Permutation test applied to lexical reconstructions partially supports the Altaic linguistic macrofamily

Alexei S Kassian,George Starostin,Ilya M Egorov,Ekaterina S Logunova,Anna V Dybo
DOI: https://doi.org/10.1017/ehs.2021.28
2021-06-01
Abstract:In this paper, we present the results of our analysis of the 110-item basic wordlists for four reconstructed and one ancient languages, the linguistic ancestors of five language families which are hypothesized to constitute the Altaic (a.k.a. Transeurasian) macrofamily: Proto-Turkic, Proto-Mongolic, Proto-Tungusic, Middle Korean and Proto-Japonic wordlists. Protolanguage wordlists were reconstructed according to strict criteria of semantic reconstruction, based on accurate semantic glossing of forms in daughter languages. Each involved form was encoded into a bi-consonantal CC-shaped sequence using the consonant class method, after which a recently developed weighted permutation test was applied. In a typical situation, our algorithm makes a small number of type 1 errors (false positive), but the number of type 2 errors (false negative) can be substantial. Our main finding is that pairs between the Nuclear Altaic taxa - Turkic, Mongolic and Tungusic - as well as the Turkic-Japonic and Tungusic-Japonic pairs demonstrate significant p-values. In some cases, this can be attributed to either ancient contacts or genealogical relationships, but at least for the Turkic-Japonic pair, a contact scenario is unlikely owing to geographical remoteness.
What problem does this paper attempt to address?