We have

Hebrew: šeš;
Russian: šestʹ;
Ukrainian: šistʹ;
Latin: six;
English: six;

Hebrew: yeš;
Russian: yestʹ;
Ukrainian: ye, isnuye;
Latin: est;
English: is;

Hebrew: ze;
Russian: se;
Ukrainian: сe [t͡sɛ];
English: this;

Hebew: še;
Russian: šo, što;
Ukrainian: ščo;
English: that (as in "say that...", "think that...").

There are more, but we can conclude, for instance that word-initial š corresponds to Slavic š or št, word-ending š corresponds to Slavic stʹ.

I know the PIE etymology of these words, but still wonder if the Hebrew words developed under influence from Greek, Hittite or whatever?

If there are no such influence, what can be done to rule out false sound correspondences when reconstructing proto-languages?

  • 15
    One or two data points is hardly enough to posit a regular sound correspondence. If there were many more of these, it would require explanation, but otherwise the obvious answer is coincidence.
    – TKR
    Commented Aug 4, 2022 at 19:31
  • 8
    The initial [j] in the Slavic words for “is” is prosthetic and quite late, in Proto-Slavic it used to be *estĭ, even in modern Bulgarian “is” lacks that [j]: e [ɛ]. How could the Greek verb esti and Hittite verb eszi produce the initial yod in Hebrew יש yeš, especially taking into account that in Hebrew יש yeš is an adverb? Those kinds of questions show the absence of preliminary research efforts.
    – Yellow Sky
    Commented Aug 5, 2022 at 7:54
  • 4
    @Anixx - Who says *h₁ was voiced? As far as I know, of the three PIE laryngeals, only *h₃ was possibly voiced, there rest being voiceless. *h₁ is reconstructed as either [ʔ] or [h]. Nevertheless, you included that Hebrew יש yeš into the question asking “if the Hebrew words developed under influence from Greek, Hittite or whatever?”
    – Yellow Sky
    Commented Aug 5, 2022 at 8:27
  • 4
    @Anixx - Again: who says that? Whose theory is that? What about this? Why is “the most common theory” not mentioned there?
    – Yellow Sky
    Commented Aug 5, 2022 at 8:36
  • 3
    where I've seen ɣ suggested as a laryngeal it's been for *h3 (for people who don't believe *o was rounded) and occasionally *h2 (although the evidence seems to point to that being voiceless). I'd never seen any suggestion that *h1 is anything but ʔ or h before seeing Andreev (1957)'s row in the table Yellow Sky linked, other than the naive suggestion that the three laryngeals match the three dorsal stop series but afaik this has never had any scholarly backing
    – Tristan
    Commented Aug 5, 2022 at 10:09

3 Answers 3


Just as you cannot compare two random species today to accurately assess their taxonomy (otherwise we would conclude all crabs form a single family when they actually form at least five distinct groups within the Decapods), we cannot simply compare two random modern languages to decide if they're related.

Instead we need to iteratively work backwards, building up our evolutionary tree, by first comparing the most similar languages and only later looking comparing more distant languages using our knowledge of the modern languages' histories.

Luckily both Indo-European and Semitic are well studied and firmly established families so we have access to information about those older stages.

  • Six:
    • Hebrew שש šeš < PS *šidṯ-um (cf Arabic سِتّ sitt, Aramaic שת šeṯ, and in the masculine with the chiastic-concord -at suffix, Akkadian 𒐋 šedištum and Ge'ez ስድስቱ sədstu both showing the original d)
    • Russian шесть šestʹ < PIE *s(w)éḱs, with the -t' being a Slavic innovation. The w is in brackets because not all branches show evidence of it, it has been suggested that the original number was in fact *wéḱs, with the initial *s of the following number contaminating it, either replacing the w as in Albanian, Balto-Slavic, Germanic, Indic (& Nuristani), and Italic, or appearing before it as in Celtic, Iranic, and Tocharian. Greek could reflect either an initial *w or *sw, and Armenian likely reflects a form beginning *suw. Other arguments in favour of a lack of an initial *s are the metathesised loan *eks₁w- in Proto-Kartvelian
  • Is:
    • Hebrew יש yeš < a form ending in ṯ (cf Aramaic אית ʾīṯ). Note also that the Hebrew is not a verb, but a particle instead
    • Russian есть yestʹ < PIE *h₁és-ti (note that the -ti here is inflectional, and not part of the root)
  • This:
    • Hebrew זה ze < a form beginning in ḏ (cf Aramaic די dī & Arabic ذُو ḏū)
    • Russian сей/сё sej/sjo (not sure where se came from) < PIE *ḱis
  • That:
    • Hebrew ש־ še < likely from a form with ṯ as it is probably related to its synonym אשר ašer (cf Aramaic אתרא ʾaṯrā, Arabic أَثَر ʾaṯar).
    • Russian что što < an extended form of PIE *kʷid. Note that unlike the Hebrew this is also used as an interrogative, which was its original sense.

