Typology of Northern Eurasia: Vocalism

Even among scholars who don’t consider ‘Altaic languages’ to be a valid cladistic grouping it is common to think of these languages as at least typologically similar. Thus, it would be useful to sketch out what these common traits are, to what extent are they found in other (vaguely) neighbouring languages, and more importantly, what differences are found between them. A good starting point would be comparing the vowel systems.

Turkic

The system usually considered most representative in naive approaches is that of Turkic, which contains in its most minimal form a perfect cube of 8 vowels distinguished by [+/-high][+/-front] and [+/-rounded]: /*a *ä *o *ö *u *ü *i *ï/.

However there are further complications, first of which being that there was also a closed *e with reflexes different from *ä in some languages, being written as i in Old Turkic and merging with i in Chuvash for example. Yet, some researchers have proposed that this phoneme is produced by an early soundlaw *ä: > *e:, however others claim that both pairs of *ä *ä: and *e *e: are distinguished.

Another proposed phoneme is *ë, a mid-back unrounded vowel, meant to explain the correspondence of Chuvash ï ~ Yakut ï ~ Other languages a, as well  as cases of a appearing in some Old Turkic inscriptions, Chuvash, Yakut, Khalaj or Mongolian where other languages have ï. Nonetheless, these correspondences are imperfect and this phoneme is not accepted by all (or even most?) Turkologists.

And the third proposed addition is *ia which is introduced to explain cases of Chuvash Ću ~ Common Turkic *Ca ~ Mongolic *Cï where Ć symbolizes  palatalization of the preceding consonant and u being the normal reflex of Turkic *a in Chuvash. [1] There is a theory that seeks to explain this phenomenon through regular breaking of *a: > *ia in Oghur, however sifting through Turkic comparative materials it is easy to find many exceptions to this rule.

Mongolic

Similar to Turkic, Proto-Mongolic vowel system is commonly denoted as /*a *e *o *ö *u *ü *i *ï/, however in most Mongolic languages ‘front’ and ‘back’ vowels are in fact distinguished by vowel height/pharyngealization/tongue root position. For example, the vowel system of Chakhar Mongolian is /ɑ ə ɔ o ʊ u i ɪ/, this is traditionally called ‘rotation’ with the idea that previously front vowels backed to their current position.

However as this backing is less phonetically plausible than the reverse direction, it is increasingly popular to posit that Proto-Mongolic was based on tongue root harmony. Yet this is difficult to reconcile with Middle Mongolian, where the spelling of *ü implies palatality in Uighur script (ui), ‘Phags-pa (eeu) and Sino-Mongolian (using characters with /y/ and /ju/ medials although plain /u/ is also attested).  Another important bit of data here is Khitan, which predates Proto-Mongolic as commonly reconstructed, that has both tongue root and (secondary) palatality contrasts according to Ōtake Masami.

And another smaller question is whether *ï is a phoneme or merely an allophone of *i occuring in ‘back’ vowel words. Even though no language preserves the distinction as phonemic, it seems more convenient to reconstruct due to cases like *čïkïn ‘ear’ and *čïkï- ‘to squeeze’ appearing with uvulars in Middle Mongol and *čïsun ‘blood’ and *(s)isün ‘soot’ having the same suffix but belonging to different classes.

Diphthongoids, including double vowels, are in Mongolic in some cases contractions of earlier *-VCV- sequences as I’ve explained earlier, currently I know of no reason to presume there were any such sequences of different origin. (I may cover supersegmentals in another post)

In a sense, Mongolic vocalism stands between Turkic and Tungusic, which should not be surprising.

Tungusic

Unlike Mongolic, I have no doubt at all that Tungusic vocalism is based on tongue root position rather than palatality. As a maximal reconstruction I would propose:/*a *ə *o *ȯ *ụ *u *ü *ị *i/, where dots below and above represent retracted and advanced tongue root respectively (caveat: some sources use the dots with the exact opposite meaning) and *ü is an actual [y] or at least [ʉ].

