The Phonology of Proto-Indo-European

C. George Boeree

Although it is impossible to know how PIE was actually pronounced before it developed into its daughter languages, linguists have attempted to reconstruct more likely possibilities. The assumption is that there is some consistency to the changes that took place (phonological "laws") in the daughter languages, and that the pronunciation can be determined by working backwards. This article explains some of the results and adds a few comments as well.


The traditional set of consonants included the following plosives:


gʲʰ gʷʰ
The plosives with ʲ are palatalized, i.e. pronounced with a slight j (English y) sound; those with ʷ are labialized, i.e. pronounced with a slight w sound. Those with ʰ are aspirated, meaning the b, d, g and gʷ are followed by a slight "breathy" h. This distinction is important in some languages (such as many in India), but is absent in English.

Because the palatal group of velar plosives (kʲ, gʲ, and gʲʰ) is very common, and the plain group (k, g, and gʰ) is very rare and may be accounted for in other ways, many now prefer a simpler set, with the plain set entirely replacing the palatals:
Notice that the b is in parentheses. That is because b is extremely rare in PIE, and may, in fact, not even truly exist. This is a very unusual situation. If any labial plosive is missing in a language, it will be the p, not the b. For this reason, the Russian linguist Diakonov suggested that the voiced set of plosives were really ejectives, that is, the sound is followed by a strong release of air from the glottis, indicated by a following apostrophe. This has led to the glottalic theory, which lists the plosives so:
t’ k’ k’ʷ
Notice that the first row of plosives have become aspirated or "breathy". We can see this in the difference between the sounds of p, t, k, and qu in spin, steep, skip, and squeeze and those in pool, tool, cool, and quo, which are actually aspirated, even though we never think about it.

It is interesting to note that this system of plosives is very similar to that of the Northwestern Caucasian languages, which may have had an early effect on PIE. Kabardian, for example, has exactly the same system.

And just to be complete, a more recent version of the glottalic theory looks like this:





I will be bold and suggest another possibility:
kʲʰ kʷʰ
gʲʰ gʷʰ
In other words, the traditional first line is unvoiced and pronounced with aspiration; the second is pronounced unvoiced without aspiration, and the last line is pronounced with voice and aspiration. This is a common set of plosives as languages go, and easily permits the near absence of the p. Also, nearly all languages that have voiced aspirated plosives (e.g. dʰ) also have unvoiced aspirated plosives (e.g. tʰ).

The problem of certain combinations of stops in roots not being represented is also simplified:
No two unaspirated stops (such as tek) in a root;
Among aspirated stops, no voiced with unvoiced (such as dhekh or thegh).
In support of this, note that my aspirated p, t, and k are retained in Armenian and become parallel fricatives in Germanic, while in most other daughter languages they are deaspirated. Plain p, t, and k of the middle line are retained in Armenian and Germanic, while they become voiced in others. And aspirated b, d, and g are retained in Sanskrit, while devoiced in Greek and deaspirated in the rest. Tokharian apparently lost all aspiration and voiced-unvoiced distinctions among plosives. All this suggests that Hittite, Tokharian, Germanic, and Armenian all left the "center" of PIE speakers before that "center" changed to the more traditional system.

In Mandarin, the sounds written as b, d, and g in pinyin are actually unvoiced, unaspirated sounds, and p, t, and k are stronger and aspirated. So, p, t, kʲ and kʷ could easily become b, d, gʲ and gʷ (which would still be differentiated from aspirated bʰ, dʰ, gʲʰ and gʷʰ). pʰ, tʰ, kʲʰ and kʷʰ would no longer need to be aspirated (or they could become unvoiced fricatives f, θ, x, and xʷ, as in Germanic). With no balancing unvoiced aspirates, the voiced aspirates would be pressed to deaspirate, even if that involves a collapse of voiced stops into a single set (b, d, g, and gʷ).

Among the velar plosives, I suggest that the distinctions between kʲʰ/kʲ/gʲʰ and kʷʰ/kʷ/gʷʰ are similar to Irish slender (lips as if to say ee) vs broad (lips as if to say oo).

In western PIE (eg. Celtic, Italic, and Germanic), the slender velars lost their slight y sound, and the broad became more clearly kw and gw. In eastern PIE (eg. Balto-Slavic and Indo-Iranian), the broad velars lost their slight w sound, and the slender velars became ɕ, ʃ, or s and ɟ, ʒ, or z. The western languages became known as centum languages (from the Latin word for 100, pronounced kentum); the eastern languages became known as the satem languages (from the Avestan word for 100). Tokharian and Greek are also considered centum languages, and Hittite may have been. Armenian and Albanian are considered satem.

Grimm's Law (regarding changes in the pronounciation of plosives in Germanic) looks a little different if one uses this scheme, and no longer requires a "cascade" of changes (wherein the first changes, from aspirated unvoiced plosives to fricatives, are the most unusual, and the final changes, from aspirated voiced plosives to unaspirated voiced plosives, occurs in most of the Indo-European daughter languages):
pʰ > ɸ > f tʰ > θ kʲʰ  > x > h
kʷʰ > xʷ > hw
t (no change) kʲ > k
kʷ > kw
bʰ > b dʰ > d gʲʰ > g gʷʰ > gw > w
Note that the Germanic words with a "p" were borrowed from non-Germanic languages, for the most part.

It is also interesting to note that as Uralic languages adopted various PIE words, the adapted them to their own speech patterns. The regular plosives (p, t, k, kw in my scheme) were the only ones they were comfortable with. The aspirated plosives they turned into fricatives, both the unvoiced ones, as in Germanic, but also the voiced ones. In addition, the Uralic languages moved the accent to the first syllable, the same as the Germanic languages. Perhaps at least a part of the origin of the Germanic languages is just PIE as attempted by a Uralic substrata.

