Hallo conlangers!

On Wed, 19 Mar 2014 21:37:40 -0400 Herman Miller wrote:

> I've previously mentioned the alternation between /d/ in some Tirëlat 
> dialects and /r/ in others in the middle of words. It turns out that I 
> have a pair of homonyms with initial /d/. Whether from an original */ɾ/ 
> that changed to /d/ in initial position, or an original */d/ that 
> changed to /ɾ/ between vowels in one dialect, it seems reasonable to 
> speculate that these two words originally had different initial sounds 
> that were affected by a merger.

Yes, that makes sense.
> If there's two original stops that merged to /d/, then either there were 
> three dental stops in early Tirëlat (counting the ancestor of /t/), or 
> there was another place of articulation involved. I don't think another 
> place of articulation is likely, especially since my idea for the origin 
> of the alveolar affricates is that they were alveolar stops in early 
> Tirëlat.
> The other option is that /r/ was originally /ɾ/ to begin with, and 
> merged with /d/ initially. Early Tirëlat would have two rhotics, /r/ and 
> /ɾ/. That could work, and that might be the best option. I've already 
> established that the voiceless /r̥/ is a more recent development.

Another possibility to consider.  But what you discuss below is
definitely much more interesting!
> If there were three dental stops, there likely would have been three 
> bilabial stops and three velar stops. I'm thinking that the third velar 
> stop could be a predecessor of /ɣ/.

This one could be the velar counterpart of the one that gave /ɾ/.
The third bilabial stop could have yielded /w/, then.

>       What if one of the three sets of 
> stops was ejective or glottalized? Then maybe there were only /tʼ/ and 
> /kʼ/, with a gap where /pʼ/ would be (or a /ɓ/ that later merged with 
> /b/?). Or one of the bilabial stops could have been the ancestor of /w/...

Missing /p'/ is quite common.  It is, for instance, also assumed
by the proponents of the "glottalic theory" of Proto-Indo-European
(PIE */b/ is very rare, and all etymologies involving it are
problematic; this is considered one - but not the only - reason why
the voiced grade may once have been ejectives).

> Supposing that early Tirëlat had one of these two sets:
> */t/ */tʼ/ */d/
> */t/ */tʼ/ */tʰ/
> what would be the most likely mapping of early Tirëlat stops to the 
> modern sets?
> /t/ /d/ /d/ (dialect A) vs.
> /t/ /d/ /r/ (dialect B)
> I think neither of /tʼ/ or /tʰ/ is likely to develop straight into /r/. 
> So the second set seems pretty straightforward: */t/ > /r/, */tʼ/ > /d/, 
> /tʰ/ > /t/. But what about the first one?

As Alex has said, */d/ is the most likely to change into /ɾ/.
Then, either */t/ or */t'/ may have evolved into a new /d/.
Perhaps the latter (*/t'/ > /d/) is more likely if */t/ was
phonetically aspirated (as such stops often seem to be in
languages with stop systems of this kind, like Georgian), and
I have heard that there are precedents of this in the Caucasus.
Also, it is the very same change assumed by the proponents of
the PIE glottalic theory.  Maybe, the ejectives first became
implosives and then voiced stops.  Or the system went something

*/t t' d/ > */tʰ t d/ > */t d ɾ/

(similar to what I fancy to have happened in the prehistory of
PIE, except that there, second-stage */d/ gave */dʰ/ instead
of /ɾ/; at the second stage, my conlang family Hesperic would
have branched off).

... brought to you by the Weeping Elf
"Bêsel asa Éam, a Éam atha cvanthal a cvanth atha Éamal." - SiM 1:1