Reconstruction (was: Grimm 's Law fact or myth: Gessman (1990))

--- In cybalist@yahoogroups.com, "fournet.arnaud" <fournet.arnaud@...>
wrote:

> From: "tgpedersen" <tgpedersen@...>

> >Obviously there are
> >the same methodological problems in reconstructing words as
> >reconstructing buildings or civilizations, but in each case what you
> >reconstruct is something which once existed, and no one is doubt of

that.

If only that were true. For counter-examples, see Pokorny. It's been
suggested that all roots not attested in three branches should be
discarded as probably being wrong.

Similarly, with Nostratic reconstructions, even if the hypothesis is
correct, a great many proposed correspondences are wrong.
Unfortunately, I think in many cases we will not be able to tell which
ones are wrong. Some of us have to learn to work with noisy data.

> One method could be to work with a one-to-one feature approach
> for example a matricial formula like P_K where P is any labial stop

and K

> any velar stop,
> be it voiced, voiceless or whatever.
> That method would sort out more words than the comparative words does,
> and in the case of some of your beloved substrates,
> we might easily identify words with the same phonemic pattern as the
> standard formulas.
> This is another approach.
> It's more quantic than newtonian.
> I suppose the orthodoxists will look upon this method with the same
> horrified gape as XIX century's physics would look upon quantics and
> relativity.

You've dropped your detection threshold - you'll always have a lot of
false cognates. You'll have to cope with the fact* that, say, 80% of
your correspondences are wrong.

*The error rate will have to be an estimate rather than a fact - not
something that unduly bothers statisticians.

Richard.