In The Genius of Kinship (2007), I hypothesized that proto-Athabascan kinship terminology had 4 self-reciprocal grandparental terms. The western (Apachean) branch of Southern Athabascan preserved the original 4-term pattern better than any Northern Athabascan language. Southern Athabascan languages demonstrate all the stages of the transformation of the original complexly differentiated pattern, with Kiowa-Apache sporting the most transformed and simplified model (see below, Table 34).
What my analysis was missing at that time was the formal phonomorphological and etymological proof that Proto-Athabascan kinship terminology indeed had 4 self-reciprocal grandparental terms, that this ancestral pattern underwent simplification and dereciprocalization in Northern and Pacific Athabascan branches and that Apachean preserves the Proto-Athabascan grandparental set better than any other branch.
Recently, Victor Golla sent me an e-mail (cc-ing Jack Ives, Michael Krauss, Johanna Nichols and James Kari) in which he attached a short unpublished manuscript “Where Does Navajo náli Come From?” providing the necessary formal support for my original typological inference. In it, he demonstrates that, in addition to Apachean, Eyak, too, featured a 4-term self-reciprocal pattern in Gen +/-2.
The Apachean reconstruction by Golla largely repeats the Eyak kinterm set.
To quote from Golla, the Eyak
“MoMo/wmDaCh root is clearly cognate with Athabaskan *-chu — no surprise — but believe it or not the FaMo/wmSoCh root *-k’i?h just as regularly corresponds to Athabaskan *-ch’?ne, and the FaFa/maSoCh root –?uh is a plausible cognate of Wailaki and Kato –?a?e ~ –?ah.”
Proto-California Athabascan (PCA) retained the 4-term pattern in Gen +/-2 but it lost self-reciprocity between the grandparent and grandchild forms. At the same time, one of Eyak grandparental terms, namely –?uh FF, mSC, is cognate with PCA *-’a?. This means that the Apachean-PCA isogloss *-nyale is a formal but not semantic innovation.
Northern Athabascan grandparental set (see below) consisting of just 2 terms for grandfather and grandmother appears to be the most derived.
Two Northern Athabascan languages have retained forms pointing in the direction of the ancestral 4-term system: 1) Wets’uwet’en (Babine-Carrier) distinguishes MM (tso’) from FM (ts’inï’) — using terms cognate with those of PCA, Apachean and Eyak; 2) Tahltan similarly attests a *chu vs.*ch’?ne contrast in the grandmother terms.
Golla concludes that the diversity of grandparental kinterms in Athabascan languages
“misleadingly suggests that the simple kin terminology found in most Alaskan and Canadian languages represents the Proto-Athabaskan situation, and that PCA and Apachean constitute an innovating ‘Southern Athabaskan’ clade of the sort that Matson and Magne posit… the Proto-Athabaskan-Eyak language must be assumed to have had a 4-term lineage-distinguishing grandparental terminology, which ‘survived intact’ in PCA and Apachean instead of being innovated there.”
The drastic simplification from 4 terms to 1 term in Kiowa-Apache stems from Plains influence.
Jubilating over these conclusions, the following is essentially what I wrote back to Golla.
There are three more areas of research where my broader typological inferences could be tested against actual phonological and lexical data.
1. Typologically the self-reciprocal equations found in some Southern Athabascan languages between “father’s brother” and “man’s brother’s son,” “mother’s sister” and “woman’s sister’s daughter,” “father’s sister” and “woman’s brother’s daughter,” and “mother’s brother” and “man’s sister’s son” could go back to Proto-Eyak-Athabascan (PEA) times as well. They would complete the self-reciprocal picture contained in the 4 grandparental terms that Golla reconstructed. There are kin terminologies in North America and beyond that have 4 self-reciprocal grandparental terms plus all aunt and uncle terms are self-reciprocal, too. Interestingly, one old unpublished Russian source on Ket lists one of those equations (koj or qo.j MoBr =MoSi = SiSo). Ives, Rice and Vajda only give MoBr =MoSi for qo.j, plus other secondary meanings, but not the self-reciprocal one.
2. Southern Athabascan sibling terminologies lexicalize Ego-Sex and Relative Sex in addition to Relative Age. Again, this is something that Northern Athabascan sibling terminologies must have lost as they only have terms marked for Relative Age. Eyak and Tlingit has Ego-Sex/Relative-Sex in addition to Relative Age, which again supports the archaism of Southern Athabascan compared to Northern Athabascan. But I’m not aware of any formal reconstructions that would test the hypothesis of progressive reduction of sibling set complexity in Athabascan languages. There are several studies of Austronesian languages that seem to support this diachronic universal.
3. Going back to the PAE grandparental terminology, it’s now time to compare it with Yeniseian. One hypothesis I have is that Ket qip ‘grandpather’ and qi.ma ‘grandmother’ are based on the same root morpheme and this morpheme is cognate with Eyak –k’i?h/Athabaskan *-ch’?ne FaMo = wSoChi. In general, I noticed that Ket takes the reduction tendency observed in Northern Athabascan languages to the extreme (e.g., Ket has one single term for siblings bisep with no terms reflecting Relative Age, Relative Sex/Ego Sex; one single root morpheme for nephew/niece and, then, only one for son/daughter), so it wouldn’t be surprising if the loss of linear distinctions that Golla observed in the treatment of grandparental terms in many Athabascan languages (Connector Sex is neutralized) is continued in Ket in the form of a neutralization of the Referent Sex distinction in the root morpheme. This is exactly what happened with grandparent terms in Kiowa-Apache.
Golla’s reconstruction of 4 self-reciprocal terms for grandparents in PEA changes the nature of the conversation among kinship theorists around the earliest form of Athabascan kin terminologies from the “Dravidian” to the “Kariera” model. It also makes PEA look clearly “Amerindian” and not “Northeast Asian” as 4 self-reciprocal terms for grandparents is not an East Asian trait. It brings up a question of EA phylogeny as Southern Athabascan looks like a conservative branch now (at least in this one dimension, potentially complemented by others in the kinship domain, as I outlined above as well as in my book). At least from a kinship systems perspective, SA doesn’t look like a recent offshoot of a Subarctic population but more of a proto-Athabascan relic that drifted southward along the western slopes of the Rockies from British Columbia to Plateau through Great Basin and into the Southwest.