Treating language contact as normal: Coalescent-theoretic modeling of the prehistory of the Bantu language family

The Bantu language family of sub-Saharan Africa is among the largest in the world by the number of languages, by geographical extent, and by the number of speakers. The expansion of the Bantu languages is an important example of large-scale language-family expansions in the history of humankind. To learn the early prehistory of the Bantu language family, one needs to disentangle the signal of the original splits and diversification from that of the subsequent language contact, known to be strong in the Bantu languages. We introduce a coalescent-theoretic model to computationally study the prehistory of the Bantu family. Our model treats language contact as a norm rather than a rare exception, in contrast to earlier computational work. Applying Approximate Bayesian Computation, we show the rates of both language change and language contact to have been so high within the Bantu family that the signal of the original diversification has been largely erased in the currently available Bantu lexical data.