One well-known division of Indo-European languages is Centum - Satem. However, my wikipedia-induced understanding is that due to the existence of Tocharian it is at best unclear whether it corresponds to an actual evolution of PIE into two dialects/languages.

But perhaps I'm wrong and many serious linguists do consider the Centum - Satem division as corresponding to dialects of PIE? For example on the (striking but maybe naive?) basis of the similarity of the word "five" in Balto-Slavic, Indo-Iranian and Tocharian languages?

Are there any other binary divisions of Indo-European languages which are considered by serious linguists to correspond to an evolution of PIE into two branches? I'm particularly interested whether anyone serious considers Balto-Slavic and Indo-Iranian to have had evolved together for a while, perhaps together with other families, but in separation from Centum languages.

  • 6
    Most IEists would support a primary division of IE into Anatolian vs. everything else. Beyond that, very little is clear.
    – TKR
    Commented Aug 10, 2016 at 22:54
  • The word for five in Russian is very similar to Greek or Latin as well, taking into account the (quite moderate) regular sound changes.
    – Anixx
    Commented Mar 23, 2020 at 18:09
  • Consider the proposed n-ary Division of English: You have to draw a line somewhere, but look, even a barrier as big as the Ocean does not stop cross pollination. What's the opposite scenario for PIE: A bunch of tribes riding and raiding each other and everyone else, who spread, without a clear concept unless the skyf ather guided them, across an area spanning from the Altai Mountains to the Carpathian and back, dying like the flies, all the while they maintain language integrity, until a winner emerges and repopulates the known world by fathering two children. Except that one has a lisp?
    – vectory
    Commented Mar 30, 2021 at 11:15

2 Answers 2


As for the centum-satem distinction, nowadays indoeuropeanists usually don't think of it as a west-east dialect division but they rather view the satem palatalisation as an innovation which took place in central IE dialects in opposition to the peripheral or outside IE dialects which did not undergo that change. In such a scenario, following the principles of grouping dialects on the basis of common innovations and not common retentions, one could only posit an existence of a Central PIE dialect, leaving the outside ones out of the equation. A binary oppposition of centum-satem, in this shape, is well embraced by modern scholars. It turns out that the central-peripheral version is more flexible and allows to take into account such peculiarities as weird centum lexemes in Slavic, and the great distance between for example Tocharian and Celtic, which in the older view should have belonged to the same dialect.

Unfortunately one can point multitude of other isoglosses, involving other language changes, which do not necessarilly overlap. This is a thing that proves a task to confidently divide PIE into two dialects very difficult if not impossible. Needless to say one also has to deal with time depth and a fact that division of PIE was not a instant event, that some groups of people left Common PIE speaking communities earlier and some later.

  • Thanks for your answer. Is the Central PIE dialect which you mentioned studied somewhere? Perhaps even reconstructed? Also I gather from your answer that indeed Balto-Slavic and Indo-Iranian had evolved together for a while, in separation from the western/southern european langauges. Do you know for how long and whether there existed a Proto-Balto-Indo-Irano-Slavic language? (perhaps in fact the Central PIE dialect you mention could be referred to as such?) Commented Aug 12, 2016 at 16:36
  • I have never seen any publications focused on the reconstruction of the specific Central dialect. The existence of such is just assumed on the basis of centum-satem data. Balto-Slavic and Indo-Iranian would have developed together just when they were part of this Central dialect.
    – czypsu
    Commented Aug 13, 2016 at 9:48
  • Woudn't it be considered interesting to study this Central PIE and its descendants before the split into Balto-Slavic and Indo-Iranian (and maybe others)? From what you say, it might be almost trivial to professional linguists, but I still find it interesting that, say, Polish, is "genetically" closer to Sanskrit than to German, if I understand you correctly. In any case the influence due to geographical proximity almost "cancels" having a relatively recent common ancestor, but still some quantitative study sure would be interesting (most interestingly - for how long did they evolve together?) Commented Aug 14, 2016 at 12:42
  • This Central PIE is extracted on the basis of the satem change. So the research would have ended in a conclusion that in the Central PIE we had this change where certain velars got palatalised. We can just add the satem change to our reconstructed PIE and we have the Central dialect. As I wrote before, other isoglosses of other language changes within PIE and its descendants do not overlap with the centum-satem isogloss, so there is no really explicit dialect to reconstruct. The notion of Central PIE is brought up just to label the satem langauges in a vague reference to spatial relations.
    – czypsu
    Commented Aug 14, 2016 at 14:48
  • czypsu, thanks for your response. I have to say though that I don't understand your dismissiveness - If you pose the existence of Central PIE, of which, at least currently, balto-slavic and indo-iranian are the two biggest groups, then I think at least the question "for how long did balto-slavic and indo-iranian evolve together before splitting" is a valid research question, and I can't find an answer, neither in your responses nor elsewhere. Independently of this point, wikipedia article on centum/satem has some suggestions of other common innovations shared between balto-slavic and iranian. Commented Aug 15, 2016 at 12:34


