Yang et al. 2024: Inferring language dispersal patterns with velocity field estimatio

[Jaska]

Sizhe Yang, Xiaoru Sun, Li Jin & Menghan Zhang 2024:
Inferring language dispersal patterns with velocity field estimation
https://www.nature.com/articles/s41467-023-44430-5

"Specifically, the inferred dispersal centre of Indo-European languages was located in the Fertile Crescent which is the earliest ancient agricultural homeland in the world (Fig. 2b). This observation favours the Anatolia origin hypothesis of Indo-European languages rather than the alternative competing hypothesis of Pontic steppe region origin."


Some initial observations:

1. They offer a model to replace the traditional “center of gravity” principle. However, this principle is only on of the many different methods to locate a linguistic homeland, and among the weakest ones at that. Attacking this one principle is not enough, because proto-languages are usually located primarily based on the comparison of words and concrete reality (paleo- and archaeo-linguistic evidence). These arguments cannot be overruled by applying this new velocity field approach.

2. They attack the tree model, because it lacks the ability to count for contacts between already diverged language lineages. However, linguists have also brains, not only family trees. Areal, contact-induced innovations are of course accounted for, and they can even be drawn in the family tree as lines between branches. Still, often they are not, because they are irrelevant for the taxonomic structure of the language family – and that is what family tree is used to illustrate. Nothing more. (An example: there have been long-lasting contacts between Saami and Finnic, but this can never blur their status as different branches of the Uralic language family. It is easy to distinguish between words descended from Proto-Uralic and words borrowed later from one branch to another. It only requires that we know the phonological history of these branches.)

Moreover, dialect continuum is no problem for the family tree: there is no law requiring always only binary branching. The examples they present, like Indo-Aryan and Slavic, are established Indo-European branches in which all the languages and dialects descend from Proto-Indo-Aryan and Proto-Slavic, respectively. This poses no problem for the family tree illustration: we just draw as many daughter branches as we can support. “The rake model” for some branches is easy to incorporate in the family tree illustration. And again, based on the phonological history, we still can support sub-branches even within Slavic or Indo-Aryan, because we can distinguish old inherited features from later contact-induced features.

3. Even the supplementary data apparently does not tell it explicitly, but it seems that their method tries to find the shortest route between the homeland and the modern location of the daughter languages. This presupposition obviously explains why they get the result that the Indo-European homeland was in Anatolia: it is located in the linear route between Europe and South Asia. However, we know very well that languages do not always spread via the shortest route but instead very randomly. For example, the route of the Indo-Iranian branch was actually spiral-like:
1. Pontic Steppe --> 2. Middle Dnieper Region --> 3. Volga-Kama Region --> 4. Southern Ural Region --> 5. Central Asia --> 6. South Asia and Near East.

We know this based on linguistic evidence and compared to archaeological and genetic evidence. No mathematical model based on a reach toward the most economical solution can even begin to challenge these results. Such a model can be instantly disproved on the ground of these scientific results and its a priori presupposition which apparently does not even allow other than very straight routes of expansion.

[Mythbuster General]

To me it's rather fascinating to see that when everyone thought the steppe origin of Indo-European languages was a done deal, in recent years the Middle Eastern origin hypothesis seems to gain new adherents in the scientific field. I welcome this, because it pushes scientists to refine even better and with more accuracy their positions.

[Jaska]

To me it's rather fascinating to see that when everyone thought the steppe origin of Indo-European languages was a done deal, in recent years the Middle Eastern origin hypothesis seems to gain new adherents in the scientific field. I welcome this, because it pushes scientists to refine even better and with more accuracy their positions.

True. But many of these challenging propositions are based on erroneous (like this one) or unreliable methods (like most wordlist-based calculations), and they do not even try to disprove the actual linguistic evidence supporting the steppe homeland. Unfortunately here more quantity means less quality.

Admittedly, Early Proto-Indo-European = Proto-Indo-Anatolian is a bit more open to interpretations, although Late Proto-Indo-European can hardly move away from the steppe any more. In any case, these methodologically weaker studies would have any value only if there were no unambiguous results achieved by reliable linguistic methods.