Mongolic *x and *g

In comparative Mongolic phonology, two sounds are usually distinguished, one, which I will write *x, disappears between vowels (presuming it existed, Nugteren reconstructs these sequences as vowel hiatuses) forming diphthongs and long vowels, while the other, which I will write *g, is preserved in all languages as an unaspirated or voiced stop or fricative.

Diachronically there are three possible sources of *x:

  1. Earlier *p, which is otherwise missing in Mongolic, evidenced by internal alternations between *b and *x (Proto-Mongolic *dexel vs. Kalmyk devl) and external comparison with Turkic and Tungusic words.
  2. Earlier *ŋ, which is not otherwise found intervocally. In internal alternations (*möxersün ~ *möŋgersün, *noxasun ~ *nuŋgasun) and perhaps some Turkic loans (cognates?), but most show *ŋg.
  3. Turkic *in loans, which was probably already a continuant at the time of loaning.
  4. Earlier plosive *g.

The fourth source is most recognized, but also the most problematic, as there is no clear conditioning environment in which *g would lenite into *x. Intervocally *g is not as common as it could be expected by comparing distributions of other plosives, but it is hardly marginal, and the distinction between *-VgV– and *-VxV– is firmly phonemic.

One explanation (due to Poppe, IIRC, but I can’t find the original paper) is that lenition operated before primary long vowels which were shortened in Mongolic, but preserved in Tungusic. This theory was later abandoned, but it seems to have been implicitly revived in Tumurtogoo’s collection of Pre-classical Uighur-script monuments.

A more promising direction sometimes taken is the idea that *g derives from earlier *k by a lenition process while earlier **g regularly turns to *x , but again the conditioning factor is unknown.

Another piece of evidence are Mongolic loans into Evenki which reflect *x as *k, with an unexpected strength mismatch, which I interpret as Evenki *g being already fricative and/or voiced while Mongolic **g > *x was still a plosive. Alternatively, maybe there is a strength switch phenomenon behind this that is also responsible for the mismatch found in the ‘stone’ word (Tungusic *ǯolo ~ Mongolic *čïlaxun ~ Turkic *diāl’ ~ Korean dol).

Despite all of this, I want to reiterate that *g and *x (and *k) are different phonemes in Proto-Mongolic and in the language of earliest Uighur monuments too (sometimes called pre-Proto-Mongolic, an abuse of terminology I disagree with). The fact that both are represented with the same pair of glyphs can be chalked to widespread homography of the Mongol script, distinguishing between *g and *x is as natural as distinguishing between *o/ö and *u/ü or *k and *g, or even more, but for historical reasons they are too often treated as the same sound.



Classification of Tungusic

There is a nice overview of previous attempts at classifying Tungusic languages in Robeets’ review of Recent advances in Tungusic linguistics (Turcologica 89), available at ResearchGate. Most of the schemes (other than Vovin 1993) work with four subgroups: 1) Evenki-Even (with Negidal and Solon), 2) Udege-Oroch, 3) Nanaic (Nanai, Ulcha and Orok) and 4) Manchu; and disagree on their genetic relations, the most common question being whether Manchu is the first group to diverge or if it belongs with Nanaic, and whether Udege-Oroch belongs with Northern or Southern languages.

There are many details of Tungusic comparative phonology which may be discussed when attempting a categorization. However, if we’re researching genetic relatedness, only those correspondences which can plausibly represent earliest branch-defining innovations are of direct importance.

From the sound laws usually mentioned (lenition/debuccalization of *p-, contraction of *-VgV- sequences, …)  the one I found most informative, is the correspondence of /i/ ~ /i/ ~ /u/ ~ /u/, reconstructed as original *ü. This is because the backing of a front rounded vowel seems less common cross-linguistical than other isoglosses offered.  By emphasizing this law I commit to a North-South model which splits the family into two equal halves, so let’s see how this might work out chronologically.

The Southern group

The Southern group couldn’t have stayed together for too long as there aren’t that many clear common innovations after the backing, here are some I can think of:

The fortition of intervocalic nasals

Correspondance of Southern /-ŋg-/ to Northern /-ŋ-/ seems very clear cut as a Southern innovation, but it is also quite common and thus a possible parallel innovation.