Here we have a similar situation as with Mongolic *ï, retracted high vowels *ụ and  *ị are likely merely allophones of *u and *i, recessively taking on retraction of other vowels in the word. In cases where there are only high vowels, they are advanced in monosyllabics and retracted otherwise, at least this is how the theory goes.

*ȯ is also not accepted by all sources, but there’s a number of fairly solid correspondence sets Evenki u ~ Even ө ~ Orok o ~ Nanai u that seem most elegantly explained through such a reconstruction.

The only diphthong accepted by all sources is *ia which has a unique reflex in all branches. Some reconstructions also have *iə, *ua, *uə, *ui and perhaps *ai. These are usually best preserved in Nanaic.

Korean

Late Middle Korean vowel system is often transcribed as /a e o wo u wu i/ supposedly representing [a ə ʌ o ɨ u i], with vowel harmony grouping low(er) a o and wo against higher e u wu and i being neutral as also suggested by some analysis of Mongolic and Tungusic. However some scholars have suggested that Korean vowel harmony is a Middle Korean innovation.

Korean also features many diphthongs with on- and offglides y and w. Alexander Francis Ratte in his dissertation on Proto-Korean-Japanese seems to consider all of these secondary (deriving from consonants and contractions of multiple syllables) except for ye which he derives from Proto-Korean *e. [2]

Japanese

It seems well accepted that Proto-Japanese vowels are /*a *e *ə *o *u *i/ , and that the more complex system of Old Japanese vowels arose through contraction of earlier diphthongs (see for example Proto-Japonic *e and *o in Eastern Old Japanese).

Old Japanese also features a limitation on vowels co-occuring in roots that is sometimes compared with vowel harmony. According to Ratte, it amounts to *ə never co-occuring with *a, *o and *u, indeed this doesn’t seem to have much in common with other cases of Eurasic vowel harmony and he explains it through an earlier assimilation of *ə around these vowels, which does appear to be a priori likelier.

Uralic

The Uralic system consists of /*a *ä *e *ë *o *u *ü *i/, frontness being the distinguishing feature between umlauted pairs and with *ï sometimes added or *ë sometimes denied.

Despite incompleteness, this seems most similar to Turkic vocalism among neighboring languages (excluding traditional reconstructions of Mongolic). Perhaps it could be speculated that Turkic vocalism is itself a product of Uralic (or specifically Samoyedic) influence.

Indo-European

Proto-Indo-European possesses a minimalist inventory of /e *o *ē *ō/, marginally and controversially  *a ā, and *i *u as syllabic glides. Considering ablaut alternations, it is probable that it previously had even fewer vowels (and if glides don’t derive from breaking of some sort).

If Indo-European is related to some of the Eastern languages (Indo-Uralic being a popular hypothesis), clearly it would have to had gone through a complete collapse of its vowel system, perhaps under the influence of Caucasian languages (although Chirikba reconstructs pre-West Caucasian with *ö and *ü).

Even though early Turks partly derive genetically from Andronovo culture populations, who presumably spoke Indo-Iranian, their reduced vowel system hasn’t left a recognizable trace.

Yeniseian

I haven’t read the literature on comparative Yeniseian phonology, but from the bits and pieces I’ve seen, it seem that Proto-Yeniseian is taken to have the same inventory as Ket: /a *e *ə *o *ɨ *u *i/ with /ə ɨ/ possibly being /ɤ *ɯ/. [3]

I am hoping that investigation of the Dene-Yeniseian connection will offer further glimpses into the past of Yeniseian itself, and hopefully into the deep prehistory of Siberia.

Yukaghir

For Proto-Yukaghir Nikolaeva reconstructs /*a *e *o *ö *u *ü *i *ï (her *y)/, a full cubic system based on frontness. However this doesn’t describe the situation as found in neither Tundra nor Kolyma Yukaghir.

This back vocalic *ï surfaces as normal i, broken ya or a as well as lowered e, which is to say, in every way except as back vocalic. Nikolaeva also mentions that “In the modern languages u is  harmonically back in only a few words, mostly from Tundra Yukaghir. Such words are normally fairly recent borrowings from Tungus”, which necessitates reconstructing *ü, which reflected as a back vowel in both languages but belongs to the harmonically “front” class.

