Re: Proto-Indo-European, glottalic theory and consonant inventories.

On Fri, 20 Jan 2006 03:33:44 +0100, Steven Williams <feurieaux@...>
wrote:

>Having reread Gamkrelidze and Ivanov's work for about
>the tenth time, I decided to skim the Wikipedia
>article
>(http://en.wikipedia.org/wiki/Glottalic_theory) on
>Proto-Indo-European, to get an idea of what the
>current state of the theory is.
>
>About midway through the article, I see what appears
>to be a more recent reconstruction of the PIE
>consonant system, which goes something like this:
>
>voiceless --> p - t - k - q - k_w
>ejective/glottalized --> (p') - t' - k' - q' - k_w'
>aspirated --> p_h - t_h - k_h - q_h - k_w_h
>
>The first thing I wondered was, who proposed this
>system, and when? It's interesting. For one, it posits
>a complete uvular system, something that's
>typologically a little marked (I mean, neither Quechua
>nor the Kartvelian languages typically show such a
>three-way contrast, right?).
>
>The article did mention that the uvular-velar contrast
>was probably allophonic, and corresponded to the
>traditional plain velar vs. palatalized velar
>contrast, so the table should be more like this, by
>that token:
>
>voiceless--> p - t - k - k_j - k_w
>ejective/glottalized --> (p') - t' - k' - k_j' - k_w'
>aspirated --> p_h - t_h - k_h - k_j_h - k_w_h
>
>Anyways, is there any actual support for this
>supposition, that PIE had an opposition between plain,
>ejective/glottalized, and aspirated stops (voicing's
>either non-existent or non-contrastive in all three
>series)?
>
>What are the 'holes' in the theory? Am I totally
>misinterpreting all this?
>
>(I know, Wikipedia's hardly a good source for such
>relatively obscure scholarly material as this, but
>until I figure out how to navigate my university's
>labyrinthine journal archives, I'm stuck with that)
The rationale for the glottalic theory is the typological unlikelihood of
the traditional inventory of stops reconstructed for IE, namely:

Voiceless: p t k^ k kW
Voiced: b d g^ g gW
Voiced Aspirated: bh dh g^h gh gWh

No language has ever been encountered that makes such a contrast among its
stop series.  Every language known with voiced aspirated stops also has
voiceless aspirated stops (the prime example being Sanskrit).  As a result,
the glottalic theory is an attempt to reconcile this typological
unlikelihood, with the plain voiced stops actually being voiceless
glottalized stops and the plain voiceless and voiced aspirated stops
actually being plain voiceless and plain voiced stops (respectively) with
aspirated allophones.  In such a way, then, IE roots that traditionally
have two voiced aspirated stops now have two voiced stops, one of which may
be allophonically (then phonemically) aspirated.  (IIRC, there isn't any
descendant language where both stops "remain" aspirated.)  Evidence in
favor of this view comes from differing "directions" of "aspiration
dissimilation": progressive (e.g. Latin) vs. regressive (e.g. Greek,
Sanskrit).  Only two IE groups that I know of, Germanic and Armenian,
underwent a process of regularizing aspirated allophones of the voiceless
stops (e.g. Grimm's Law in Germanic).  However, it is very possible that
the Celtic languages underwent such a process as well, albeit only with
*/p/; as that phoneme is lost in the attested Celtic languages, it is very
possible that its development was something like */p/ > */p_h/ > */f/ >
*/h/ > */0/.  There are many existing languages that have underwent similar
changes, e.g. Arabic (*/p/ > /f/), Japanese (*/p/ > /h/), and High German
(*/p/ > /pf/, */p/ > /f/).  (There's an Altaic language that also lost
*/p/, but I forget which one.  Mongolian?  Manchu?)

- Rob