Who introduced the notation (vowels with inverted breve below) for Proto-Indo-European laryngeals and when?

Proto-Indo-European has been reconstructed with so-called "laryngeal" consonants, spelled *h1, *h2, and *h3. These were lost in branches other than Hittite but left traces on adjacent short *e as well as compensatory lengthening in descendant languages. Based on how they color *e, it's likely that these sounds were fricatives [x], [χ], and [xʷ], corresponding to the velar, uvular (formerly "plain"), and labiovelar plosives. Occasionally when these are syllabic, I've seen them spelled with the schwa indogermanicum: *ə1, *ə2, and *ə3.

But lately I've been seeing reconstructed PIE with *e, *a, and *o with half-circles under them. I'm guessing that these correspond to syllabic allophones of laryngeals *ə1, *ə2, and *ə3. Am I right? What source introduced this orthography and why? Is it aesthetic, to make PIE look more like "words" and less like mathematical formulae, for the same reason that Pinyin uses accent marks instead of tone numbers? Or is there some new insight into the reconstruction of how the laryngeals were pronounced?

And how do you type these, other than copying and pasting from elsewhere?

  • You can probably type them with the Unicode combining inverted breve below (U+032F) or some other combining character, combined with the schwa letter. I'm sure TIPA has some way to do it too.
    – tsleyson
    Commented Nov 16, 2014 at 21:09
  • Can you link to some articles which use the specific notation you mention?
    – Valandil
    Commented Jul 23, 2017 at 8:49
  • 1
    @Default Anixx's posting history in the PIE tag uses the e, a, and o with inverted breve below. The other two notations (h and schwa) are used in Wikipedia's article about laryngeals and (presumably) the articles that it cites. Commented Jul 23, 2017 at 18:54

3 Answers 3


Typing these characters is fairly straightforward, if you have an appropriate keyboard (or can customize yours): they're simply the lowercase Latin letters <e a o>, followed by the character U+032F Combining Inverted Breve Below. This second character can't be copied and pasted on its own, since it's combining, but it should be available in any decent IPA keyboard (such as this online one). In TIPA (the standard way to type linguistic symbols in LaTeX), you can use the \textsubarch macro (\newcommand{\htwo}{\textsubarch{a}} or the like).

The reasons for using these letters are a bit more complicated.

First, I'll provide a basic summary of what the laryngeals actually are. (OP obviously already knows this, but it makes the answer more useful to others.)

It's generally agreed that many Proto-Indo-European phonemes could appear as either consonants or vowels. In particular, *r *l *m *n *j *w could all appear in both ways. This is why it's sometimes claimed that PIE had only two vowels (*e *o) when various reconstructions appear with *i *u: the *i *u were the vocalic versions of *j *w, which are seen as consonants first and foremost.

A PIE root, at its most basic, consists of two groups of consonants. For example, the root *bʰ_r means "carry", while the root *l_jǵ means "bind". In between these two consonants would be placed either a short vowel *e or *o (short grade), a long vowel *ē or *ō (full grade), or nothing at all (zero grade). And in the case of nothing at all, one of the consonants would often end up becoming vocalic to make it pronounceable.

But linguists found some roots which didn't seem to fit this pattern. Some of them had no consonant at the end, like *d_ "give", while others had no consonant at the beginning, like *_d "eat". And the ones with no consonants at the end, didn't seem to follow the normal ablaut patterns. "Give", for example, only ever appeared with *ō, which was never short or replaced with *e.

The current view is that these changes were caused by "laryngeals" (which have nothing to do with the larynx but the name has stuck). There were (at least) three of them (*), *h₁ *h₂ *h₃ = *e̯ *a̯ *o̯. When they appeared next to an *e, they would "color" it into their respective vowel. If they appeared after a vowel and before another consonant, they would also make the vowel longer. Then, in every language besides Hittite, they would all disappear. So the root for "give" wasn't actually *d_, but *d_h₃ = *d_o̯.

