5 Jan 2008

The origin of the Indo-European uvular stop (traditionally the "plain, non-palatalized stop")

This is just a continuation of my jovial traipsing through modern reinterpretations of Indo-European phonology (see Reinterpreting the Proto-Indo-European velar series) where I explain that the traditional reconstruction of a palatal series violates the principle of markedness and that it's high time we modernize the theory for the 21st century. So the following will proceed by just assuming that the "non-palatal stop" series are uvulars which are aligned with *h₂, and that the "palatal" series (*ḱ, , *ǵʰ) is the unmarked "plain stop" series (*k, *g and *gh). This makeover has a bucketload of implications that, in tandem with Glottalic Theory (or rather my version using unaspirated, semi-voiced stops rather than ejectives) and proposals by Jasanoff concerning a new PIE grammar, radically but necessarily departs from the old look and feel of Proto-Indo-European (PIE). Based on a modification of Glottalic Theory, for example, we can push our reinterpretation further so that the traditional stops *ḱ, , *ǵʰ, *k, *g and *gʰ are rewritten as *k, *k:, *g, *q, *q: and , respectively.

As I suggested before, in this new way of thinking, the presence of uvularity itself is the cause of "vowel colouring". So both uvular stops (the original "non-palatal series" of *k, *g and *gʰ) and *h₂ (a uvular fricative[1]) colour *e to *a in the same fashion. Now, people who are versed in PIE linguistics may think that I'm sniffing glue by seemingly not accounting for words like *ker- "to cut". However, Jasanoff reinterprets thematic verbs like *bʰer- as originally athematic roots (thus *bʰēr-) which originated from subjunctive forms in *-e-. This accounts, for example, for the post-IE aorist in *bʰēr-s-, of which the Anatolian branch shows little trace. Likewise then, we might posit that *ker- is underlyingly an athematic root *qēr- (both with an update of plain *k to uvular *q, and athematization of this root à la Jasanoff). Additionally, it has already been remarked by others that vowel colouring does not affect long vowels in PIE, only short vowels[2]. We can extend this same rule to uvular stops as well. So while *qēr- would be pronounced as /qɛ:r-/, its reduced form *qer- however would be pronounced /qær-/ with colouring. We would also need to reinterpret traditional *kap- "to grasp, to seize" as *qep- with the same vowel colouring, which illustrates how true phonemic *a may be confused with coloured *e /æ/ in these instances.

In the traditional model, consonant clusters oddly prefer palatal *k (in opposition to markedness). However, we can see that the reinterpretation that I'm promoting has the advantage of doing away with palatalization in all these instances. There would still be some clusters with uvular *q as in *qreu- "raw; flesh" (traditional *kreu-) however the new model finally explains the distribution. As I refer to above, I would prefer a modified Glottalic Theory whereby it is both acknowledged that a) some stage of Pre-IE must have contained ejectives (to explain away PIE's curious phonotactic constraint against two voiced stops in the same syllable as a rule forbidding two ejectives in the same syllable) and b) PIE itself did not have ejectives but only unaspirated, semi-voiced "fortis" stops (the traditional stops *d and *g), thereby explaining the complete absence of inherited ejectives in its daughter languages while retaining the explanatory power of Glottalic Theory.

Now, that this is all out of the way: What is the cause of this uvularization in the first place? Nostraticists, working with the flawed palatalized model of yore, were in effect sent down a wild goose chase for a very long time. Afterall, the onus was up to them to explain the source of an effectively non-existent palatalization in Pre-IE. So a new model would shake things up, not only for PIE but for Nostratic as well. So far, I've stuck with my original explanation for uvulars. In my theory, uvulars became phonemicized and distinct from plain stops once Syncope (i.e. the loss of most unstressed vowels) kicked in. Before this time, *k/*k:/*g and *q/*q:/ were merely allophonic variations of plain *k/*k:/*g which alternated depending on the neighbouring tautosyllabic vowel. In my "Mid IE stage", a low vowel *a lowered *k to /q/ by assimilation while *k remained /k/ neighbouring *e. Since there are only two vowels, *a and *e, in my theory, it's all pretty straight-forward. I date the few roots with uvulars in onset clusters as post-Syncope (that is, dated to the Late IE Period). So *qreu- can be etymologically understood as a Late IE deriviative of *qēr- "to cut" that was coined after *k and *q became phonemically distinct.

I'll rant on some more about this crap later. There's a heckuvalot more to talk about.

NOTES
[1] Read Fox, Linguistic Reconstruction: An Introduction to Theory and Method (1995), p.179 (see link here): "The fact that *h₂ causes a-colouring would suggest that it is pronounced far back in the mouth - perhaps a uvular or pharyngeal sound [...]". Also Campbell, Historical Linguistics: An Introduction (2004), p.43: "For example, vowels are often lowered before uvular and pharyngeal consonants [...]" (see link).
[2] Note Ringe, From Proto-Indo-European to Proto-Germanic (2006), p.14 (see link): "But the laryngeal had no effect on the *o of *h₂ḱh₂owsiéti '(s)he's sharp-eared', so far as we can tell, nor on the of *ḗh₂gʷʰti '(s)he's drinking' [...]". (Personally, if I was ruler of the linguistic universe, I would rewrite *h₂ḱh₂owsiéti as *ḥkḥousyéti and *ḗh₂gʷʰti as *(ʔ)ḗḥgʷti.)

UPDATES
(Jan 05 2008) Egad, I stated "PIE itself did not have ejectives but only aspirated, semi-voiced 'fortis' stops" and I of course meant unaspirated fortis stops. Sorry, this topic is complex enough without me confusing everyone. I've corrected this error above.

