Very characteristic of the Germanic
languages - such as English
, and Swedish
- is an alternation
of vowels: plurals like man ~ men, goose ~ geese
, strong verbs like sing ~ sang ~ sung, write ~ wrote ~ written
, adjective with noun like long ~ length, hale ~ health
, and verb with noun like sing ~ song, bear ~ birth
Some of these have parallels in other Indo-European languages, such as Greek and Sanskrit, and can be traced back to their common ancestor, and to some extent even further back. They are fundamental to the vowel system, and the grammar, of Proto-Indo-European.
The word ablaut is also sometimes used for sound changes in other languages, not related, but of a similar character, and can be used of consonants.
Discovery of ablaut
In the eighteenth century Friedrich Klopstock
used the word Umlaut
(sound change) for all these Germanic changes, but in 1854 Jacob Grimm
, in a comparative
study of the Indo-European
languages, made an important distinction. The man ~ men
and long ~ length
types of umlaut
were confined to the Germanic languages, and had an explanation in terms of sound change
. But the sing ~ sang ~ sung
and sing ~ song
alternations were paralleled in all the Indo-European languages. Grimm called this deep alternation Ablaut
He arranged the Germanic strong verbs into classes. Modern English still contains pure examples of most of them, though many that existed in Old English have since developed differently, or become weak. We no longer say help ~ halp ~ holpen, though in German it's still helfen ~ half ~ geholfen. The classes are called the ablaut series, and the specific vowel in the pattern is called the ablaut grade.
Class I goes i-o-i, including drive, ride, rise, write.
Class II goes oo-o-o, including choose, shoot.
Class III goes i-a-u, including begin, drink, shrink, spring, swim.
Class IV goes ea-o-o, including bear, break, shear, steal.
Class V goes i-a-i, including give.
Class VI goes a-oo-a, including shake.
Class VII goes o-e-o, including grow, know.
By comparison, Ancient Greek has a verb dérkomai 'to see', perfect dédorka 'I have seen', aorist édrakon 'I saw'. The crucial alternation here is der ~ dor ~ dr. This pattern of e, o, and zero kept coming up. Compare also légo 'I say' and lógos 'word'. In Sanskrit it's been partly obscured because all of e o a changed to a, but a similar thing may be seen in present bharati 'bears', participle bhrta 'borne', and babhâra 'bore' with lengthened vowel.
It has also been much mutated by subsequent changes in other languages. The Germanic pattern sing ~ sang is an example of e ~ o alternation, with e affected by the neighbouring n and the change from o to a being a general change in Proto-Germanic. The form sung is originally zero grade, with a vowel reinserted to make it pronounceable.
What gradually emerged over the nineteenth century was that a lot of ablaut variation could be explained by three basic processes. One is the alternation between e and o. One is the vowel dropping out altogether. And one is the vowel lengthening to ê and ô. As these occurred in all the daughter Indo-European languages, they must have been active and grammatically important in the parent Proto-Indo-European (PIE).
The earliest work on PIE had been on the assumption that Sanskrit was the oldest. This was misleading, and so was the fact that the otherwise excellent ancient Indian grammarian Panini had the grades wrong: traditional Sanskrit morphology began with the zero grade bhr, and expanded it to the so-called guna grade bhar and the vrddhi grade bhâr. Grimm and his colleagues worked out the guna grade was primary.
Creation of vowels
Of course PIE had more vowels than just e
. Or did it? Grimm helped solve the consonant problem with Grimm's Law
, but despaired at ever understanding the history of the vowel
s. But over the course of the century some remarkable theories were put forward and gradually accepted.
One was the theory of resonants or sonants: see my write-up there for details, but in brief this was the idea that some sounds could act as both consonant and vowel, like l, m, n in bottle, bottom, button forming a syllable by itself. If the semivowels y or w got into one of the situations where the vowel had dropped out, it would be pronounceable if it turned into a vowel, so *bheydh- in zero grade, i.e. without the *e, appeared as bhidh-. Likewise *klew- 'hear' in zero grade is *klu-. This explained the existence of i and u as forms of e.
The laryngeal theory
The other theory was proposed to explain instances of a
where the existing theory would prefer e
. Most reconstructed root
s had consonants on either side, like *pet-
'fly' and *bheydh-
'bid'. These intractable a
didn't: they were roots like *ag-
'drive' and *dô-
'give'. In 1879 Ferdinand de Saussure
made the very bold conjecture that there had once been some kind of consonant there, which had changed the nature of the vowel and then disappeared. So *ag-
was really *xaeg-
was really *dexo-
. You needed at least two, one for a-colouring and one for o-colouring.
These elements were soon dubbed laryngeals because of a guess they might resemble so-called laryngeal sounds that did exist in Semitic, and they became a very useful hypothesis. It was not until 1927 that proof was obtained. Jerzy Kuryłowicz showed that Hittite, which had been deciphered in 1915, contained some kid of laryngeal consonant h in places where Saussure's theory predicted them. The exact number required for the theory, and how they were actually pronounced, are still in dispute. These days they are normally notated something like H1, H2, etc.
So some a and o come from e near a laryngeal; most i and u come from e plus a sonant consonant. Other vowels arose in late PIE or the daughter languages when the syllabic sonants developed vowels to make them easier to say. Going deeper into the prehistory of PIE, the only unexplained alternation was e ~ o.
A reason was proposed by Adolf Holtzmann
as early as 1844: he thought variable accent
was responsible. Unfortunately the two languages with the best evidence of accent, Greek
, disagreed. Clearly, Proto-Germanic
and Old Latin
had both obliterated their original pitch accent
at some stage and changed to a strong initial stress accent
; so they didn't seem to be able to help.
In the 1870s Verner's Law was proposed, and Verner showed that evidence of variable accent could be obtained from Germanic, even though no known Germanic language had it, because it explained an alternation in the consonants. With this information it could now be seen that Greek had changed more: légo 'I say' and lógos 'word' disguised an earlier stage when é was used when accented and o away from the accent.
The modern interpretation of the situation is still up in the air. Sonants and laryngeals are accepted, and we know what the later stage of PIE looked like, with some confidence. At that late period it had five short vowels i e a o u and five equivalent long vowels, and perhaps a schwa. Ablaut was no longer a phonetic process but had been co-opted to mark different forms of verbs grammatically: o-grade for the singular of the perfect tense, for example. This much is clear from reconstruction.
There are three features that are older than this level: the e ~ o alternation, the vowel ~ zero one, and the short ~ long one.
What is not clear is how deep into prehistory we can go. This is long before writing. Here is one popular theory. Originally there was only one vowel, which we can write e. This sounds bizarre, but has a parallel in the modern North-West Caucasian languages, though they're unrelated to Indo-European. It's not that there's only one vowel sound as such, but that all the shades of sound can be explained by the influence of the neighbouring consonants. So in *bheydh- the y would make the e more of an i-sound, while in *H2eg- the H2 makes the e more a-like.
This stage of the language had a strong stress accent. Some unstressed vowels disappeared. That explains the vowel ~ zero alternation.
Next it changed to a pitch accent. So instead of changing the quantity of vowels, accent changed their quality. The accented vowel was é and the unaccented one became o.
Then some consonants dropped out, specifically the laryngeals, and the vowel lengthened in compensation. Compare English light, which got its long vowel when gh dropped out. So *doH3- became *dô-.