Unlike *ü, *e and *ö are found as actual front vowels in both Tundra and Kolyma, yet Nikolaeva mentions that *e was written with a or o in the earliest attestation, and while she interprets this as indicating a lower /æ/, it suggests a central /ə/  to me.

Chukotko-Kamchatkan

Fortescue gives /*a *æ *e *ə *o *u *i/ for Proto-Chukotko-Kamchatkan with recessive vowel harmony (umlaut?) based on ‘dominant’ *a *e *o against ‘recessive’ *æ *i *u and *ə being a reduced vowel acting different from full vowels.

Eskimo-Aleut

Proto-Eskimo-Aleut had /*a *ə *u *i/, the simplest (outside of Proto-Indo-European) of the vowel systems discussed here. Perhaps this is a result of a collapse of a Chukotkan-like system as happened in Kerek and Alutor, but this is baseless speculation.

Nivkh

Nivkh possesses an ideal height harmony system: /*a *ə *o *u *e *i/ which is reflected basically as such in all dialects, although actual harmonic alternations themselves have been lost in most cases.

Ainu

Most Ainu dialects apparently have simple vocalism, /*a *e *o *u *i/, but in a 1992 paper Vovin reconstructs a more complicated system with four height levels, front rounded vowels, a back unrounded *ï and two low vowels. I do not know how accepted these ideas are among other scholars.

Sino-Tibetan

Old Chinese is reconstructed with /a ə o u e i/. Marc Miyake presents an interesting idea that Old Chinese had an opposition between high and low vowels and that pharyngealization in late Old Chinese was caused by low vowels in presyllables which were lost.

Tangut also had a tense/lax distinction which might also be related to the Chinese situation, but Tangut materials are too opaque for me to grok.

Conclusion

By now, everyone interested in this type of stuff heard of the idea of advanced tongue root harmony for Proto-Mongolic [4]. Yet, it’s acceptance hasn’t been universal. In a relatively recent paper Janhunen defends original palatal harmony not only in Mongolic but Tungusic, Korean and Nivkh by noting

There are, indeed, good arguments, both internal and external, for assuming that vowel rotation was a process that was active in Northeast Asia, and that started relatively late, reaching different languages and language families in different times..

To me, it seems more inherently more plausible that a rotation happened in reverse direction, and quite unacceptable to overrule the testimony of attested language by what I suspect to be a projection archetypal Ural-Altaic languages based on familiarity. There must be something beyond that, but I am yet to hear these arguments.

Footnotes

[1] Some sources would reconstruct other *iV diphthongs such as *io or *iu, but if diagnostic palatalization in Chuvash appears in these positions as well, perhaps we would be better off reconstructing palatalized consonants.

[2] Even if Japanese and Korean may not form a valid genetic grouping, Ratte’s work seems to grasp at something real, presumably due to shared history between Korean and Japanese, just as comparisons of Turkic, Mongolic and Tungusic offer insights into their prehistory that would be impossible otherwise.

[3] This is Vajda’s analysis. There are alternative takes, covered by Vajda here, but his appears most convincing.

[4] One of the earliest appearances of this theory is this paper, written by an author with a primary focus on phonology.

