7 Dec 2007

The Great Pre-IE Centralization

I'm influenced easily by commenters. I'm weak. No, it's all good because I've been recently influenced in a good way :) The other day, someone commented on Proto-Indo-European (PIE) phonology, regarding something that might have occured several millenia back in Pre-IE. Since I just finished a rant on Nostratic, I may as well add more to this conjectural fun.

One interesting thing that many have noticed is the few vowels that PIE contains compared to the number of consonants. PIE may have five vowels, however *i and *u can be shown to be merely vocalic reflexes of consonants *y and *w. For example, we can see that the zerograde forms *bʰuh₂- and *likʷ- of roots *bʰeuh₂- "to become" and *leikʷ- "to leave" behave in the same way as words with zerograde roots with resonant consonants like *bʰr̥- and *gʰʷn̥- from *bʰer- "to carry" and *gʰʷen- "to kill". To add, *a is so rare for some reason that some feel it doesn't exist at all (although I think this is an overly extreme position). Only *e and *o are productive in ablaut patterns and as a result it's suspected that it all comes from a centralized vowel system of and *a[1]. My view is that there was indeed an earlier system like this and that *a is rare because most instances of former *a had shifted to *o at a very late date in PIE's history, leaving behind a residue of instances of *a in words like *daḱru "teardrop" and *ǵʰans- "goose". These rare words would serve as witnesses to the earlier stage of the language.

However, if PIE originally did have a centralized vowel system, it implies all sorts of other wonderful things. Such a system, using only a few vowels and many consonants, is also typical of all Abkhaz-Adyghe languages nearby. The Abkhaz-Adyghe (AbAd) group is also known as "North-West Caucasian" or "NWC" because its languages center around the north-west region of the Caucasus Mountains which are flanked by Russia and eastern Turkey. We know that these languages have been there for uncountable ages. If the Nostraticist Allan Bomhard is correct in suggesting that AbAd and a very early form of IE were in contact with each other, is it also possible that their vowel systems evolved together as well?

The interesting thing about the phonological patterns that PIE and AbAd exhibit is that they could have easily evolved from a system with more vowels and less consonants by way of a common process of assimilation. In AbAd, palatalization (the act of pronouncing something with an added "y"-like quality) and labialization (the act of pronouncing something with an added "w"-like quality) are standard features in their sound systems. In PIE, only labialization is a phonemic feature of the language[2]. If we pronounce the sound /u/ (as in "ooooh"), we should notice that we're rounding our lips (i.e. labialization). This can be thought of as a quality that isn't necessarily loyal to /u/ but can also be transferred to neighbouring consonants in a word. So if you say /gu/ (as in the word "googoo"), when do you start rounding your lips? Chances are, you start rounding your lips while you're pronouncing /g/. This might be called anticipatory labialization because you're anticipating the next sound, so subconsciously you begin to round your lips ahead of schedule. This sort of anticipation is a universal tendency for speakers around the world to do.

So what does anticipation have to do with pre-PIE and pre-AbAd sound systems? When you say /gu/, it's practically unavoidable that you're really saying [gʷu] and this is how easily a plain stop can become a labialized stop. Similarly, a sound like /gi/ can easily become /gʲi/ through the same process. This process alone doesn't make a new phoneme however since the English sounds [g] and [] are still just allophones of a single phoneme /g/. They are said to be non-distinctive or non-contrastive because there is no contrast between /gu/ and /gʷu/ in the language, and the accompanied lip-rounding of /g/ is predictable when neighbouring the vowel /u/ in English.

Nonetheless, assimilation is no doubt the first step in creating the rich consonant systems found in PIE and AbAd. The second step involves the reduction of all previous vowels to a mere handful. Once this happens, labialization and palatalization can automatically become phonemic because the previous triggers for the assimilation (i.e. the original qualities of the neighbouring vowels) become erased due to vowel mergers. As a simple hypothetical example, if a language began with a 3-vowel system (/a/, /i/ and/u/, let's say) and all the vowels were suddenly reduced to /ə/, then we might automatically have a contrast between /gʷə/, /gʲə/ and // (from earlier /gu/, /gi/ and /ga/). Suddenly we have three consonants (u-coloured //, i-coloured // and plain /g/) where there was formerly only one (plain /g/). Since we'd have a contrastive triplet, we could then say validly that the three sounds are contrastive and thus phonemic. This is probably how the sound systems of these two language groups evolved.

