Advice on Sound Changes

[evmdbm]

If you have an inflected ur-conlang and you want to bring it on a few hundred years, do you guys just take the base citation version of a noun (say the nominative singular) and apply the sound changes you choose to that or do you apply the sound changes to all the different forms of the noun of which there could be several if say there are four cases and two numbers, or even more cases and numbers (there are 15 in Finnish!)? Depending on what you do with final consonants or vowels you could get some strange results.

The background is I was messing about to create some different, but related, words for naming languages and realised that say if you took
drimpaq (bridge) in Vedreki, where the accusative singular suffix is -a, dative -et and genitive -em, and ran all those four words through and dropped the last consonant you'd get drimpa -aqa, -aqe and -aqe. That's fine except any other dropped final consonant would just multiply declension patterns.

It would seem even worse with verbs because any moderately inflected verb (or insanely inflected one like Ancient Greek) would come out with literally zillions of new patterns once you start dropping final sounds.

In the end (using zompist's cheat in LCK) it is enough for now that Nakhese comes out as trempa and Katradheen as draempeqh!

[mèþru]

Usually one would do the second. Analogy or exceptions to sound change can undo/prevent some of the damage.

[Ares Land]

If you have an inflected ur-conlang and you want to bring it on a few hundred years, do you guys just take the base citation version of a noun (say the nominative singular) and apply the sound changes you choose to that or do you apply the sound changes to all the different forms of the noun of which there could be several if say there are four cases and two numbers, or even more cases and numbers (there are 15 in Finnish!)? Depending on what you do with final consonants or vowels you could get some strange results.

Insane numbers of paradigms are entirely naturalistic.
Go check that article: http://mvtabilitie.blogspot.com/2008/09 ... nding.html for an idea of how bad these things can get.
Some of my favourite quotes: "describing the Old Irish verbal system as 'complex' is like referring to the Arctic as 'somewhat chilly'."
"By around 900AD, dissention was growing in the ranks at having to write and speak this horrific language"
"The most recent introduction to the old language, David Stifter's Sengoidelc, is a hard-core philological tome, and yet introduces cheerful cartoons of sheep dressed in Gaelic costume specifically to cheer the reader up."
I'd love to build something that complex, but I'm afraid I'll never have the insane amount of patience that'd be necessary.

Anyway. Back to your question, here's my own method, FWIW, using verbal morphology as an example:

1. I try to pick a representative number of verbs. Somewhere around 50 works nicely, and list their paradigms in a spreadsheet.
2. I apply sound change to the whole set.
3. With naturalistic sound changes, it's entirely possible to get 50 different paradigms. It's time to take note of patterns.
4. First I'll look for root alternation. This should really only occur in the most common verbs, so I make an educated guess at what these are and forgo root alternation for the rest.
5. Alternatively, maybe there are patterns to root alternation that can be described in a morphological rule!
6. Now just looking at the affixes, there should 15-20 patterns. Again, I try to make a guess at which might be the four or five most common ones (running changes on a few other verbs just to check).
7. For the rest of your lexicon, just like in a natural language, you should have a few criteria that lets you know which paradigm to use.

Another tip: the citation form in the parent language is not necessarily the most useful (Romance languages mostly used the accusative, for instance).

[alice]

Insane numbers of paradigms are entirely naturalistic.
Go check that article: http://mvtabilitie.blogspot.com/2008/09 ... nding.html for an idea of how bad these things can get.
Some of my favourite quotes: "describing the Old Irish verbal system as 'complex' is like referring to the Arctic as 'somewhat chilly'."
"By around 900AD, dissention was growing in the ranks at having to write and speak this horrific language"
"The most recent introduction to the old language, David Stifter's Sengoidelc, is a hard-core philological tome, and yet introduces cheerful cartoons of sheep dressed in Gaelic costume specifically to cheer the reader up."
I'd love to build something that complex, but I'm afraid I'll never have the insane amount of patience that'd be necessary.

Wait for it...

Pfft. I've been making conlangs way more complex than that ever since I was eight years old.

[Pabappa]

My main conlangs are at two extremes in their approach to this problem. With Pabappa, massive analogy occurred, and while the language still has many inflections, all verbs and nouns are regular and there are no separate declensions to learn . But in Poswa, there was no analogy at all, so every root has four separate forms, and knowing one won't tell you much about the other three. I have yet to make a language that occupies the middle ground where most natlangs are. However, I think even the extremes are viable , should you decide you want a language like Old Irish or one with perfectly regular morphology.

[linguistcat]

I'm working on something similar myself but starting with Old Japanese. Throw in vowels being lost in both verb roots (sometimes) and the inflections (mostly regularly, but depends on if the root dropped any vowels), and I'm getting a lot of weirdness even before factoring in that even Old Japanese had 4 regular paradigms and 4 irregular ones from what we can tell. I'm probably going to use a lot of analogy in the end, personally. But it might be fun to have all these different quirks in my language I'm making.

[evmdbm]

I guess the Nakhese would probably just assume trempa was feminine and decline it Acc trempu, Dat trempu, Gen trempem (or trempe once the -m is dropped) in the singular. I think my main concern once I started mucking around with the SCA was to create something naturalistic. As it stands there are no irregular verbs in Vedreki, which doesn't strike me as naturalistic. Sound change appears the only way to generate irregulars, but I'd rather not generate a maze of apparent single instances or a case where the patterns are as incomprehensible as Old Irish - let's face it at some point I need to explain it all to you lot and understand it myself.

I've only got 200 or so words in total, so maybe I'll just muck around and see what happens. Better to do it now rather than when I have to work backwards like Verdurian to Cadinor.

[zompist]

Ars Lande's method is excellent. I'd probably be a little lazier, but I'd definitely run several verbs and nouns through their full set of forms and see what happens. I love a set of sound changes that drastically messes up the system.

[zyxw59]

I like to run words with all their forms thru one at a time, until I see the relevant patterns and I can confidently decide what declension they belong to just by looking at the citation form.