21 thoughts on “Typology of Northern Eurasia: Vocalism

  1. I would add to this that IE, Uralic and Sino-Tibetan are clearly older than most other families here. Many of their individual branches show interesting deviations from the proto-language that could add some more detail to the big picture. A few examples from the first two:
    Samoyedic: overall shape of the PU system retained faithfully, but *i and *u partly reduce to *ə; a distinction *a / *å is introduced (surfaces in Nenets-Enets as a palatalization contrast, elsewhere as a labiality contrast)
    Mansi: reorganization into a two-height system with a general length contrast and not much of a roundedness contrast, basic organization as *i *ii *ïï *u *uu; *ä *ää *a *å *aa. Used to be reconstructed as a bigger system, but most of this is reanalysable by recognizing a larger system of velars (e.g. *küü : *kuu can be re-mapped as *kuu : *quu, where the former only occurs in loans from Komi and Russian; *kü : *kä from PU *kü : *ki can be IMO re-mapped as *kʷä : kä).
    Permic: major reshuffling, vowel harmony eliminated in favor of a central vowel series: at minimum with *ɨ *ə, per most researchers including also a roundedness contrast with *ʉ (survives in dialectal Udmurt) and *ɵ (no general contrast with *ə retained, but some lects merge them as [ɵ] rather than [ə]).
    Indo-Iranian: “second” vowel collapse into *a *ā *ai *au *ə *i *ī *u *ū (some people claim that *ī *ū and some *ā were still *iH *uH *aH, but as stated this seems incompatible with the fact that the first two still trigger RUKI); diphthongs later shift to new /eː oː/ pretty much universally; all sorts of umlauts later on especially in the Eastern Iranian–Nuristani–Dardic zone, with some fairly complex modern inventories
    Tocharian: massive vowel rotation, including *ĕ > *i and subsequent unconditional high vowel reduction; mid vowel system simplified to just *ə (usually transcribed *a, and real *[a] as *ā, per Indic script customs)

    You hint at pharyngealization contrasts in earlier Yukaghir (which I’ve suspected too). I also wonder if things like Tocharian or Permic could be reanalyzed more straightforwardly with the help of this “technology”. E.g. in the latter *i *ɨ *ʉ *u → *i *ị *u *ụ?

    Liked by 1 person

    1. Yeah, that’s something I should have noted. Thanks for making it more complete.

      I’m not seeing what you mean about Tocharian, should pharyngealization apply only to non-high vowels (like in Tungusic maybe)?

      I welcome any attempt to make Permic vocalism less complicated. Are there any known vowel assimilation phenomena that could be explained through pharyngealization?

      A thing that seems to reoccur in these systems is the ‘neutral i’ or at least ‘recessively front/high i’. What’s the state of this in Uralic languages (that have anything resembling harmony)? I know at least Hungarian and Finnish have neutral /i/ and /e/ but I’m guessing from what I know about Proto-Uralic that this is secondary.

      Like

  2. Permic has a fair bit of metaphony going on that doesn’t appear to be explainable as simply regular old i-umlaut or a-umlaut, e.g. *ä-ə > *ɨ but *ä-ä > *ɛ; *o-ə > *ɵ but *o-a > *u. I don’t have any ready-made theory to pitch though, I only know that something is off (but yeah that’s kind of obvious to begin with).

    With Tocharian one thing I have in mind is that *e seems to be unstable in pharyngealization-based systems and tends to merge with either *ə or *i, but maybe this is a wrong door to bark down.

    On neutral /i/ — Hungarian gets it through a merger of earlier *i and *ɨ (< PU *e and *ë; PU *i goes to OHu †ɪ > /e/). A similar but not identical merger occurs also in Nganasan: *e and *ï > /ɨ/ (further > /i/ in palatalizing environments). Finnic may have gone through this too, but I have an alternate theory in the works where it’s rather the harmony itself that gets disrupted here by some unrelated processes.

    Erzya synchronically basically has CV harmony, which only ends up looking like vowel harmony since stem-final palatalized consonants have usually developed due to preceding front vowels. So, schematically, *ECV > *EĆV > EĆe — but also *AĆV > AĆe when there are original Uralic palatals around.

    Southern Mansi had harmony, but arguably got rid of /i/ entirely at one point! [i iː] arise from /ɛ eː/, partially obscured by fronting of *ëë and shortening of *ää. Also, SMs was heavily Tatar-influenced, so I’m skeptical on treating its harmony as inherited. The alleged collapse of harmony in the rest of Mansi is very clean, with no traces of gradual attrition as can be found in varieties like Southern Khanty, Western North Estonian, or Veps.

    Harmony is very thorough in Eastern Khanty with no neutral vowels. I think /i/ consistently counts as front also in all of Mari; and also in Kamassian, which I only recently found out has vowel harmony. This despite how it has a “Hungarian-esque” merger of PSmy *i and *ï. Other secondarily fronted vowels like *åj > /e/ or *åə > /ü/ trigger front harmony too.

