1 post tagged “vowel harmony”

One might not realise it, but typologically the Indo-European verb is actually incredibly strange. Depending on your reconstruction the Aspect is marked in two or three ways, while actually marking it once would be enough and is enough in the majority of languages.

To illustrate this let me give you some examples.
*h₁és-mi 'I am' a athematic present 1.sg
*wóid-h₂e 'I know' a perfect 1.sg

Not only does the vocalism *é/ó mark whether the verb is a present or a perfect, also the ending *-mi and *-h₂e mark that the verb is a present or a perfect. This double marking is superfluous, and most languages try to get rid of such features, or don't develop it at all.

It get's even more extreme when we look at reduplicated presents and perfects. These verbs mark their aspect three times. For example:
*dʰi-dʰéh₁-mi 'I place'
*de-dórḱ-h₂e 'I saw'

Typo of reduplication, type of grade in the root and the ending mark the aspect of the verb! This is overkill to say the least. The overkill is so bad, that it makes these forms highly suspicious.

I'm not one to quickly suggest a completely different reconstruction for actual indo-european. I'd sooner accept that the currently reconstructed form were in fact 'true' at some point in time, but their unlikelihood is a strong indication to a likelier system in a not too distant past.

For example I am quite confident that the stop series T/D/Dʰ is correctly reconstructed, at least for the more eastern dialects. But this doesn't take away the fact that this system is so unlikely that it must have been something else first (for example Glen Gordon's Voiceless/Creaky voiced/Fully voiced system is quite nice).

Well the same goes for this double/triple marking of the verbal aspect. I have no doubt that this was once true, it's way to widely attested to argue the existence; nevertheless this can't have been the original situation, it's just too unlikely.

So this begs the question, how did it work? Several options come to mind. First the, in my opinion, most blatantly obvious one is that the *e vocalism in the verb is actually a form of i-umlaut. the *i of reduplication would then actually be an umlauted *ə, while the *e we find in the verb root iself is actually an *ä (i-umlated *a, phonetically probably [æ] before merging with *e). Now the question is how did this umlaut occur.

The first origin of this proposed 'i' to cause the i-umlaut that comes to mind is the *-i we find in the present endings. There's a problem with this though, we do not find this *-i ending in past tense of the present (that sounds odd, remember the 'present' is an aspect not a tense in Indo-European, it's a horrible misnomer actually). For example
*dʰi-dʰéh₁-mi in the past tense would become *(h₁e-)dʰi-dʰéh₁-m.

Nevertheless we still find this 'i-umlaut' even though there's an absense of the *i. This is still possible, one could propose that the i-umlaut was reanalysed as a marker of the present aspect. And then this i-umlaut was spread through the paradigm, even in places where techinically it wouldn't even exist.

Whether one finds this convincing, is another story. I can definitely see it as a possibility, but would not dare to call it proven, until we'd compare Pre-Proto-Indo-European with another proto-language and can confidently say that those two are related. Right now it's only a 'tendency' that can be noted. I like the elegance of the explanation though.

There's still problems with this though, as we'd also want to see a very consequent i-umlaut in the instrumental singular. But would such a thing be preserved if only one case would affect it?. There's several other cases where such an environment would occur that would cause an i-umlaut. But none of them would be as commonly occuring as in the present aspect. So therefore it's imaginable that it would only survive in the present aspect. But because of this, it makes it really hard to find undisputable proof for the idea.

Then again, if the *-i was the particle that brought about the i-umlaut, we would have to see this as quite a recent development, as this present marker is probably a later addition to mark the present. We can deduce this from the form we find in the aorist, which takes the endings without *-i (so called secondary endings), while the Aorist probably wasn't a past tense (although it is in Greek and Classical Sanskrit), but rather a perfective or 'focus' aspect, which is by no means a past tense.

Another explanation that Glen Gordon proposed in his lovely PDF of Pre-Indo-European changes is that due to Altaic influence Indo-European went through a short period where it had Vowel Harmony. The sheer coolness of this theory makes it incredibly appealing. Often though, 'cool' ideas are incredibly exotic, and thefore highly unlikely. I know of know language that 'gained' vowel harmony through intensive contact with a Vowel Harmonic language.

Maybe we could point out Old Norse as having done that. Due to exsessive i and u umlauts, words have ended up having either only back vowels, or front vowels rather often. In some way we could definitely consider this vowel harmony Except that it's regressive (The vowels of the suffix influence the whole word, while usually vowel harmonic languages have the word that influences the vowels of the suffix). Maybe we could see this appearance of Regressive vowel harmony as having been influenced by the Finno-Ugric languages with Progressive vowel harmony that were spoken in this region.

