Research Focus Area

Evolution of Human Languages

The Dene-Caucasian Hypothesis

John Bengtson, Association for the Study of Language In Prehistory
George Starostin, Russian State University for the Humanities

This collaborative project deals with the development and testing of the Dene-Caucasian (also called Sino-Caucasian) hypothesis on a deep genetic connection between certain language families and language "isolates" of Eurasia and North America.

The hypothesis was originally formulated by Sergei Starostin in the 1980s: using the traditional comparative method, he had made a serious attempt at demonstrating that the North Caucasian, Sino-Tibetan, and Yeniseian language families should all be grouped into one single macrofamily that he called Sino-Caucasian. Existence of such a deep taxon, or, at least, parts of it, had already been proposed earlier by some adventurous linguists (such as Alfredo Trombetti, Karl Bouda, Edward Sapir and others), but it took Starostin's work to scientifically demonstrate that these families shared a significant amount of basic vocabulary, and that these lexical comparisons could not be explained away as chance resemblances because they were connected by systematic phonological correspondences ("sound laws") between and among the three families.

A detailed description of Sino-Caucasian historical phonology and a large accompanying glossary have been completed by Sergei Starostin not long before his premature demise in 2005; they remain formally unpublished, but both manuscripts are available online. Today, Starostin's research is continued by his son, George Starostin, whose experience in comparative Yeniseian and Sino-Tibetan linguistics (particularly Old Chinese) allows him to verify and modify elements of the proposed protolanguage system with enough competence.

Also around the late 1980s - early 1990s, Sergei Starostin's colleague, Sergei Nikolayev, made a similar comparison between the Na-Dene family (as proposed by E. Sapir) and Starostin's Sino-Caucasian macrofamily. After this the term "Dene-Caucasian" came into use to describe the expanded taxon. In the late 1980's American linguist John Bengtson also began to work on the Dene-Caucasian hypothesis, focusing mainly on Na-Dene, Burushaski, and Basque. His early work on Basque generated a good deal of discussion, much of which was published in the new journal Mother Tongue (1995). Since then, he has worked towards establishing a much more rigorous system of correspondences that ties in Basque with the rest of Dene-Caucasian, with several extensive papers published on the subject.

The "Dene-Caucasian" question has recently seen a small surge of public interest in the light of Edward Vajda's work on a possible deep-level connection between Na-Dene and Yeniseic languages that echoes the hypothesis earlier suggested by Merritt Ruhlen on the basis of "mass comparison", attempting to support it through more robust historical arguments (Vajda, however, remains agnostic about the other members of Dene-Caucasian). The collective opinion of the EHL team on Vajda's research is that it represents another important step forward in demonstrating the reality of Dene-Caucasian, but also that much of his evidence, based on highly problematic and subjective elements of internal Yeniseic reconstruction, cannot qualify as diagnostic or demonstrative. In addition, concentrating on a binary connection between Yeniseic and Na-Dene leaves out important connecting links that inevitably skew the picture (comparable to reconstructing Indo-European based on a Celtic-Iranic comparison, while leaving out evidence from all other branches); a more inclusive assessment of the material suggests that it is actually Sino-Tibetan, not Yeniseic, that is the closest branch to Na-Dene (a hypothesis originally proposed by E. Sapir, but unsubstantiated until now).

Ongoing research by Bengtson, G. Starostin, and some of their colleagues shows that the most promising model of Dene-Caucasian would classify the entire stock into three "binary pairs": North Caucasian / Basque, Yeniseic / Burushaski, Sino-Tibetan / Na-Dene. This is based primarily on lexicostatistical assessment of the data, yielding several important shared lexical innovations for each pair, as well as on isolated morphological isoglosses (of secondary importance). All of the three pairs are then "tied up" on a higher level, suggesting an approximate date of splitting around 12,000 BP. It is hoped that the final word on Dene-Caucasian will be a large comparative grammar and dictionary, published some time in the future.