(I wish I could get top-quoting to work on this phone).

The starting point is Khangaþyagon, which has a phonology similar to English or German. One of the things I want to do to it is simplify consonant clusters, and that's a natural source for my new laterals. Once they're there, I thought it would be fun to play with them a bit. 


Sent from Samsung Mobile

 Alex Fink <[log in to unmask]> wrote: 

On Tue, 7 Jan 2014 21:17:10 +0000, Pete.bleackley <[log in to unmask]> wrote:

>The ancestor language has /w/ /l/ and /j/ phonemically. One wave of sound 
>changes creates /5/ and /L/ from clusters, and another turns them into 
>allophones of /w/ and /j/ respectively. When the laterals merge, their non-
>lateral allophones are unchanged.
>That's as much as I've worked out so far.

Do you insist on /5/ and /L/ having phonemic existence before the allophonic stage?  That seems considerably more unlikely to me than doing it the direct way, creating them in a subphonemic split.  If you were doing it out of some sense of necessity, worried it's impossible to get glides to spontaneously lateralise without a lateral input, you needn't worry, as it can happen: for example Hopi had the change *_w_ > _l_  before a (nonhigh?) back vowel.  I think I've seen a similar change for /j/ but can't call one to mind now (except that Jeff Burke used it in his Central Mountain family and presumably had an antecedent in mind), but Arapaho-Atsina at least did *_j_ > _n_ initially.