Detecting contact in language trees: a Bayesian phylogenetic model with horizontal transfer

This paper introduces contacTrees, a Bayesian phylogenetic model designed to integrate language contact and horizontal transfer into the study of language evolution. The model efficiently infers both the phylogenetic tree of a language family and contact events between its clades. Simulation studies demonstrate contacTrees' accuracy in reconstructing language family history, including contact events, and highlight the systematic errors that can arise from ignoring contact. A case study on Indo-European languages confirms contacTrees' ability to identify known contact events and loanwords, providing a statistically superior fit to the data compared to traditional phylogenetic methods. The improved model significantly refines estimates of tree height, bringing them closer to established estimates. This work bridges a gap between theoretical and empirical models of cultural evolution, offering crucial insights, especially for less-documented language families.

Nico Neureiter, Peter Ranacher, Nour Efrat-Kowalsky, Gereon A. Kaiping, Robert Weibel, Paul Widmer, Remco R. Bouckaert