Back to Glen's Vowel Harmony. He describes this occurence like this Proto-Indo-Aegean development like this:

A brief period of Vowel Harmony, an isogloss shared with Pre-Proto-Altaic to the southeast,
may be warranted to explain the source of qualitative ablaut of *ə and *a between active and
stative conjugations. Presumably, the active conjugation which used personal endings
exclusively with *ə-vocalism (i.e. 1ps *-əm, 2ps *-əs and 3ps *-ə) and the stative
conjugation which used personal endings exclusively with *a-vocalism (i.e. 1ps *-ax, 2ps *-
at, 3ps *-a) acquired their distinctive vocalisms in the stem by way of regressive vowel
harmony emanating from their contrastive personal endings.

Note that, indeed, this is also Regressive Vowel Harmony, just like Old Norse, while Proto-Altaic has Progressive Vowel Harmony. I am not completely sure about the credibility of Glen's reconstructions, I have not yet ventured even further back than Pre-Indo-European. What I do wonder is where these proposed vowels that caused the Regressive Vowel Harmony went.

The Indo-European endings for the ə-vocalism are *m, *s, *t which go back to an earlier *mə, *sa, *ta. The last two might be innovations of Indo-European but the first person singular is *mə and not *əmə as this would ruin the Syncopation.

The same goes for the endings that caused the a-vocalism they are: *h₂e. *th₂e and *e. These go back to earlier h₂ə, *th₂ə and  *ə. Only *ə goes back perfectly to an earlier *a. *ax would have an expected result *h₂. The second person singular ending is clearly built of the *t that Glen reconstructs + *h₂ from the first person singular ending. But from the forms *-ax and *-at I failt to see where those *a's went, and where those *ə's came from.

Hopefully, Glen will shed some light on this in the comments box, so keep your eyes peeled ;-).

Whether this Vowel Harmony truly developed is very dependant on how likely you think it is that Regressive Vowel Harmony arises by areal influence of Progressive Vowel Harmony*. Or even whether a language can 'gain' Vowel Harmony from areal influence at all.

After reading Glen's PDF, and his idea of Vowel Harmony, I came with an idea that originally motivated me to write this lengthy post.

In Proto-Indo-Aegean there used to be an *i and an *u which were lost due to vowel centralisation, Glen formulates it as follows in his PDF:

Due to influence of neighbouring Pre-Proto-Abkhaz-Adyghe to the east, the original 4-vowel
system of *a, *ə, *i and *u collapsed into a centralized system of only *a and *ə. In open
syllables *i and *u were diphthongized to *əi and *au respectively while in closed syllables,
they became *ə and *a respectively. Note that at this stage, while *i and *u may have
continued to exist on the phonetic level as allophones of diphthongs *əi and *au, they were
now only variants of *y and *w on the phonemic level.

From reading this was, if *ə can come from *i, then wouldn't it be possible that the i-umlaut took place much earlier than in Indo-European itself, namely in the Pre-Proto-Aegean stage. Glen indeed does reconstruct endings with *i in this period giving some credibility to this idea. On the same page as mentioned above he replies to this suggestion though. I see no reason why I'd reformulate it, while he explains it best, so I'll quote:

Yes. I presume you are pointing to a "pre-Indo-Aegean" stage then? I call that stage "Proto-Steppe", the ancestor of Indo-Aegean, Boreal (Uralic, EA, Chukchi-Kamchatkan) and Altaic. In the Proto-Steppe stage, I'd reconstruct transitive personal endings *-im (1ps), *-it (2ps) and *-i (3ps) plus intransitive endings *-uh (1ps), *-un (2ps) and an endingless 3ps. Since Steppe *u > IAeg *a and Steppe *i > IAeg *ə via Centralization, your i-umlaut idea almost works and was something that I initially thought of.

However, since Centralization and the concommitant transfer of labialization from the vowel to the neighbouring consonant is naturally heavily dependant on vocalism, I would have to oppose i-umlaut at the Proto-Steppe stage unless you can find a different formulation of Centralization that works to explain the rise of labialized consonants in PIE.

If this ablaut preceded Centralization, we would find verb stems alternating between labialized (eg. *kʷ) and non-labialized consonants (eg. *k). This is why I opt for vowel harmony *after* the dissolution of the Proto-Steppe speech area and after the Centralization rule had already taken place, circa 8500 BCE or so. This hypothesis is in alignment with Allan Bomhard's views in Indo-European and the Nostratic Hypothesis (1996).

So there you have it, a very lengthy, and an inconclusive survey of the theories that fly around in my head about explaining the odd occurence of the vocalism in the Indo-European verbal roots, and on top of that an explanation why it is so important to try and explain this odd occurence due to its highly unnatural superfluency to mark the Aspect.

*I just realised, that maybe I'm all wrong, and Pre-Proto-Altaic originally had Regressive Vowel Harmony. This would definitely surprise me, as the Altaic languages that I have a bit of knowledge of namely Turkish, but mainly Classical Mongolian, both have Progressive Vowel Harmony (I'm not a big proponent of adding Japanese to the Altaic family, and besides that, it has no obvious Vowel Harmony at all). If for some kind of bizarre twist of history Altaic shifted from Regressive Vowel Harmony to Progressive Vowel Harmony, I believe that this does add a bit more plausibility to the 'areal influence' theory.