Proto-Turkic Consonants

« previous post | next post »

I seldom announce the publication of Sino-Platonic Papers on Language Log, but this one, although seemingly highly esoteric, will actually be of interest to many readers.  Aside from numerous Turkic tongues, among other languages and groups it touches on, the following are mentioned:  Mongolian, Tibetan, Chinese, Japanese, Tocharian, Uyghur, Bulgar, Tatar, Bactrian, Tungusic, Celtic, Dravidian, Yeniseian, Samoyedic, Chuvash, Latin, Italic, Prussian, Slavic (various languages), Sanskrit, Kitan, Hungarian, Xiongnu (Appendix 2 is a list of Xiongnu words surviving in Altaic languages), Circassian, Caucasian, Avar, Dingling 丁零, Khotanese Saka, Sogdian, Khwarezmian, Old Persian, Middle and New Persian, Pashto, Ossetian, and numerous Iranian languages, Yuezhi, Koguryŏan (Korean).


Sino-Platonic Papers is pleased to announce the publication of its three-hundred-and-twenty-fifth issue:

"On *p- and Other Proto-Turkic Consonants," by Orçun Ünal (Göttingen Academy of Sciences and Humanities, Göttingen, Germany)


To my first teacher in Mongolian

Claus Schönig (1955–2019)


The present study takes as a starting point the question of whether Proto-Turkic had an onset *h- or *p- and aims at reconstructing its consonantism. The answer to the initial question is searched for in the fourteen Turkic lexical loans of adjacent languages such as Mongolic, Kitan, Yeniseian, and Samoyedic. At first sight, the data provided by these loanwords seem ambiguous. However, once it is demonstrated that both the daughter languages of Proto-Turkic, namely Proto-Bulgar Turkic and Proto‑Common Turkic, had the historically unattested initials *d2 and *ń-, these data can be taken to point to the existence of *p- in these languages as well as in Proto-Turkic. The discussion is extended with the question of rhotacism and lambdacism. As regards the rhotacism, Proto-Turkic is assumed to have two rhotic consonants, phonologically denoted as */r1 r2/. The lambdacism, on the other hand, turns out to be a tougher problem. Based on several lexical borrowings into and from Turkic, a further consonant */t2/ is posited for Proto-Turkic. This consonant, originally of affricate and probably later of fricative pronunciation, yielded /š/ in Common Turkic and /l/ in Bulgar Turkic. Thus, the Proto-Turkic consonantism is reconstructed as having a series of consonants */t2 d2 r2/ that underwent serious changes in historical Turkic. Finally, */k2/ is added to this series to explain the correspondence of k- and vocalic onset between some Turco-Mongolic cognates. In addition, significant sound changes in the prehistory of Turkic are dated through external evidence.

Key words: Proto-Turkic, onset consonant, lexical borrowing, consonantism


This and all other issues of Sino-Platonic Papers are available in full for no charge.

To view our catalog, visit


  1. Guillaume Jacques said,

    April 2, 2022 @ 10:48 am

    Concerning *r1 / *r2 and *l1 / *l2, our article "Turkic kümüš 'silver' and the lambdaism vs sigmatism debate" could have been useful:

  2. Mehmet Oguz Derin said,

    April 2, 2022 @ 2:29 pm

    The use and presentation of external data is a significant step, and (for anybody wondering but not having the other resource handy) Proto-Turkic consonant inventory seems to differ from the table in Lars Johanson's Turkic (2021). Great share!

  3. Jenny Chu said,

    April 2, 2022 @ 9:40 pm

    Goodness, why do you hesitate to announce the publication of Sino-Platonic papers on Language Log? That's fascinating and I had no idea that existed!

  4. Chris Button said,

    April 3, 2022 @ 11:00 am

    This lambdaism v sigmatism debate came up recently on language log.

    I am firmly in the camp that kümüš must have had an original lateral coda.

    The connection with Chinese 銀 is interesting. Alternation of -l and -n codas is attested across the lexicon, so it’s perhaps not surprising for a Wanderwort like this. The medial -r- is a problem only if should actually be reconstructed. Sometimes (as with 茶 “tea”, which we’ve discussed before on LLog), it’s not justified. This may well be such a case, and I’m tempted to invoke Pulleyblank’s use of Bhat’s observation about clusters with k- causing retroflexion or associated reflexes. Here we have km- giving ŋ- with a reflex as if from ŋr-.

  5. Chris Button said,

    April 3, 2022 @ 11:03 am

    *a vocalic reflex, that is

  6. Andreas Johansson said,

    April 4, 2022 @ 3:30 am

    Why did we end up with "Common Turkic" as the name of a subset of Turkic?

  7. David Marjanović said,

    April 4, 2022 @ 4:41 pm

    Probably because its descendants are very easily recognizable as related, much like most of Romance, while Chuvash is Uncommon Turkic – in the early 19th century, IIRC, it was sometimes thought to be a Uralic language with Turkic influence rather than vice versa.

  8. R. Fenwick said,

    April 6, 2022 @ 1:26 am

    I've yet to read the full paper, but I'm already wonderstruck by the proposed Turkic etymology of Common Circassian *daʁá "fat, oil" (stress is reconstructible on the final mora in view of the preservation of the final vowel in Temirgoi дагъэ). There are various other Turkic loans in Common Circassian—e.g. *pʰəɮə́ "elephant", *maɮə́ "sheep", *ħábɮa "village quarter", *andə́qːə "hyena"—but to the best of my knowledge, nobody has ever proposed such an origin for *daʁá. It'd be a fascinating signal of early Oghur influence in Circassian, where most loans have previously been attributed to Kipchak varieties.

RSS feed for comments on this post