Re: YACQ: Plausibility of a sound change

Jörg Rhiemeier scripsit:

>Tommaso R. Donnarumma <trd@...> writes:
>
> > The passage from glottidalized to voiced is, I think, rather simple:
> > if the speaker releases the glottal occlusion prematurely, then you
> > have a (partially) voiced consonant.  I believe that this kind of
> > anticipatory phaenomena is rather common...
>
>This change, however, goes through an intermediate plain voiceless stage
>(or so I think),
For what I know, this isn't correct.  If glottalized stops go to
voiced stops through anticipated release of the glottal occlusion,
there's no intermediate stage with voiceless stops, although there
can be an intermediate stage where the newly appeared voiced stops
are only partially voiced (i.e., there is a small delay between
the moment the oral occlusion occurs and the moment when the vocal
chords start vibrating). This is, however, irrelevant:  different
languages may  have different delays (or no delay at all) of voice
in voiced consonants.

>so the most likely result if glottalized stops go to
>voiced is that they take the plain voiceless stops with them:
*If* glottalized stops first merge with voiceless stops, there's
no other way.  But this is only one of the possibilities...

>glottalized     -> plain voiceless -> voiced
>plain voiceless -> plain voiceless -> voiced
>
>But this is what is intended to have happened in R. A., anyway.
>The aspirated stops then lost their aspiration because they were the
>only voiceless stops left.
Yes, this is a good alternate description for the evolution of
R. A...

> > What was wrong was the passage from ejective to aspirated.  I did
> > find a tricky explanation for it, but that's what it was:  a tricky,
> > and rather unbelievable, explanation.
>
>Hmmm, this probably requires some indirect mechanism (again via plain
>voiceless).  But we have to get the old plain voiceless stops out of the
>way, lest we lose all contrasts in Kluna, and restore them afterwards.
>Perhaps the following happened:
>
>plain voiceless -> voiced          -> plain voiceless
>glottalized     -> plain voiceless -> aspirated
>aspirated       -> aspirated       -> aspirated
>
>I.e., the plain voiceless stops become voiced, and then the glottalized
>stops plain voiceless in a "chain shift".  Then the plain voiceless
>stops merge with the aspirated series, and finally, the voiced stops
>move back to plain voiceless.
>
>Does this make sense?
This surely makes sense to me.  There is just one possible problem,
though, but that's my fault:  when I said I wanted to track Kluna and
the Ripe Apple Language back to a common proto-language, I really meant
a reconstructed, unattested proto-language.

Now, the kind of interpretation you propose, although plausible when
working forward from the proto-language, is not cogent when working
backwards from the attested languages.  In fact, your proposed intermediate
layer (voiceless/voiced/aspirated) is sufficient to explain the Ripe
Apple Language too -- and, indeed, this is exactly what was suggested by
Nik Taylor.  I fear that, in absence of data from a third language (which,
for the time being, I don't plan to add), there would be no need to
postulate a series of ejectives in the proto-language...

Anyway, your suggestion is good food for thought.  I could also decide to
initially work with the voiceless/voiced/aspirated system, and then come
back to the plain/glottalized/aspirated system when I'm ready to add a
third language to this family (in this case, Kluna and the Ripe Apple
Language would form a subtree together)...

I'll keep you posted as to my decisions.  In the meanwhile, many thanks
for your input!

Happy conlanging,

Tommaso.