    Liked by 1 person

    1. Perhaps *-ə went through Finno-Permic *-i causing palatalization at some stage but then backed to *-ɨ (together with *ü and *e) and caused centralization and/or ATR thenceforth. At least that’s the pattern I think I see alt-tabbing frantically between the FrathWiki page and Sammallahti.

      I suppose this ‘unstable e’ could corroborate ‘neutral i’ being a minor attractor state if true.

      Interesting, I forgot that *ë goes to Hungarian i, that could explain how alleged *ë words in Turkic could show Hungarian i ~ Chuvash ï but ~ medieval Volga Bulgarian a.

      Like

  3. I’m not really on board the current standard reconstruction with unstressed *-i, exactly because this doesn’t seem to regularly trigger palatalization of anything in any of the numerous central branches that delete it — unlike Finnic, where *-i of any origin triggers palatalization left and right when lost. But this is a bigger discussion than I can go into here. Suffice to say that IMO it could still have been a close vowel such as [ɨ], instead of a completely featureless schwa.

    Like

      1. If you mean the mostly unexplained initial voicing going on in Permic: no, that’s never passed my mind. Usually explanations for this have been sought from the consonant environment, and I figure some cases are just loans from IE that never lost their voicing (so probably in parallel into Permic and relatives). The examples are distributed a bit too widely across all the different vowels for that to seem any kind of promising.

        Near-minimal pairs like *künčə > *gɨdž ‘nail’ but *küntə > *kɨd ‘smoke’ are also going to be a pain to explain in any fashion at all.

        (+ATR voicing happens “in reverse” in Armenian dialects, probably not coincidentally right next door to Oghuz. I think I’ve seen an example of the Turkic vowels-to-stops direction from Bantu sometime, but alas I don’t know where to even begin to possibly relocate that.)

        Like

  4. First of all, thanks for a bunch of papers I’ll need to read!

    Another proposed phoneme is *ë, a mid-back unrounded vowel, meant to explain the correspondence of Chuvash ï ~ Yakut ï ~ Other languages a, as well as cases of a appearing in some Old Turkic inscriptions, Chuvash, Yakut, Khalaj or Mongolian where other languages have ï. Nonetheless, these correspondences are imperfect and this phoneme is not accepted by all (or even most?) Turkologists.

    At the top of Marek Stachowski’s academia.edu page, there’s a long work on the vowels of Yakut, and buried in there are Yakut-internal explanations for (apparently) all the Yakut ï ~ other languages a cases.

    regular breaking of *a: > *ia in Oghur

    That would have to go through an *[æː] stage, wouldn’t it? And that would surely count as a front vowel.

    Middle Mongolian, where the spelling of *ü implies palatality in Uighur script (ui), ‘Phags-pa (eeu) and Sino-Mongolian (using characters with /y/ and /ju/ medials although plain /u/ is also attested).

    I wonder if the use of front-vowel signs is really meant to indicate -RTR instead. After all, the Hunmin Jeong-eum calls i “with unretracted tongue”, ə o u “with slightly retracted tongue” and the other three “with retracted tongue”.

    Middle Korean transcriptions of Mongolian argue strongly against a front-back interpretation, as demonstrated in Ko Seongyeon’s work, particularly his thesis.

    As a maximal reconstruction I would propose:

    What about long vowels and diphthongs? (You’ve almost omitted all mention of length for Turkic, too.)

    Considering ablaut alternations, it is probable that [pre-IE] previously had even fewer vowels

    Given that ablaut is shared with Afro-Asiatic, West Caucasian and reportedly Kartvelian, I don’t think an explanation for its origin should be sought within the separate prehistory of IE. However, there’s increasing evidence that all the long vowels can be explained as compensatory lengthening after consonant losses, either word-finally (*-o-mj > *-ō “thematic 1sg ‘present'”, at a pre-PIE stage when syllabification was different from that of PIE) or in reduplications that left two identical consonants in the same syllable. I’ll look for my source later. – Fascinatingly, PIE may have had a contrast between |i u| and |j w| at the morphophonemic level despite lacking it at the phonemic level, judging from a few *e ~ *i alternations.

