Re: Deriving vowel harmony diachronically (was Re: Can realism be retro-fitted?)

On Jan 21, 2007, at 6:23 AM, Jörg Rhiemeier wrote:

> Hallo!
>
> On Sat, 20 Jan 2007 21:06:43 -0600, Eric Christopherson wrote:
>
>> How did you go about deriving vowel harmony? (Jörg, same question!) I
>> have some ideas about it, but I'd like to see how others go about it.
>
> OK.  Pre-Proto-Albic (Proto-Indo-Albic, or whatever you may want to
> call it) had only three vowels: *a, *i and *u.  What happened is that
> the vowel features [+open], [+front] and [+round] became
> autosegmental,
> i.e. they bound to morphemes rather than to segments.
[snip rest of description of Albic autosegmentality & harmony]

Your system seems really elegant to me. The only question I have is
how the vowel features become autosegmental in the first place. I
suppose that could be due to analogy, or maybe long-range vowel
assimilation (umlaut).

> Now to your scheme.
[snip]
>> - Because of the relative abundance of -uCu and -oCu (and possibly -
>> QCu) forms, coupled with the relative scarcity of -uCi and -oCi forms
>> (and nonexistence of -QCi forms), stems with back vowels and final
>> consonants *other* than /p/ or /w/ will analogically adopt the -u
>> ending.
>
> I'm sorry but I can't follow.  Are /p/ and /w/ so overambundant that
> they occur more frequently than all other consonants combined?
> I'd expect the -oCi and -uCi forms to be more frequent than the
> -oCu and -uCu forms.  Or does the abundance of /i\/ and /@/ have
> to do with this?
No, /p/ and /w/ are not that abundant. That's one potential problem.
(Actually I suppose I could create a bunch of new roots ending in /p/
and /w/ ;) )

Nor are original /i\/ and /@/ that abundant. But within the domain
of /p/ and /w/-final roots, the existence of original back rounded
vowels together with original central vowels which became back
rounded means that the /opu/ and /upu/ types are more common (again,
within the domain of /p/ and /w/-final roots) than /epu/ and /ipu/.
So I might need to backtrack a little bit, and use the *relative*
abundance of the /opu/ and /upu/ types as something to build from.

On the other hand, my limited understanding of analogy is that it can
generalize from even a relatively rare pattern. I think an
overabundance of exemplary forms does help, but isn't strictly
necessary. Still, I was using the supposed abundance of /p/ and /w/
forms as justification, which perhaps I should not do. I will think
about this some more.

Thanks for your comments!