Knowing that and considering the close geographical proximity of Proto-IE and Proto-AbAd to each other, it's then tempting to suspect that they evolved together this way because of areal influence. Their similarly evolved sound systems may provide some proof that they were in contact with each other at an early date. For me, I think that this is a serious possibility but it would have to have occured long before PIE itself which is said to be dated to about 4000 BCE. So I would suggest a date of 8000 BCE or so when the reduction of the vowel systems of both proto-languages would have begun to occur. (There's still more to talk about on this subject alone. So many topics so little time! Ugh.)

Others may not be convinced of any of this or my suggested date, and that's fine. This is conjecture afterall. Much more needs to be determined before any of this can be stated with certainty but it's always healthy to ponder every once in a while beyond what we currently know in order to encourage new avenues of research and exciting discoveries in comparative linguistics.


NOTES
[1] The idea that the alternation of *e and *o in Proto-Indo-European ablaut (e.g. *pódm̥ "foot" [acc.sg.] / *pedós "of the foot" [gen.sg.]) comes from an earlier ablaut of and *a was proposed as early as 1965 by Pullyblank. See Bomhard/Kerns, The Nostratic Macrofamily: A Study in Distant Linguistic Relationship (1994), p.76.
[2] Read Reinterpreting the Proto-Indo-European velar series to understand why I think PIE in fact doesn't have palatalized phonemes and how PIE's antiquated orthography can be deceiving.

8 comments:

  1. I don't think it's at all necessary to put this vowel reduction that far back in history. Once there's a tendency of vowel reduction (especially outside of accent), these things tend to spread like a wild fire.

    I'm a strong proponent of this theory, because I simply refuse to believe that Indo-European once was a one vowel language. All signs seem to point towards this, but the absolute uniqueness of such a system makes it hard to believe it did. I therefore believe that it's all vowel reduction (You might like to know that you made me change in opinion in this case ;) ).

    But like I said, 4000 years earlier than PIE doesn't seem necessary at all. Maybe 1000 years is already extremely lenient.

    Just as a thought, might the surviving *a's be due to Para-IE dialects. So of the same family, maybe variants that didn't have a vowel reduction losing it's contrastive a?

    Hmmmm.

    (Also, damn, Blogger no longer allows me to leave a link to my blog on my name.)

    Well then:

    My Blog - Visit it! (Or don't, I know some of you will anyway.)

    ReplyDelete
  2. Phoenix: "I don't think it's at all necessary to put this vowel reduction that far back in history. [...] But like I said, 4000 years earlier than PIE doesn't seem necessary at all. Maybe 1000 years is already extremely lenient."

    You haven't explained why you feel this way. However I can certainly explain why I feel that this Centralization event took place at a very early date. My reasoning involves the otherwise intractable nature of e/o ablaut.

    Certain ablaut patterns in IE are evidently more recent than others. It's easy to see that alternation between *e and nil (eg. *ʔés-ti vs. *ʔs-énti), for example, is merely the product of an earlier shifting stress accent and so this ablaut pattern is very intuitive. Then I would also say that some instances of e/o ablaut such as *pódm/*pedós are just a bit older (from late Mid IE *pát:am/*pat:ása). However the source of much of e/o ablaut is simply not transparent at all (eg. *wóide "she knows" vs *wéidos "appearance").

    This, I feel, is in itself an indication that we're dealing with an archaicism. Only by first reinterpreting e/o ablaut as earlier ə/a ablaut do we get closer to its transparency.

    To explain fully, I'd have to write another long article but for now I'll just sum up my current solution: PIE's e/o ablaut originates from early pre-IE vowel harmony which would have been an isogloss shared with pre-Altaic... circa 8000 BCE. Discuss.

    Phoenix: "[...] I simply refuse to believe that Indo-European once was a one vowel language."

    Good for you. As you can see, I don't buy it either but it's not just because such a vowel system is unlikely but also because the details of e/o ablaut can't be easily wished away like this. At least two vowels are required to explain the origin of ablaut in PIE because, as I state above, I'm suspecting the solution involves early vowel harmony.

    ReplyDelete
  3. I'm wondering if Pre-IE had a three-vowel system, namely /i a u/. Later on, /i/ caused palatalization of preceding consonants and /u/ caused labialization of them. This could have also happened to coda consonants. At that point, palatalization and labialization could be understood as consonantal features. The vowels would meet somewhere around [1] (in X-SAMPA notation), which could then lower to [@] and still be distinct from [a].

