The Altaic Hypothesis revisited

« previous post | next post »

"Altaic: Rise and Fall of a Linguistic Hypothesis", NativLang (9/28/19) — video is 12:29; extensive discussion after the page break

A lot of Language Log regulars are mentioned.  One name I missed is that of Roy Andrew Miller (1924-2014), author of Japanese and the Other Altaic Languages (Chicago: University of Chicago Press, 1971).


Comments by specialists

J. Marshall Unger:

I saw this a while back.  Not too bad.  He should have emphasized that Martine Robbeets’s Transeurasian is an effort to start fresh, this time including Korean and Japanese as witness languages and all the new data collected since Poppe, and “rebuild” the “Altaic” reconstruction.

Marcel Erdal:

My IMPRESSION is that Turkic and Mongolic are NOT related genetically, but:

1) The fact that Starostin/Dybo/Mudrak and others wrote lots and lots of nonsense does not disprove the genetic hypothesis.

2) The terrible Drevnetjurkskij Slovar' (on which Russian research on this topic is generally based) and  Clauson's dictionary are outdated. Most Old Turkic texts were only edited during the last decades and what can be learned from them about the lexicon is at present available only to 3 or 4 scholars.

3) Serious work on Chuvash etymology is only now getting published and MUCH more needs to be done; that is essential for these matters.

4) With Khitany now emerging from the mists and the 6th century "Para-Mongolic" inscriptions, Proto-Mongolic will have to be reconsidered; there were some moot questions about it even before these recent developments.

In many cases it's not clear whether loans and influences went from some early stage or variety of Turkic to some early stage of Mongolic or the other way around, so ways of CONVERGENCE and SHARING still need to be worked out.

Peter Golden:

I remain an agnostic, awaiting further developments, but, in essence, concur with Marcel. We certainly need a new Clauson (and not just because mine is coming apart at the seams from use). I am using András and Árpád’s WOT volumes daily and one would like to see more on Chuvash. I consult EDAL (Starostin, Dybo and Mudrak) and choose what is useful for a particular problem, without necessarily accepting their overall view. 

Shimunek’s Languages of Ancient Southern Mongolia and North China is another step forward, albeit with its share of idiosyncrasies. He subtitles his book the ”Serbi or Xianbei Branch of the Serbi-Mongolic Language Family,” which he prefers to “Para-Mongolic” – I find the latter term perfectly acceptable, but I am an historian with philological interests and not a linguist – and can plead “innocent” if pressed…or a gun is held to my head. What does the group think of the term “Transeurasian”?

I accept the notion of borrowing and convergence, but as someone who studies nomads, have often wondered how precisely that occurs? The nomadic economy requires people to live in small groups (4-5 households most often) spread out over considerable territory. When, where and under what circumstances does borrowing and/or convergence take place? Having grown up in a multilingual home, in a crowded city (New York) in a neighborhood where it was the rare older person who spoke English WITHOUT an accent and I heard (and acquired) languages on the street, I can fully appreciate borrowing and convergence in an urban setting with people living cheek by jowl. I can recall Puerto Rican Spanish sprinkled with curse words in Yiddish (šmok and poc, being the usual favorites) or calques of American English phrases into Spanish, e.g., tómalo suave “take it easy” or more often denoting simply “be cool” or “chill out” as the kids would say today. Chicano Spanish in California has such gems as “estoy watchando”  ("I am watching”).

I can understand how titles and names for luxury goods can be borrowed, but why the word for “stone”? Cf. Turk. taš, Mong. čilaġun. And how does that occur in the steppe?

I grew up surrounded by a host of languages in what had been a predominantly Irish neighborhood (everyone one in the parental generation spoke with a brogue) and was becoming Puerto Rican (it’s now largely Dominican). There was a large Greek community (all of the American-born generation had retained fluency in their ancestral tongue), Jewish communities ranging from refugees from Germany to those coming from Eastern Europe (speaking a host of languages), Italians, Poles, even a few Swedes et al. On my street there was a group of Chinese families that even fielded their own street hockey team (we played hockey on roller-skates, dodging cars and flying pucks :). An interest in languages came naturally…if one paid any attention.

