Dene-Yeniseian Language Family: Evidence for a Back-Migration to the Old World?
The 2012 Dene-Yeniseian Workshop took place on March 24 at the University of Alaska Fairbanks. Since the seminal presentation by the West Washington University linguist, Edward Vajda, of morphological and lexical evidence relating the small Yeniseian language family from Western Siberia and Na-Dene (Eyak-Tlingit-Athabascan) languages in North America, the annual Dene-Yeniseian Workshop deserves everybody’s attention.
Both the program and the videos of most of the talks from the 2012 workshop are available. The following two presentations are, in my opinions, the most interesting.
Geography, Demography, and Time Depth: Explaining How Dene-Yeniseian Is Possible
Archaeology, human genetics, linguistics, and folklore studies often yield what seem to be conflicting evidence of human prehistory when in fact they tell different parts of the same story. This presentation unites evidence from disparate branches of the study of human prehistory in a new synthesis to explain how the Dene-Yeniseian linguistic hypothesis fits with what other branches of science have revealed about the peopling of the Americas from North Asia in the late Pleistocene and Early Holocene.
Applying Computational Phylogenetic Methods in Evaluation of the Dene-Yeniseian Hypothesis
Gary Holton and Mark Sicoli
Linguists have for a century suggested a common linguistic heritage for Na-Dene and Yeniseian, but only recently has a case been made using traditional methods of historical reconstruction (Vajda 2010). Our work responds to challenges to use alternative methodologies to evaluate this hypothesis (Campbell 2011, Diamond 2011) and contributes to a growing body of literature using bioinformatic methods to push the time depths at which linguistic comparison can test hypotheses of prehistory (see also Dunn et al. 2005, Gray and Atkinson 2003). In this paper we examine the Dene-Yeniseian hypothesis using computational phylogenetic methods applied to two recently compiled data sets. The first data set is composed of typological and morphological characters, while the second is composed of lexical characters coded for cognacy. We fit trees to these data using both Bayesian and neighbor-joining network models. In both cases the degree of tree-like structure within Northern Na-Dene is minimal, suggesting a long history of diffused innovations (cf. Krauss 1973, Hill 1978) and a possible center of radiation (Sapir 1916). The typological dataset in particular shows a significant tree-like signal at higher levels that supports the Dene-Yeniseian hypothesis, with a thought provoking twist. The phylogenies suggest an ancient heritage for Yeniseian in Beringia with the relationship representing a terminal Pleistocene back migration.
The Dene-Yeniseian proposal has created an unusual, intriguing and stimulating situation in many ways. It is a top-down theory of ancient dispersals driven not by archaeology or genetics, but by linguistics. It puts a strain on comparative method in linguistics as it suggests that traces of common origin can still be discernible after 10-12,000 years of language separation. It bridges the gap between traditional historical linguistics and long-range comparison (as Vajda said once, he drew inspiration from both short-range and Greenbergian methods) and between synchronic typology and diachronic reconstruction. Finally, it constitutes the first proposal for a language relationship across the Bering Strait that has serious evidence behind it and that consequently commanded international and interdisciplinary attention.
Vajda’s key finding is the strong isomorphism between cross-linguistically rare multi-part (stem + a string of grammatical elements) verbal templates found in Na-Dene and Yeniseian languages (see below, from Vajda, Edward. 2012. “The Dene-Yeniseian Connection: A Reply to G. Starostin,” Journal of Language Relationship 8: 140-141).
It is complemented by some 100 lexical etymologies, many of which pertain to “forest lifestyle” (birch bark, canoe, etc.), with some strong sound correspondences behind some of them (the emergence of tones in Ket from glottalic or plain consonantal codas in proto-Dene-Yeniseian).
