26

On Perfectly Good-Looking Morphological

Comparanda and Their (Sometimes, However,
Lacking) Significance for Hypotheses of Language
Relationship
Some Marginal Footnotes on the (Still Ongoing?) Altaic Debate

Stefan Georg

In recent years, few (if any) papers which touched on the debate about the pos-
sible mutual relationships between Turkic, Mongolic, Tungusic, and possibly
also Korean and Japanese, failed to mention (often in the opening paragraph)
that this ‘perennial’ discussion is ‘still ongoing’ with ‘unabated vigour’. But is it?
It may indeed be the case that the vigour which once characterized this
debate may have abated to some degree, and that nowadays fewer dissident
voices (on ‘Altaic’ as a valid family, that is) make themselves heard—but this
may also be my very personal impression, possibly due to selective perception.
What has certainly not abated, is the stream of publications which declare this
debate done and dusted, and of course decided in favour of the assumption of
genealogical (divergent) relationship of these (and only these) five language
families. The, so far, last milestone of this is certainly the publication of the
impressive volume Robbeets and Savelyev 2020.
And when I call it impressive, I hasten to add that it is certainly also impor-
tant, well-conceived and executed and certainly indispensable for decades to
come, and that its editors and contributors earned with it the well-deserved
gratitude of all practitioners of Altaic linguistics and the philological disci-
plines which—though each of them well able to live and strive alone—con-
stitute this large field.
This does not mean that there is nothing to be unhappy with in this hand-
book (and for which handbook of this size and scope could such a thing
ever be said?), and it may not come as a major surprise to anybody famil-
iar with this field that this—still—concerns the very question of the mutual
relations between the languages concerned. Quite expectedly, the central chap-
ters about genealogical classification and comparative matters defend Altaic
as a genealogical grouping, and they also do offer linguistic data to underpin
this.

© Stefan Georg, 2022 | doi:10.1163/9789004499966_027 Stefan Georg - 9789004499966

Downloaded from Brill.com02/09/2023 12:02:07PM
via Stockholm University
on perfectly good-looking morphological comparanda 421

In the few pages which follow, I want to revisit just one of the realms in which
the proponents of the genealogical hypothesis claim considerable progress
to have been made in the last decades, namely the realm of verbal morphol-
ogy. One of the editors, M. Robbeets, has devoted considerable effort to the
demonstration that the languages in question do indeed possess a core of truly
shared morphological elements (mostly suffixes), the very existence of which
is claimed to be sufficient to end this almost bicentennial debate (according
to my way of thinking, it begins with Schott 1836, with everything published
before this, including the often quoted, but rarely read, Strahlenberg, belong-
ing at best to its prehistory, but this is open to discussion), now once and for
all with Altaic again vindicated (and this time for good). She devoted a whole
book to this (2015), but her list of cognate suffixes may already be found (with
later adjustments in details, some new entries and some eliminations in the
course of time, and some rather drastic changes in terminology) as early as
2007 (Robbeets 2007a, b) and is repeated in Robbeets 2020 in tabulated form
only—below, I reproduce what I perceive as its gist, in shorthand form (it is not
my intention to omit or distort anything of substance here, so the interested
reader remains referred to its authoritative presentation in their original pub-
lications, and especially the book-length discussion of these items in Robbeets
2015):

table 26.1 Proposed morphological comparanda of Turkic, Mongolic, Tungusic, Korean and
Japanese (“Altaic”)

No. Recon- Label Represented in As e.g. in

struct

1 *ana- negation, negative J, K, Tung, Nanai ana ‘negative noun’ (190)

verb (171–191)a Mongb
2 *ə- negation, negative Tung, Mong, Tk Ewenki e- ‘negative verb’ (193–
verb (192–202) 194)
3 -lA- manipulative (213– Tung, Mong, Tk Mong. altan ‘gold’ → alta-la- ‘to
227) gild’ (221)
4 -nA- processive (227– J, K, Tung, Mong, Nanai jolo ‘stone’ → jolo-na- ‘to
239) Tk petrify’ (234)

a All page numbers in this table refer to the, sometimes lengthy, treatment(s) of these elements
in Robbeets 2015.
b Robbeets mentions (but does not illustrate or discuss) a Mongolian [an-] negation, no men-
tion or discussion of this in Robbeets 2015.

Stefan Georg - 9789004499966

Downloaded from Brill.com02/09/2023 12:02:07PM
via Stockholm University
422 georg

table 26.1 Proposed morphological comparanda (cont.)

