Funny W

[Rankin, Robert L]

>> Actually more than one kind of consonant can be involved, but h and ? are the ones that pull disappearing acts and remain the most likely candidates in those cases where no other conditioning factor can easily be identified.  As I say n the handbook article, one has to be very careful not to use such things as "finagle factors".

> This argument assumes that there is an extra consonant involved.  If so, a laryngeal might be most reasonable.  But postulating an extra consonant that has since disappeared looks like a finagle to me.  
 
Well, we KNOW that ?/h do have the necessary obstruentizing effect on adjacent w/r.  There are good examples of r?  >  t? and rh  >  th in several morphemes.  This leaves possible pigeon holes for the mirror image sequences *?r and ?w.  Both laryngeals have appeared and disappeared numerous times in the various languages over time.  No one has ever written the book that would catalog these changes, so we're not on the firmest of ground here.  There has been both epenthesis and metathesis of glides in Siouan, not just in the one rule or constraint fits all of synchronic phonology, but multiple times in multiple environments over several millennia.
 
> If *W and *R were nasally-released stops, then I don't think we need anything extra.

That's essentially just what we've done in positing *W and *R opposing *w and *r  --we've added a feature to differentiate them phonologically.  You've picked the feature [nasal] to do that, and I don't think that's unreasonable (some languages have mb and nd as reflexes).  Others of us have essentially left that feature "blank", and that is what the upper case letters signify.  Additionally, we find that the "consonantizing" feature added to *w/*r is often assimilated from an adjacent consonant in a certain number of cases.  This makes us suspect that there was probably some "disappeared" consonant responsible in the unexplained cases, and this leads us back to the laryngeals . . . full circle.  What I'm saying is that the reconstructions W/mb/wC/Cw along with R/nd/rC/Cr are all in some ways less than satisfactory and essentially notational variants.
 
Bob