Initially “approved” by a committee of respectable linguists (see The Dene-Yeniseian Connection, edited by James Kari and Ben Potter. Anthropological Papers of the University of Alaska 5 (1-2), 2010), the Vajda proposal has recently been criticized by Lyle Campbell, from a traditional comparativist perspective, as well as by George Starostin (“Dene-Yeniseian: A Critical Assessment,” Journal of Linguistic Relationship 8: 117-152, 2012), from the point of view of the putative larger grouping, Dene-Caucasian, of which Yeniseian and Na-Dene may be part. Both Campbell (“Review of The Dene-Yeniseian Connection, ed. by James Kari and Ben A. Potter,” International Journal of American Linguistics 77 (3): 445-451, 2011) and Starostin (building on a critical note by an exert in languages with synthetic morphology, Andrej A. Kibrik [“Transitivity Indicators, Historical Scenarios, and Sundry Dene-Yeniseian Notes,” The Dene-Yeniseian Connection, 2010]) called into question the plausibility that such a complex, synthetic morphological structures as the multi-part verbal templates could survive for 10,000-12,000 years. At the same time, there is a general consensus that these templates must derive from more analytical forms for tense, etc. concatenated together via grammaticalization to form a complex whole. If some parts of the template (e.g., tense markers) show similarity between Yeniseian and Na-Dene, while others do not, the conclusion must be made that the grammaticalization occurred independently in two language families. The first concern seems to be a no-win situation for Vajda because Campbell distrusts not just complex morphological forms but also simple monosyllabic roots that Vajda adduced as part of the Dene-Yeniseian lexical corpus. In both cases, convergence supposedly looms large. But this logic makes convergence look like an omnipresent mystical force affecting language without regard for the kind of material it operates on and this makes it from a legitimate source of doubt into a target of doubt. The second concern may be dispelled if it turns out that the synthetic template (in a somewhat simplified form) is in fact a retention from a deeper phylogenetic level (bi-partite templates are widely attested in North American Indian languages as well as in some Sino-Tibetan languages, the latter being part of the Dene-Caucasian hypothesis) but that the “slots” were filled up through grammaticalization multiple times in Yeniseian and Na-Dene without affecting the overall template structure. The relative independence of structural semantics and discrete lexical forms can be frequently observed in kinship terminologies – the same underlying structural-semantic pattern may yield itself to multiple waves of lexical replacements. If the larger hypothesis of a musical protolanguage is correct and if synthetic morphology (typical for New World, Papuan and Caucasus languages, see map below for complex verb morphologies, from Vajda’s talk) historically precedes analytical morphology, then the Yeniseian verb is a relic of this ancestral state in Western Siberia.
Starostin is also puzzled by the fact that there are no suggestive similarities between Yeniseian and Na-Dene pronouns, while there are similarities in the “forest lifestyle” vocabulary. Pronouns constitute a closed lexico-grammatical set, which is not easily susceptible to borrowing, that long-rangers often use as a strong proof point for such macrophyla as Dene-Caucasian or Amerind. Forest lifestyle words, on the other hand, are not part of the basic vocabulary of a language and are subject to frequent borrowing. Similarities in pronouns presumably exist between Yeniseian and the “western” Dene-Caucasian languages (Burushaski and North Caucasian) but not between Na-Dene and Yeniseian. To summarize, Starostin (2012, 137) offered a fanciful and somewhat defensive assessment of the Vajda lexical effort:
“Vajda’s ‘regular correspondences’ are not, or, at least, not yet properly ‘regular’ in the classic comparative-historical sense of the word. Most of them seem to be based around one ‘psychologically impressive’ example, which is then backed up by 1-2 supporting comparanda that are usually weaker from either the semantic, or the phonetic, or the distributional points of view, but still manage to produce an aura of ‘regularity’. The same could be extended to his treatment of the verbal morphology, where a tiny handful of intriguing isomorphisms are surrounded by an impenetrable sea of assumptions and highly controversial internal reconstructions that create an illusion of systemic reconstruction where there really is none.”
Starostin’s counter-argument against the unique Dene-Yeniseian taxon is not based on a bulletproof comparative evidence because pronouns (and Campbell is always very keen on emphasizing this, including a personal communication that I had with him in 2005) DO GET borrowed (Thomason, Sarah G., and Daniel L. Everett. 2001. “Pronoun Borrowing,” Proceedings of the Berkeley Linguistic Society27: 301-315) and they are subject to forces of convergence (Nichols, Johanna. “Why ‘Me’ and ‘Thee’?”Historical Linguistics 1999: Selected Papers from the 14th International Conference on Historical Linguistics, Vancouver, 9-13 August 1999. Pp. 253-276. John Benjamins, 2001). And it is true, as Campbell noted (and Vajda admits) that some of the most striking forest lifestyle matches between Yeniseian and Na-Dene (e.g., the ‘birchbark’ word) are indeed negated by the fact that in Yeniseian the corresponding word is a borrowing from Selkup. So, between the lack of bulletproof evidence for pronoun stability and the existence of evidence for borrowing in the Yeniseian forest vocabulary, neither the Dene-Caucasian, nor the Dene-Yeniseian proposals enjoy a firm footing.
The overall consensus seems to be that a) Vajda did succeed in “planting” formal linguistic “seeds” for a phylogenetic connection between Yeniseian and Na-Dene (the majority of linguists seem to concur); b) but these “seeds” need to be “watered” to produce more and better evidence (Campbell, Starostin) and c) they may represent retentions from the Dene-Caucasian level and independent Yeniseian-Na-Dene innovations (Starostin). In the current version of the Dene-Caucasian phylogeny (based on lexicostatistics performed on the 50-word Swadesh list), Na-Dene is closer to Sino-Tibetan as two “easternmost” parts of the phylum and Yeniseian is closer to the western members of the phylum, Burushaski, North Caucasian and Basque (see below).
