Horizontal transfer of linguistic features as an integral part of the evolutionary history of language families - Phylogenetic reconstruction of linguistic tree-based networks

• Applicant: Dr. Lina Herbst
• Subject Area: Applied Linguistics, Computational Linguistics
• Term: since 2021
• Project identifier: Deutsche Forschungsgemeinschaft (DFG) - Project number 461279234

Today more than 7000 different languages from more than 200 language families are spoken worldwide. The evolutionary relationships within a language family are often illustrated and studied using a linguistic phylogeny, or language family tree. Based on linguistic features this allows inference of the internal structure and the time-depth of a language family. However, linguistic features are not always “inherited”, they can also be acquired horizontally through language contact.Non-treelike evolution is also common in evolutionary biology. Scholars have represented these horizontal relationships within phylogenetic networks. Broadly speaking, a phylogenetic network is a tree with horizontal edges and can provide a more appropriate representation of language relationships. By allowing linguistic features to move horizontally, a better and more accurate representation is possible. However, in historical linguistics network reconstructions are still in their infancy. While great progress has been made in synthesizing vertical family genealogical models and contact-induced change, no standard has been established so far. Some approaches amount to a mere visualisation of the data, as they only display the conflicts in the data while not indicating the actual transferred features. More advanced approaches fail to infer the timing of branching events, i.e. of the divergence dates of subgroups and the origin of the language family. For that, rooted and timed phylogenetic networks need to be inferred.The proposed project aims to incorporate horizontal transfer into phylogenetic models to infer rooted and timed linguistic networks. These networks will be based on a birth-death process, which will be adapted to account for horizontal transfer events. The resulting birth-death-transfer process will be integrated in a Bayesian phylogenetic framework, such that knowledge about the past can be included e.g. on the depth of a sub-family.This approach will be validated with a thorough simulation study and application to languages for which transfers are agreed upon by linguistic experts. After successful validation, the process will be applied to the Polynesian language family, which has been suggested to have a highly reticulate history with many linguistic transfers. We will detect the timing and location (within the phylogeny) of linguistic transfers among the Polynesian languages and identify which linguistic features were transferred. We will further evaluate and quantify the susceptibility to transfer of different aspects of language from the grammar to the lexicon.Implemented as a package within the open access, cross-platform BEAST2 software framework and accompanied by a hands-on tutorial, the method will be readily applicable by other scholars.

• DFG Programme: WBP Position