dive (was Re: Sos-)

> > (At least the last three steps propose actual conditions. The
> > initial six-way split does not seem to be even falsifiable.)
>
> Assumptions made in linguistic reconstructions rarely are;

Soundlaws quite frequently are. A proposed soundlaw *k > t in English for example would be easily falsified by examples like "can", "wake", "king", "ask".

> the best strategy for evaluation is a secondary one for Popper and resembles Occam: most forms explained with fewest assumptions.

And "unconditional split" is a fairly big assumption.

> > > and get
> > >
> > > *sá:-/*sák-/*sáp-/*sank´-/*samp´-
> > > *sú:-/*súk-/*súp-/*sunk´-/*sump´-
> > >
> > > which would solve the 'Suomi' mystery

> > Several possible etymologies exist, there's no need to posit a
> > yet another one
>
> You don't really want me to comment on that, do you?

Why wouldn't I?
If we have X number of fairly plausible etymologies, adding one that seems less plausible is not progress.

If you can argue it to be more plausible than it appears at face value, then by all means go on.

> > (at least one involving all sorts of hypothetical forms).
>
> Reconstructions are hypothesis. I thought you knew?

Reconstructions within extant families are one layer of hypothesis. I'm referring to the substratal layer you're proposing underneath it, which does not appear to be a reconstruction in the comparativ sense.

> If you are asking for which forms it purports to explain

Yes, that's exactly it.

> look at 'sump' here:
> http://runeberg.org/svetym/0995.html

I don't see a need to assume any velar variants if that's all we are explaining.

> > > and on the connection to supposed PIE *pen- "swamp"
> > > http://tech.groups.yahoo.com/group/cybalist/message/63881
> > > Funky.
> >
> > _Bagno_ with voiced stops and a common Slavic distribution quite
> > clearly cannot be a loan from Finnic.
>
> Of course it can if the Slavic language family has a Finno-Ugric substrate.

Undemonstrated speculation.

> The alternative is that both loaned from an unrelated substrate, ie the ar-/ur- etc language.

False dichotomy. I'm going with the default alternativ of "unrelated".

> > > Instead a loan *paN- ? Cf.
> > >
> > > 1 pin´: (Sal. pinli), pl. pi`n´n´&^D (neu: sùomli, pl. -st)
> > > finne (finnländer);
> > > s. pin´-mo:, pi`n´n´&^mìez.
> >
> > Transparently a loan from _Finn_. This provides no new insight.
>
> No exactly transparent. The one thing which points in that direction

Perhaps I should've said "transparently one is a loan from the other".

> is the root -i- vs. Tacitus' -e- in Fenni, which makes it likely the word participated in Gmc *-en- > *-in-.

Another thing is that this word only occurs in Livonian.

> > BTW note that there's no a/u "alternation" in "salt" in Uralic.
> > *a is regularly reflected as Komi /o/ ~ Udmurt /u/, and the
> > latter under some conditions
>
> And they are?

I've only seen the change mention'd in literature in passing (and I've yet to read the latest treatise on Permic historical phonology).

> Okay, so there is a/u alternation in Komi and Udmurt, under some conditions.

Um no, I said there's a development u > ï in Udmurt in some words in some dialects (no alternation, no /a/, and not in Komi).

> > Connecting this to "salt" has semantical problems, /ï/; long *a:
> > regularly > *o: in Finnic.
>
> Don't try connecting them in Finnic then.

You managed to mess my quote somehow. That vowel part went with the previous section.

The semantic gap between "island" and "salt" does not seem any smaller in other languages.

> The a/u-alternation I am referring to is that of the ar-/ur- language; I suspect it arises from denasalization of a nasal vowel -aN- (cf. -aN- > -u- in Russian).

That has no relation to denasalization; it's a~: > o~: > o: > u: > u, plain old long vowel raising.

> The nasalization might have been 'sweated out' in auslaut as an -n# which was then rhotacized > -r#,

Is this a hypothetical or evidenced rhotacism?

> which means there must have been a word or morpheme boundary after the -ar-'s and -ur-'s Kuhn found in river names etc.

So you've now turn'd an apparent ar/ur alternation into a hypothetized aN/uN alternation. What did this accomplish, other than the addition of some extra assumptions?

John Vertical