    For Proto-Yukaghir Nikolaeva reconstructs

    Nikolaeva took “Proto-” literally as “the first”, i.e. “the oldest stage that can be reached by any method of reconstruction”. That is quite distinct from the less etymological but more widespread meaning of “Proto-“, i.e. “the last common ancestor of the attested varieties”: Nikolaeva arrived at her “Proto-Yukaghir” by applying a whole lot of internal reconstruction on what others would call “Proto-Yukaghir”. Frankly, internal reconstruction has way too many degrees of freedom to be used like that.

    Most Ainu dialects apparently have simple vocalism, /*a *e *o *u *i/, but in a 1992 paper Vovin reconstructs a more complicated system with four height levels, front rounded vowels, a back unrounded *ï and two low vowels. I do not know how accepted these ideas are among other scholars.

    Boer’s book on Japonic tonology completely trashes that reconstruction, IIRC.

    Tangut also had a tense/lax distinction which might also be related to the Chinese situation, but Tangut materials are too opaque for me to grok.

    Two papers on this are forthcoming.

    Finnic may have gone through this too, but I have an alternate theory in the works where it’s rather the harmony itself that gets disrupted here by some unrelated processes.

    The Balto-Slavic substrate?

    Like

    1. At the top of Marek Stachowski’s academia.edu page, there’s a long work on the vowels of Yakut, and buried in there are Yakut-internal explanations for (apparently) all the Yakut ï ~ other languages a cases.

      Yeah, it’s controversial, I’d like to see some stats on the phenomenon before deciding either way.
      An interesting datum is that such words show i in Hungarian loans from Old Chuvash, which I reckon would make the distinction about as old as the Orkhon monuments, although this still doesn’t guarantee that it isn’t secondary.

      That would have to go through an *[æː] stage, wouldn’t it? And that would surely count as a front vowel.

      Maybe it could go through something like [ɑ:] > [ɤɑ] > [ɯɑ] > [ɨɑ] > [iɑ]. I don’t find the theory convincing either way.

      I wonder if the use of front-vowel signs is really meant to indicate -RTR instead. After all, the Hunmin Jeong-eum calls i “with unretracted tongue”, ə o u “with slightly retracted tongue” and the other three “with retracted tongue”.

      Middle Korean transcriptions of Mongolian argue strongly against a front-back interpretation, as demonstrated in Ko Seongyeon’s work, particularly his thesis.

      Yeah, we had this discussion before. I’m sold on the tongue-root theory, but I think it has its fair share of problems. I should probably do a post on them at some point for the sake of future discussions.

      What about long vowels and diphthongs? (You’ve almost omitted all mention of length for Turkic, too.)

      I intend to make a separate post for prosody, this one is long enough as it is.

      Given that ablaut is shared with Afro-Asiatic, West Caucasian and reportedly Kartvelian, I don’t think an explanation for its origin should be sought within the separate prehistory of IE. However, there’s increasing evidence that all the long vowels can be explained as compensatory lengthening after consonant losses, either word-finally (*-o-mj > *-ō “thematic 1sg ‘present’”, at a pre-PIE stage when syllabification was different from that of PIE) or in reduplications that left two identical consonants in the same syllable. I’ll look for my source later. – Fascinatingly, PIE may have had a contrast between |i u| and |j w| at the morphophonemic level despite lacking it at the phonemic level, judging from a few *e ~ *i alternations.

      I agree. It is attractive to think of PIE as a language coming from the North Eurasian typological area into the Caucasian area and experiencing certain adaptations as a result. My goal is to investigate to what extent does this area exist and what are its characteristics.

      I feel increasingly skeptical about our understanding of PIE, particularly the idea that all words originally belonged to an ablauting root paradigm. But the field is so vast that, despite dipping my toes in it, I couldn’t even start to form a coherent critique.

      Nikolaeva took “Proto-” literally as “the first”, i.e. “the oldest stage that can be reached by any method of reconstruction”. That is quite distinct from the less etymological but more widespread meaning of “Proto-“, i.e. “the last common ancestor of the attested varieties”: Nikolaeva arrived at her “Proto-Yukaghir” by applying a whole lot of internal reconstruction on what others would call “Proto-Yukaghir”. Frankly, internal reconstruction has way too many degrees of freedom to be used like that.

