Why do Ancient Greek words have "εί" from PIE "e"?

Ancient Greek κείρω <- PIE *(s)ker-.

  • This is a good question, but it might be a better fit for Latin SE, which also does Greek questions.
    – TKR
    Commented Dec 15, 2021 at 17:14
  • 2
    They don't, as ει was just an orthographic convention to indicate long /e/.
    – LjL
    Commented Dec 15, 2021 at 18:36
  • 2
    @LjL Why should that matter? Apart from the fact that that ει doesn’t go back to just *e with no further conditioning doesn’t change the fact that, with some conditioning (like a following yod), Greek does have ει in places where PIE had *e. How ει was pronounced is completely irrelevant to that; it wouldn’t change anything about this question if it were [au] or [iø]. Commented Dec 15, 2021 at 23:36

1 Answer 1


[Mods feel free to migrate to Latin SE if appropriate.]

This is part of the "first compensatory lengthening", a set of regular Greek sound changes involving the loss of PIE/Proto-Greek *y and *s.

In this case, the full PIE proto-form was *ker-y-ō, with the present-tense formant *-y-. The sound change in Greek is:

[e|i|u][r|n][y] > [e:|i:|u:][r|n]

That is, when a vowel e/i/u preceded a resonant r/n which itself preceded *y, then the *y was lost and the vowel was lengthened. (There may have been an intermediate stage in which the resonant became a palatalized geminate, but this is uncertain.)

The spelling ει, as LjL notes, doesn't actually represent a diphthong here, but the long vowel [e:]. Some cases of ει go back to real diphthongs ("genuine diphthongs"), but others stand for the result of compensatory lengthening ("spurious diphthongs"); similarly for ου. The reason ει and ου were used for both types of sounds is that the original diphthongs underwent monophthongization, so that e.g. [ei] became [e:].

The change above is part of a larger set of changes that also involve other vowels and resonants, as well as *s in place of *y. For example, a sequence like *-ory- underwent metathesis rather than lengthening, ending up as οιρ, e.g. *morya > μοῖρα. To make things more complex yet, different Greek dialects had different outcomes for some of these sequences, so that Attic κείρω corresponds to κήρω and κέρρω in other dialects.

  • The last fact is part of an argument that supposed metathesis in Attic comes from palatalization and compensatory lengthening, then dialectally differing reanalysis via loss of palatality vs. unhooking palatality from the consonant. Also note the outcome of Vly.
    – user6726
    Commented Dec 15, 2021 at 19:30
  • @user6726 Yes, we discussed this on Latin SE recently. Do you know if there are good parallels for the supposed depalatalization and degemination? latin.stackexchange.com/a/16966/40
    – TKR
    Commented Dec 15, 2021 at 20:17
  • Could you please give an example of [|i|u] > [|i:|u:] Commented Dec 17, 2021 at 16:46
  • 1
    @КузнецовАнатолий Sure: *krin-yō > κρῑ́νω, *barun-yō > βαρῡ́νω (those hard-to-make-out marks are supposed to be macron-acute combinations), and many other verbs in -ινω -υνω.
    – TKR
    Commented Dec 17, 2021 at 17:32
  • @TKR thank you! Commented Dec 18, 2021 at 8:33

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.