Re: swallow vs. nighingale, SWALLOW

--- In cybalist@yahoogroups.com, "Grzegorz Jagodzinski"
<grzegorj2000@...> wrote:

> Instead of believing in non-existing rules (prove if I am wrong),

I'd rather

> believe in "exceptional" rules - in out example, that -l- can yield -r-
> WITHOUT a rule. If finding rules had been so simple, nobody would

not have

> denied that Greek and Latin words for swallow or leech are cognates.

But

> some deny!

<snip>

> I do not believe Pokorny, Beekes and other Neogrammarians at all -

but I

> understand that if anybody believe in unexceptional rules, this one

MUST

> deny the cognacy because no rules can be found. Instead, I prefer

believe

> that some rules function only partially, and some phonetic changes

are not

> caused by any rules at all. And my believe is in high consistency with
> facts.

But we must remember that admitting 'exceptional' rules reduces the
quality of an explanation, as does introducing ad hoc rules. There
ought to be some probabilistic formulation of this, but the evaluation
would be based on Bayesian statistics and the a priori distributions
are complex. (I hope there is a way of measuring ad hocness.)

For example, false cognates do arise - arguably Pokorny is full of them!

Lacking evaluation techniques - and I think the problem is in
correctly deriving the distributions, as linguistics seems to hold a
great fascination for mathematicians - we must resort to rules of
thumb, such as a neogrammarian explanation being significantly better
when all other things are equal.

Richard.