This is self-indulgent of me, but this is how I presented the laryngeal theory to my poor Historical Linguistics students in 2002.
Saussure (1879): let’s look at Ablaut in proto–Indo-European:
- e:o:Ø
Greek patéra, eupátora, patrós
[father.ACC, of.good.father, father.GEN]
- eR:oR:R̩ where R is a resonant (jwrlmn):
- R=w:
Greek eleusomai, eiléːloutʰa, éːlutʰon
[I.will.come, I.have.come, I.came]
- R=l:
Greek stélloː, stólos, éstalmai
[I.send, sending, I.have.been.sent]
[a: epenthesis and reflex of *ə]
- ē:ō:ə
Greek títʰeːmi, tʰoːmós, Latin facio
[I.put, heap, I.make]
- ā:ō:ə
Greek pʰaːmí, pʰoːnéː, pʰásis
[I.say, voice, utterance]
- ō:ō:ə
Greek dídoːmi, dôːron, Latin datus
[I.give, gift, given]
Now hang on:
e o Ø
eR oR R̩ (j> i, w > u, l> l̩ > (a)l)
ē ō ə
ā ō ə
ō ō ə
Yuck. eR:oR:R̩ is just e:o:Ø put in front of R.
Why can’t ē:ō:ə be as nice? Why can’t it be… eə:oə:ə > ē:ō:ə ?
That way, we still have e:o:Ø + something, and a simple monophthongisation to wrap it up.
If we’re lucky, we might be able to account for ā:ō:ə and ō:ō:ə the same way.
Furthermore, Proto-IE phonotactics is typically (s)C(R)e(R)C:
- bʰer-‘carry’, k̂ei-‘lie’, kers-‘run’, pel-‘thrust’,
spen-‘stretch’, sreu-‘flow’, lewk-‘light, bright’, melg– ‘milk’,
plek-‘plait’, dʰwer-‘door’
But some roots are CV or VC—often with the same suspicious long vowels:
- aĝ-‘lead’, aug-‘increase’, dō-‘give’, ed-‘eat’,
gʰrē-‘grow, green’, gnō– ‘know’, od-‘smell’, stā– ‘stand’
So in Ablaut, ē, ā, ō are kind of behaving like eə, oə (former diphthongs.)
In phonotactics, ē, ā, ō are kind of behaving like eC or Ce (where C can include j, w—so eC, Ce can still be a diphthong)
Saussure’s solution: there is something there: a ‘sonorant coefficient’, *X
With X, we can say:
e o Ø
eR oR R
eX oX X > ē ō ə
aX oX X > ā ō ə
oX oX X > ō ō ə
(cf. eR:oR:R̩, e.g. ew:ow:w̩ [= u])
We can also say:
- dō– ‘give’ <**doX= CVC
- gʰrē-‘grow, green’ < **gʰreX= CRVC
- stā– ‘stand’< **staX= sCVC
Only one more catch: where do a and o come from in the left column in the first place? (eX; aX;oX). Why can’t we just make do with e:o:Ø for all rows?
Hey! Let’s posit two sonorant coefficients:
eA oA A> aA oA A > ā ō ə
eO̬ oO̬ O̬ > oO̬ oO̬ O̬ > ō ō ə
(to which others later added E:
eE oE E > eE oE E > ē ō ə)
A is ‘something’ to which e assimilates, giving a
O̬ is ‘something’ to which e assimilates, giving o
E is ‘something’ that leaves e alone
When A, E, O̬ drop out, compensatory lengthening or something for a, e, o; schwa for Ø.
If we have assimilation going both directions, we can now say:
aĝ– ‘lead’ <**Aeg = CeC
ed-‘eat’ < *Eed = CeC
od-‘smell’ < O̬ed = CeC
What’s this buying us?
- Consistent phonotactics: all stems are CVC (plus or minus R and #s).
- Makes sense of all kinds of Ablaut: it all boils down to e:o:Ø + something.
- Ablaut like ā:ō:ə now makes sense: it’s actually originally eA:oA:A.
Saussure tells the world…
… and noone cares. Who’d buy this bunch of abstract algebra? Who cares what the phonotactics of proto-Indo-European are? What’s wrong with just saying ā ō ə? What’s the evidence that these phonemes ever even existed?
Saussure dies 1916. Around that time Hittite is deciphered: turns out to be Indo-European, though distantly related to other IE languages.
‘water’: Greek hýdoːr, Gothic watō, Old Church Slavonic voda, Gaelic u(i)sce, Albanian ujë ,proto-IE *wedōr
… Hittite watar.
1927: Jerzy Kurylowicz notices something:
Latin/Greek Saussure Hittite
mālum ‘apple’ < *meAl- maḫl-
plānus ‘flat’ <*pleA- palḫiiiš
ōs ‘bone’ < *Oes- ḫastai
anti ‘against’ < *Aent- ḫanti
argēs ‘white’ < *Aerg- ḫarkis
esti ‘he is’ < *Ees- es-
A and O correspond to Hittite ḫ.
(E is no more attested in Hittite than in the rest of Indo-European)
Semiticists had noticed that the sonant coefficients, assimilating e to a or o, were behaving like Arabic laryngeals
… ḫ is a laryngeal.
So EAO must have been laryngeals: h1 h2 h3
Saussure was so right, even he would have been surprised.