International audience
In this presentation, I will show the results of a small-scale cluster analysis performed on agroup of 10 related Bantu languages. Bantu languages are spoken from Cameroon to Kenya inthe North, all the way to South Africa, covering a big part of the African continent. They aregrouped together into zones, indicated with a capital letter, and subgroups, indicated with two-digit numbers (Guthrie 1971). This reference system is based on geography and does not makeclaims about genealogical relations between languages (updates to the original system havemaintained this referential character, see e.g. Maho 2012). However, languages within asubgroup can actually be related, as is the case with the languages of our study.The subgroup that we are concerned with here is A80, spoken at the far North-Westernend of the Bantu area (Cameroon and neighboring countries); more specifically, the 10languages in this group for which enough data is available (Cheucle 2014, Grimm 2021,Vermeir & Allassonnière-Tang, forthcoming). Even though these 10 languages (Bekwel,Gyeli, Kol, Koonzime, Kwasio, Makaa, Mpiemo, Njyem, Shiwa and Sso) are clearly related,as becomes apparent when studying the vocabulary, it is also clearly possible to add depth inthe analysis by identifying subgroups of languages that are more closely related. The numberof subgroups and how the 10 languages are divided amongst them optimally can be calculatedusing cluster analysis. Our analysis is based on a list of 556 concepts for which at least sixlanguages of the corpus have a unique translation, providing a more in-depth study than theones cited by Cheucle (2014, pp. 17-27). We used the Levinshtein method to calculate thephonological distance matrix between words, resulting in a hierarchical clustering. The resultsshow a 0.8 positive correlation with the hierarchy proposed on Glottolog (Hammarström et al.,2024), with only slight differences in the lower subgroups. Subgroups clearly share distincttypological features, such as complete affrication (Gyeli, Kwasio, Shiwa, Sso), syncope(Koonzime, Njyem), and complex codas (Kol, Makaa).