Ket (ISO 639-3 ket) is spoken along the middle Yenisei River in Russia's Krasnoyarsk Krai, and native speakers now remain only in the villages of Kellog, Surgutikha and Maduyka. In 1926, 1,225 Kets spoke the language natively. By 1989 the figure was 537. Russia's 2020 census, carried out in 2021, counted 1,088 ethnic Kets, and 153 of them reported speaking Ket. Reports from field linguists put fluent speakers below 30 by 2024. Ket has prefixing verbs with noun incorporation and SOV word order, and it has neither /p/ nor /ɡ/. Descriptions of its tone range from no tone at all to eight contrastive tones.
Ket is the last living member of the Yeniseian family. Its closest relative, Yugh, died out in the 1970s. Kott, Arin, Assan and Pumpokol, which were spoken further south and recorded in 18th- and 19th-century wordlists, had disappeared earlier still. Before 2008 the family had no demonstrated relatives anywhere, and proposals linking it to Burushaski, Sino-Tibetan or North Caucasian never found general acceptance.
The 2008 proposal
At a symposium at the University of Alaska Fairbanks on 26–27 February 2008, Edward Vajda of Western Washington University presented about ten years of work. He argued that Yeniseian is related to Na-Dene: Tlingit, the extinct Eyak, and the Athabaskan family, which runs from Koyukon and Gwich'in in Alaska to Navajo and Apache in the US Southwest. His evidence was mainly morphological: about ten matching prefix and suffix positions in the Ket and Proto-Na-Dene verb templates, plus more than 110 proposed cognate morphemes and a set of sound correspondences. One correspondence explains Yeniseian tone: in his account, Ket's tonal contrasts arose from glottalized consonants at the end of the syllable, and Na-Dene languages still have such consonants.
Michael Krauss, Jeff Leer and James Kari, all Na-Dene specialists, responded favourably. So did the historical linguists Eric Hamp and Johanna Nichols, and Bernard Comrie endorsed the method. The symposium summary also placed Haida outside the family, although Haida had sometimes been grouped with Na-Dene. The papers were published in 2010 as The Dene–Yeniseian Connection, edited by Kari and Ben Potter (Anthropological Papers of the University of Alaska, new series 5).
| Ket | IPA | Meaning |
|---|---|---|
| уль | ʊʎ | water |
| сюль | ɕʊʎ | blood |
| боʼк | bɔʔk | fire |
| кеʼт | kɛʔt | person |
| тыʼсь | tɨʔɕ | stone |
| аʼт | aʔt | bone |
| ӄусь | qʊɕ | one (inanimate) |
| ӄак | qak | five |
The objections
The symposium's press release was headlined "Linguists demonstrate Siberian-North American link". The specialists who reviewed the 2010 volume did not treat the question as settled. Lyle Campbell, writing in the International Journal of American Linguistics in 2011, concluded that the connection could not yet be accepted, and he has stayed doubtful since. In 2012 George Starostin examined the data in the Journal of Language Relationship and found that many of Vajda's correspondences were not regular in the strict comparative sense, but he concluded that something in the comparison should not be abandoned. In his reply in the same issue, Vajda described Dene–Yeniseian as a hypothesis that the families are related, not yet a worked-out family tree.
Genetics does not decide the question. Most Ket men sampled in genetic studies carry Y-chromosome haplogroup Q, which is also common among Native Americans. That shows a shared ancestry between populations, but a population can change languages, so it is not evidence about the language family.
Which way did the migration go?
In 2014 Mark Sicoli and Gary Holton ran a Bayesian phylogenetic analysis in PLoS ONE. They used Ket, Kott, 37 Na-Dene languages and Haida as an outgroup, coded for 116 phonological and morphological features, of which 90 were informative. Models that set Yeniseian apart from a Na-Dene clade fit the data significantly worse. From this they argued for a spread out of Beringia, with the ancestors of the Kets moving back into Siberia. In 2020 Igor Yanovich argued that the study was methodologically flawed: its typological data were too few to produce a tree that did not depend on the model's prior assumptions, so the direction of migration does not follow from it. If the hypothesis holds, it links languages a long way apart. Measured between the map's points, Ket is about 4,900 km from Koyukon and about 8,900 km from Navajo.