On Wed, 7 Aug 2002 6:47:4 +0100
> En réponse à "H. S. Teoh" <[log in to unmask]>:
>
>
> I'd say, don't worry about Latin letters right from the start; get
> your
> phonology going first, and *then* consider what's the most
> straightforward
> way to write them in Roman letters.


As almost every trying of mine was a failure, I give you how I do with sabyukà. Maybe other people work that way.

First I tried to decide roughly the basic patterns of the language (agglutinating, synthetic, case-marked, ergative, accusative and so on...). I think when you go on hollidays, it is interesting to have an idea of where you want to go, rather than take your car and drive it blindly. Having a rough idea of what you want to get can help to avoid mistakes. An error I made *several times* in a preceeding trying was to design the vocabulary (and I had spent so much time on it, deriving roots from IE) and a phonological system whithout having any idea of what the grammar would look like. I experienced a lot of problems, because the phonological system prevented me from constructing the morpho-syntactic rules I had in mind. For example, if you want an agglutinating language, with rules of assimilation rather than epenthesis (even if they don't exclude each other), a good point to start with , especially when like me, it is more or less your first attempt, could be to have a rather simple ph



onological system. However, if you're sure that two vowels will never merge together, and you'll have systematically an epenthetic consonant ( [t,n, ?] for example), then you can deal with a greater number of vowels.

I tell that because I used to try to design first the phonology, vocabulary and then the grammar, and every time it was a discouraging failure. The only way that seems to work properly with me is to do like this. I don't mean this is a better method, but this is another way to go ;).

Julien