(*) Some scholars say there were more, maybe as many as five, but this theory isn't widely accepted. I haven't seen any modern reconstruction with less than three.

Writing the laryngeals as *e̯ *a̯ *o̯ certainly does seem more elegant, especially by analogy with *j *w sometimes being written *i̯ *u̯. But it's unclear how good this analogy is.

When the consonants *j and *w are left on their own (in the zero grade or for other reasons), they appear as *i *u. Other consonant-vowel pairs seem to have done the same. This is quite uncontroversial: for instance, *nj-sd-os "nest" surfaces with an /i/ sound in most descendants (see Latin nīdus).

However, when the laryngeals were left on their own (i.e. with no adjacent vowels), they disappeared in almost all descendants. Greek alone shows the reflexes /e a o/. Latin, for instance, shows /a/ for any laryngeal left alone between two consonants, while "proper" PIE *e *o within roots survive. To me, this is convincing evidence that *h₁ *h₂ *h₃ were not the consonantal equivalents of *e *a *o.

It's possible that they did have vocalic equivalents (sometimes called the Schwa Indogermanicum and written *ə₁ *ə₂ *ə₃), but it's also possible that they didn't: the branch that eventually became Greek might have inserted some sort of epenthetic vowel next to laryngeals, which then got colored. But it seems clear that, if they did have vocalic equivalents, they were not the same as the vowels *e *a *o.

As such, I wouldn't recommend using these symbols. They're certainly more elegant and easier to read than *h₁ *h₂ *h₃, but the false analogy with *i̯ *u̯ seems too dangerous. Instead, they might perhaps be written *ĕ *ă *ŏ? But I don't believe this has ever been done.

  • 1
    I’d say it’s pretty universally acknowledged that the ‘vocalic equivalents’ of the laryngeals arose from epenthetic schwa-like vowels being inserted before or after the laryngeal. More importantly, though, you don’t seem to actually answer the question here – who came up with the e̯a̯o̯ notation? When was it first used? The logic behind it is fairly obvious, but says nothing about who thought of it. Commented Apr 25, 2020 at 6:55
  • But the feminine suffix -h₂ bacame -a in all branches? Or not?
    – Anixx
    Commented Feb 24, 2021 at 20:01
  • @Anixx Isn't the feminine suffix usually reconstructed as *-eh₂, with a vowel that then gets colored by the laryngeal?
    – Draconis
    Commented Feb 24, 2021 at 20:02
  • @Draconis -eh₂ or -h₂ are just ablaut (or thematic) grades of the suffix as I see it. Some words had it while others did not (consider gwenh₂)
    – Anixx
    Commented Feb 24, 2021 at 20:05
  • also many had -ih₂ feminine suffix.
    – Anixx
    Commented Feb 24, 2021 at 20:06

As I look back on this with the benefit of hindsight, I see that I wrote a lengthy answer about the benefits and drawbacks of this notation, but didn't say anything about where it came from.

As far as I know, this notation (e̯ a̯ o̯ for *h₁ *h₂ *h₃) is part of Anixx's personal transcription system, used in their answers here. I've never seen it used elsewhere.

  • I think I may have seen it elsewhere, but can't remember where. I know I've seen h<sub>e</sub>, h<sub>a</sub>, h<sub>o</sub> much more widely, and that lacks most of the disadvantages you raised to Anixx's system, whilst still having it's advantages
    – Tristan
    Commented Feb 23, 2021 at 10:20
  • ugh, looks like sub doesn't work in comments, and underscores will just italicise stuff. You get the idea though
    – Tristan
    Commented Feb 23, 2021 at 10:20
  • 1
    @Tristan The main advantage the breves have that subscript vowels don't is readability, imo; whether it's a number or a vowel, the subscript requires me to look closer rather than recognizing the symbol immediately with a quick glance.
    – Draconis
    Commented Feb 24, 2021 at 20:05

As you mentioned with "syllabic", in PIE some consonants could be considered semi-vowels, that is they could become phonetic vowels beween consonants (vocalize). For instance, u̯ -> u, i̯ -> i Some researchers also suppose that r, l, n, m (sonorants) could also vocalize into r̥, l̥, m̥, n̥.