It's thinkable that the Tochariens either caused the split together with, or under preassure from, non-IE groups, or that they remain as trace elements (in the particle physics sense).

Imagine for example a wide dialect continuum, an incurser ripping right through, polarising one half that propels the other half far into the east, maybe T dragging S along with them in the wakes, never changing its own trajectory and polarity due to momentum, propelled ever further by repulsive forces.

The models East-West and Center-Periphery are not mutually exclusive, anyhow, e.g. in a radial diagram (using polar coordinates). Consider for example a shaded lamp, or the flame of a camp fire that has a well define center as well as top-bottom. That's an apt metaphor, because we only see shadows on the wall of a cave, in the platonic sense.

I don't want to go as far as to suggest a string theory of linguistics involving branes instead of branches. But dialect continuum is a common term of art, and wave-model an illustruous idea. In this context, comparisons to the particle-wave duality from physics have been drawn in the literature, indeed very dramatic.


There is a seizable ammount of apparent doublets in the PIE-Lexicon as reconstructed, some of which are better explained diachronicly. This is however difficult, if synchronic derivation on the PIE level remains a possibility, in light of ignorance, as much as obscure, naturalized and thus indescernable loans can hardly be accounted for, that would otherwise imply contact, but instead suggests unbroken descent through erroneous reconstruction (as finding a phoneticly plausible cognate with however far fetched semantic derivation is possible 50% of the time, and if not, a shape-shifting monstrum of a phoneme like *gwh- may be conjured).


Based on the phonetics of the Glottalic theory after Gamkrelidze and Ivanov, an apparent grouping of Germanic, Armenian and Hittite has been proposed (Gamkrelidze 2010: In defense of Ejectives for Proto-Indo-European). Kammerzell (1999: Glottaltheorie) gives diagramms for a general overview. Kortlandt put some of the theory to good use for Slavic accentology (cf. Dercksen in the introduction to their dictionary). Other's argue--vehemently--that this theory is chasing ghost images.


A recent, indecicive comment on branches and Tocharien's position in the "tree" (as opposed to "gammut", e.g.) is from Michaël Peyrot (2019: Indo-Uralic, Indo-Anatolian, Indo-Tocharian)

it [Tocharien] is often assumed to have been the second branch to split off after Anatolian, as in the tree reconstructed by Ringe, Warnow & Taylor (2002: 87; see figure 13.1, next page).

Several authors have argued that Tocharian was the second branch to split off […]. Yet, the evidence is not overwhelming, and, strikingly, many authors strongly differ in the arguments they adduce for this position of Tocharian in the family tree.

