Testing methods of linguistic homeland detection using synthetic data

Abstract

Two families of quantitative methods have been used to infer geographical homelands of language families: Bayesian phylogeography and the ‘diversity method'. Bayesian methods model how populations may have moved using a phylogenetic tree as a backbone, while the diversity method assumes that the geographical area where linguistic diversity is highest likely corresponds to the homeland. No systematic tests of the performances of the different methods in a linguistic context have so far been published. Here, we carry out performance testing by simulating language families, including branching structures and word lists, along with speaker populations moving in space. We test six different methods: two versions of BayesTraits; the relaxed random walk model of BEAST 2; our own RevBayes implementations of a fixed rate and a variable rates random walk model; and the diversity method. As a result of the tests, we propose a hierarchy of performance of the different methods. Factors such as geographical idiosyncrasies, incomplete sampling, tree imbalance and small family sizes all have a negative impact on performance, but mostly across the board, the performance hierarchy generally being impervious to such factors. This article is part of the theme issue ‘Reconstructing prehistoric languages'.

Inferring language dispersal patterns with velocity field estimation

“…To validate the feasibility of our LVF, we sourced 1000 simulated datasets from Wichmann and Rama’s work 44 (see details in Supplementary Notes section 2 ). Each dataset consisted of 20 simulated language samples with 306 binary-coded linguistic traits generated by a given phylogenetic tree.…”

“…Apart from the phylogeographic approach, we also compared the LVF to the other three phylogeny-free approaches. They are the diversity (DIV), centroid (Centr), and minimal distance (MD) approaches 44 , 55 (see details in Methods). These approaches rest upon completely distinct theoretical foundations from the LVF and phylogeographic approach.…”

“…It resembles the geographic projection of the phylogenetic tree that facilitates outlining the language dispersal trajectories in geographic space. In simulated validations, we verified the computational effectiveness and robustness of our LVF using 1000 simulated datasets where the dispersal patterns are given 44 . Specifically, the effectiveness of the LVF was validated based on the difference between estimated and given patterns, and its robustness was evaluated under different parametric settings.…”

Cross-linguistic conditions on word length

“…For the purpose of reducing the extension of a genus [17] to a manageable single location, we have chosen as a representative location that of the language within the genus having the smallest average distance to the other languages in the genus (with the choice being made arbitrarily when only two languages were available or in other cases of ties). This minimal-distance location has been shown to give a surprisingly good approximation to a language family homeland [18].…”

Proceedings of the 5th Workshop on Research in Computational Linguistic Typology and Multilingual NLP

“…Trimmed alignments could either form the basis of a more extensive annotation of phonetic alignments in a computer-assisted setting (List, 2017b), or they could serve as the basis of extensive cross-linguistic, typologically oriented studies devoted to the regularity of sound change and sound correspondence patterns. For example, correspondence patterns have already been used in typological studies investigating the history of pronoun systems in South America (Rojas-Berscia and Roberts, 2020), or for studies with simulated data that use phonetic alignments to construct artificial cognate sets (Wichmann and Rama, 2021).…”