In the spirit of Joseph Greenberg, Vajda believes that, while only linguistic evidence counts as proof of a linguistic grouping (in this case, Dene-Yeniseian), it is important and instructive to contextualize a linguistics-driven hypothesis with the help of data derived from other disciplines. In his 2012 workshop presentation, Vajda builds an interdisciplinary case for the unique connection between Na-Dene and Yeniseian using linguistic, molecular genetic (Y-DNA and mtDNA) and archaeological arguments. He observes that the speakers of Ket, the last surviving member of the Yeniseian family, are unique in Siberia in having Y-DNA hg Q1, the most common haplogroup in the New World, at high frequencies. The other haplogroup in Ket is hg C, which is very common in Na-Dene speakers.
Basically, the distribution of hg Q and hg C in Kets and Na-Dene is inverted – Na-Dene are rich in hg C, which sets them apart from other American Indians, and poor in hg Q, while Kets are rich in hg Q but have some hg C. Since hg C is rare in the Americas outside of Na-Dene, the co-presence of hg Q and hg C is Kets creates a strong similarity to Na-Dene, albeit with inverted haplogroup frequencies. Vajda, however, rejects the direct connection between hg C in Kets and and in Na-Dene because they do not belong to the same sublineage. Instead, Vajda proposes that the ancestors of Na-Dene picked up hg C during their migration to the New World after the break down of the Dene-Yeniseian connection.
On the mtDNA side, Vajda observes that, while Na-Dene are very homogeneous, with the majority of lineages belonging to Amerindian hg A2, Ket mtDNAs are very diverse. This diversity, however, is not an indication of Ket antiquity as a separate population but rather of its position at the crossroads of gene flows between Western and Eastern Eurasia. Kets had originally carried only hg A2, just like Na-Dene, but they absorbed hgs U and F from an earlier Siberian substratum (“Baikal fisher-folk”) and then topped it off with Selkup-derived hgs H and C. It is rewarding to see that Vajda does not simply declare the greater antiquity of Kets vs. Na-Dene because of the former’s superior levels of mtDNA diversity but actually reconstructs – in a manner reminiscent of historical linguistics – the proto-Yeniseian mtDNA lineage composition by filtering out subtrate and admixture effects on the Ket gene pool. In the end, Vajda considers proto-Yeniseian gene pool as consisting of Y-DNA hg Q1 and mtDNA hg A2 (see below).
But Vajda admits a gross contradiction in his otherwise very cogent argument when he reconstructs an essentially Na-Dene mtDNA pool for proto-Yeniseian (hg A2 at high frequencies), while attributing hg C3 in Na-Dene to a massive gene flow from Siberia. He argues that Na-Dene represent a separate migration into the Old World because of the high concentration of hg C3 in their Y-DNA pool. But he forgets that hg C3 is widely attested in the Americas from Eskimos to Northwest South America, which made geneticists believe from the very beginning that just one migration brought both hg Q and hg C to the America. Most importantly, hg A2 attested in Ket is widely believed to represent a back-migration from the New World to Siberia (see the visualization below, from Tamm et al. 2007).
Since Yeniseian mtDNA pool reconstructed by Vajda is likely derived from the New World, it means that proto-Yeniseians originated somewhere in the New World where Na-Dene languages are currently found (Alaska or the Northwest Coast of Canada). But based on Y-DNA, Vajda postulates a migration of proto-Na-Dene from Siberia. The lack of a coherent population genetic argument basically invalidates Vajda’s attempt to map his linguistic hypothesis and genetic data onto Siberian archaeology. (In a general sense, in the absence of ancient DNA, any argument for an association between languages and genes with specific archaeological cultures looks speculative and futile.)
It is much more logical to reconstruct both mtDNA hg A2 and Y-DNA hgs Q1 and C3 at moderate frequencies, with some additional lineages at low frequencies, for proto-Dene-Yeniseian and hypothesize a migration of proto-Yeniseians (with the A2, Q1, C2 composition only) from the New World, with subsequent increase in the frequencies of Y-DNA Q1.
The issue of a back-migration of proto-Yeniseians from the New World is directly addressed by linguists Gary Holton & Mark A. Sicoli in their 2012 workshop presentation. They applied Bayesian phylogenetic and Neighbor-Joining methods to discover that Yeniseian typological variation was a subset of Na-Dene typological variation.
Although I generally distrust Bayesian phylogenies, the fact that Holton & Sicoli openly argue and provide evidence for a back-migration of Yeniseian from the New World to Western Siberia is revolutionary and definitely reflects the new research Zeitgeist around the problem of the “peopling of the Americas.” In addition, Holton & Sicoli’s method produced a novel model of Na-Dene internal subgrouping, in which Southern Athabascans split before Northern Athabascans. I am very intrigued by this latter model because I had arrived at the same conclusion by analyzing Na-Dene kinship terminologies (see this section in The Genius of Kinship). Southern Athabascans appeared to be markedly more conservative in their kin terminological structures than Northern Athabascans. Interestingly, Navajo is the most polysynthetic of Na-Dene languages.