All of these languages share a highly complex prefixing verb structure in which tense and mood markers are interdigitated between subject and object agreement markers. The morphological hallmark of the family is a series of prefixes found directly before the verb root that raise or lower the transitivity of the verb word. These prefixes, traditionally known as "classifiers", derive historically from a combination of three distinct classes of morphemes and are not found in any other Native American language family. The phoneme system contains a large number of dorsal consonants as well as a general absence of labial obstruents. In the historical phonology there is a widespread tendency, observable across many Athabaskan languages, for phonemic tonal distinctions to arise from glottal features originally found at the end of the syllable. The glottal features in question are often evident in Eyak or Tlingit. These languages are typologically unusual in containing extensive prefixation yet being SOV and postpositional, features normally associated with suffixing languages.
Proposals of deeper genealogical relations involving Athabaskan–Eyak–Tlingit
A genealogical connection between the Tlingit, Eyak and Athabaskan languages was suggested early in the 19th century, but not universally accepted until much later. Haida, with 15 fluent speakers, was originally linked to Tlingit by Franz Boas in 1894. Both Haida and Tlingit were then connected to Athabaskan by Edward Sapir in 1915. Linguists such as Lyle Campbell today consider the evidence inconclusive. They have classified Haida as a language isolate. In order to emphasise the exclusion of Haida, Campbell refers to the language family as Athabaskan–Eyak–Tlingit rather than Na-Dene. In 2010 Jeff Leer published extensive primary materials on what he calls PAET.
In 2008, Edward Vajda of Western Washington University presented evidence suggesting that the Na-Dene languages might be related to the Yeniseian languages of Siberia, the only living representative of which is the Ket language. Key evidence by current comparative methodologies includes homologies in verb prefixes and also a systematic correspondence between the distribution of Ket tones and consonant articulations found in Athabaskan–Eyak–Tlingit. Vajda's paper has been favorably reviewed by several experts on Na-Dene and Yeniseic languages, including Michael Krauss, Jeff Leer, James Kari, and Heinrich Werner, as well as a number of other well-known linguists, including Bernard Comrie, Johanna Nichols, Victor Golla, Michael Fortescue, and Eric Hamp. The conclusion of this seminar was that the comparison with Yeniseic data shows that Haida cannot be classified in a genealogical unit with Athabaskan–Eyak–Tlingit.
According to Joseph Greenberg's controversial classification of the languages of Native North America, Na-Dené is one of the three main groups of Native languages spoken in the Americas. Contemporary supporters of Greenberg's theory, such as Merritt Ruhlen, have suggested that the Na-Dené language family represents a distinct migration of people from Asia into the New World that occurred six to eight thousand years ago, placing it around four thousand years later than the previous migration into the Americas by Amerind speakers; this remains an unproven hypothesis. Ruhlen speculates that the Na-Dené speakers may have arrived in boats, initially settling near the Haida Gwaii, now in British Columbia, Canada. Linguist Edward Sapir suggested that the Sino-Tibetan languages are related to Na-Dené. A fringe hypothesis by Sergei Starostin suggested that Na-Dené may belong to the much broader Dené–Caucasian superfamily, which also contains the North Caucasian languages, Sino-Tibetan languages, and Yeniseian languages. This proposal is rejected by nearly all current linguists.
PAET, PAE and PA stand for Proto-Athabaskan–Eyak–Tlingit, Proto-Athabaskan–Eyak and Proto-Athabaskan, respectively.
To prevent cluttering the table, phonemes in the PAET, PAE and PA columns are not asterisked.
Leer doesn't reconstruct the PAET affricates, and. Judging from their rarity, he assumes they may be attributable to the resolution of former consonant clusters.
In Athabaskan and Eyak, sibilants can be diminutive variants of shibilants. In Tlingit, on the other hand, shibilants might sometimes be diminutive variants of sibilants. These correspondences are in parentheses.