18 Nov 2008

PIE Uvulars: A revised solution of their origin

I've been frought with stress this past week due to something ridiculous that happened to me. I'll recount my ordeal in the very next post coming up. At any rate, after deep relection, I think I have an awesome way to rework my theory to account for the uvular phonemes in Proto-Indo-European (PIE) much better than I had before and with the least amount of added complexity. As usual, you may want to consult my continually updated document Diachrony of Pre-IE to get a grasp of what I'm suggesting. Listen to this latest crazy idea of mine.

I would like to now propose that distinct uvular phonemes had already existed by the end of Old IE (OIE) when unstressed vowels merged phonetically to schwa. They were, as I stated before, initially produced by allophonic differences dependent on the neighbouring vowel. Originally at the Proto-Indo-Aegean stage (before 7000 BCE), a velar sound (ie. any of *x, *xʷ, *k, *kʷ, *g̰, *g̰ʷ, *g, or *gʷ) neighbouring the low vowel *a acquired an allophone with a [+low] quality (ie. *x → [χ], *xʷ → [χʷ], *k → [q], *kʷ → [], *g̰ → [ɢ̰], *g̰ʷ → [ɢ̰ʷ], *g → [ɢ], or *gʷ → [ɢʷ]). I've already mentioned that the Mongolic language, Khalkha, exhibits the same alternation. There are also the examples of Even and Yakut that are both undergoing similar processes of phonemicization of uvulars as I describe for earlier stages of Pre-IE[1]. So when unstressed vowels merged in OIE, the nature of the uvularization automatically became obscured.

However to add to this idea, I also propose that Indo-Aegean's Decentralization of the inherited vowel system hadn't caused merger of former accented to *a just yet. Rather, the two vowels must have remained distinct for a while in OIE until phonemicization of uvulars was complete.

With these revisions come some interesting changes to my views concerning some important roots and their prehistoric etymologies. For example, the well-known PIE root for "dog", *ḱwon-, might then ultimately originate from Proto-Steppe *kə-huni "tamed canine" (not *ka-huni, as I believed before), thus becoming Indo-Aegean *kəxʷanə and then MIE *kaχʷána due to Penultimate Accent Shift (PAS). The vowel in that example, not being a low vowel, didn't uvularize the preceding word-initial velar stop to *q-, although the following laryngeal was uvularized by the second vowel. To explain another example, PIE *kreuh₂- "raw flesh", we must reconstruct MIE *qaréuxa- to account for it with a distinct uvular stop at the beginning to yield later PIE *k-. If this was a native term used in the earliest stages preceding PAS, then only *a may be prescribed in the first syllable in order to explain the later uvular, thus we should presume earlier *kárəuxə-.

This also has an impact on Proto-Semitic (PSem) loans that I identify in my online pdf. With the allowance of uvulars at this stage of cultural and linguistic contact between PIE and PSem, the interaction between the two will have to be revised slightly. For example, PSem participle *māšiʔu is now more understandably converted to MIE *mésɢ̰a- (> PIE *mesg- "to dip in water") with uvular stop *ɢ̰ because it would have been the closest approximation possible to a word-medial glottal stop for an MIE speaker. I maintain that word-medial glottal stops did not exist in the language at this stage.

I'll save my solutions concerning the possible geneses of the poorly reconstructed PIE particle *[ǵ/g](ʰ)[e/o] and the mystery verbal extension -g- for a later post.

[1] Fortescue, Language Relations Across Bering Strait (1998), p.72 (see link) explains that uvularization of velars neighbouring low and/or back vowels is quite linguistically natural.


  1. Over the past several days, I came up with about the same idea on my own. Given this, I also wonder if this means that *h1 was [x] and *h2 was [X] (X-SAMPA for a voiceless uvular fricative). This would point to an allophonic variation of an original **[x] before non-low and low vowels, respectively. As for *h3, it may have been labialized, although other labial(ized) phonemes do not seem to cause o-coloring (at least not consistently).

    What I really find interesting here is that those roots with "aberrant a-vocalism" may not be aberrant at all!

  2. Rob: "As for *h3, it may have been labialized, although other labial(ized) phonemes do not seem to cause o-coloring (at least not consistently)."

    True, and I've noticed this. However, my explanation is that, unlike other labialized phonemes in PIE, *h₃ was an aspirate and therefore particularly prone to erosion compared to the more acoustically salient labialized stops: *kʷ, *gʷ and *ghʷ. So one plausible path to eventual erosion in non-Anatolian languages would have been /hʷ/ > /ɦʷ/ (voicing) > /ʷ/ (nullification with residual labialization).

    So what I'm suggesting is what I like to call "disassociated labialization", or rather labialization remaining as residue a while after the phoneme to which its attached has eroded to null. This leaves the labialization to wander to a neighbouring phoneme, as if a lost puppy looking for a new home, such as the vowel *e, presumably colouring it at first to front-rounded /ø/ or mid-rounded /ɵ/ before completely merging with back-rounded *o. If I'm not mistaken, this initially allophonic labial colouring of *e by *h₃ had started before the dissolution of the PIE speech community.

  3. Also note that it's believed that laryngeals failed to colour long .

  4. If I'm not mistaken, this initially allophonic labial colouring of *e by *h₃ had started before the dissolution of the PIE speech community.

    We might not want to forget that in a good portion of the Indo-European languages *eh₂ and *eh₃ actually have the same reflexes. The 'lost puppy' effect of the labialisation really only seems to have had effect on Latin and Greek. (And without a doubt, but not speaking out of experience also Armenian).

    One might want to place the colouring of the laryngeals in a more wide Indo-European time, where dialects where already heavily divergent. And some places the labial aspect of *h₃ is lost completely, and behaves like *h₂ Which is as Rob points out somewhat to be expected since the other labial consonants don't colour either. While in other places this *h₃ clearly left a mark.

  5. Okay, so now uvularization depends not on PIE or even OIE, but Proto-Indo-Aegean vowel distinctions, which were long since reduced out of existence. The obvious question, then, is: what do you have to base those on? If it's purely on their effects on PIE consonants, you don't seem to be gaining anything but more complication over a reconstruction that has uvulars to begin with.

    In other words, what does the Aegean half of your Indo-Aegean reconstruction tell us, or is there even anything presentable yet?

    On another topic, I still think null would be a more natural substitution for a glottal stop in a loanword than a voiced dorsal stop of any kind.