In many of these cases we see the Hebrew ש š or ז z coming from an earlier dental fricative ṯ or ḏ, where the Russian ш š or с s comes from a dorsal stop ḱ or kʷ. The correspondence sets needed would be:

  • Six: Semitic *dṯ ~ PIE *ḱs
  • Is: Semitic *ṯ ~ PIE *s
  • This: Semitic *ḏ ~ PIE *ḱ
  • That: Semitic *ṯ ~ PIE *kʷ

Correspondences between coronal and dorsal stops, when both languages have plenty of both classes are simply not plausible, and the fact we also have inconsistencies in voicing makes it even worse.

Regardless, in all cases we see the forms diverging as we go further back in time rather than converging. This is exactly the behaviour we would expect to see of coincidences, and the opposite of the behaviour we expect to see of genuine connections where the similarity becomes more apparent as we move backwards in time (up the linguistic family tree) and forms converge on the original common ancestor.

Even if these correspondences were plausible though, and we ignore the fact that the similarity disappears as we compare older forms, four apparently similar words is nowhere near enough to establish or even suggest a direct relationship.


what can be done to rule out false sound correspondences when reconstructing proto-languages

The main thing that is commonly done is to not directly compare random words from the modern languages, instead you systematically reconstruct earlier sub-branches. You would not compare Russian and English, you would compare West Germanic languages to reconstruct proto West Germanic, etc. (of course we can also look more directly at Old English, Old High German, Gothic, Old Norse, but often we don't have old texts). Likewise you build towards proto-Germanic and proto-Slavic. When you do so, you look for regular sound correspondences whereby regular sound changes can account for the development of the proto language into the relevant daughters (attested or reconstructed). That would yield the discovery that English that and Russian što (← Proto Slavic čьto) are not from the same root even within Indo-European. You would do the same thing with any putatively related word in Hebrew. You would look at related Semitic languages to see if there are similar-looking words with related meanings, to determine where Hebrew še came from (again, ignoring the fact that there are millenia of historical documents for Hebrew).

If you are willing to engage in random comparisons of words without concern for the systematicity of relationships, I don't think there is any way to guard against making false connections.

  • I included English "that" only to give the meaning. The English cognate to Russian što is what
    – Anixx
    Commented Aug 4, 2022 at 19:06
  • This is self-contradicting. You don't compare "randomly" but "systematically". What System?! You do look for "regular sound correspondences" but not Russian niet / English *not*—meaning, you look it up in the book and the only reconstruction you get to do is retracing the steps and leaps that the author have taken, which usually means Sanskrit, Greek, the Roman empire¹. None of which answers why there are apparent correspondances which even more notable figures have observed, or how to falsify results. The snark is right on point, however, because that's the kind of polemics that'll due.
    – vectory
    Commented Aug 9, 2022 at 18:56
  • [1]: and very rarely a broader view involving the Levant, or the Caucas for that matter. And then you have to trust the specialists. There are literally dozens of them! I am apparently doing footnotes in comments now. I am so sorry.
    – vectory
    Commented Aug 9, 2022 at 19:10

The example of 'six' is interesting. It has been noted for over a hundred years that the word for 'six' (and for that matter 'seven' as well) are very similar in Indo-european and Afro-asiatic languages. It is sometimes proposed that these are borrowings from proto-semitic into proto-Indo-European but I suspect this is a guess. I would suggest if you look at the others, the further back into history you go the more different the hebrew precursors will look to the Slavic ones.

  • 3
    seven is relatively close, but whilst six has sometimes been suggested as linked to PIE it's completely baseless. Even a cursory glance at Semitic languages other than Hebrew reveals that Proto-Semitic must have had a word ending in *dṯ, which bears no real resemblance to the PIE *ḱs. The argument about whether the PIE word even originally contained an initial *s destroys the similarity entirely
    – Tristan
    Commented Aug 5, 2022 at 10:53
  • @Tristan well, is not ṯ the same sound as "th" in English "thick"? In that case it is quite close to s.
    – Anixx
    Commented Aug 5, 2022 at 11:30
  • yes, but the d & ḱ aren't even remotely close
    – Tristan
    Commented Aug 5, 2022 at 13:01
  • If *H1 ~ *y is so out of the question, how do you explain pronominal *yos as it's usually related to *h1e- ~ *ís? These are thought to be related, dunno how
    – vectory
    Commented Aug 9, 2022 at 9:58
  • the further back into history you go the more different the hebrew precursors will look to the Slavic ones. How low can you go? Not only do you have to jump through hoops to get there, but the very question is "what can be done to rule out false sound correspondences when reconstructing proto-languages?". Ie. how do you safely go back so far without falling for accidental correspondance that is very likely no coincident at all if the speakers were in close contact. It's simple: regular changes identify borrowings in the presumably related languages, then the rest just is QED.
    – vectory
    Commented Aug 9, 2022 at 19:17

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.