Further down, Peyrot follows Kloekhorst (2008) through barely seven points as evidence for a Indo-Anatolian divide. This is questionable though, first of all because they have a tendency to assume Hittite retained more archaisms, leading to the proposal of a common innovation in the others, and thus the appearance of a solid core PIE. Second, the individual examples are tailored to support the assumption. This isn't necessarily wrong if it's done systematicly. However, seven individual lexical examples are not yet a system. Whether the points are particularly strong, I am not aware of any criticizm either way; though Peyrot would likely have mentioned it after eleven years, if anything of substance turned up. #7. "horse" might be particularly instructive, if they rode out immediately after breaking the horses in the Steppe for the first time; Als, the Kurgan Hypothesis is just a hypothesis; Kroonen et al (ref?) evaluated theories after genetic evidence found south of the Caucasus promised new perspectives; Further, supposed eastern connections to horse words in e.g. Mongolian and Chinese require more research (attempted by Blazek and Schwartz 2017), precisely because Macro-Altaic is a burned bridge in the wider community by now. #1. "disappear" / "die" is pretty close to "forget"

farāmūšīdan cognates with Lithuanian užmiršti/pamiršti, Latvian aizmirst/piemirst, old Armenian moṙanam, Tocharian A märs- all of them meaning to forget

(reddit with discussion and link to helsinki.fi)

This might look derivative, especially with the here not further to be specified preverb. Peyrot seems to put a lot of weight on stems, as is evident further down:

One of the frequently cited Hittite-Tocharian matches is Hitt. eku-zi / aku- ‘drink’ ~ Toch.AB yok- ‘drink’ < *h₁egwh- (Pinault 2006: 93). Although Anatolian and Tocharian are indeed the only two branches in which this verb is found, and most other branches have reflexes of the more common *peh₃- ‘drink’, this etymon is difficult to use as an argument for the Indo-Tocharian hypothesis. Most importantly, even though Anatolian and Tocharian are the only branches in which *h₁egwh- is attested, reflexes of this root are also found in Lat. ēbrius ‘drunk’ and Gr. νήφω ‘be sober’.

[emphasise mine]

That might take a second to sink in.

A pro pos: One problem of the comparative method is that reanalysis carried out with technical perfection should appear like a proper root to be thus indescernable. The other problem is that the selection of comparative material is subject to intuition, and that this intuition is frequently misleading--an argument more often leveled at long range comparisons (e.g. Japanese-Hebrew)--and has to withstand falsification. The Greek looks easily associated with *peh3-; The Latin is reminiscent of brew. On the other hand, German einen heben "to drink [liquor], chug [bear]" and Humpen "jug" could be compared, the same way that Upper German n. Fetzen 1 "buzz [from liquor]" 2 "jug" might compare to *peh3---not reliable. They further treat tentative Indo-Uralic reconstructions, hinging on "the inclusion of Proto-Samoyedic *e̮- among the cognates" [diacrit omitted] in meticulous detail, which, they admit, rests yet on too many unknowns. Thus there is no solid limit to the reconstruction. It defies believe how far one would have to go. Usually the grammatic typology is deemed more reliable; Comparative grammar is lagging far behind the lexical reconstruction however. Just to stay with lexis for a moment longer:

a) I had mistakenly considered toxic, intoxicated (viz inebriated), as I hadn't noticed the gloss for "PFU *toxi-", that is "‘bring’", which would make the comparison unlikely, but, as I had at the same time considered that Humpen "jug" would remind of En. hump, Ger. humpeln "to limp", further bump, and fuzzy memories of roots in *bh-, as well as jug that is of uncertain origin with one variant that reminds of goble, goblet, I had nearly set out to see how PFU *t and *j should relate, and ... there it is: "from ... τοξικόν (toxikón) [φάρμακον (phármakon)] ("poison for use on arrows")", which "Pokorny (1959) connects it to the Greek root φαρ- as in φάρος (pháros, “plough”) and φάρυγξ (phárunx, “throat”), from a Proto-Indo-European *bʰer- (“to cut, pierce, scrape”) ..."; that's uncertain, but suits me, because *bʰer- is of course also glossed "carry" (cf. to bear). Cracking the Jackpot! Lat. fero "to bring" supposed from the same root, is a highly synthetic, suppletive paradigm with tol- in perfect; *bher- has a doublette in *dher- ... The point being, the third problem of the comparative method is that there are no rigid laws for semantic derivation. That's not actually a problem, rather the strength well of language to beginn with. However, the comparative method makes no promise to disentangle the ensuing mess. Rather, many etymological dictionaries try to isolate only the most stable roots. Naturally, dialectal differences are due to instabilities in the system and therefore barely tangible in its rigid framework. The exclusionary premisses that Mycenean Greek had no relation to Finnic is likewise intangible: it's simply not a question of linguistics. Meanwhile, Finns are belittled who try working out evidence of Greek culture in old Finnland. b) no less speculative, it should be doubtful whether above mentioned "Tocharian A märs-" (to forget) etc., if it rather belonged with √mrs

