Miguel:
> We have to distinguish between several possibilities:
> (1) *-we-/*-ye- medially in a root or suffix
> (2) *w-e/*y-e across a morpheme boundary
> (3) root-initial *we-/*ye-
> (4) suffix-initial *-we/*-ye
>
> In cases 1/2, the normal outcome is full-grade *yé/*wé, but
> usually with variants *í/*ú. In case 3, the outcome is
> full-grade *wé/*yé, in case 4 the outcome is *í/*ú.
Hmm, this theory is immediately complex and the above changes
are absolutely unintuitive, the likes I've never seen before.
Congratulations, you win the prestigious SayWhat?! Award for
"Most Hard to Grasp Theoretical Construct".
> This also immediately explains the t/s alternation: an
> original stressed *ú labialized a following dental, while
> unstressed *u did not.
I thought we could one day agree (silly I know) that *t/*s
alternation is merely the product of an exceedingly early
change of final *-t to *-s (before the case endings that
we see later were resolved). This places the change in a
more analytic stage of early IndoTyrrhenian and explains
why 2ps *-s is sibilantized (expected *-t mirroring *tu)
while 2pp *-te is unaffected. Clearly because *-t > *-s.
So we really have:
eITyr ITyr proper
nom. *-wa-t-sa > *-wat-sa > *-wat-sa
acc. *-wa-t-am > *-watam > *-wat-am
gen. *-wa-t se > *-was se > *-was-ase
These are forms with a stative ending *-et inherited
from ProtoSteppe *-it. As we can see, *s is found in oblique
cases suggesting that the change of *-t > *-s occurred
_while_ cases were still being postposed as particles.
Thus *-t is found in all cases except the only existing case
endings at the time which operated on the subject and object.
Simple.
This above idea predicts that there should be both inanimate
stems with just *-es AND animate stems with alternating *-t/*-s-.
Lo and behold.
> The accent _had to be_ on the vowel before the *t.
Nonsense. Your speech can only be two minutes. You've exceeded
your time limit so just accept the award and move on. There
are many other people to honour today.
> *swésor- (obl. *swésr-) is not a typical example. The word
> is acrostatic for some reason
The reason is that it ends in *-sor-. Stems in *-or-, *-tor-,
*-on-, *-oi-, etc.... basically any stem with an ending of the
form *-(C)V(C)-. The theme here is, ironically enough, the
thematic vowel (the V in the pattern). Even *nepot- was sucked
into this paradigmatic pattern although **-ot- isn't even a
suffix and **nep- doesn't really exist in IE.
> Perhaps it's because the word is a compound of *swé- "own",
> and *ser-/*sor- "woman",
I think he's got it! A round of drinks for everyone! :)
> I'm OK with that (since the lengthening took place _before_
> zero grade, and plural *-s cannot be separated from the
> nom.pl. ending *-es, it may simply be that there was still a
> pre-zero-grade vowel *-Vs in the plural affix, blocking
> Szemerényi lengthening,
Not blocking. Simply that the plural suffix *-es always had a
syllable and was unreduced when Szemerenyi Lengthening applied.
Normally, *-es becomes *-&s according to the Syncope rule.
Remember, *a > ZERO and *e > *&, right? However, the plural was
too much at risk of merger with the nominative singular which
was now *-s by Clipping (and sometimes it appeared as homophonous
*-&-s in thematic stems!) and so *-es for good reason needed to
remain to contrast properly with singular *-s.
However, not so with resultant accusative plural *-m-&s which
was not at risk of any merger. The dropping of *& here to
produce expected *-m-s was by analogy with the zerograde singular
*-m which did not contain a vowel. Lengthening also did not
apply because it could not affect syllabic *m -- There is no
such thing as **m: at any stage of IE.
Again, simple. It's just not true that any *s causes lengthening.
This rule isn't dependent on phonetics. It's dependent on
morphophonology.
= gLeN