Richard Wordingham wrote:

--- In Nostratica@yahoogroups.com, "H.M. Hubey" <hubeyh@...> wrote:
>
>
> Richard Wordingham wrote:
>
> > --- In Nostratica@yahoogroups.com, "H.M. Hubey" <hubeyh@...>
wrote:
> > > The *PIE sounds and the changes in the daughter languages have
> > > not been agreed upon.
> >
> > I think you'll find the basic correspondences are pretty well
agreed
> > upon.  The arguments are about origins (i.e. in pre-IE), details
and
> > phonetics.
>
>
> Yes. What if there is a different set of sounds for *PIE ?

This is what I had in mound when I mentioned arguments about
phonetics.  The most striking example is the 'glottalic theory' -
that what is normally reconstructed as, for example:

/t/ ~ /d/ ~ /dH/

was phonetically

[t] ~ [t'] ~ [d]

where the middle item in the series is ejective.

> > > There should be quantitative metics e.g. like R^2 for
> > > statistical fitting.  Among the factors
> > > 2. how many words are explicable with these rules
>
> What good does it do to have 13 rules to explain 13 words?

Well, if we have 4 segments per word, that's 52 segments explained.

I made no comment because I assumed you were considering how to
penalise ad hoc rules introduced to resolve a single word.

> > > 3. how many starred forms are needed
> >
> > I don't understand this factor.  Please expand on what you have in
> > mind.
>
> They posit PIE words from which attested ones are derived.
Furthermore,
> like the
> old elements table there are holes and they too have to be filled
in
> with starred forms.
> The more you have to create the worse your system.

PIE is not attested; it is inferred.  Therefore all PIE words should
be starred.  However, the starring is therefore redundant, and often
omitted.

Obviously they have to be starred; the question is how many?

> I can probably create a protolanguage for any 3 languages e.g.
Chinese,
> Latin
> and Swahili (Swadesh 100 list) if I was allowed to create any
number of
> protowords
> and as many sound change rules I wanted. Clearly, there are better
> systems and worse
> systems and we want to know which is which.

In this case, you will have protowords wherever the reconstruction
implies cognates, so my first thought is that it will have up to 100
starred forms.  Are you suggesting that "don't know" should be a
permissible 'reconstruction'?

It's a very good thought experiment, though.  Numbering the meanings
alphabetically by English gloss, one could make a reconstruction by
taking words no 1, 4, 7,... from Chinese, 2, 5, 8,... from Latin, and
3, 6, 9,... from Swahili.  No sound changes required!  That
reconstruction needs to get a very poor score.

That should get a poor score, and that is why we need some principled and
objective way of scoring things.

> > > Richard Wordingham wrote:
> >
> > > > Another example is the pair
> > > > *pah2ur- and *h2ngni- of PIE words
> > > > for 'fire'.

> Now, to create *PIE all the languages that are presumably IE were
> selected and
> then the protoforms constructed.

You know, I trust, that that wasn't how it was done, though the end
result is now probably the same.  Dempwolff's reconstruction of Proto-
Austronesian (PAN) (now regarded as merely a reconstruction of Proto-
Malayo-Polynesian) is probably a better example, but I am open to
correction.  Robert Blust has an interesting trick for finding new
PAN forms - he looks for possible forms with the same final syllable
as other, known roots and a related meaning.  It yields results! 
(These final syllables have been dubbed 'roots', but their status as
such is disputed.)

Again, the problem is about being able to compare the various alternatives
and reaching some objective conclusion. Even if there are several ways of
scoring, at least we would know that SystemA got score 99 on X-metric and
76 on Y-metric etc.

> To create Nostratic the same things have to
> be done. That means Turkic yan (to catch fire), yak (to set on
fire)
> cannot be
> ignored especially when it is obvious that this yak> ag is
possible, and
> that -ni
> is likely a suffix from Georgian/Kartvelian or one of the CAucasian
> languages
> e.g. the Abkhazians name for themselves is either Apswa or Apsni.
The latter
> may be the name for them given by Georgians. If there is a homeland
for
> fire-worship it is in the South Caucasus, where there were natural
fires
> because of surface oil and leaking gas. And these words show up in
that
> region.

*h1ngni- is what would emerge from studying accepted IE languages,
though *h1egni- would fit Latin ignis and Sanskrit agni-.  (Indeed,
Pokorny gives umlauting *h1egni- and *h1ogni-.)

That is not the main problem. The main problem is how much evidence from outside
IE should count in the reconstruction of Nostratic. One does not construct IE on the
basis of Icelandic and Sinhalese only. If a third language is taken into account some of
the postulated sound changes would have to be modified. If we add a 4th language
more changes are needed. Hence, if evidence from outside IE is considered for Nostratic
then IE might have to be changed.

> This is just a simple example of problems to be encountered. There
has to be
> an objective and principled way to judge the various proposals.

Richard.




To unsubscribe from this group, send an email to:
Nostratica-unsubscribe@yahoogroups.com



Your use of Yahoo! Groups is subject to the Yahoo! Terms of Service.

-- 
...
M. Hubey
...
hubeyh@...
http://www.csam.montclair.edu/~hubey