Juha Janhunen:

Well, it happens that exactly a word for 'stone' has entered from Greek to Western Europe via Latin, e.g. French pierre: 

From Middle French pierreOld French pierre, from Latin petra, a borrowing from Ancient Greek πέτρα (pétra).

However, the "Altaic" comparison of Turkic tash (< *tïxash) with Mongolic ciluxa(n (< *tïlu-ga-n) and Tungusic jolo (< *jola), not to mention Koreanic tolh and Japonic i-si, does not fill the requirements of regular sound correspondences. In my opinion, all these words have a different origin, so there is no "Altaic" word for 'stone', nor was any of these words transmitted as a borrowing between these languages. By accident, some of these items begin with a *t-, while some of them contain a medial *-l-, but this does not mean that they are cognates. 

As for Transeurasian, I am pretty sure that I proposed that term to Martine Robbeets in a discussion with her at a conference in North America in the 1990s. I suggested that she should approach the "Altaic" comparisons with caution and for this reason use a more neutral term. However, it seems that she just adopted the new term without changing the framework. She may, of course, not remember that we ever talked about this.

Peter Golden:

Many thanks, Juha.

Given my first name, I was well aware of the origins etc. I still do not understand the socio-linguistic context for such a borrowing.

The taš etc. explanation makes sense (and I recall seeing something to that effect previously), but it does require a goodly number of coincidences, not impossible, but…

Marcel Erdal:

A wonderful new Old Uyghur dictionary will appear perhaps in 3-4 months in Germany –  960 pp. without quoting text passages and without data from Middle Turkic and modern languages! Concerning Chuvash, Clára Agyagási's Chuvash Historical Phonetics, which appeared last year, is a huge step forward but has important shortcomings; I am at present writing a detailed review of that, with suggestions and lots of details. This is delaying my reedition of my Old Turkic Word Formation, which will hopefully be ready in a few months. The discussion around Mongolic is in full swing; lots of question marks are a good sign for future research. What about that still unedited brâhmî inscription, Sasha?

András Róna-Tas:

It is impossible, or almost impossible to prove that something is not existing. You can present facts which  if they are existing, they may exclude that other supposed facts are existing. But in linguistics such counter facts are very rare if existing at all. You may show that the arguments, which were cited  in favor of certain facts, are inconclusive or unacceptable or simply wrong. But in all these cases you cannot evade sober judgment.

I never claimed that the so-called Altaic languages are or are not genetically related. What I claim is that most of the facts and arguments cited in favor of the genetic relationship of the Altaic languages turned out to be borrowings and arguments in favor of being borrowings. This does not exclude that in an even earlier period they may have been the offspring of  the same language family. When I asked Roy Andrew Miller, when did the common Japanese-Korean-Tunguzic-Mongolic-Turkic proto-language exist, he answered: "I never thought about that". Nevertheless his criticism on Robbeets was annihilating. By the way, scholars, surely we all, are more critical of the products of others than of our own.

I do not think that the languages of the nomads are basically and structurally different from the languages of sedentary people („Turanic” languages). Nomadism is a relatively recent development, has many different forms, and is always living in symbiosis with agriculture. In all armies. also in the great nomadic armies, people were recruited or forced irrespective of their original mother tongue. The other places of gathering different people were the monasteries. The same is true for the caravans. Even small moving groups meet each other in a year more times than people living in isolated small villages. So borrowing or copying (Johanson) may have solid socio-linguistic backgrounds in case of nomads as well.

From the Altaic field more and more new material is coming forth. Nobody could or would have predicted that gender will be one of the important features of the Para-Mongolic Khitan. It may be a secondary development, but even then…

So Altaic studies remain a promising field even if there does not exist an Altaic linguistic family.

So now let us hear from the Language Log chorus on this most important topic.


