This is a stratchpad for an idea I've had floating around for a long time.
I'm trying to put together a conlang where each morpheme has several possible allomorphs, whose distributions are fairly unpredictable.
First I'm creating an ancestor language. This is a pretty regular, agglutinating language, with a simple phonemic inventory.
i a u
Syllable structure is [C]V[L][n] Where L is an approximant.
Here's a list of roots that I've generated using this phonology. I'll be defining them a few at a time, with the assumption that the speakers are a bronze-age city state.
In the meantime, I'll also be putting together the morphology, syntax, and corpus.
While the phonemic inventory is simple, there's going to be a lot of allophony. Once I've got a decent sized corpus written, I'm going to convert it into a phonetic (rather than phonemic) transcription.
The next step is Cry havoc and let slip the dogs of sound change. I'm going to apply sound changes that break the allophony, turning different allophones into independent phonemes, or merging realisations of different phonemes. At this point the morphology should become a lot more fusional and unpredictable.
--PeteBleackley 08:09, 14 September 2007 (PDT)