Lexicostatistical classification of the Nubian languages and the issue of the Nile-Nubian genetic unity
Mikhail Vasilyev (Russian Presidential Academy, mvhumanity@gmail.com); George Starostin (Russian Presidential Academy of National Economy and Public Administration, mvhumanity@gmail.com)
Journal of Language Relationship, № 12, 2014 - p.51-72
The paper focuses on a statistical anomaly found within the lexicostatistical matrix for the Nubian group of languages. Earlier, the discovery of this anomaly by M. Bechhaus-Gerst has led to a radical revision of the established classification for this group, in which the original “Nile-Nubian” branch, including Kenuzi-Dongolawi and Nobiin, was dissolved, with Nobiin re-defined as one of the two primary branches of Nubian. Here we apply two different phylogenetic methods to the lexicostatistical data of Nubian languages: the first one is a slightly modified variant of the UPGMA algorithm as used in the StarLing software package, whose results are in better agreement with the traditional classification, and the second is a statistical method of “least average deviations”, which yields the same classification as that of M. Bechhaus-Gerst. Subsequent comparative-historical analysis of the situation explains the apparent controversy between the resulting trees and proves the importance of additional etymological argumentation in resolving the observed contradictions.
Keywords: Nubian languages, lexicostatistics, phylogenetic methods, method of least average deviations, substrate theory