On 3/20/2014 6:36 AM, Alex Fink wrote:

> [d] would have even more of a propensity than [t] to develop into [4]
> (after all, [4] is voiced, and it is common for voiced stops to be
> shorter than voiceless too).  So the most straightforward mapping of
> the first set would be [d]>  [4] (>  /d/ or /r/), and then I suppose
> [t]>  [t], [t_>]>  [d].  In fact, your mapping of the second set
> would most likely pass through this mapping of the first set as an
> intermediate stage.

That makes sense.

> Anyway, with the stop series developing in parallel, either of these
> options with a gap at proto-Tirelat [p_>] would yield a pre-modern
> Tirelat labial series of [p B] (or so), with a gap at [b].  Perhaps
> [B] could develop back to [b], though if [f v] were already present
> it seems much more likely that [v] would ensnare [B] (if it hadn't
> been [w] to start with or whatnot).  So it seems easier not to have a
> gap there.  Maybe your option of filling it with [b_<] instead is the
> cleanest, since [b_<] doesn't have far to develop to modern [b], and
> its presence in the series could have been the wedge that
> precipitated the development from ejective (through implosive?) to
> voiced in the other PoAs.
> Also to consider is what the third member of the stop series that
> yielded [ts) dz)] became.  ([z] ?)

It could be.

pʰ	p	p
ɓ	ɓ	b
p	b	w

t̪ʰ	t̪	t̪
t̪ʼ	t̪ʼ (ɗ̪)	d̪
t̪	d̪	d̪ ~ ɾ

tʰ	tˢ	ʦ
tʼ	tʼ (ɗ)	ʣ ? (hmm, that doesn't really seem likely)
t	d	z ?

kʰ	k	k
kʼ	kʼ (ɠ)	ɡ
k	ɡ	ɣ

h	h	-
ʔ	ʔ	-

But ... now I'm thinking that maybe the alveolar /d/ is what turned into 
/ɾ/ in some dialects, and the dental /d/ turned into /ð/.

Or the affricates could have been derived from palatals. */c/ > /ts/, 
*/ɟ/ > /dz/, and the absence of any /cʼ/ or /ʄ/ seems reasonable.