Selected reading


[h.t. John Rohsenow]


  1. John Swindle said,

    December 10, 2020 @ 11:32 pm

    Ancient Tungusic shamans are believed to have spoken Trance Altaic. Kazakhstan has Transuranic elements.

  2. R. Fenwick said,

    December 10, 2020 @ 11:56 pm

    @J. Marshall Unger:
    He should have emphasized that Martine Robbeets’s Transeurasian is an effort to start fresh, this time including Korean and Japanese as witness languages and all the new data collected since Poppe, and “rebuild” the “Altaic” reconstruction.

    I admit, most of what I've seen regarding Martine Robbeets's "Transeurasian" appears to be less rebuilding and more rebranding. Much of it is too uncritical of the Moscow school (viz. Starostin, Dybo and Mudrak's EDAL) without accounting for numerous valid critiques of that work, and as such, begs the Altaic question without doing nearly enough to dispense with the null hypothesis of large-scale interborrowing championed by the Altaic skeptics. It strikes me as rather like trying to prove the unity of (to use an example more familiar to me) an Ibero-Caucasian family without even confirming if the controversial North Caucasian grouping is a valid phylogenetic node. I note also that several reviews of Robbeets's work have been unenthusiastic at best and caustic at worst, levelling a range of criticisms at it including misquotation of views from opponents, omission of crucial sources, failure to adhere to its own methods, failure to cross-check actual language data, and an ignorance of key issues in Robbeets's own speciality of Japonic. (For said reviews, see e.g. <a href=";?Knüppel 2006, Miller 2007, Kara 2007, Vovin 2009, Georg 2009, Vovin 2011, Georg 2013, and Alonso de la Fuente 2016.)

    (Obligatory disclaimer: I'm not an expert in any of the families involved in Robbeets's "Transeurasian" – my speciality is Ubykh (North-West Caucasian) – but I have a pretty good grasp on the broad strokes of Turkic, and the multiple complex strata of Turkic loans in Ubykh have necessitated deep diving into Turkic etymological materials, which in its turn has required I at least familiarise myself with the long-range proposals.)

    Tangentially, am I the only one who finds the name "Transeurasian" pretty problematic? Even taking the non-Eurocentric one of the two possible senses, the Europe part of Eurasia is represented in "Transeurasian" only by the late western expansion of Turkic; the putative "Transeurasian" phylum doesn't cover nearly enough geographic extent, language families, languages, or speakers for the name to be accurate in any useful sense. (Compare Trans-New Guinea, a phylum which at least does extend across more than 90% of the island's length and – depending on whose form of the phylum one accepts – may include as many as 50-60% of the thousand or so languages spoken there.)

  3. R. Fenwick said,

    December 11, 2020 @ 4:25 am

    [I'm including both of R. Fenwick's comments because they are sufficiently different and equally valuable not to remove one. The reason the first one didn't appear promptly is because it was still in the queue for monitoring in the middle of the night when everybody at LLog headquarters was asleep.]

    My earlier comment appears to have been bumped, probably because I tried to include a bunch of links to citations. Trying again without the links.

    @J. Marshall Unger:
    He should have emphasized that Martine Robbeets’s Transeurasian is an effort to start fresh, this time including Korean and Japanese as witness languages and all the new data collected since Poppe, and “rebuild” the “Altaic” reconstruction.

    As Juha Janhunen appears to suggest above ("However, it seems that she just adopted the new term without changing the framework"), most of what I've seen regarding Martine Robbeets's "Transeurasian" doesn't suggest much "starting fresh" at all, and seems to me to be less about rebuilding Altaic from the ground up than it is about rebranding old Altaic work with a shiny new Transeurasian identity to make it easier to swallow. Much of Robbeets's Transeurasian is built fairly uncritically upon the pro-Altaic Moscow school, and thereby begs the Altaic question, not doing nearly enough to dispense with the null hypothesis of large-scale interborrowing that lies at the core of the question of these languages' relatedness. Without that, it strikes me as rather like trying to prove the unity of (to use an example more familiar to me) an Ibero-Caucasian family without even confirming if the contentious North Caucasian is a valid phylogenetic node.

