Re: backwards conlanging

On Sun, 26 Nov 2000, Muke Tever wrote:

> > From:    Yoon Ha Lee <yl112@...>
> > Subject: backwards conlanging
> >
> > I've read several conlang sites that suggest starting with an
> > ancestor-language and then deriving conlangs from the ancestor-conlang.
>
> (...which is something I find quite useful, as I personally find my evolved
> conlangs to feel smoother than my ex nihilo ones...)
I can imagine.  :-)  I love the idea; it makes a lot of sense and it lets
you create families.  The problem is I don't know enough about deriving a
language in this way, and all the conlang sites that do suggest this are
terribly vague about possibilities.  (Langmaker.com has a neat table of
common consonant changes, but doesn't say a darn thing about likely vowel
changes.  I'm still reading up on phonetics/phonology and historical
linguistics so I can get a handle on the processes involved.)

> > Does anyone have suggestions for a related-but-faux conlang technique for
> > same?
>
> The way I slightly understand it, one way to reconstruct a protolang from
> just one language might be... rrr..  sort of like the idea of 'underlying
> form'.  If, say, an affix has one appearance in some stems, and another in
> other stems, one might try to reconstruct an "original"/"underlying" form
> that in your modern language produces two forms due to sound change but in
> the protolanguage was only one, and perhaps find a way to reverse-engineer
> that kind of sound change elsewhere.  Er.  Or not.
I'll see if I can get that to make sense.  I did derive much of the
vocabularly for Chevraqis from Arakis, but I realized that the phonology
for Arakis was a) nonsensical and b) Really Yucky to my ears.  So
basically I've got some 50 root forms sitting around that are wasted work
because I have to start from scratch.  I also really badly want Chevraqis
to remain a case language and sound changes really seem to screw with
that.  I'll try to find some time to fool around with the sound changes
program from Zompist.com and see if I can come up with something that
doesn't make hash of the work I've already done.

I wish I'd thought of the method *before* I started work on Chevraqis.
<wry g>  Now I know....

YHL