Sean Whalen wrote:
> No, rt can become rs. no matter what the
> environment; it's just optional. Later r > 0 before
> s. or T and s. > s^
>
> m@...^a-/m@...@ta- < mrto- and mas^ya- < mer-t-yo- but
> m@...@Tyu- < mr-t-yu-
>
> compare Sanskrit
>
> mrta- and martya- and mrtyu-
To imitate your style -- No, it isn't just optional. Mayrhofer's
explanation of the rt/s^. alternation as conditioned by the place of
accent has generally (though perhaps not unanimously) been accepted by
the field. <m&s^.a-> is a compositional variant that developed in
_posttonic_ positions:
*n.'-mr.to- > *ámr.ta- > Av. am&s^.a-
But the addition of a suffix or an enclitic could result in a shift of
accent:
*amrtá-ta:t- > Av. am&r&tata:t-
Your other examples simply conform to the rule: the accent of Skt.
mártya- is initial and that of <mr.tyú-> suffixal. Where's your free
variation?
Note that the same conditioning is responsible for Av. -hrp-, -hrk-
versus pretonic -r&p-, -r&k- (cf. v&hrka- 'wolf' < *wl.'kWo-). Since
<s^.> is found instead of expected *hrt, the actual posttonic
development of *-r.t- is evident: the rhotic portion of &r < *r. became
devoiced, and since the following voiceless stop was homorganic with it,
the two segments easily coalesced into a voiceless postalveolar
fricative. No such coalescence was possible in heterorganic sequences.
> No, in Avestan r. > ar before a consonant in the
> same syllable or word-finally. Later r. > @r
> syl-final and Vr > Vr@ at the end of a syllable so we
> can see:
>
> mrz.dika-
>
> mrz.-di-ka mr-z.di-ka
> marz.-di-ka
> z^ z^
> m@...^di-ka
> m@-r@-z^di-ka
>
> Both variants of "pity" (compare Sans. mr.d.i:ka-
> "favor" with metathesis of z. giving long i).
>
> There's plenty of evidence for this so as^a- must
> come from metathesis of H2 (there are many such cases
> with H in Sans. and Av.).
More likely *h2r.-tó- (adj.) vs. *h2ár-to- (noun) with some
cross-contamination in Indo-Iranian.
Piotr