    Further, others have published reviews ranging from unenthusiastic at best to caustic at worst, levelling various criticisms including misquotation of opponents, omission of crucial sources, failure to adhere to its own methods, failure to cross-check actual language data, and ignorance of key issues even within Robbeets's own speciality of Japonic. Honestly, Merritt Ruhlen's On the Origin of Languages was the last time I remember seeing so many separate authors go so hard on a single work.

    (Obligatory disclaimer: I'm not an expert in any of the families involved in Transeurasian – my speciality is Ubykh (North-West Caucasian) – but I have a pretty good grasp on the broad strokes of Turkic; Ubykh's multiple complex strata of Turkic loans have necessitated deep diving into Turkic etymological materials, and that in its turn has required I at least familiarise myself with Altaic and further hypotheses and the associated literature.)

    Finally, in response to Peter Golden's question to the floor (on the term "Transeurasian"), for my part I'm not a fan at all; it's misleading and inaccurate. Even taking the non-Eurocentric one of the two possible senses it could carry, the European part of Eurasia is represented in "Transeurasian" only by a comparatively late Turkic expansion during the historical era, and only into a small part of Europe at that; indeed, one could make the argument that Transeurasian isn't even as trans-Eurasian as, say, Indo-European is. For mine, the putative Transeurasian family doesn't cover enough geographic extent, language families, languages, or speakers to justify its name in any of those senses. The Trans-New Guinea phylum, by way of comparison, does at least extend across more than 90% of that island's length, and includes between 40% and 60% of the thousand or so languages spoken there, depending on whose model one accepts (much work remains to be done).

  4. R. Fenwick said,

    December 11, 2020 @ 10:04 am

    Ah, apologies for the double-up. I'd thought some sort of comment-cutting algorithm had been responsible because I'd actually seen the comment appear at first, but subsequently disappear within some seconds, hence the follow-up comment (reworked in spots for little other reason than clinical anxiety). I didn't mean to imply an impatience at the good folk of LLog not working during their sleep, and am sorry if that was what came across. I'm very grateful for the great work done at the LLog HQ – keep it up!

  5. Jim Unger said,

    December 11, 2020 @ 11:31 am

    Victor took my brief, informal comment from a personal e-mail and posted it here. I merely wanted to express my understanding of Robbeets's research outlook. Personally, I still maintain the skeptical attitude I took in my chapter "Japanese and what other Altaic Languages?" in Phil Baldi's 1990 Trends in Linguistics volume (Mouton). I think the proto-Korean-Japanese reconstruction as lately revised by Alexander Francis-Ratte (amending earlier comparisons by Martin, Whitman, and others) makes it unreasonable to deny a genetic relationship. Robbeets thinks pKJ was the earliest branch of the Transeurasian to break away, but it seems to me that we need to test whether or not a Macro-Tungusic (Benzing's Tg languages plus K and J) can be rigorously worked out before making that claim. As I wrote in a different place in the Baldi volume, "language history is not a big empty canvas to be filled with broad brushstrokes and daring, sweeping lines, but a broken mosaic that must be reassembled methodically, bit by bit."

  6. Sally Thomason said,

    December 11, 2020 @ 11:54 am

    Peter Golden asks how a lot of borrowing could occur when the people were nomads living in groups of 4-5 households. Unless they intermarried exclusively within those 4-5 households, there must have been many opportunities for transfering linguistic features. And weren't there also trading relations beyond the 4-5-household living groups?