Outside of Germanic (and, to an extent, Armenian), the changes look something like this (including the centum / satem distinctions):
pʰ > p
tʰ > t
kʲʰ > k / ɕ (etc.)
kʷʰ > kw / k
t > d
kʲ > g /  ɟ (etc.) kʷ > gw / g
bʰ > b
dʰ > d
gʲʰ > g /  ɟ (etc.) gʷʰ > gw / g
Sanskrit, of course, retains the original aspirations for the first and last row. Next, here are examples of PIE words (with my versions in parentheses) and their English descendents:
*pōds, *ped (pʰōts, pʰet)
      - foot
*tréyes (tʰréyes)
      - three
*ḱwon (kʲʰwon)
      - hound
*kʷid, kʷod (kʷʰit, kʷʰot)
      - what
 - *déḱm̥t (tékʲm̥t)
      - ten
*ǵénu, ǵnéus (kʲénu, kʲnéus)
      - knee
*gʷih₃wo (kʷihʷwo) (alive)
      - quick
*bʰréh₂ter (bʰrehtʰer)
      - brother
*dʰugh₂-tér (dʰukʲhtʰer)
      - daughter
*ǵʰans (gʲʰans)
      - goose
*gʷʰerm (gʷʰerm)
      - warm

 Before moving on to the vowels, I must mention the rest of the consonants believed to have been a part of PIE: s, m, n, l, r, j (sometimes written i̯ or y), and w (sometimes written u̯). These have remained remarkably stable in most of the daughter languages.


Just as the consonants have different possible systems, so do the vowels. The traditional system postulates five vowels, which can be short or long (e, o, a, i, u, ē, ō, ā, ī, ū). In addition, there are 12 diphthongs (ei, oi, ai, eu, ou, au, ēi, ōi, āi, ēu, ōu, āu). Also included in many interpretations is the central vowel ə, possibly pronounced like the a in ago.

When it became clear that Hittite (as well as the other Anatolian languages) was in fact a relative of the other Indo-European languages, a new way of looking at the vowels arose: The basic vowels in this system are e, ē, o, ō (with a possible a and ā), plus vowel versions of w and j (i.e. u and i). The evidence from Hittite suggests that a set of three laryngeals (consonants pronounced somewhere in the back of the mouth), labelled h₁, h₂, and h₃, affected the pronunciation of these vowels. Notably, the laryngeals altered the pronunciation of e, changing it to a when next to h₂ and to o when next to h₃, and lengthening the vowels when they follow them. When a laryngeal occured between two consonants, it would have been pronounced as ə.

Suggestions as to what these laryngeals may have sounded like include the following:
h₁: [ʔ] [h] [ç]
h₂: [ħ] [ʕ] [ʜ] [ʢ] [χ] [x]
h₃: [ʕʷ] [ʢʷ] [ɣʷ][ xʷ]
In my (amateur) opinion, the laryngeals may have been pronounced as hʲ, h, and hʷ. I make these suggestions because Hittite was most likely to retain an h where h₂ (and sometimes h₃) had been and, in a few daughter languages, you can find i where h₁ had been and u where h₃ had been. I would expect that the hʲ sound would move the vowel towards i and the hʷ  towards u, thereby altering ə to e and o respectively. The progression, then, from pre-PIE (PIE before Hittite split off from the rest) to Balkan PIE might look like this:
h₁e – hʲə > e
h₂e – hə > a
h₃e – hʷə > o

eh₁ – əhʲ > ē
eh₂ – əh > ā
eh₃ – əhʷ > ō
(I prefer to use ə with laryngeals, to allow movement to the front for e, to the back for o, and down for a.)

Indeed, in ancient Greek if nowhere else, laryngeals in these positions become e, a, and o, respectively. In other daughter languages, they are all represented by a single vowel or they disappear altogether.

If you look at the Kabardian language (of the Northwest Caucasian family), the whole range of expressed vowels (i, e, a, o, and u) are actually reflective of underlying diphthongs with a or ə plus w or j, similar to the effects of laryngeals. Just looking at the origins of the long vowels:
/ǝw/ > [u:]
/aw/ > [oː]
/ǝj/ > [iː]
/aj/ > [eː]
ha init, ah elsewhere > [aː]
It has been suggested that very early PIE was influenced by this family of languages (and may even have had only two basic vowels - a and ə!).

I also suggest that, when a laryngeal occurs between consonants or at the beginning or end of a word before or after a consonant, these three breathy laryngeals were whispered unvoiced vowels:
hʲ > i̤/e̤
h > a̤
hʷ > ṳ/o̤

All systems include the possibility of using l, m, n and r as syllabic consonants which function like vowels. These are written l̥ n̥, m̥, and r̥, and are pronounced a little like they often are in unstressed endings in English such as those in little, mitten, and bitter. In the laryngeal approach, these too can be affected by the laryngeals:
iH > ī (H represents any of the laryngeals)
uH > ū
l̥ = lə
l̥l = əl
l̥h₁ = ləhʲ > lē
l̥h₂ = ləh > lā
h₃ = ləhʷ > lō
The same pattern applies to n̥, m̥, r̥. Note, however, that these distinctions are only truly found in Greek. In all other branches of PIE,  l̥H > l̥: > Vl/lV, that is, the presence of a laryngeal after a syllabic consonant only lengthens the syllable, and the daughter languages add a vowel (V) before or after the l, n, m, or r.

Once again, we will never know how these prehistoric people spoke. But it does make for an interesting linguistic puzzle!

© 2014, C. George Boeree