Hallo!

On Tue, 30 Oct 2007 16:14:47 +0100, Henrik Theiling wrote:

> Hi!
> 
> Jörg Rhiemeier writes:
> 
> [...]
> 
> > So why does the GMP not account for these "effects"?
> 
> Aha!
> 
> The GMP currently cannot distinguish between morphological endings and
> normal stem ends, because I was too lazy to program that.

The GMP *should not* distinguish between morphological endings and
normal stem ends (where they are word-final), because real sound
changes *do not make such distinctions*.  You are trying to simulate
something that *just doesn't happen* in natlangs.  If a final -m
goes away, for instance, it does so no matter what kind of morpheme
it is part of.

> To do it properly, i would have to introduce some kind of morpheme
> separator to be able to state rules like: the -t in endings drops
> (e.g.  in the 3rd person singular) and the vowel in endings becomes
> [log in to unmask]  I haven't done that yet because it's a lot of work and there's a
> trick around it that works just the same.

No, you don't need "some kind of morpheme separator", because the
sound changes don't care the least of the morphological structure
of the word.  Invoking a "morpheme separator" would be *unnatural*.

> That vowel reduction is currently implemented in the GMP by simply
> shifting the last syllable's vowel to -e.  This means that there are
> restrictions on the use of the GMP: the Perl script only works as
> expected if the last syllable is indeed an ending.  E.g. it cannot
> distinguish whether -uum should become -u-@m or -u:m.

As I have said above - sound changes don't make a difference there.

> You said yourself: there are two things to account for: sound shifts,
> and morphological deterioration.  I do the latter before shifting,
> carefully selecting something that produces a result identical to what
> would be produced if the deterioration was handled right within the
> GMP somewhere.

That way you have it backwards.  Morphological deterioration in
Romance languages (as well as in English, and in any other language
that shifted from synthetic to analytic grammar) is *caused* by
sound changes: it's sound changes that erode suffixes to the point
of indistinguishability.  They start with fully inflected word forms,
hence such word forms, and not extracted roots, ought to be the input
of your GMP.

Think of sound changes as natural forces (wind, rain, etc.) eroding
a building.  The natural forces don't make a difference whether a
particular stone is the keystone of an arch, a step in a staircase,
the top of a pillar, or whatever architectural configuration.
To them, it is just a stone.  Of course, it matters for the building
which function the stone has; but the eroding forces make no difference.
That is how sound change works.  *Sounds* are changed, no matter
what they encode.  Of course, the sound changes will trigger further
changes in the structure of the language (such as the collapse of
the case system), as the erosion of stones in a building will trigger
further changes in the structure of the building (such as the collapse
of an arch).

> > What happens in natural evolution of languages are *sound changes*;
> > a proper simulation of language change simulates the sound changes,
> > and does *not* resort to the kind of highly unnatural "technical
> > tricks" you are using in Terkunan.
> 
> I am using the technical tricks only because there is a restriction on
> when my current implementiation of the GMP works as expected.
> 
> These tricks are really, I promise, only for doing what you want: good
> simulation.

The whole problem is that, as I already said, you are trying to
simulate something that never happens in natlangs.

> >> In short: I think I do exactly what you want me to do.
> >
> > No, you don't.
> 
> Yes, I do. :-)

Well, we have talked past each other here.  The fault lies not so
much in *how* you simulate language change, but rather in *what*
you simulate.  You fancy a change of the language that selectively
clips off grammatical endings - that kind of change is unnatural.
What is natural are sound changes which, starting with fully
inflected words, erode final phonemes and syllables and thereby
cause the loss of the grammatical endings.

> > Ray and I have pointed out what is wrong with your approach.  Is it
> > so hard to understand?
> 
> It seems so.  But we're coming closer to understanding each other.

Yes, we are coming closer.  My fault lay in assuming that you were
trying to simulate natural sound changes as discribed in historical
linguistics textbooks.  You are trying to simulate something different :)

And after all, it is *your* conlang.  I just try to help you build
a more plausible Romance language by pointing out mistakes you made.
If plausibility and naturalness of the changes to your language are
not your goal, my critique shall be void.

... brought to you by the Weeping Elf