    We certainly see evidence of labialization from Pre-IE in the labiovelar series. However, a couple issues remain. First, do we see evidence of labialization anywhere else? There seem to be instances of labialized coronal consonants, but nothing in the bilabial category. Either the labialization was absorbed there or the bilabial stops were lenited. Another question is whether we can reliably distinguish between labialization caused by an earlier */u/ and labialization caused by syncope. With the velars, a distinction is made between */k_w/ and */kw/. It would seem to me that the former came from */ku/ and the latter from */k@w/.

    What's nice about the "centralization theory" is that it explains how */a/ did not undergo qualitative Ablaut. This is because the Ablaut vowel would be properly */@/, the *high* vowel in the two-tiered vertical vowel system. It also suggests that many instances of */a/ in IE are not the result of borrowing or laryngeal coloring; rather, they were simply inherited.

    - Rob

    ReplyDelete
  4. Follow-up:

    Another interesting thing about the "centralization theory" is its treatment of onsets. Zero phonemic onsets with high vowels pre-centralization would lead to labial or palatal onsets afterwards. That can help explain the abundance of */w-/ and */y-/ in IE.

    - Rob

    ReplyDelete
  5. Ooooh, I get it now. It took me a while of pondering to figure out what you mean exactly. I think I got it. You're saying, in a nutshell, that *a didn't participate in the vowel reduction and that it was only the more closed vowels like *u and *i that merged to *ə (schwa).

    My system is a little different since I start out with four vowels /a ə i u/. The schwa *ə first merges with *a and then the high vowels *i and *u merge to *ə. Only *u causes labialization and this labialization only operates on stops (and possibly sibilants) in the onset and coda of the first syllable of the word (i.e. the stressed syllable).

    I explain e/o ablaut in terms of subjective/objective differences in vowel harmony. Thus, after the vowels are reduced, we'd have 1ps objective *CəC-əm (1ps *-əm from *-im) versus 1ps subjective *CaC-ah (1ps *-ah from *-uh). This implies that the original vocalism of all IE verbs was lost to Centralization. If I recall, Bomhard had already remarked that the original vocalism of IE verbs was lost although his system is different than mine.

    However, there's a flaw I notice in your own point of view. Since the overwhelming majority of verbs contain ablauting *e/*o, it implies that there was an overabundance of verbs in *i and *u beforehand, and that makes no sense. Furthermore, between a choice of ablauting verbs and plain verbs, you'd think that the natural tendency would be for ablauting verbs to be in the minority in favour of regular verbs!

    Notice that in my system, all verbs began to ablaut because of vowel harmony and eventually the a-grade (later becoming o-grade) would become productive as a signal of stativeness. There's a natural morphological and semantic connection between stativeness, intransitive verbs and subjective conjugation worth exploring here.

    ReplyDelete
  6. Rob: "Zero phonemic onsets with high vowels pre-centralization would lead to labial or palatal onsets afterwards. That can help explain the abundance of */w-/ and */y-/ in IE."

    If you're claiming that there was an earlier contrast between, say, *(y)i- (> *yə-) and *ʔi- (> *ʔə) then I supposed I can accept that as a possibility, although I think rather that initial vowels *i- and *u- yielded *ʔə while semivowels were already in existence beforehand to supply the language with semivowels after the reduction of the vowel system.

    ReplyDelete
  7. My system is a little different since I start out with four vowels /a ə i u/. The schwa *ə first merges with *a and then the high vowels *i and *u merge to *ə. Only *u causes labialization and this labialization only operates on stops (and possibly sibilants) in the onset and coda of the first syllable of the word (i.e. the stressed syllable).

    I thought it might be of interest for you to learn that something very similar happened in Tocharian B (once again I never know too well for Tocharian A ;) ).

    Both *i, *e and *u merged into ä (which is /ə/).

    But instead of the *u leaving behind labialisation, the *i and *e left behind palatalisation. The *e palatalises all consonants. The *i fails to palatalise the PT *w and *kʷ. (Which is one of the things I love about Tocharian. An e that palatalises more consonants than i!)

    just for completeness I'll add the shifts of the other consonants:
    *h2e > *a > TochB a
    *o > TochB e.

    So as you may have noticed the short vowel system forms a three value central vertical vowel system. Not dissimilar to your proposed Indo-European vowel system, or the Caucasian languages.

    I'll probably write a little article about the Tocharian vowel system on my blog soon.

    My blog (I guess I'll make a Blogspot blog soon, and just give a link to vox, similar to what you did.)

    ReplyDelete
  8. You might want to check out John Colarusso's work. He's done extensive research in North-West Caucasian languages, and suspects a relation between PIE and NWC. At least, that's what I read in his "Nart Sagas of the Caucasus". I'm skeptical of the relationship, but I agree that there are definite areal similarities.

    ReplyDelete