The story of Na-Dene
01
From 'Athapascas' to Na-Dene
The Athabaskan core of the family was recognised early. In 1836 Albert Gallatin, working from vocabularies gathered by fur traders and explorers, grouped the languages of the northern interior under the name 'Athapascas', after Lake Athabasca. By the 1850s scholars had noticed that the Navajo and Apache vocabularies collected in the Southwest matched these northern lists closely, and later in the century the Hupa and other Athabaskan languages of California and Oregon were added. The resemblances were so plain, in basic words and in the elaborate verb, that nobody seriously doubted them. The French missionary Émile Petitot published a large comparative dictionary of the northern languages in 1876, and John Wesley Powell's classification of 1891 recognised 'Athapascan' as one of the families of North America.
The wider family was a harder case. In 1915 Edward Sapir argued that Tlingit and Haida were distant relatives of Athabaskan, pointing to the shared order of verb prefixes and to a scatter of similar stems. His colleague Pliny Earle Goddard, the leading Athabaskanist of the day, was unconvinced, and published his doubts in 1920. The matter rested for decades on Sapir's authority rather than on worked-out proof.
The proof came in stages. Eyak, barely known until Frederica de Laguna visited the Copper River delta in the 1930s, was shown by Michael Krauss in the 1960s to be the sister of Athabaskan as a whole, which gave a second witness against which to test Tlingit. Krauss and Robert Levine then argued that Haida's resemblances were borrowings or chance, and most specialists now leave it out. Later, Jeff Leer worked out regular sound correspondences between Tlingit and Athabaskan-Eyak, in verb prefixes as well as stems. Since then the relationship of the three branches has been generally accepted, though it is a deep one.
02
Homeland and Spread
The Athabaskan languages are most varied in interior Alaska, the Yukon and northern British Columbia. By the principle Sapir set out in 1916, that a family's homeland is most likely where its branches are most diverse, Proto-Athabaskan was spoken somewhere in that region, perhaps two thousand years ago or a little more. Eyak and Tlingit, both spoken on the Gulf of Alaska coast, suggest that the older Proto-Na-Dene community lived on that coast or in the interior behind it, but its date is unknown; most specialists think in terms of several thousand years.
The southern outliers are the clearest evidence of movement. In 1936 Sapir showed that some Navajo words make most sense as northern vocabulary adapted to a new environment, a case of linguistic palaeontology for a migration rather than a homeland. The Pacific Coast languages reached Oregon and California about a thousand years ago, and the Apachean languages the Southwest around 1300 to 1500 CE. The route of the Apachean ancestors is disputed: some archaeologists see their traces in the Great Basin, including perishable finds from caves near the Great Salt Lake, while others favour a route along the western Plains.
Archaeology in the north is difficult, because Athabaskan material culture was largely made of wood, bark and hide, and linking particular sites to a language is rarely secure. Genetics has added a new line of evidence. In 2012 David Reich and colleagues found that Na-Dene-speaking groups carry ancestry from a second stream of migration from Asia that most other Native Americans lack, and a 2019 study by Pavel Flegontov and colleagues linked that ancestry to Palaeo-Eskimo populations. How these genetic events relate to the spread of the languages is still an open question, since genes and languages need not travel together.
03
Prefixes, Classifiers and Shapes
What strikes a learner first in any Na-Dene language is the verb. Almost everything grammatical is expressed by prefixes, stacked in a fixed order before a verb stem that comes last in the word. A single verb can mark subject, object, aspect, mode, number and adverbial meanings such as 'up', 'back' or 'into the fire'. Suffixes are few. Next to the stem sits a small prefix that linguists call the 'classifier', which in fact adjusts the verb's voice and valency, for example making a verb causative, passive or reflexive. The classifier is found in all three branches and is one of the strongest pieces of evidence that they are related.
Athabaskan languages are also known for their classificatory verbs. Navajo, for example, has about a dozen different stems for 'handle, give, carry' or 'lie', depending on the shape and nature of the object: one for a solid roundish thing such as a stone, another for a slender rigid thing such as a stick, others for flat flexible things such as cloth, for contents of an open container, and for a living being.
The sound systems are large and consonant-heavy. Most languages have ejective consonants, a voiceless lateral ł and lateral affricates written tł and tłʼ, and Tlingit and Eyak have a full uvular series. Vowel systems are small. Tone developed after the family broke up: in Athabaskan it grew out of an earlier glottal 'constriction', and Tlingit acquired its tones separately. Word order is typically subject–object–verb, with postpositions rather than prepositions. Nouns take possessive prefixes: Navajo shikeeʼ 'my foot' but bikeeʼ 'his or her foot', and Hupa whixeʼ 'my foot'. All the languages are now written in Latin-based practical alphabets, and some northern communities still use syllabics adapted from Cree.
04
Open Questions
The first question is the family's edge. Joseph Greenberg's 1987 classification of the Americas kept Haida in Na-Dene and treated Na-Dene as one of three great stocks of the continent. Specialists in the languages, including Krauss, Levine and Leer, concluded that the Haida evidence does not hold up, and Haida is now usually classed as an isolate. A few scholars still argue for a remote connection.
The second is the Dene-Yeniseian hypothesis. Edward Vajda's proposal, presented at a symposium in Fairbanks in 2008 and published with responses in 2010, links Na-Dene with the Yeniseian languages of Siberia through parallels in the verb prefixes and a set of proposed cognate stems. It drew serious attention, and some historical linguists regard it as the most promising long-distance proposal involving the Americas. Others, notably Lyle Campbell, judged the sound correspondences too few and the semantic matches too loose. Michael Fortescue and Vajda set out a fuller case in 2022, and the debate continues. Earlier and broader 'Dene-Caucasian' proposals, which would add Sino-Tibetan, Basque and the languages of the Caucasus, are rejected by most linguists.
Inside the family, the Northern Athabaskan languages resist a tree: features spread across neighbouring languages so freely that clean subgroups are hard to draw, and it is unsettled from which part of the north the Pacific Coast and Apachean groups came. Tlingit poses its own puzzle. Its correspondences with Athabaskan-Eyak are regular in some parts of the vocabulary but irregular in others, and Leer suggested that Tlingit may have absorbed a second, unrelated language in its early history. How old Proto-Na-Dene is, and whether its speakers came from Asia after the other ancestors of Native Americans, depends partly on how these questions are resolved.