  7. Penglin Wang said,

    December 11, 2020 @ 8:48 pm

    It is premature to declare the fall of the common Altaic theory pioneered by G. J. Ramstedt and N. N. Poppe. By Altaic I mean to include Turkic, Mongolic, and Tungusic groups. We should go through the accumulated Altaic scholarship to see what pros and cons are there in the theory. Among the works of Altaicists, those of Poppe had been of the most general and fundamental significance in inheriting what has been done in the past and ushering what should be done in the future by describing Mongolic and advancing the Altaic theory. As I wrote somewhere in my publications, the Altaic genealogical unity is reasonable and tenable. However, Altaic comparative studies had proved inadequate for the aimed task–the contradiction of theorization with linguistic connections was overwhelming and many interested students yearning for a better method to resolve the issues. Notably, the two main pillars–the Ramstedt-Pelliot’s hypothesis and sigmatism/zetacism–in support of the Altaic theory that provide focal points for the development of sound correspondences are wrong. Henceforth, the validity of the Altaic theory cannot be assessed by relying on the wrong organization of sound changes or correspondences. The Ramstedt-Pelliot’s hypothesis (or the so-called Ramstedt-Pelliot’s law) refers to the sound change of *p > f > h > 0. Since sound change and its possible resultant correspondences are a core subject, not just one of many things that are to be discussed, in comparative studies, I have been working to question the validity of the two pillars and presented the Altaic sound change of k/q > h > 0 and of rhotacism/lambdacism with a host of examples.

    The common assumption in Altaic studies is that a sound change intended to support the genetic unity of Altaic languages can be hypothetical reconstructions, which are partial in the sense of not involving all the groups or leaving aside a group. One notable example is the study of Middle Mongolian (MMo) h- and its eventual exclusion of the Turkic group out of the Altaic loop. We do not necessarily agree with this school of though and some other adherents of the Altaic theory that it is the hypothetical reconstruction that really matters. Among most of the interested researchers, Altaic linguistics has largely been about down-to-earth matters such as gathering data, analyzing data, and producing research findings. Most objective researchers who work on Altaic in comparative terms would be able to agree that Turkic, Mongolic, and Tungusic go stringed together in the domain of change from k/q to h and that neither can be well understood without the other.

    Words are made up of a variable combination of phonetic segments that are essential part of phonological system of a language. We now understand that Turkic, at least in its old stage, was conservative in word-initial position occupied by q/k, while Mongolic and Tungusic were relatively innovative. Phonetic segments of a language vary greatly in their functional load in certain positions at a given period of historical development. By going through ДТС (Дreвнетюркский cоварь (Old Turkic Dictionary), 1969), I found the q/k in Old Turkic very active in word-initial position. In ДТС with 643 pages for word entries, the words beginning with k have forty-four pages (288-332), and those beginning with q occupy seventy-six pages, together making up a total of 120 pages presenting 18.7 percent of the pages for entries in the dictionary. In contrast, the words beginning with h and ḥ receive only four pages (197-200) filled in by largely foreign, Arabic, loanwords. There seems a few native Turkic words such as hana ‘mother’ and hata ‘father’ alternating with ana and ata, respectively. This systematic distribution of the relevant consonants signals the recalcitrance of the Old Turkic (OT) q/k to erosion to the fricative h or the like. As a further phonotactics, in some crosslinguistic borrowing the foreign word beginning with h- presented itself with q. For example, Chinese huma (胡麻) ‘Hu hemp, Hu flax’ and hongdou (紅豆) became OT quma and qundu in the same sense (ДТС 465, 466). But this is not a clear-cut indication of the phonotactics because we see Chinese huashen (化身) ‘incarnation’ was taken as χuašin ? in Old Turkic (ДТС 638). The way Old Turkic responded to Chinese huma is similar to that of Dagur kima from Chinese huma.