Actually nobody knows how the laryngeals were pronounced. I encountered authoritative sources that claimed that laryngeals could not be vocalized at all and other ones which claimed that laryngeals in fact vocalized at least between consonants. Indeed it is difficult not to vocalize the laryngeal in pa̯tēr. The issue is complicated by the fact that syllabification rules are largely unknown for PIE.

Consider for instance the word a̯eu̯a̯os "grandfather". It can be syllabified as [aewaos] or as [xeuxos]. In support of the first theory speaks the fact that other Eurasiatic and Nostratic languages had a similar word for father (ʔab/ʔaw/ʔabbā/abu in Afro-Asiatic, appe in Uralic, ap in Dravidian).

Consider also the PIE word a̯epa̯ "water". As you may know that p often corresponds to kw in various languages (optics/ocular, quintic/pentagon, wolf [f < p]/Russian volk, Latin vesper/ Russian vecher [ch < kw]). As such the word a̯epa̯ seems strikingly similar to the word "a̯eq̆a̯" which also means water. But a̯eq̆a̯ is sometimes considered a borrowing from non-Indo-European (but related) languages. So if a common ancestor of those languages and Indo-European (say, Eurasiatic) had the word, it could be reflexed in PIE as a̯epa̯ and in another offspring as aqua (and then borrowed in this form to PIE). Thus the laryngeal a̯ likely corresponds to just /a/ is sister families.

There is also another theory. It claims that at least in some words the laryngeal a̯ corresponds to /t/. This could be dialectal or due to borrowings from sister families. Back to the word for grandfather a̯eu̯a̯os, there is also a PIE word teuta̯ "kin", "clan". As you can see, it corresponds to a̯eu̯a̯os up to the gender. And it gave a lot of reflexes meaning "grandfather", "father" or "elder relative". Consider English "dad", Russian дед "grandfather", дядя "uncle", тётя "aunt", Ukrainian and South Russian тятя "father" etc. Semantically, the word meaning "kin", "clan", or "tribe" could evolve as "grandfathership", a group of people headed by a "grandfather".

From this point of view the word for water a̯eq̆a̯ "flowing water body" would be related to the word teq̆os "flowing water" (which again corresponds up to a gender), compare for instance, Russian течь "to flow", "leak".

A third example being the PIE words a̯ecsom and tecsom, both meaning "sharp tool, chopper".

Note also that in PIE the feminine suffix -a̯ also stood for collective number and for abstract nouns. Thus, if teq̆os is a stream, then teq̆a̯ would mean "a set of streams", "flow", "river", compare nebhos "cloud" (the -os is suffix here rather than ending) -> nebhesa̯ "set of clouds", "sky" (Russian "nebesa"). Similarly, if teutos is a relative then logically teuta̯ is set of relatives, kin, clan.

  • 11
    It should be noted that Anixx refers to "Eurasiatic" and "Nostratic", which are theories which are not universally accepted.
    – Colin Fine
    Commented Nov 18, 2014 at 22:26
  • 8
    I don't see how this addresses the question, which is about the origins of the notation a̯ etc. (which I've only seen in Anixx's answers on this site, btw), not about the phonology of PIE laryngeals.
    – TKR
    Commented Nov 19, 2014 at 21:24
  • 4
    Also "South Russian" is blatantly politically charged: no linguist refers to Ukrainian as "South Russian". See this deleted question by the same poster.
    – jogloran
    Commented Nov 19, 2014 at 22:52
  • 3
    @jogloran I meant Ukrainian AND South Russian, not that Ukrainian is South Russian of course. South Russian has тятя, Ukrainian has тато, Rusyn also has a similar word.
    – Anixx
    Commented Nov 20, 2014 at 4:00
  • 4
    I agree with @TKR, the answer fails to address the question
    – Darkgamma
    Commented Jan 12, 2015 at 18:37

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.