I have come across a controversy about the [a] vowel sound in PIE. I have noticed in most reconstructed PIE words the [a] sound is not present, even when it seems to be present in most of the daughter languages, and it only seems to be reconstructed for some rare words. I have looked it up and it appears there is a debate among scholars on whether there was an [a] sound or at all. What are the reasons for this?

Also, are there any historically attested examples of [e] and [o] becoming [a] in IE languages?

3 Answers 3


The evidence for a phonemic *a in earliest PIE (before Anatolian and Tocharian broke off) is extremely lacking, it only seeming to be present in a small number of words that are likely borrowings, often present in only a small number of branches.

That said, a-colouring from *h2 is attested in all branches, and so *a ought to be reconstructed for earliest PIE, just only as an allophone of *e rather than as a phoneme in its own right.

Late PIE (after Anatolian and Tocharian broke off) seems to have for the most part either lost or merged the laryngeals, meaning this earlier *a allophone had become phonemic.

For your second point, it's worth remembering that in reconstructions, the labels *e, *o, *a (and indeed any other symbols at all that are in use) are purely for convenience and should not be taken too literally as representing the sounds represented by those symbols in the IPA, and their exact realisation is very much open to debate (in particular, there is debate over whether *o was even rounded in most instances).

So, we shouldn't be looking for instances of [e] & [o] becoming [a], but front & back non-high vowels becoming low unrounded vowels.

Spontaneous rounding and unrounding of non-high back vowels (especially lower vowels) is extremely common (as an example in the historic era, look at the American unrounding of the LOT vowel), so *o becoming *a (and vice versa) are completely unsurprising.

Lowering of mid front vowels is also not unusual (as a recent example, the DRESS vowel /ɛ/ in English has become variously [æ] & [ɐ] in California & Northern Cities English respectively). Schwas (another candidate for the sound of PIE *e) are also very frequently lowered. Likewise, non-high front vowels often become non-high back vowels which can then unround for the same reasons mentioned above. As you can see, *e becoming *a also isn't especially odd, even if it is less expected than it is for *o.

From a PoV of looking at the vowel system as a whole, 5-vowel systems are pretty stable, and 3-vowel systems are both pretty stable, but 4-vowel systems are not, so once *o and *a merge, that may provide enough pressure to encourage the *e to merge too (or else for there to be another split, returning back to a 5-vowel system).

  • 2
    Thank you, that was a brilliant answer and cleared up the matter mostly to me. Jul 30, 2021 at 13:25
  • Just a few more questions in relation to what you said. What do you mean by "H2 colouring in all IE languages"? I thought the laryngeal were only found in Anatolian and have been lost everywhere else? Also, I meant attested examples of e's and o's becoming a's. I have looked through many IE families and I have seen a sounds turn into e and o, but not e and o into a. Sometimes it can be in part of word and sometimes all of it e.g. Tocharian A Macar, Pacar > Macer, Pacer. The "car" becomes "cer"; Sanskrit Ka > Bengali Ke, Punjabi Ker; Sanskrit Sa > Bengali Se, Oriya Se. Jul 30, 2021 at 13:51
  • the laryngeals are only found explicitly in Anatolian (with the possible exception of a small number of preservations in other branches), but all branches show h2 colouring an adjacent *e to *a
    – Tristan
    Jul 30, 2021 at 14:27
  • and I give you explicit examples of e- & o-type vowels becoming a-type vowels in the answer...
    – Tristan
    Jul 30, 2021 at 14:28
  • There are explicit traces of *H2 also in Armenian, Kurdish and some Iranian languages.
    – user23769
    Jul 30, 2021 at 14:42

The general consensus is that, phonetically, early PIE had a low vowel of some sort (which could broadly be called [a]), but phonemically, it did not have a phoneme /a/ (or it had a marginal one at best).

The reason for assuming a phonetic [a] is that various words have a consistent low vowel across several daughter languages, like in "before, fore, front" (Latin ante, Greek anti, Sanskrit anti).

The reason for not assuming a phonemic /a/ is that, for the most part, it doesn't seem to be necessary. When we see those consistent low vowels, the laryngeal theory explains them as an underlying /e/ adjacent to /h₂/. And indeed, Hittite shows a direct reflex of /h₂/ in the "front" word: ant- "face, fore". (Reconstructed PIE phonemes are generally written like *this rather than /this/, and that's what I'll use for the rest of the answer.)

Why not assume a phonemic /a/? Because this "laryngeal-coloring" theory very nicely explains certain inconsistencies in ablaut (the way the main vowel within a PIE root changes in different forms). For most roots, there are three common ablaut grades: ∅, e, and o. But certain roots seem to replace the e-grade with something else: sometimes ō, sometimes ā, sometimes ē. And these same roots do something weird with the ∅-grade, which varies by language; in Greek, the ∅-grade of those roots show o, a, and e, respectively (Greek didōmi "give", dotos "given"; histāmi "stand", statos "stood"; tithēmi "place", thetos "placed"); in Latin, the ∅-grade is always a (datus "given", status "stood", factus "done").

If we assume laryngeals, these inconsistencies completely disappear. Those roots are now reconstructed as *deh₃, *steh₂, *dheh₁; the weird long vowels are the result of a laryngeal after *e; the weird short vowels are the result of a laryngeal on its own without a vowel next to it. And the ablaut is back to the classic ∅ vs e vs o, same as any other root.

And once we've added the laryngeals to the theory, well, it turns out they can explain basically all instances of phonemic *a. Some linguists still reconstruct *a as a marginal phoneme in words like *aból "apple", while others explain even these instances away as *h₂eból. At this point, adding *a to the theory as a full-on phoneme makes it more complex without adding much explanatory power, so Occam's razor suggests we should leave it out.

  • This was a very detailed explanation. It was very technical for me when you started to about Ablauts and grades. But I sort of get it. Please correct me if I am wrong, the Laryngeal basically replaces the *a phoneme, in order to account for the irregularity of deriving a, e and o from the root PIE vowel as it appears to behave erratically in different form. Is this correct? Curiously then, before the Laryngeal theory was adopted, how was deh3, steh2 and dheh1 reconstructed? Jul 30, 2021 at 18:31
  • @LinguistEnthusiast Before the laryngeal theory caught on, they were reconstructed as irregular roots *dō *stā *dhē which didn't follow normal ablaut patterns.
    – Draconis
    Jul 30, 2021 at 18:32
  • A related question to Laryngeals. How do we know they are definitely a feature of PIE, and not just a substratum influence in Antatolian from an Arabic language which are known for Laryngeals right? Jul 30, 2021 at 18:33
  • 1
    @LinguistEnthusiast I'd recommend asking that as a new question. The short answer is that we see indirect evidence of them in many different branches, even if Anatolian is the only one with direct attestation. (Also, Arabic is much later than the Anatolian languages; you'll want to look at Akkadian, which lost most of its "gutturals".)
    – Draconis
    Jul 30, 2021 at 18:41
  • Thanks Dranconis. You have been very helpful. When you say "normal ablaut patterns" do you mean like Latin always preserves the one sound *a in all forms of the word. This is regular? Jul 30, 2021 at 18:45

To your last point: In Indo-Iranian IE *e and *o regularly become “a”, *ē and *ō regularly become “ā”. The only notable exception is that *o in open syllables becomes long “ā”.

For example: Greek hepta = Vedic sapta- = Persian haft. Greek oktō = Vedic aṣṭā = Persian hašt.

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.