    The phonological evidence for the Old Turkic q variation toward χ is in the feature relative to back vowels. The χ or x, depending on the style of transcription by individual researchers, seems to be a fricative. ДТС uses χ, and Clauson (1972) has x. To study sound change, it is necessary first to go through the data in relevant languages. Because a sound change may be a lasting phonological process over time, there could be a series of supporting examples. In ДТС (635-637) I found phonetic variation of q in the direction of χ such as qalï ~ χalï ‘if’, qïz ~ χïz ‘girl’, qamağ ~ qamuğ ~ χamağ ‘all’, and qïtay ~ χïtay ‘Kitan’. These variations constitute pieces of solid evidence for the origin of MMo h- from k/q, which has nothing to do with the *p. Unfortunately, to the best of my knowledge, no one in Altaic studies had done such a work to shed light on MMo h-. In short, the change from k/q to h or χ is a predictable and attested change. It operated first on phonetic variation, already present in Old Turkic. The similar variation took place in Old Turkic as well between q and ğ, e.g., OT qïsqa ~ qïsğa ‘short’ (ДТС 448) and some other ones. The emphasis on Old Turkic may not seem obvious at first, but an example can demonstrate the point. Suppose you support the common Altaic theory advocated by Ramstedt and Poppe. If you follow them in reconstructing the *p- to connect the three groups without considering its consequence, do you think the hypothetical Mongolic form *pulağan ‘red’ can fulfill its purpose to support the Altaic unity in the face of OT qïz- ‘to redden’? Not all. Instead, MMo hulağan came from *hulğan (from *hus-ğan) through lambdacism occurred in the postvocalic position in comparison with OT qïz-. The rhotacism/lambdacism goes beyond Altaic. To exemplify this change, just consider how Old Turkic alma ‘apple’ derived from Arabic asmār ‘fruits’ (for details, see

    In my articles (Wang 1993, 1992) I dealt with the MMo h- by restoring the place and role of Turkic group in the Altaic sound change from k/q. I took note of the most elementary dimensions of the sound change of h- occurring in MMo and its continuation in modern Mongolic languages and then moved to those running through Turkic, Mongolic, and Tungusic.

    Pursuit of scholarship and truth may be disrupted with prejudice and misjudgment. My approach to MMo h- met with criticism by Janhunen (1999). As I publish my research findings, I welcome criticism as far as it is of scientific essence and fairness. Janhunen’s (1999:116, 118) article about MMo h- includes two threads of observation as follows:

    “the whole development from Pre-Proto-Mongolic to Modern Mongolic may be reconstructed as *p > *f > (*)h > 0”.

    “All of this is now common knowledge, and any radical reinterpretation of the situation, such as that proposed by Wang Penglin (1993), can only be based on an insufficient familiarity with the linguistic facts”

    As far as the article goes, these are conclusive remarks of what Janhunen is talking about. It does not, however, say anything concerning the linguistic facts underlying the origin of MMo initial fricative h- that Janhunen held to be at the heart of Mongolic historical linguistics to the exclusion of Altaic comparative studies. The whole development? Not really. As Janhunen rejects the common Altaic theory, although he has adopted the Altaic sound change proposed by the Altaicists, he has to transplant it from Altaic to Mongolic. Therefore, his effort fits very poorly and misleadingly.

    Holistic capturing sound changes so as to generalize concerning their characteristics requires considerable expertise and skill. A genuine familiarity with the linguistic facts must be manifested with concrete presentation of linguistic data drawn from the concerned languages. While Janhunen as a reconstructionist has enthusiastically pursued his work, he hardly asked himself what ‘familiarity with the linguistic facts’ means and what ‘the linguistic facts’ are. In this particular case, no single linguistic fact was presented to support the hypothetical *p. So the issue of ‘familiarity’, that is, whether or not the reconstructionist is really familiar with what they are claiming to be familiar, hardly ever arises. For Janhunen at the moment of his writing, ‘familiarity’ meant to ‘imagination’ at worst and ‘reconstruct’ at best, and ‘linguistic facts’ were an endless series of the hypothetical reconstructions with the signs * throughout his seventeen pages. To his initial unfair calling out me for ‘an insufficient familiarity with the linguistic facts’ was added the still further bias by keeping deadly silent to his ‘familiarity’ when Janhunen swiped out his favorite *p- in his 2003 work by actually adopting my research findings, for he lost courage to explain why he abandoned his favorite stand and to identify the source of his ‘familiarity’. This is not to suggest that such a bias vitiates the results: it does not appear to do so because Janhunen was no longer able to enjoy selling and repeating his *p-. Rather, it is to remind that the kinds of professional concerns and responsibility researchers have about innovative approaches and their sources have not been the same as those of laymen.


