Jesse Bangs skrev:
>> Admittely I cannot think of parallels, but m < *w / #_{r l} makes sense to
>> me as a sound change. Fortition of some sort would not be so strange in this
>> position, as the original clusters violate the sonority hierarchy. At least
>> it seems better than m < *b, where there really is no motivation for the
>> change. (I'm presuming plain *w and *b stay put during this.)
> 
> I concur. I don't think that you need much more reason that this: *w >
> m / _[r/l] sounds like a perfectly good rule to me! The nasalization
> can be rationalized as necessary to preserve sonority.
> 

I dunno.  An approximant turning into a nasal *in one go*
seems unlikely to me, though I can imagine a series

     w > B > b > mb > m

That would presumably take many generations, but it
gives ample fodder for dialects!  Heck, even w > j
is attested, and then you can have the parallel
series from j to J and a branch-off from J\ to d and n
and n > N > Ng > g.

As an off-list diacronic conlanger friend once said:
"When in doubt (about a turn of development) create
a dialect!", and as YRC said, given enough time any
sound can turn into any other sound (probably not
entirely true...)

/BP 8^)>
-- 
Benct Philip Jonsson -- melroch atte melroch dotte se
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
  "C'est en vain que nos Josués littéraires crient
  à la langue de s'arrêter; les langues ni le soleil
  ne s'arrêtent plus. Le jour où elles se *fixent*,
  c'est qu'elles meurent."           (Victor Hugo)