However there are also plenty of cases of /-nd-/ corresponding to /-n-/ which is more meaningful and likely related to the velar analogue, but pretty problematic as both /n/ and /nd/ between vowels appear to be phonotactically permitted in all Tungusic languages. For what it’s worth, EDAL reconstructs words such as *ŋin(d)a-(kün) “dog” with an *-nd-. There are also some cases of variation between /-m-/ and /-mb-/, but they are too chaotic for me to make sense of.

Presumably someone wrote about all this already and collected the relevant data, but I haven’t read it, Tsintsius doesn’t go into much detail in her Comparative phonetics, and I also didn’t find it in my cursory readings of Doerfer’s Mongolo-Tungusica.

Contraction of  *-VgV

This one is sometimes mentioned as an argument for grouping the Udege-Oroch group together with the Southern, as it also found there, however this change is quite common, especially if assume that *g was medially a [ɣ] already in Proto-Tungusic as it is in Evenki and Even (but not in Negidal), so I wouldn’t be surprised if it happened in all three branches independently.

Palatalization of *-rK

All languages except Even and Evenki have something going on with these clusters, so perhaps this concerns an earlier or a later period.

Nanai at least has a very clear change of *-rK– into –jK– (there are similar phenomena in some Northern languages, but they have a general iotation of *r) and Manchu has complete affrication. Ulcha and Orok  show Č and T respectively, however they feature a more general change of *-RK– > –-/-RT-,  although resonants other than *r do not disappear in this position.

Then it could perhaps be possible to reconstruct *rK > *jK for Common South Tungusic, however the fact that Udege, Oroch and Negidal iotated *r probably indicates that this is an areal feature or that Proto-Tungusic *r was already somehow close to [j] by articulation and changed in several languages independently.

The Northern group

In the north:

  1. *ü became *i and *ui became*
  2. *p– became *h
  3. *x– disappeared
  4. Intervocallically *p and *b became *w
  5. Possibly *ua became *
  6. Perhaps *-nd- became *n as discussed above

None of these changes are unique, but they are more numerous than common Southern Tungusic innovations, so this appears unproblematic. However, we glossed  something important over merely by using the term “North Tungusic”, geographically Udege and Oroch are spoken to the south of Nanaic languages, this can be clearly seen on this Ethnologue map which sadly doesn’t display China, but if it did it would be clear that Nanai continues down Sungari (Songhua) river. There are some theories that the Udege could have come from the north, but historically, during the Jurchen empire, they were found even further to the south and later migrated north to their current position. (See Nikolaeva’s A Grammar of Udihe for more details)

Where then was “North Tungusic” spoken? We could probably do worse than look at major waterways (or a more detailed map in Russian) when considering expansions of ancient populations. Due to diversity found in the Amur region, I would prefer to place Proto-Tungusic here rather than near Baikal or in Manchuria as it has been suggested previously.

If we start from the Amur, it seems clear that the Evenki group could have followed Amur upstream to Siberia with Negidals looping back to Amur via Amgun river and Solons entering Manchuria by following the Argun. The Manchu group must have followed Songhua all the way to Yalu river on the Sino-Korean border where Jurchens are first mentioned, and deep into Manchuria by Nen river. Nanaic languages expanded down the Amur with the Oroks finally reaching Sakhalin. The situation with Udege-Oroch is less clear, and if don’t accept their hypothetical migration from the north, “Common North Tungusic” as such would be impossible.

How then did these isoglosses come to be? I believe the key lies in Nanai dialects of China and the Kur-Urmi region. Here is a recent presentation discussing these languages by Andreas Hölzl, even if they are traditionally considered dialects of Nanai, a lot (I previously had the impression that it was the most although it appears to be more complicated) of their vocabulary has features common to “North Tungusic”.

Hölzl considers these languages as mixed and he cites Kazama’s opinion (which he disagrees with) that one of these languages, Kilen is a “a “missing link” between Ewenic and Udegheic”.  Under the Amuric hypothesis, it is exactly this view that appears most elucidating.Basically this is the plot as I’d have it:

The original Proto-Tungusic speakers spread across the Amur basin, most importantly splitting at the Songhua river delta. After this a subdialect develops on Songhua and down the Amur which will develop into “South Tungusic”, with the ancestors of Evenks and Evens to the north of  it, and the Proto-Udeges to the south. But before long, a backflow from the north or the south or both enters the Songhua delta territory again along with the Kur-Urmi region, splitting the Proto-Manchu group from Proto-Nanaic and creating something like a Common “North Tungusic” area that linguistic innovations could be transmitted through. Not only between Evenki and Udege groups, but possibly also to Nanaic.

