Indo-Iranian

[Vidhyanath Rao]

>    The Proto-Indo-Iranian phonetic values of the IE palatals thus were
> [ts'], [dz'], [dz'h], or, in a simpler notation *c', *j', *j'h with an
> acute accent on the affricate to indicate prepalatal position (as
> opposed to the hacek on *c^, *j^, *j^h denoting a more
> central palatal position of the younger palatals).

Is it correct to take this to be reason why *k'/g'/gh'+*t becomes s.t.
(retroflex) in Sanskrit but  *k/kw +*t etc become kt?

Does this also have something to do with the peculiarities of Sanskrit
ch (usually described as aspirated c)? It seems that in some dialects it
was pronounced as cs' (ch is written cs' consistently in some manuscript
groups). Also in sandhi, t+s' becomes cch (the tradition followed in
printed editions, but not universal in manuscripts) and according to the
grammarians, in some dialects, s' became ch after any stop, but in some
others it never happened. People have argued for the last 100+ years
about which is older.

[The usual explanation for ch is sk' -> k's -> ks. -> ch, the last a
``prakritism''. I find it hard to understand how other ks. escaped this
fate.]

BTW, PIE morpheme final *k' becomes k when followed by s. (from *s) in
several cases, the most common being from *drek' (ta:dr.k, adra:ks.i:t)
etc, in some others it seems to change in extant texts (RV viks.u vs
vit.s.u from Panini/upanishads on).
----

Returning to the general questions: How widespread is the merger of ruki
s + *t with *k'+t etc, in particular in Nuristani, and what happens
there to tk'?

I am also curious about one objection raised by Sihler to the retention
of occlusion in Nuristani. This is the assimilation seen in Sans. s'as'a
etc, which is found in Nuristani as well. Hamp's reply to Sihler does
not seem to address this. [I thought that this can be explained as Indic
influence combined with mapping based on subconcious awareness of sound
equilances. But the latter seems to be strongly rejected, to judge by
another thread in this list.]

 ---