    Clauson, Gerard. 1972. An Etymological Dictionary of Pre-Thirteenth-Century Turkish. London: Oxford University Press.
    ДТС = Дreвнетюркский cоварь (Old Turkic Dictionary), 1969. Лeнингpaд: Наука
    Janhunen, Juha. 1999. Laryngeals and pseudolaryngeals in Mongolic Problems of phonological interpretation. Central Asiatic Journal 43.1:115-131.
    Janhunen, Juha. ed. 2003. The Mongolic Languages. London. Routledge.
    Wang, Penglin. 1993. On limitations of Ramstedt’s Hypothesis concerning the Middle Mongolian initial h-. Contacts Between Cultures–Eastern Asia: Literature and Humanities, Volume 3, edited by Bernard Hung-Kay Luk, 380-384. Lewiston: The Edwin Mellen Press.
    Wang, Penglin. 1992. On the origin of the Middle Mongolian initial h- and the motivation for its loss. Archív Orientální 60.4:389-408.

  8. ohwilleke said,

    December 12, 2020 @ 7:11 pm

    "how does that occur in the steppe?"

    Multiethnic coalition military forces were attested at least from the Huns through the Mongols and were very likely present in some Bronze Age armies of nomads as well.

    Also, many early Indo-European cultures practiced long distance bride exchange between patrilineal tribes and it isn't unlikely that ancient proto-Turks and proto-Mongols might have done so as well.

  9. Victor Mair said,

    December 13, 2020 @ 7:09 am

    From Alexander Vovin:

    Further on stone.

    As Juha has already said, there is a problem with regularity of correspondences.

    I can further add that Japonic and Koreanic words have additional problems: Middle Korean :torh (with a rising tone) clearly indicates that the word was disyllabic *torVk, which is a nominal derivation of the verb tor- 'turn around' 'rotate', 'spin', Rolling [stone] in other words. Eastern Old Japanese has osi [əsi] and osu [əsu] 'rock', and the Northernmost Ryukyu dialect of Kikaijima has esi (althhough this is sometimes contested. KhT inscription (early 7th c.) has ǰilo that rules out a comparison with Turkic taš.


    It is a misnomer. All 'Altaic' families originated in Asia, and Turkic and Mongolic language found in Europe are results of recent migrations.

    Roy A. Miller

    Except for his criticism of EDAL and Robbeets, Miller's research is problematic, including Japanese and the Other Altaic Languages. There are too many ad hoc hypotheses unsubstantiated by data, and too many inaccuracies starting with data citation and interpretation. I proudly keep my copy of it on my linguistic SF shelf together with EDAL, Robbeets, etc.

    Andrew Shimunek

    Andrew started with a brilliant MA thesis, and after that it was a huge slide downwards. The reason is that he came under the influence of Beckwith. Thus he uses Beckwith Chinese reconstruction, which is not a reconstruction at all, at least not in the sense of Karlgren, Pulleyblank, Baxter/Sagart, Zhengzhang Shangfang, Li Fang-Kuei, Coblin, etc. Consequently, Shimunek's Chinese reconstructions are not to be trusted. This is further aggravated by a garbled periodisation of the Chinese Language history. The main advantage of this book is that it conveniently collects almost all non-Kitan data on para-Mongolic in one place, but his treatment of Kitan itself is highly questionable. On recent developments in the Kitan linguistics I highly recommend 清格尓泰,呉英喆,吉如何 2018. 《契丹小字再研究》3 volumes. 呼和浩特:内蒙古大學出版社。as well as numerous articles by András Róna-Tas.

  10. Victor Mair said,

    December 13, 2020 @ 7:11 am

    From Peter Golden:

    Dear Sasha,

    Many thanks for your most useful comments. András wrote a critical review of Shimunek published in Archivum Eurasiae Medii Aevi 24 (2018): 315-335, which is followed in vol. 25 (in press) with addenda/corrigenda to the review and solutions to technical problems connected with Kitan script. Microsoft Word for Mac and Microsoft Word for Windows do not always “play well together” and in ”trans-platforming" from PCs to Macs the Khitan glyphs became Chinese. I am following the growing literature on Kitan (András, Kane and others) with great interest.

