Max Planck linguists settled the Indo-European origin debate with a 161-language tree

For over 200 years, where Indo-European — the family behind nearly half the world's languages — began was disputed. Two theories dominated: the Steppe hypothesis, placing the origin in the Pontic-Caspian Steppe about 6,000 years ago, and the Anatolian or farming hypothesis, tying it to early agriculture about 9,000 years ago. Previous phylogenetic analyses reached conflicting ages because their datasets were inaccurate and inconsistent and their methods handled ancient languages poorly.

Researchers at the Max Planck Institute for Evolutionary Anthropology assembled an international team of more than 80 language specialists to build a new dataset of core vocabulary from 161 Indo-European languages, including 52 ancient or historical ones, with rigorous protocols for coding lexical data. They then ran recently developed ancestry-enabled Bayesian phylogenetic analysis that could test whether ancient written languages such as Classical Latin and Vedic Sanskrit were direct ancestors of modern Romance and Indic branches rather than just modern leaves on the tree.

The analysis estimated the family to be about 8,100 years old, with its five main branches already split by around 7,000 years ago, and the chronology held across a wide range of alternative models and sensitivity analyses. Tree topology and split dates pointed to an ultimate homeland south of the Caucasus, with one migration reaching the Pontic-Caspian and Forest Steppe around 7,000 years ago and later expansions entering Europe with Yamnaya and Corded Ware-associated cultures around 5,000 years ago.

Neither the Steppe nor the farming hypothesis was entirely right; the two became stages of one history. Ancient DNA showed the Anatolian branch emerged not from the Steppe but from near the northern Fertile Crescent, and the language tree matched. The team published the hybrid hypothesis in Science in 2023, offering it as a step from mutually exclusive scenarios toward a model integrating linguistic, archaeological and genetic evidence.

Fix the data, include the ancestors, and one tree decides

Earlier analyses disagreed mainly because input data were inconsistent and coding was sloppy, so rebuilding the wordlist under strict protocols removed the source of the conflict.

Including 52 ancient or historical languages let the model test real ancestry — Latin as ancestor of Romance, Vedic Sanskrit of Indic — instead of making modern languages carry the whole story.

Balanced sampling across the family stopped a few well-studied modern branches from dominating the tree's shape and age.

Ancestry-enabled Bayesian analysis put ancient written languages in their true positions, and sensitivity analyses showed the 8,100-year date was robust rather than an artifact of one model.

Cross-checking the tree against ancient DNA turned two mutually exclusive theories into one hybrid migration route supported by linguistics, archaeology and genetics.

When rival theories produce contradictory models, fix the shared inputs: balanced data, rigorous coding, and treating ancient forms as ancestors collapsed a 200-year dispute into one robust tree.

The results were announced on July 27, 2023 and published in Science, credited to P. Heggarty and colleagues. The release drew a Hacker News thread on July 30, 2023 with 233 points and 147 comments. Senior author Russell Gray called the hybrid hypothesis the resolution of the 200-year enigma, and Wolfgang Haak framed it as a step toward a model integrating archaeological, anthropological and genetic findings rather than a finished answer.