PIE √mrs- √mors- √mers- (a.) ‘falsch, umsonst, vergeblich’ [ie.: false, for nought, to no avail -- vectory]

PIE *mŕ̥sē

RV. mŕ̥sā(adv.) ‘umsonst, vergeblich’

AV. mŕ̥sā(adv.) ‘in falscher Weise’



PIE *morsei-

OHitt. maršei-(vb.1.) ‘falsch (betrügerisch) werden’

(Pyysalo et al, Proto-Indo-European Lexicon, The generative etymological dictionary of Indo European languages, with further references)

How that in turn related to √mr- might be a matter of grammar, after all.

šīdan (cp. farāmūšīdan above) is a verb in modern Farsi itself, for example. Similarly, *bher- and *dher- might simply contain a common element; Whether fara-, En. fore- and similar can count as a stem is debatable (cf. e.g. Georg Dunkel 2014, Cyril Bosch 2013).

We can further compare to miss, Ger. vermissen "to have lost, to desire, to miss (smb.)", to dismiss, Lat. mitto "to send", all PIE *meyt- (“to change, exchange, trade”) [wiktionary], further OE mete "decide", to meat [a decission], further Ger. ausmisten "to clean out", Mist "dung, trash", Miete, Misthaufen "heap [of hay, dung]" (Lat. meta "conic structure"), reasonably so if backed r modulo sonority gives high vowels, i, y. We can compare Ger. fallen "to fall, to die [in battle]", fehlen "be missing", Fehler "mistake", falsch "false" (Lat. falsus), etw. fallen lassen wie eine heiße Kartoffel (to drop something like a hot potato), Befehl "command, order" (viz message), and feilschen "to barter" (viz exchange, trade) for analogy, or tauschen "exchange", täuschen "deceive", tuscheln "to whisper (conspiratively), further untertauchen "to dive under" (to lay low, disappear), or En. to loose, lose, Ger. (ver)letzen "to hurt, damage" (lose a limb), Leid "suffering", Beleidigung "insult", ledig "lone, alone", further los "up, away", lösen "solve", let's (and all that jazz which is apparently far from related, lex, letter, Erlass, lassen, leave, Leiche "corpse", belegen (mit einem Bann, Fluch), lenis, lead), or verzeihen "apo-logy", verziehen "retreat" (already dissimalar in oldest evidence), verzichten "to pass, forefit, dismiss", Zeche, Zuchthaus etc. etc. The level of convergence in this is immense).

Kloekhorst rather argues that the sense "to die" were not found in the Anatolian branch, because it hadn't yet existed, though it's existance in many other branches is possibly borrowed and a common innovation for PIE/Anat. not even that secure. I'm might be wrong on that, but I'm sure the car died on me is a secondary anthropomorphism, and you are dead (to me) either an empty threat, figurative speach, or a judgement call. As said, it's pretty much useless to argue on that level, if there's hardly any expectation of ehh, regular semantic shift. Romans deemed social death a suitable metaphor, economies in general have appreciate the labour potential of prisoners more valuable than capital punishment, and Gökli Tepe was decorated with human skulls, but that says very little, as godkings don't die, they arrange for the ultimate ascend.

Nevertheless, Anatolian ist commonly assumed to be the most distant branch, and these proposals need to be put out to advance the theory.

  1. A matter that rather belongs with comparative etymology is god, vs deus, theos, and bog, the latter in Slavic, Iranian and Sanskrit, as far as I know.

Similarly, words for king should be instructive. The root *Hregs is clearly PIE for what it's worth, but the history of the term rex needs details. Hittite didn't have it as such. Ragnaröck is not clearly recognizable as King of kings anymore either.

  1. Thracian language is so barely known that it's not even clear which branch it should be.

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.