No. Recon- Label Represented in As e.g. in

struct

5 *(-)ki- ‘do, make’, iconic J, K, Tung, Mong, Man. jor ‘sound of some animals’
(239–246) Tk → jor-gi- ‘to chirp, twitter, hum’
(242)
6 *-mA- inclination (246– J, K, Tung, Mong Udihe xoton ‘city’ → xoto-mo- ‘go
255) to the city’ (253)
7 *-gA- inchoative (255– J, K, Tung, Mong, OTk. ada ‘danger’ → ada-k- ‘to be
266) Tk or come into distress’ (264)
8 *-ti- causative (276– J, K, Tung, Mong, OTk. bedü- ‘to be/become big,
292) Tk great’ → bedü-t- ‘to make grow,
increase, rear’ (291)
9 *-pU- reflexive-anticausa- J, K, Tung, Mong, Ewenki soli- ‘to mix up’ → solip-
tive (292–301) Tk ‘to become mixed up’ (298)
10 *-dA- fientive (301–308) J, Tung, Mong, Mong. nere ‘name’ → nere-de-
Tk ‘give a name’ (305)
11 *-rA- anticausative (309– J, K, Tung, Mong, OTk. süči- ‘be sweet’ → süči-r- ‘to
315) Tk become sweet’ (314)
12 *-gi- creative, anti- J, K, Tung Jap. tat- ‘stand’ → tat-e- ‘erect (tr.)’
causative (315–324) (< *-(C)i-) (317)
13 *-rA lexical nominal- J, K, Tung, Mong, OTk. kes- ‘cut’ → kes-er ‘adze’
izer (w/variants) Tk
(339–361)
14 *-mA lexical nominal- J, K, Tung, Mong, OTk. ör- ‘plait’ → örüm ‘some-
izer (w/variants) Tk thing knitted’ (377)
(361–379)
15 *-n lexical nominal- J, K, Tung, Mong, Mong. nisü-n ‘fly-cv’ (393)
izer (w/variants), Tk
converb (379–396)
16 *-x/ka resultative lexical J, K, Tung, Mong, Manchu je- ‘to eat’ → je-ke ‘eaten’
nominalizer (396– Tk (403)
416)
17 *-sa resultative lexical J, Tung, Mong, Mong. ide- ‘eat’ → idesi ‘food’
nominalizer (417– Tk
435)

Stefan Georg - 9789004499966

Downloaded from Brill.com02/09/2023 12:02:07PM
via Stockholm University
on perfectly good-looking morphological comparanda 423

table 26.1 Proposed morphological comparanda (cont.)

No. Recon- Label Represented in As e.g. in

struct

18 *-i/-Ø nominalizer (455– J, K, Tung, Mong, OTk. tög- ‘to pound, crush’ → tög-
466) Tk i ‘cleaned/crushed (cereal)’ (464)
19 *-x/kU nominalizer, infini- J, K, Tung, Mong, Mong. yabu- ‘to go’ → yabu-qu
tive (466–481) Tk ‘the process of going’ (476)
20 -Ø imperativec J, K, Tung, Mong, Mong. yabu- ‘to go’ → yabu-Ø
Tk

c Not discussed in Robbeets 2015.

I will not comment on Robbeets’ reconstructs for her proto-language, nor will
I, here, discuss the legitimacy of the lower-level reconstructs these are based
on (i.e. whether Proto-Turkic, Proto-Mongolic, etc. do have these morphologi-
cal elements in the first place, and whether everything which is known from
their attestation, function, history etc. lends itself to a meaningful compari-
son between the constituent branches of Altaic and a fortiori to the recon-
structs presented here); further, I will not say anything on Robbeets’ functional
labeling of these forms (which underwent quite considerable changes over
the years). For the argument at hand, I will treat these comparisons as argu-
menti causa legitimate, though, pending further analysis, I do not necessarily
think that they (all) are. I acknowledge, though, that Robbeets devoted a whole
book (2015) to the justification of these and accept that her, often painstak-
ing, discussion of each and every one of these elements deserves (and should
receive) detailed scrutiny on its own. The column ‘as in …’ serves illustrative
purposes only and is not intended to distract from (or paint an unnecessarily
unfavourable picture of) Robbeets’ discussion of all these elements in her orig-
inal publications.
Since I do not know, whether the Altaic/Transeurasian languages are gene-
alogically related or not,1 I will admit at this point that, if they are, a list of
morphological comparanda like the one in the table above (or some parts of

1 But I do not think that I have to hide my continuing skepticism towards the hypothesis that
they are—I think that they are not, and I do think that there are very good reasons for thinking
like this, but this, of course, does not translate to anything which may be labelled as knowl-
edge.