5 comments:

  1. Nice blog post, Glen! I do have a question, though. Why does Jasanoff reconstruct long vowels in the thematic conjugation's ancestors (e.g. *bhe:r- instead of *bher-)?

    - Rob

    ReplyDelete
  2. In a nutshell, from what I gather, it has to do with explaining away Narten presents, sigmatic aorists and thematic verbs all at once. So, on the pattern of 3ps. *stēuti "they praise" vs 3pl. *stewn̥ti and based on s-aorists which have lengthened grade (e.g. *wēǵʰ-s-), we can get rid of thematic verbs such as traditional *bʰer-e- by reanalysing them as originally subjunctives rather than presents. The true presentive forms would have had lengthened grade and so Narten presents like *stēuti are presumably leftovers from the days of yore. Also read here where I've highlighted (Anttila, Greek and Indo-European Etymology in Action: Proto-Indo-European *aǵ- (2000), p.129).

    The story goes that subjunctives, having an inherent 'future' meaning were easily repurposed for conveying the present indicative (read here). Over time, the subjunctives forms would come to replace most of the original Narten presents, leaving some residue left over, and I've written already about how this change of subjunctive to indicative could come about in terms that English speakers might wrap their noodles around better. Essentially it's a change of semantics from "I would go" (future potential) to "I will go" (future certainty) to "I am going" (present certainty).

    ReplyDelete
  3. Awesome, I feel like I'm actually coming closer to understanding what you're trying to say.

    One question though. You explain the Uvulars as secondary developments of the velar series (through lowering). Wouldn't it be odd, in a system to have
    Labials stops, Dental stops, Velar stops, Labio-velar stops and then... a uvular fricative?

    It could of course be a velar fricative followed by your Pre-IE *a. But we'd expect to see velar fricatives around too then right?

    *h1 is without much possible arguing a glottal stop. and *h3 is quite clearly something labial, maybe a labio-velar fricative. For the completeness of the systems we would really like *h2 to be a Velar fricative right?

    /s x xw/ Only a missing labial fricative to make a perfectly symmetrical system.

    Hmm. Okay, I have no clue if this is any longer a question. But you undoubtedly see what I'm getting at. The main question I guess is: "What would you like to do with *h2, and the seemingly unnaturalness of a uvular fricative within a labial-dental-velar-labio-velar system?"

    ReplyDelete
  4. Phoenix: "One question though. You explain the Uvulars as secondary developments of the velar series (through lowering). Wouldn't it be odd, in a system to have Labials stops, Dental stops, Velar stops, Labio-velar stops and then... a uvular fricative?"

    Yes, it probably would, but I don't think that way. Prior to being a uvular fricative, I believe that *h2 was simply a velar fricative /x/, aligned then with /k/, not /q/. I believe that uvularity as a distinctive phonemic feature was acquired after Syncope and so this is when *h2 was pushed back.

    It's not strange for velar /x/ to merge to uvular /χ/, is it? In this way, it probably gave *h1 motivation to weaken to /h/ in mediofinal positions (note the vocalisation of syllabic laryngeals which is rather difficult for a glottal stop but easy for /h/ to do).

    I admit, what I'm suggesting about the evolution of laryngeals is not an easy path from A to B. It involves some changes that probably only make strong sense once I draw a diagram.

    Phoenix: "*h1 is without much possible arguing a glottal stop. and *h3 is quite clearly something labial, maybe a labio-velar fricative. For the completeness of the systems we would really like *h2 to be a Velar fricative right?"

    Man, I like your questions! In PIE proper, I think that *h1 = initial /ʔ/ and mediofinal /h/; *h2 = uvular /χ/; and *h3 = /hʷ/ but later voiced /ɦʷ/ in "PIE 2" (hence Sanskrit pibati < *pipɦʷeti).

    However, in Mid IE (MIE) before Syncope, I propose thus far that *h1 was purely a glottal stop in all positions; that *h2 was /h/ ~ /x/; and that *h3 was /hʷ/ ~ /xʷ/.

    So... MIE *h2 aligns with *k (/k/ ~ /q/); MIE *h3 aligns with labialized *kʷ (/kʷ/ ~ /qʷ/). In Late IE and PIE itself however, *h1 is aligned with velar *k; *h2 with uvular *q; and *h3 with labialized velar *kʷ. The relationship between laryngeals and dorsals change over a millenium.

    In other words, for fricatives we have /s sʷ x xʷ/ in MIE. In PIE we have /s h χ hʷ/, with no velar fricative, only uvular. Note that French has /s ʁ/ and no /h/ or /x/ whatsoever. So you should really take it up with the Parisians :)

    ReplyDelete
  5. In other words, for fricatives we have /s sʷ x xʷ/ in MIE. In PIE we have /s h χ hʷ/, with no velar fricative, only uvular. Note that French has /s ʁ/ and no /h/ or /x/ whatsoever. So you should really take it up with the Parisians :)

    Actually it's also reminiscent of Dutch. I'm not sure of how aware you are of Dutch dialectology. But standard we tend to describe our g and ch as being /x/. But this is a highly idealised representation of the truth. In fact in my, and many north-western dialects (which is the prevalent dialect on television and other media) /χ/ is a lot more common. Here it aligns with the regional r which, like French is pronounced /ʀ/, and in some speeches it's even /ʁ/ (and due to word final-devoicing we find some dialects where word final r and g are not distinguishable :D)

    It's interesting how, the /x/ came to align itself with a resonant, rather than the velar stop (and it's also odd that our /g/ at some point shifted to a velar voiced stop and later to a voiceless uvular stop of course).

    But if you realise that this is a natural development, we can be sure to say that a *h2 that served as the corresponding fricative for both k~q due to allophonic distribution, it's indeed not that weird that it got the fixed value *h2 once the q phonemisised.

    ReplyDelete