Proto-Indo-European Ablaut (Apophony)

[Rin]

How did the PIE ablaut system develop? In other words, and more generally, what sorts of changes would lead a non-apophonic system to become apophonic?

[chris_notts]

Well, I don't know specifically how the PIE system developed, and I'm not sure anybody does. But the most common way to get vowel alternations is from vowels being influenced by following vowels. For example, if the following vowel is /i/ then the preceding vowel may be fronted and/or raised. Then when the /i/ which triggered the change is lost (a common process word-finally), the only marker of the distinction is the change in the root internal vowel.

See the following on Wikipedia for example:

http://en.wikipedia.org/wiki/I-mutation
http://en.wikipedia.org/wiki/A-mutation
http://en.wikipedia.org/wiki/U-mutation
http://en.wikipedia.org/wiki/Germanic_umlaut

I should point out it's unlikely that every single verb or noun has the right environment for this kind of change to occur. In languages where apophony is the normal marker for some distinction, it's likely that originally it only applied to a subset of nouns or verbs and was generalised by analogy.

[WeepingElf]

How did the PIE ablaut system develop? In other words, and more generally, what sorts of changes would lead a non-apophonic system to become apophonic?

The answer is, unfortunately, that the answer is not known, at least not for sure. Otherwise, it would be found in the handbooks. The main factor seems to have been accent positions, though. In my humble personal opinion, Pre-PIE had a penultimate word accent which, as the lengths of the inflectional endings varied between zero, one and two syllables, hit different syllables in different forms of the word. The accented syllable developed e-grade, the unaccented ones developed o- or zero-grade. The details, however, are very complex, and I haven't found out and don't really expect to find out. I don't seriously expect to get mentioned in future handbooks of Indo-European as the scholar who cracked that riddle

In one conlang, I developed freakish alternations of roots and affixes from a penultimate accent in an agglutinating protolang. Just a few ordinary regular sound changes, and the result is a maddening amount of alternation and even seeming suppletion. Probably something similar happened in PIE.

[Whimemsz]

In one conlang, I developed freakish alternations of roots and affixes from a penultimate accent in an agglutinating protolang. Just a few ordinary regular sound changes, and the result is a maddening amount of alternation and even seeming suppletion. Probably something similar happened in PIE.

That's quite similar to what's happened in Potawatomi and some southeastern dialects of Ojibwe. Ojibwe-Potawatomi have stress rules that divide words into iambic (W+S) bisyllabic feet counting from the beginning of the word (with some minor complications like that long vowels can never be weak, nor can vowels in word-final syllables). Potawatomi, and later Odawa and Eastern Ojibwe, have deleted all metrically weak vowels. Now, since the addition of person prefixes changes which following vowels fall in weak or strong positions (because feet are counted from the beginning of the word rightward), this leads to enormously complex morphophonological alternations, complicated further by various resolutions of consonant clusters.

For example, in non-syncopating dialects of Ojibwe, "shoes" is makizinan, and "his shoes" is omakizinan. In syncopating dialects, these are respectively, mkiznan and makzinan. In non-syncopating dialects, "chief" is ogimaam and "my chief" is nindoogimaam; in syncopating dialects these are gimaam and ndoogmaam respectively. In non-syncopating dialects, "I have a cold" is nindagigokaa, while "she has a cold" is agigokaa; in syncopating dialects, these are respectively ndakokaa and gikaa (< ndaggokaa, gigkaa). Now, often these alternations are being analogically simplified in ways other than by the development of some sort of Afroasiaticky or IE-y consonant-root+vowel-alternation system, but it could easily have been analogized in that direction instead.

[Wattmann]

In one conlang, I developed freakish alternations of roots and affixes from a penultimate accent in an agglutinating protolang. Just a few ordinary regular sound changes, and the result is a maddening amount of alternation and even seeming suppletion. Probably something similar happened in PIE.

Thank you, Weeping Elf, you have made my life far easier now - I wondered where did the alternations come from.

[Rin]

Thanks, guys. I think I understand a bit better.