Stefan Georg - 9789004499966

Downloaded from Brill.com02/09/2023 12:02:07PM
via Stockholm University
424 georg

it) could be part of a meaningful argument to the end of demonstrating this.

But before this has happened, it may be useful, indeed it is necessary, to ask,
whether this list is, in itself, specific enough to be significant, i.e. whether it
describes a unique set of morphological elements, common to the languages
at hand, but not to others, which are not part of the hypothetical genealogical
grouping it is meant to vindicate.
I do not think that it is, since a rather casual look at one (to be fair: two rela-
tively closely related) language(s) from the Volga region (and from the uncon-
troversial Uralic language family), Cheremis (or Mari) and Mordvin, reveals
almost the whole roster of these suffixes (or reasonably, sometimes perfectly,
similar suffixes) with comparable or, again, sometimes identical, functions in
these languages as well, to wit: see Table 26.2.2
If these affixes in Cheremis (and Mordvin) exist and have the functions
illustrated here, and I eagerly await the demonstration that they do not, the
similarities of form and function they display with the Altaic affixes (and recon-
structs) on Robbeets’ list can only3 mean:
a) these commonalities are indicative of ‘Ural-Altaic’ as a genealogical
grouping;
b) the Cheremis/Mordvin affixes were borrowed from somewhere else (say,
from Turkic languages with which these languages, or some of the higher
genealogical nodes they belong to—‘Volga-Finnic’, ‘Finno-Ugric’ or
even Uralic—have been in intense contact with in the past;
c) the commonalities between the two lists are fortuitous resemblances
without any historical significance at all.
Robbeets herself (2014: 203) excludes the first possibility, and so do I. She is
quite clear about this, saying that, concerning any such comparison of Altaic
with Uralic data, “we are unable to test the assumed sound correspondences
(…) against regular sound correspondences established on the basis of lexical

2 I hope I will be forgiven for omitting detailed references here; most of these elements
can easily be found in the specialist literature—for my modest purpose I used Vasil’ev,
Savatkova and Učaev 1991, Wichmann 1953, Lehtisalo 1936, and Alhoniemi 1993. Items #9
and #18 are from (Erzya-)Mordvin, the other ones from Cheremis, with no attempt to dif-
ferentiate the two major variants of this language, though most examples are from Eastern
Cheremis.
3 I am confident that the (logically possible) conclusion that, because of these observations,
Cheremis (and/or Mordvin) would ultimatively have to be taken out of Finno-Ugric and
Uralic and joined to Robbeets’ Altaic (or, then, Transeurasian, a term, which, ironically, then
could indeed be somewhat legitimate, at least in this writer’s humble opinion) as its sixth
member does not really need any comment here.

Stefan Georg - 9789004499966

Downloaded from Brill.com02/09/2023 12:02:07PM
via Stockholm University
on perfectly good-looking morphological comparanda 425

table 26.2 Cheremis (and Mordvin) comparanda to “Altaic” morphological markers

No. Cheremis (or Mordvin) example

1 ana negation with the 1st p. pl. of verbs

2 o- < *e- negative verb, ‘not to be’
3 kürtńö ‘iron’ → kürtńö-la- ‘to clad with iron’
4 ꞵij- ‘to be straight’ → ꞵij-ana- ‘to become straight’
5 rošt ‘noise’ → rošt-k- ‘to emit (some) noise’
6 molo ‘different’ → mole-ma- ‘to change, become different’
7 laꞵəra ‘dirt’ → laꞵər-ɣa- ‘to become dirty’
8 pur- ‘to go in, enter’ → pur-ta- ‘to bring in, make go in’
9 (ńeja- ‘to see’ → ńeja-v- ‘to be seen’)
10 lüm ‘name’ → lüm-ẟa- ‘to name, to call’
11 ert- ‘to pass (time)’ → ert-ar- ‘to spend (time)’
or
il- ‘to live’ → ələ-ž- ‘to come to life’ (if < *-r- ?)
12 ere ‘clean, pure’ → erə-k-ta- ‘to clean’
13a —
14 kolə- ‘to die’ → kolə-mo ‘dead’
15 tolə- ‘to come’ → tolə-n ‘come-cv’
16 ꞵüẟəl- ‘to wrap’ → ꞵüẟəl-ka ‘wrapped’
17 moẟa- ‘play’ → moẟəš ‘game’
18 (kul- ‘to die’ → kul-i ‘dead’)
19 palš- ‘to help’ → palš-ək ‘help’
20 tol- ‘to come’ → tol ‘come!’

