Bayesian phylolinguistics infers the internal structure and the time-depth of the Turkic language family

Abstract

Despite more than 200 years of research, the internal structure of the Turkic language family remains subject to debate. Classifications of Turkic so far are based on both classical historical–comparative linguistic and distance-based quantitative approaches. Although these studies yield an internal structure of the Turkic family, they cannot give us an understanding of the statistical robustness of the proposed branches, nor are they capable of reliably inferring absolute divergence dates, without assuming constant rates of change. Here we use computational Bayesian phylogenetic methods to build a phylogeny of the Turkic languages, express the reliability of the proposed branches in terms of probability, and estimate the time-depth of the family within credibility intervals. To this end, we collect a new dataset of 254 basic vocabulary items for thirty-two Turkic language varieties based on the recently introduced Leipzig–Jakarta list. Our application of Bayesian phylogenetic inference on lexical data of the Turkic languages is unprecedented. The resulting phylogenetic tree supports a binary structure for Turkic and replicates most of the conventional sub-branches in the Common Turkic branch. We calculate the robustness of the inferences for subgroups and individual languages whose position in the tree seems to be debatable. We infer the time-depth of the Turkic family at around 2100 years before present, thus providing a reliable quantitative basis for previous estimates based on classical historical linguistics and lexicostatistics.

Biocultural diversity of common walnut (Juglans regia L.) and sweet chestnut (Castanea sativa Mill.) across Eurasia

“…Linguistic evolution of the six major cognate sets ( *KVrV , *ŋuńV‐ , *a‐/an‐gōza , *HwV́rƛ̣V , *ḳV̆rḳV , *ṭVɫV ) referred to walnut across Eurasia. The spatial and time distribution of walnut terms was derived using the consensus language trees with approximate estimation of the divergence times computed between languages of a Dravidian family (Kolipakam et al., 2018), b Altaic phylum‐Turkic family (Mikic et al., 2011; Savelyev & Robbeets, 2020), c Uralic phylum‐Finno‐Ugric family (Honkola et al., 2013), d Indo‐European phylum (Chang, Cathcart, Hall, & Garrett, 2015, in accordance with steppe theory), and e Kartvelian phylum (Koryakov, 2002) included into the putative Eurasiatic macro‐phylum as proposed by Pagel, Atkinson, Calude, and Meade (2013). f Afro‐asiatic phylum‐Semitic family (Kitchen et al, 2009), and g the putative Dene‐Sino‐Caucasian macro‐phylum (Van Driem, 2008) including Basque (Valdiosera et al., 2018), Burushaski, North‐Caucasian phylum (Koryakov, 2002), and Sino‐Tibetan phylum (Sagart et al, 2019) were also included in the reconstruction…”

Triangulation supports agricultural spread of the Transeurasian languages

“…44 ) lists (Supplementary Data 2). The Turkic and Tungusic basic vocabulary included is based on a revision of recently published datasets 45,46 . Cognate coding is supported by an inventory of basic vocabulary etymologies and sound correspondences across the Transeurasian languages presented in Supplementary Data 2.…”

Phylogenetic signal and rate of evolutionary change in language structures

“…The combination of reconstructed states and the rate of change of particular features can allow further research to contextualize the rates in time, if there is enough information on the age of the proto-language. For example, the age of Proto-Turkic was estimated to be around 2100 years before present [ 5 ]. In the first step, one extracts the features reconstructed as ‘present’ in Proto–Turkic (95% probability of being ‘present’ or higher).…”