>> TET -> /o/ (foveo, toga)
>> TEWT -> Z (s'ubháyate, fuga)
>> TERH -> /o/ (goné:, janáyati [laryngeal not dropped in the
>> causative?], cennan)
>> STERT -> /o/ (spondeo, spondé:)
>> SRET -> Z (hr.zaya-, sleg)
>> TWEH -> Z (kuéo:, phué:)
Jens:
> The parameter by which loss of the infixed element is governed is
> not easy to make out.
I see a pattern clear as day.
Simple revelation: An IE speaker was not a grammaticist. They spoke
their language as they perceived it, not as it was deemed correct
as it is so rigidly defined by some of the more anal-retentive
linguists. So, when we look at *genh- for example, whether the
analysis is "correct" or "incorrect" in the sense of the actual
etymology of this root/stem, we must analyse it as *gen-h- for the
purposes of grasping the pattern above.
In other words, *genh- is treated like a derivative of **gen-.
Likewise, *spend- is treated as an n-infixed derivative of **sped-.
However, a root with a form like **gHweh- would not be able to
be analysed as **gHwe-h- quite obviously since it violates the
CVC-root structure. Rather **gHw-eh- would be more appropriate
and we'd expect **gHuh-eye- as the causative. Indeed.
So what's the rule? It looks to me like the following:
1. Judge whether the vowel-grades of the perceived root
and extensions, if any
2. If the root appears zero-graded, make the extension zero too
3. If the root appears full, give it o-grade unless the coda
of the root is *w, in which case give it a *u-grade just
to throw off Jens and Miguel
There. So now everything makes sense once we open ourselves up to
the PIE mind and divide up the verb stems into two distinct patterns
of _perceived_ root-and-extension-pairs.
Pattern 1: CVR-(C-) -> CoR-(C-) (sometimes *Cu-(C-))
TET -> TET- -> /o/
TERH -> TER-H- -> /o/
STERT -> STER-T- -> /o/
TEWT -> TEW-T- -> Z
Pattern 2: CR-VC- -> CR-C-
SRET -> SR-ET- -> Z
TWEH -> TW-EH- -> Z
The only question remains as to why expected *ou becomes *u but
we know this to be a recent sound change in the latter half of
Late IE.
= gLeN