Dirk Elzinga replying to my post:
>This is how I did Tepa as well. But I've noticed that in natural
>languages there will be certain phonotactic sequences that appear
>relatively frequently, while others, though possible, appear
>infrequently or not at all (the accidental gap syndrome). I didn't get
>that quite right in Tepa when I generated the list of possible Tepa
>roots; my list had all phonotactic sequences for mono- and disyllabic
>roots equally represented. (I think LangMaker has a way to encode
>numerically prominent phonotactic sequences.)


Yes to all of the above; I wasn't specific enough.  Frequencies had to be
adjusted after generating the list (supposedly the program could do that,
but it wasn't cooperative).  Aesthetically displeasing forms were
deleated.....

>When mapping meanings to roots, I found myself unconsciously selecting
>certain types of roots over others (eg; roots ending in -u are not as
>common in Tepa as roots ending in other vowels, though there is not
>phonotactic prohibition against it). This process has given Tepa a
>perhaps more natural feel. >

I did that too, perhaps more consciously in at least some cases-- I decided
early on the /p/ and /f/ would be rare. More or less unconsciously, however,
words with final /p/ started grouping themselves into a naturally
"pejorative" class.  I do think I overdid the palatals, overlooking the fact
that /ya-/ and /yu-/ are very common prefixes, and /-S/ is the neuter pl.
But I'm satisfied, and believe it has a natural feel too-- all thoses /S/'s
certainly move it away from its Indonesian model.

Monosyllabic Gwr is being generated by hand; but its proto-language is
something of a nightmare.