      The purpose of reconstruction is to explain variation among cognates to a satisfactory level, she obviously thought her model was regularized by typological considerations, but I believe these were mistaken.

      Boer’s book on Japonic tonology completely trashes that reconstruction, IIRC.

      Good to know.

      Two papers on this are forthcoming.

      Looking forward to scratching my head and comparing Middle Chinese reconstructions on Wiktionary.

      Like

    2. Given that ablaut is shared with Afro-Asiatic, West Caucasian and reportedly Kartvelian,

      The Caucasian parallels may have a point, but I’m sure it’s much too early to say anything solid about PAA vocalism. Semitic could be a part of the same areal bundle as PIE ~ PWC ~ Kartvelian, esp. when the rest of the family does not really agree: weaker ablaut in Berber that may or may not be through major vowel reduction and/or schwa epenthesis, harmony gone wild in Central Chadic, relatively stable monovocalism in at least parts of Cushitic, and Egyptian just seems like a big philological can of worms…

      (Or, alternately, just for kicks: I once did a count and Bomhard’s dictionary of Nostratic actually has more of both Semitic-IE and Semitic-Dravidian than Semitic-other-AA comparanda. Since he also proposes exactly the same phoneme inventory and pretty much exactly the same grammatical elements for PN and PAA, this could well be inverted so that it’s not that Afrasian is actually a part of Nostratic — but that Nostratic is a branch of Afrasian.)

      The Balto-Slavic substrate?

      Not even close, but for now I think I’ll keep the details under the wraps.

      Liked by 1 person

      1. Huh, I knew that Afro-Asiatic was in a bad shape, but I wasn’t aware that templatic morphology wasn’t in, I figured “morphology is all we have” included typology as well.

        Like

  5. Old Japanese also features a limitation on vowels co-occuring in roots that is sometimes compared with vowel harmony. According to Ratte, it amounts to *ə never co-occuring with *a, *o and *u, indeed this doesn’t seem to have much in common with other cases of Eurasic vowel harmony and he explains it through an earlier assimilation of *ə around these vowels, which does appear to be a priori likelier.

    At the same time, though, the vowel correspondences to Korean assumed/corroborated in this Ratte paper are interesting; they might mean that the Old Japanese lack of harmony, like the Modern Korean one, is the result of a few mergers in an originally harmonic system, or conversely that vowel harmony was interpreted into Korean as an areal/substrate feature.

    (At this occasion I’d like to register my loathing for the Yale transcription of Korean. It’s completely internally consistent, and yet so constantly misleading. I’d rather spend the proverbial morning to learn Hangeul!)

    Like

    1. in this Ratte paper

      And the others, of course. But do you know where I can find his thesis? It’s not on his academia.edu page that you’ve linked to.

      Like

    2. Or maybe this is merely how Japanese adapted loans from a harmonic language.
      This paper on Ainu as Altaic has some interesting lookalikes, it might be of interest: https://scholarspace.manoa.hawaii.edu/handle/10125/9913
      Of course it goes without saying that the majority of them are probably spurious in every sense.

      I used to hate Yale transcription too, but it sorta grew on me.

      Here’s Ratte’s thesis: https://etd.ohiolink.edu/pg_10?2975821986945::NO:10:P10_ETD_SUBID:113782#abstract-files

      Like

      1. Or maybe this is merely how Japanese adapted loans from a harmonic language.

        Would be a pretty strange layer of loans, though, and a very early one.

        This paper on Ainu as Altaic has some interesting lookalikes, it might be of interest:

        Thanks, I’ll try to take a look at it… but not soon, it’s a whole thesis!

        Here’s Ratte’s thesis:

        Awesome, thanks!

        Like

  6. In a relatively recent paper Janhunen defends original palatal harmony not only in Mongolic but Tungusic, Korean and Nivkh by noting

    I’ve now read the paper. The argument for rotation seem to be the most common mergers, which can be at least as easily explained the way Ko Seongyeon does it.

    Like

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

This site uses Akismet to reduce spam. Learn how your comment data is processed.