conventional methods in historical linguistics do not employ an explicit optimality criterion to evaluate evolutionary language trees. These methods cannot quantify uncertainty in the inferences nor provide an absolute chronology of divergence events. Previous attempts to estimate divergence times from lexical data using glottochronological methods have been heavily criticized, particularly for the assumption of constant rates of lexical replacement. Computational phylogenetic methods from biology can overcome these problems and allow divergence times to be estimated without the assumption of constant rates. Here these methods are applied to lexical data to test hypotheses about human prehistory. First, divergence time estimates for the age of the Indo-European language family are used to test between two competing theories of Indo- European origin - the Kurgan hypothesis and the Anatolian farming hypothesis. The resulting age estimates are consistent with the age range implied by the Anatolian farming theory. Validation exercises using different models, data sets and coding procedures, as well as the analysis of synthetic data, indicate these results are highly robust. Second, the same methodology was applied to Mayan lexical data to infer historical relationships and divergence times within the Mayan language family. The results highlight interesting uncertainties in Mayan language relationships and suggest that the family may be older than previously thought. Finally, returning to biology, similar tree-building and model validation techniques are used to draw inferences about human origins and dispersal from human mitochondrial DNA sequence data. These analyses support a human origin 150,000-250,000 years ago and reveal time dependency in rates of mitochondrial DNA evolution. Population size estimates generated using a coalescent approach suggest a twophase human population expansion from Africa. Potential correlations between human genetic and linguistic diversity are highlighted. I conclude that there is much to be gained by linguists and biologists using the same methods and speaking the same language.