What we have in pausal forms are two separate issues:
1. Why there is elongation at all (or restoration to the original
vowel, in your scenario)?
2. Why the epenthetic/ elongated (or restored) vowel varies?
You offered an explanation to the second point. I doubt it mainly
because of historical improbability of such restoration and the lack
of Secunda evidence. In your shame'u example, ayin was lost by the
time of Masoretes, but could influence them choosing e over a.
My question, though, relates to the first point: whether the choice of
vowel is etymological or phonetical, why there is elongation at all?
And that, I venture, can be explained by uptalking (high rising
terminal). From the anecdotal evidence, this speech defect is
relatively common among Jews. Perhaps not incidentally, it surfaced in
the US first in Los Angeles among middle class youngsters, where
Jewish presence was high.
Tendency to uptalk liturgical text would be only natural, since
cantors may feel shy when narrating the divine text. And uptalk is
commonly blamed on speaker's insecurity.
If the Masoretes thought to counter the uptalk, they had to relatively
reduce the final vowel. Since that would create schwa in many cases,
they opted for indirect relative reduction, increasing the previous
vowel instead.
Vadim Cherny
P.S. And sorry for mishandling email header in the previous message.
---
Randall Buth wrote:
> Well, there are lots of cases where the Masoretic vowels preserve
stuff
from earlier stages of the languages. As for the Secunda, it is 'a'
witness, but not always the most reliable for the best preservation.
(for that matter, Greek had lost phonemic length by that time.)
As for your a+u = o hypothesis, you
left out the `ayin, so there is no expectation for reduction.
And note some pharyngeal sister-pausals:
laqaHu 'they took'
and nasa`u 'they travelled'.
Neither one has an etymological
'i/e' vowel, and both have the expected 'a' forms.
shame`u 'they heard', on the other hand, reflects its 'stative'
'i/e' etymology. Chalk one up for the Masoretes.
A hypothesis needs to cover wide data.