09-08-03 01:11, H.M. Hubey wrote:

>
>
> Piotr Gasiorowski wrote:
>> Actually, the Pashto development was {PIr. -t-, -þ-, d} > ð > l, as in
>> the following (safely reconstructed) words:
>>
>> (PIE *p&2ter- >) *pitar- > *piðar- > pla:r


> I notice that 2 of 3 are reconstructed, thus
> *pithar > pitar
> *pithar > pla:r

I confess I can't see what you mean.


>>
>> (PIE *kWetwores >) *caþva:r- > *caða:r- > calor
>
>
> Again, this is what they call "tavsanin suyunun suyu" in
> Turkish. (Cream of cream of rabbit soup).

No. We get from *kWetwores to <calor> or from *dek^m. to <las> by
applying regular sound changes -- not ones reconstructed ad hoc for
these particular words but ones that have been established independently
and are confirmed by numerous examples. It isn't much different from
physics: the observation of numerous stones falling allows you to
predict how any other stone will fall under similar conditions.

The PIE reconstructions themselves were not arrived at with the help of
Pashto evidence. People who first proposed them knew precious little
about Pashto. To reconstruct *kWetwores and *dek^m. it's enough to
compare only some of the IE languages. The fact that Pashto fits into
the picture only confirms that it's also an IE language.

> If the reconstructions are going to be taken as rules for more
> reconstructions what purpose
> would it serve to find "general" principled rules.

Rules are derived from the comparison of related forms. By examining the
data we discover recurrent systematic correspondences and draw
conclusions from them. The circularity involved is not of the vicious
kind. If further forms confirm our rules, that's all the better; it
simply means that we are on the right track. If they don't, we have to
find out what's wrong and either abandon our earlier analysis or find a
plausible explanation for the irregularities. Contrary to what you seem
to suggest from time to time, linguists often choose the former option.

Reconstruction is in fact not something done for its own sake but a kind
of controlling mechanism. In order to prove that two forms are cognate,
you must be able to reconstruct a protoform and a pair of derivations
consistent with what has been established so far. A comparative analysis
of related languages is _required_ to produce a logically consistent and
linguistically natural reconstruction. This simply guarantees that the
whole construction hangs together.

Piotr