a I was not able to find any reasonably good(-looking) -r-suffixes in Cheremis or Mord-
vin with the required function(s)—I did manage to find some rather weak or bad
looking ones, though—, so I am happy to leave this cell empty.

comparison.”4 I concur, but I do have to stress here that the claim that every-
thing is fine with intra-Altaic lexical comparison and the sound-correspon-

4 A very similar list as the one given here, then based on Robbeets (2007a, b) alone (and since
then slightly updated to catch up with the the various changes Robbeets’ list underwent until
its most recent version in 2020), was presented on the occasion of the symposion Altaique
ou pas?, organized by Guillaume Jacques and Anton Antonov on 10 Dec 2011 in Paris, where
M. Robbeets and the writer of these lines had been invited to exchange their thoughts on
the matter. It has also found some circulation in the form of an informally distributed piece

Stefan Georg - 9789004499966

Downloaded from Brill.com02/09/2023 12:02:07PM
via Stockholm University
426 georg

dences based on these cannot go without the comment that this, indeed, is
well-known as the position of Robbeets’ and the Moscow-based school of Altaic
comparativism she is heavily indebted to, but that the corpus of lexical com-
parisons, which underlies this position, has met with too many objections from
the scholarly community to be regarded as anything in the way of the commu-
nis opinio of the field.5
Between possibilities b) and c), further scrutiny may (and certainly will) be
able to tell b)-cases apart from c)-cases, but I do not think that this is necessary
at this point, because the list, as it stands, illustrates rather straightforwardly
that (if, again, ‘Ural-Altaic’ is to be excluded):

The interplay of areal interaction between unrelated languages and pure

chance-similarities is (and in this case was) able to produce practically
the same picture of allegedly historically shared morphological compara-
nda as the one presented in Robbeets’ works (2015, 2020) as evidence for
the genealogical relationship of Turkic, Mongolic, Tungusic, Korean, and
Japanese.

In other words: If the comparanda found for Cheremis/Mordvin and Altaic are
not sufficient (or not at all indicative) for grouping these two entities together in
the genealogical sense, and I (and very likely M. Robbeets, too, if I am allowed to
say this here) think they are not, the ones found within five-term-Altaic are not
either. Borrowing and chance alone were, then, able6 to produce this roster of
superficially comparable/similar elements, the list of 20 allegedly ‘shared’ ele-
ments between the Altaic/Transeurasian languages is therefore, as it stands,7
inconclusive.
Did I disprove the validity of ‘Altaic/Transeurasian’ with this short and rather
dry exercise? Of course, I did not.
Everything this ‘exercise’ is meant to emphasize is the fact that, after all those
decades, even centuries, of heated debate, even the latest (and quite painstak-
ingly formulated) version of this hypothesis was and is not able to get beyond

of gray literature under the working title of The poverty of Altaicism—it could be that this
statement of Robbeets’ constitutes a—however faint—echo to this, but maybe it’s only this
writer’s vanity which makes him prone to think so.
5 I cannot go further into this here, but cf., i.a. Georg 2009.
6 And, while I admittedly think they did, I only say here that they could have been able to do so.
7 It goes without saying that the very shortness of Robbeets’ list made the task of compiling
this control list relatively easy, and a substantial increase of good morphological compara-
nda for the former might eventually render this argument obsolete, but please note that the
emphasis here lies on good.

