Topic description / abstract:
The Bantu language family, covering most of subsaharan Africa and comprised of about 500 languages, is notorious for featuring enormous horizontal flow of linguistic features. This makes it difficult to reconstruct the evolutionary tree of the family, representing faithfully the early history of the different large groups of the Bantu. Computational phylogenetic methods are regularly employed in linguistics, including for the Bantu, but they do not allow us to properly account for horizontal flows. Of course, precise inference becomes extremely difficult when horizontal flows are factored in. In this project in progress, we build a simplified model of the Bantu expansion with horizontal flow, and study whether we can decide between different major historical scenarios using Approximate Bayesian Computation. We use both linguistic and genetic evidence, planning to combine them in a joint language-and-genes model to increase the statistical power. We find it important to not just attempt to answer the "Bantu question", but also to learn the practical limits for statistical inference from linguistic data in our inference framework in general, conditional on the size of data realistic to obtain, the possibility of errors in data collection, and a reasonable treatment of horizontal flow.
(work by Silvia Ghirotto^1, Patricia Santos^1, Andrea Benazzo^1, and Igor Yanovich^2;
^1: University of Ferrara, ^2: Tübingen University)
Igor Yanovich holds a PhD in theoretical linguistics from the Massachusetts Institute of Technology (2013). He has worked both in "classical" and computational historical linguistics, and is currently engaged in several interdisciplinary projects aiming to uncover the human past, and involving statisticians, geneticists, anthropologists and linguists. Igor is currently the leader of a DFG-funded junior research group at the University of Tübingen. He held postdoctoral fellowships at the Alexander-von-Humboldt foundation, the Philosophy department at Carnegie Mellon University, and the DFG Center for Advanced Study "Words, Bones, Genes and Tools".