Meanwhile Manchu-Jurchen drifted away from the rest of Tungusic, but that is a story for another post.

Hüis Tolgoi

Lately there has been some news about the Hüis Tolgoi inscription. In particular that it was interpreted as Mongolic.
For some background, Hüis Tolgoi inscription is a stele discovered in Mongolia (here is the is the exact location at Google maps) in 1975 featuring vertical Brahmi writing. Lately it was a subject of an expedition which produced the following four draft papers (available at
The Khüis Tolgoi inscription – signs and sounds  – by Dieter Maue
Khüis Tolgoi inscription – On the discovery, the whereabouts, condition of the stones, and our expedition – by Mehmet Ölmez
The Historical context to the Hüis Tolgoi inscription – by Etienne de la Vaissière
Interpretation of the Hüis Tolgoi Inscription – by Alexander Vovin


The authors are likely to be familiar to readers that keep up with “Altaic” and Asian linguistics in general. Despite great work they did here and elsewhere, I’d like to express my doubts regarding the interpretation of the language as (para)-Mongolic. These considerations are probably nothing new for the authors, but I’d like to publicize them anyways.

The first objection I have is to the idea that velars and uvulars are distinguished in writing and designate vowel harmonic class of the word. This would seem like a good idea because qağan, darqan and digīn (my transcription), recognizable Central Asian titles found in the inscription, exhibit three different guttural sounds. However, in the whole text ğ is found only in qağan, which I’d say makes it likely to be a foreign sound in the language of HT, perhaps a voiced velar fricative. The other problem is that no (front-vowel) k is found in the text. It appears to me a simpler solution to presume that we’re dealing with a single pair of velars that don’t correspond to any vowel harmonic distinction.

Yet my proposition is not without problems either. Why is q written with a special glyph instead of Brahmi k, and why does it correspond to Uighur Brahmi back q? My solution is that /k/ was spirantized in the language of HT, which would also explain why qağan and darqan are written in Karakhanid with Arabic   /x/. Of course this would mean that the glyph transcribed as x cannot be /x/, but I have no better suggestions, perhaps it is another fricative or consonant vowel combination like /ku/, which would square well with Vovin’s idea that -x could be the Mongolic nomen futuri.

As I find the idea of the text possessing an “Altaic” typology doubtful, I also have to question the decision to read all ambiguous initial (r)u- signs as /u-/, although this particular constraint does seem to have a wider distribution, so it might as well turn out to apply to the language of HT as well. Another unexpected typological feature is initial consonant clusters found in two words (kranyaguń and dro) which are strictly prohibited in both Mongolic and Turkic, perhaps they could be results of a syncope, as suggested, or maybe graphical devices to signify some vowel or consonant distinction, however I’d say they surely decrease the likelihood of the inscription being Mongolic.

As to Vovin’s idea that the language of HT is closer to Middle Mongolian than Khitan or Xianbei, it seems somewhat peculiar to me, because I’d expect from a much older language (HT precedes MM by around 600 years) to be closer to the common ancestor of the family and thus other older varieties. Instead, the reconstruction presented here is more phonologically innovative than MM, it features apocope, syncope, metathesis and contraction of (some) MM *-VxV- sequences into monopthongs. Of course this isn’t impossible, Khitan appears to be more innovative than MM on some points, yet it does again significantly reduce the likelihood of this reconstruction.

An overarching problem with Vovin’s glosses is that most convincing ones have cognates that are first attested in Turkic, this includes töre- ‘be born’, uka- ‘understand’, te– ‘say’ (not even attested in Mongolian!), katun (likely Sogdian), dörö ‘law’, tayi (from Turkic tap-) and jula ‘torch’. Others may have dubious ad hoc sound changes such as metathesis of vowels or random changes in vowel quality.

Morphology seems to be the strong point of this reconstruction. Seeing -ba, -ju, -n, -x, -un and -d, one really can’t help but think of Mongolic, but with liberties taken on every level of the reconstruction, incidental matches can happen. Even if the text looks somewhat plausible, I certainly wouldn’t put the matter “beyond reasonable doubt”.

All in all, I’m feeling kinda pessimistic if this is the best that top researchers in the field can get from this inscription, although I understand that this is merely a draft and that there are more Brahmi inscriptions to decypher. Presumably the language won’t turn out to be Turkic or Indo-European, or researchers would have identified it as such already, but I am certain that further investigation will illuminate the history of languages of the region.