[Rin]

I remember reading somewhere that pre-PIE might have been like Semitic with consonantal roots. How probable is this?

[Basilius]

About as probable as anything else.

[Pole, the]

I remember reading somewhere that pre-PIE might have been like Semitic with consonantal roots. How probable is this?

But then, where are the consonantal roots from? Most probably from ablaut.

[Chagen]

I remember reading somewhere that pre-PIE might have been like Semitic with consonantal roots. How probable is this?

I don't think it had consonantal roots, but I remember reading that it didn't have infinitives.

[Drydic]

I remember reading somewhere that pre-PIE might have been like Semitic with consonantal roots. How probable is this?

It's a horrible misunderstanding of the PIE ablaut system. There's a vague resemblance, yes, but that's just it: a vague resemblance. (and don't even get me started on the Semitic system...)

I don't think it had consonantal roots, but I remember reading that it didn't have infinitives.

Wrong. PIE doesn't have a single reconstructible form for the infinitive, yes, but show me an early post-split IE language that doesn't have infinitives. The category existed, it just had multiple possible forms...which is, in fact, exactly the situation attested in Vedic Sanskrit, multiple forms with identical meanings.

[Basilius]

I don't think it had consonantal roots, but I remember reading that it didn't have infinitives.

Wrong. PIE doesn't have a single reconstructible form for the infinitive, yes, but show me an early post-split IE language that doesn't have infinitives. The category existed, it just had multiple possible forms...which is, in fact, exactly the situation attested in Vedic Sanskrit, multiple forms with identical meanings.

Seems to be mostly correct. Also, looking at all the similar formations - but not quite identical - used as infinitives in distant IE langs, I often get the impression that infinitives just inflected for some more categories (with different forms surviving in different branches), or something.

(Although Vedic infinitives weren't identical in meaning, that bit was wrong.)

[pun]But I'm afraid the majority view is currently exactly what Chagen quoted.[/pun]

[Drydic]

(Although Vedic infinitives weren't identical in meaning, that bit was wrong.)

Care to back this up? Seriously; my sources on Vedic state that they were mostly interchangeable, and that they didn't gain differentiated meanings until later on, in Classical Sanskrit.

[Basilius]

(Although Vedic infinitives weren't identical in meaning, that bit was wrong.)

Care to back this up? Seriously; my sources on Vedic state that they were mostly interchangeable, and that they didn't gain differentiated meanings until later on, in Classical Sanskrit.

To begin with, Vedic infinitives had inherent cases...

[Drydic]

Better keep going, since that doesn't detract from my point in the slightest.

[Basilius]

OK.

Care to back this up? Seriously; my sources on Vedic state that they were mostly interchangeable, and that they didn't gain differentiated meanings until later on, in Classical Sanskrit.

I don't think this is correct.

My sources say that Classical had just one form of infinitive, in -(i)tum. Despite being formally (and historically) an accusative form, it was used also in contexts where an ordinary noun would be put in some case other than accusative.

Vedic was quite different. Its numerous infinitives were frozen case forms, and generally forms with different inherent cases weren't interchangeable. That is, one had to use a dative infinitive to say "in order to do X", an accusative infinitive in contexts like "want to do X", a genitive/ablative infinitive in contexts like "prevent from doing X", etc. That case did matter is especially clear when the infinitive phrase has an argument put in the same case as the infinitive.

It's only the concurrent forms having the same inherent case that can be (and often are) said to be interchangeable, but my personal gut feeling is that even this is based on lack of data (and/or of in-depth research) rather than anything else. It would be more accurate to say that we just don't know what conditioned the choice of (e. g.) one of the numerous dative infinitives in a particular context.

(The above relates specifically to Vedic proper; it would be quite probable that some relatively late texts still classified as "Vedic" could violate the older rules of usage in all imaginable ways; at any rate, a transitional phase of that sort would look natural given the ultimate unification in Classical. But this is just my speculation trying to explain the statement you quoted - I haven't read anything on the specific topic of the language of Brahmanas etc.)