Posted by Mark Rosenfelder (64.16.210.2) on November 29, 2000 at 15:12:41:
In Reply to: Re: Secret History of Verduria V: The novels you can't read yet posted by Mark Rosenfelder on November 20, 2000 at 18:00:52:
Around 1992, I decided to learn all about historical linguistics and re-do all the Eastern languages.
I started by reading Theodora Bynon's Historical Linguistics. I learned a lot, but perhaps the most important thing was to learn what I'd been doing wrong in previous amateur attempts: I didn't know about the regularity of sound change. Sound changes don't apply sporadically; they apply across the board, in every word that has the triggering environment.
(Theoretical note: Well, actually, they start sporadically, affecting certain words. But the sound change keeps on hitting more and more words, and generally it ends up getting them all. So if you look at a sound change in progress, you see wide variation and even confusion; but if you look at a historical sound change, it looks exceptionless. And in any case the regularity principle remains an excellent guideline: it makes you keep looking for subtle regularities that underlie apparent exceptions. You learn more that way.)
The regularity hypothesis simplifies creating multiple daughter languages, largely reducing it to a problem of finding a nice set of sound changes and applying them to a lexicon (I use a program to do this).
That's fine for working forward (e.g. Cadhinor to Barakhinei), but trickier for working backward (e.g. Verdurian to Cadhinor). I worked through the vocabulary word by word, thinking about the possible parent forms (with good rules, there's usually a choice). (I don't have a program for this sort of multiple-outcome reverse engineering, but I knew the rules well enough to do it by hand.) Sometimes I just didn't like the possible Cadhinor form, or (worse yet) the Verdurian form was simply impossible to generate given the rules. In such cases I could tweak the rules, or change the Verdurian form, or borrow the word from some other source (I figured that would add verisimilitude anyway).
At the same time I revised a good deal of the vocabulary. I had an embarrassing number of words that were direct steals from English or Russian-- e.g. antelop for 'antelope', now changed to gudun. I also had lots of long unanalyzed words-- a rare unfixed one is lelitsala 'art'. There was some idea of deriving this from elir-dhalec 'life-enriching', but unmotivated phonetic distortions like that are no longer allowed! I got rid of many of these forms by using derivations instead; others were divided properly into morphemes. (E.g. shrifta 'knowledge', diverging oddly from shrifec 'know' is now explained as incorporating a Cadhinor collective suffix -ta.)
I also complicated the morphology. Probably anyone who's tried learning Verdurian curses me for this-- or ignores those parts-- but I'm rather proud of this, since it's a naturalistic feature of real languages. Most of the complications come from the regular sound changes: e.g. the irregular 1s form lagao 'I get' preserves the g that was fricativized in the infinitive lazhec. (If you think that's bad, you'll really hate Barakhinei.)
A particular challenge was to provide some sort of historical justification for the verbal inflections of Verdurian, which exhibit teasing hints of regularity. I had no restrictions on the Cadhinor and Proto-Eastern forms, of course, but I did want the two sets of sound changes to start with something more regular and end up with the Verdurian forms (perhaps tweaked a little by analogy). This proved to be surprisingly difficult. The best I could do was to unify the three conjugations in one (here), and that system could hardly be more arcane... though it's any worse than many real-world examples.
Much of what I learned during all this ended up distilled into the Language Construction Kit.
I know nobody will follow this advice :) but I will say that it's easier to do it right the first time. Writing Verdurian, I still have to check the dictionary all too often to make sure I'm not remembering the old form rather than the revised one.