    I have known Chris Beckwith for many years and have read his works with an appreciation of some elements and more than a little head scratching at others. His insistence on the old Wade-Giles system of transcription of Chinese rather than Pinyin (Shimunek follows him) is one of his idiosyncrasies, for which he has never made a convincing case. It is not an end of the world issue, but says something about Chris. As a non-Sinologist who uses Karlgren, Pulleyblank, Schuessler, Baxter/Sagart, Kroll and Coblin to get some sense of the OC, LH, MC (EMC and LMC) reconstructions of what is masked by the Chinese renderings of foreign names (Turkic in particular), I cannot speak with any authority about Beckwith’s system of reconstruction. He has never, to my knowledge, published a substantial work on it. I am aware that it differs from the others. I duly note all the forms when given. However much I am tempted, I am not about to begin the serious study of Chinese at my age. My wife, a native of Shanghai, has been after me to do so for the nearly 60 years that we have been together. My standard excuse is that the Georgian verb destroyed part of my brain. Keşke bilseydim! Fortunately, she helps me get through some parts of texts.

  11. Victor Mair said,

    December 13, 2020 @ 7:14 am

    From Peter Golden:

    I read Chris’s works with great interest and appreciation for the boldness of a number of his ideas and his erudition. As Victor writes, his ideas "are perceptive, worthwhile, and stimulating” — and to be read with one’s critical faculties working full-time. We have published several of his articles (dealing with the ethnonyms Tabghach, Türk etc.) in Archivum Eurasiae Medii Aevi. Although I may not find all of his conclusions on the mark; they make me think and question my own views, not to mention those of others. That is all to the good. Chris and I have had ongoing discussions on what lurks behind 阿史那 Ashina, the name of the ruling clan/tribe of the Türks. He published his latest views on the subject in the Festschrift for me. I still have my doubts regarding his conclusions, the basic thrust of which was expressed in his writings decades ago, but I read the article with pleasure and have already cited it in writings that touch on the subject – as one of a number of possibilities. I have never had a discussion with Chris (and they have been in places ranging from Bloomington, Indiana to Paris) that I have not enjoyed and from which I have not learned something.

    This is a wonderful discussion and I look forward to learning more.

  12. Victor Mair said,

    December 13, 2020 @ 7:15 am

    From Alexander Vovin:

    I think that Beckwith is a very interesting historian (as far as I can judge not being one myself — some of his books are very interesting reading, imho), but when he starts to talk about historical linguistics, whether it is Chinese, Japanese, Turkic, Mongolic, etc., it is methodologically simply not acceptable and it is further aggravated by the corruption of data. His insistence on Wade-Giles is just one of the most benign parts of his approach. I think that his choice is political. Well, I have no love for PRC lost, but I trust that Pinyin is more convenient and more logical.

    I am well aware of András's review of Shimunek. There is also one by Rykin and I am slowly preparing my own, where I will mostly address Sinological problems.

  13. Chris Button said,

    December 13, 2020 @ 11:41 am

    Of the names above, the only one I'm aware of who could make a completely valid claim to being both a historian and a linguist is Edwin Pulleyblank.

    The reason Pulleyblank never published an Old Chinese lexicon like his one for Early Middle Chinese and later stages is because he never felt there was enough evidence. That's admirable, but it's also a shame for people trying to come to grips with his approaches today. I think he did want to publish at least some kind of final statement on Old Chinese, but that unfortunately was not to be.

    For me, the only certain thing about Old Chinese reconstruction is that one day–when the whole mess has been largely figure out–someone is going to look back and realize Pulleyblank had, in one form or another, already figured a substantial portion of it out. His approach just won't seem so boundary-pushing whenever that time in the future may be.

RSS feed for comments on this post