Three-Way Contrast of Secondary Articulations in PIE

[David L. White]

> On 17 Mar 2001, David L. White wrote (quoting my previous message):

>>> The analysis of the obstruents as occurring in three series, to wit, plain,
>>> palatalized, and labialized, is unremarkable.

>> What is another example?

> There are several NW Caucasian languages for which this is the parsimonious
> analysis of the obstruent system; besides Kabardian/Adyghe/Circassian, I know
> of Abaza/Abkhaz and Ubykh.

        I will have to investigate.  Abstract parsimony can sometimes lead
to less than real "solutions".

>>> Presumably, your confusion stems from my comments on Kuipers' re-analysis
>>> of "plain":

>> Practically speaking, "plain" would have to be "[a]-colored", to stay out of
>> the way of consonants that were [i]-colored or [u]-colored.  Otherwise, with
>> normal coarticulation, "plain" velars (to take the most obvious case) would
>> develop de facto [i]-coloring near /i/.  Of course, this would not apply to
>> a language without /i/, but Old Irish does have /i/.

> The examples have all been drawn from languages which lack /i/ and /u/, to
> support an analysis of PIE as lacking /i/ and /u/.  However, the re-analysis
> of plain (that is, non-palatalized and non-labialized) obstruents as being
> somehow [a]-colored, in order to allow an analysis of the vowel system as
> lacking any full-specified vowel at all, quite simply violates phonological
> universals.

        It's the "in order to" part that is the problem.  I agree that
analyzing a language as having no vowels (in part by leaping from "one
vowel" to "no vowel") is seriously problematic.

> (This objection leaves aside the simple fact that the analysis missed the
> contrast between /a/ and /I-/, so that reducing the "one-vowel" system in
> this way was already wrong.)

> I was unaware that we were discussing Old Irish--the Subject: header
> refers to three-way contrasts in PIE.

        Whether Old Irish had a three-way contrast is an issue.  Some say it
did not, and I have always wondered whether they were right.
My point is that to speak of "plain" consonants existing in such a system is
in a sense non-sensical, because "plain" in effect means "as in a language
without secondary articulations", which in turn means "as in a language with
ordinary co-articulation", but a language with secondary articulations could
not possibly have ordinary coarticulations, since overriding ordinary
coarticlation for phonemic effect is what languages with secondary
articulations do.
Such languages can, for example, produce back velars before front vowels,
and front velars before back vowels.  The very term "plain", as applied to
velars, woud in effect mean "tending to be backed before back vowels and
fronted before front vowels".  In a language with secondary articulations,
this really could not happen.  So where one velar series was "[i]-colored",
and another was "[u]- colored", a third would almost have to be something
like "[yeri]-colored" or "[y]-colored" (IPA value) or (in terms of F2)
"[a]-colored",  in order to stay out of the way of the other two.  But such
colorations are not, to my knowledge, known.  So I think regardless of other
errors, Kuiper's instinct that a 3rd quality would have to be "[a]-colored"
rather than "plain" was fundamentally correct.   I hope all that makes
sense.

>> If part of the claim is that [a]-coloring does not exist, I agree, but
>> assert that it would have to for the sound-system commonly posited for Old
>> Irish to be viable.

> I'm afraid that my Old Irish background is so weak as to be non-existent,
> so I cannot evaluate this claim.  How do you come to this conclusion?

        As above.  What is "plain" /k/ in a sequence /ki/ supposed to have
been that would be different from palatalized /k/ in the same sequence?
Likewise what is "plain" /k/ supposed to have been in the sequence /ku/ that
would be different from labio-velarized /k/ in the same sequence? I suppose
we could say that the palatalized and labio-velarized versions would come
out more like [kyi] and [kwu] respectively, but according to my
understanding secondary articulations are not ordinarily so strong, and to
regard them as equivalent to following [y] or [w], though it may be a
convenient first approximation for learners, is a mistake.  I suppose it's
possible though, and if I have to "stand corrected" I will, not for the
first (or last) time.  But I am distressed at how casually "plain"
consonants are posited for such a system.  The phonetics would not be that
un-problematic.

Dr. David L. White