Hallo!

On Mon, 29 Oct 2007 18:17:56 +0100, Henrik Theiling wrote:

> Hi!
> 
> Jörg Rhiemeier writes:
> >...
> > The problem, Henrik,
> 
> Please add 'i think' more often, because given your absolutive style
> of writing, it becomes increasingly hard for me to stay calm.
> 
> But I know you mean well, so I will answer calmly.

Oh, I'm sorry.  I didn't want to say that "you have a problem".

> > is that you appliy the GMP to *artificially
> > extracted roots* - units that would *never* exist for themselves
> > in the natural evolution of the language.
> 
> Wait, please.  I think you are mixing up layers of the construction
> now.  It is vital to distinguish the tool from the goal.

The tool ought to be appropriate to the goal, otherwise you end up
trying to fix a wristwatch with an axe :)  If you want a language
that evolved from Latin by naturalistic sound changes, you have to
simulate these sound changes - by applying a GMP to *words*, not just 
artificially extracted roots.  Is that so difficult to understand?

> In the red boxes, I speak as a conlanger.

Sure.  A naturalistic conlang involves two levels of reality:
the imagined reality in which the conlang is a natlang, and
the actual reality in which it is a conlang.  These must indeed
be kept separate.

>       This layer is obviously 
> utterly unnatural.  Of course, since Terkunan does not exist as a
> natural language, but is a conlang.  The red boxes show which
> algorithms and methods I use.
> 
> But please consider that these technicalities have a reason.  They are
> not there for the sake of being technical.  They are there because I
> use the computer as a tool to produce a certain result.
> 
> The *method* is unnatural, of course, but what it simulates is not.

I wonder, why apply an unnatural method if you don't want unnatural
results?  This way, you arrive at an unnatural language.  What you
get looks like a Latinate auxlang with strangely distorted morphemes,
not like an actual Romance language.

> The method is meant to simulate exactly what you require: a
> deterioration of the endings.  That's why I use -/@m/ as an accusative
> ending: for most words, it is just a speed up: if you read the GMP
> thoroughly, you will find that for most words, using the original -am,
> -um, -em produces the same result, because the reduction of final
> vowel is taken care of.  For the rest of the words, I alter the input
> to simulate effects the GMP currently does not account for.
> 
> In short: I think I do exactly what you want me to do.

I slowly see what you are getting at.  Your nouns look good to me.
I can easily understand how the accusative ending is lost and the
bare stem (minus the thematic vowel) remains.  That is what happened
in French, after all.  The verbs, however, still look wrong to me.
Why do your verb forms evolve from the *supine*, a form used very
little in Latin?  I'd rather use something like the 3rd person
singular present indicative as the starting point.  Here is a
point-by-point critique of your verb construction:

> Verbs usually derive from either the present stem or the supine stem
> (or the future participle stem)

Why these little-used forms?

> with the reduced stem vowel found in  compounds (for facere, use fic-,
> not fac-).    

Why?  This looks rather implausible to me.

> If the supine stem is a regular extension (+ optional vowel + t)
> of the present stem, neglecting a possible stem vowel change and
> neglecting a possible drop of stem-final glides, and if no consonants
> fuse (for vidēre, use vis-, not vid-), and if the present stem vowel
> is followed by at least a consonant (for īre, use it-, not ī-), then
> the present stem is used including a potential vocalic/glide ending.     

This is a rule of the sort that is used in auxlangs to extract roots
from verb paradigms, but doesn't resemble the way languages change
naturally.  Sound changes are *insensitive to the morphological
structure of the words affected*.

> If the supine stem ends in the regular -t ending, this -t and,
> if present, also the preceding vowel are stripped. The -t is not
> stripped if the stem vowel would then be final.  

See above, and the question is, why the little-used supine?

> To the resulting stub, an /-əm/ ending is added, and this is then
> sound shifted with the GMP. 

This is an artificial process that doesn't look like a plausible
development in the language.

> Often, prefixes are sound shifted separately so that compound verbs
> are compounded in Terkunan, too.   

Hmmm.  That strikes me as rather unnatural.  As I said above, sound
changes are insensitive to the morphological structure of words.
That's the reason why they produce irregularities.  Of course,
analogy often steps in to restore regularity.  But your method
doesn't look as if it properly simulated that.

I still maintain the position that your construction method is
unnatural, and that it would make *much* more sense to apply the
GMP to the Latin words *directly*, rather than to root elements
extracted by the rather artificial rules you give in your yellow
boxes.

... brought to you by the Weeping Elf