Re: [tied] tt/st/ss

--- In cybalist@yahoogroups.com, Piotr Gąsiorowski <gpiotr@...> wrote:

>
> Sean Whalen napisał(a):

> > There's no reason to assume a PIE stage of tt>tst.
> > This would mean that d(h)t>tt already, but Baltic and
> > Slavic show that devoicing before voiceless stops
> > hadn't occurred yet (among others); Sanskrit shows
> > dht>ddh.
>
> Why?
>
> (1) You can have affrication independently of devoicing, e.g. -d(H)-t-
> realised as *-dz(H)-t-

But Iranian has dht>ddh but no loss of s/z between obs.

(or, as an alternative, restoration of voicing

> through analogy in some branches).

Why would this analogy only work on voiced aspir. at a time when it
makes it look like there's no PIE dzH (that is, no d+t > s+t ana> d+t

> t+t and so on in Iranian)?