Stefan Georg - 9789004499966

Downloaded from Brill.com02/09/2023 12:02:07PM
via Stockholm University
on perfectly good-looking morphological comparanda 427

the enumeration of morphological comparanda (in the realm of the verb, said
to be quite important for and indicative of any genealogical hypothesis), which
can, rather effortlessly, also be found in languages not covered by the initial
claim.
Again, I do not know, whether the Altaic/Transeurasian hypothesis is wrong,
but it seems to be inevitable to state that, so far, all attempts to demonstrate its
validity have been unsuccessful.
Let me close this short contribution with some rather personal remarks.
When I took up my studies of Altaic languages and linguistics back in 1984
or so, I began as an ardent believer in this relationship. The critics I read back
then simply made me angry, and I thought, in my youthful arrogance, that they
simply needed to learn the Indo-Europeanist methodology I was about to learn
at that time myself, and they would see! No, it turned out that it was me who had
to ‘see’, and I had to change my mind on this. But after a period of utter disap-
pointment, after contemplating to give up on all these languages and to study
Tibeto-Burman or Finno-Ugric instead, I realized that this disappointment was
not a nuisance—it was a finding.
And I cannot emphasize strongly enough, that it was mainly one person, one
true teacher, who made me realize this—that there was something wrong with
‘Altaic’ in the first place, but that this, at the same time, was and is nothing to
be worried about, that it was not a loss at all, or, if a loss, then only a loss of an
old error, the very stuff scientific progress is made of.
Though I, during the more than three decades since then, sometimes used
the epithet (originally coined as a snide) ‘Anti-Altaicist’ for myself (at times
with some pride, I admit), the teacher I am talking about never did so, and
he certainly will not approve of being addressed with this name today. He
never declared Altaic (in any of its extant versions) to be an outright absur-
dity, and he never went as far as calling it finally and utterly disproven. On
the contrary, all the skepticism he showed towards this and other hypothe-
ses was always palpably and only driven by his unquenchable thirst to know
and to always know more about the historical processes (processes of language
history among them) which shaped the ethnolinguistic makeup of the vast con-
tinent of Eurasia. Such an attitude will simply never be content with any ‘truth’,
which is simply proclaimed, rather than sought for, and it will always be per-
fectly happy that, whatever the term ‘truth’ may actually refer to, it can only be
approached asymptotically.
It should come as no surprise to anybody that the teacher I am talking about
here was and is—András Róna-Tas.

Stefan Georg - 9789004499966

Downloaded from Brill.com02/09/2023 12:02:07PM
via Stockholm University
428 georg

References

Alhoniemi, Alho. 1993. Grammatik des Tscheremissischen (Mari). Hamburg: Helmut

Buske Verlag.
Georg, Stefan. 2009. Review of: Martine Robbeets: Is Japanese related to Korean, Tungu-
sic, Mongolic and Turkic? (Turcologica 64). Wiesbaden: Otto Harrassowitz.
Bochumer Jahrbücher für Ostasienforschung 32: 247–278.
Lehtisalo, Toivo. 1936. Über die primären ururalischen Ableitungssuffixe (Suomalais-
Ugrilaisen Seuran Toimituksia lxxii). Helsinki: Suomalais-Ugrilainen Seura.
Robbeets, Martine. 2007. How the actional suffix chain connects Japanese to Altaic.
Turkic Languages 11/1: 3–58.
Robbeets, Martine. 2007. The causative-passive in the Trans-Eurasian Languages. Tur-
kic Languages 11/2: 235–278.
Robbeets, Martine. 2014. The Japanese inflectional paradigm in a Transeurasian per-
spective. In: M. Robbeets, Martie and Bisang, Walter (eds.), Paradigm Change. In the
Transeurasian languages and beyond (Studies in Language Companion Series 161).
Amsterdam and Philadelphia, 197–232.
Robbeets, Martine. 2015. Diachronic Verb Morphology. Japanese and the Transeurasian
Languages (Trends in Linguistics 291). Berlin: DeGruyter Mouton.
Robbeets, Martine. 2020 = Robbeets, Martine and Savelyev, Alexander (eds.), The
Oxford Guide to the Transeurasian Languages. Oxford: OUP.
Schott, Wilhelm. 1836. Versuch über die tatarischen Sprachen. Berlin: Veit.
Vasil’ev, V.M.; Savatkova, A.A. and Učaev, Z.V. [Васильев, В.М.; Саваткова, А.А. и Учаев,
З.В.]. 1991. Mарийско-русский словарь [Cheremis-Russian dictionary]. Joškar-Ola:
Marij Kniga.
Wichmann, Yrjö. 1952². Tscheremissische Texte mit Wörterverzeichnis und grammatika-
lischem Abriss (Apuneuvoja suomalais-ugrilaisten kielten opintoja varten v). Hel-
sinki: Suomalais-Ugrilainen Seura.

Stefan Georg - 9789004499966

Downloaded from Brill.com02/09/2023 12:02:07PM
via Stockholm University