Sean Whalen wrote:
> When you said "following the accent" I thought you
> meant directly following; I would have phrased my
> rebuttal differently and included different examples.
But yes, I really meant what you'd thought I did (see my self-correction
concerning PIIr. *a-mr.'ta-). An *r followed by a voiceless stop was
devoiced when it stood directly after the accent. The correlation
between the location of accent in Vedic (reflecting, in most cases, the
PIIr. state of affairs) and the distribution of <s^.> vs. <r&t> in
Avestan was already known to Bartholomae more than a hundred years ago.
There is no devoicing if the the accent is not on the syllable
containing the rhotic, for example, in compounds with unaccented *harta-
as the second element: cf. <anar&ta-> 'falsehood' < *n.'-harta- (as for
the placement of accent in privative compounds, see below).
> Why does n. have the accent? Do you mind if I use n_
> for syllabic n, etc., since we're using . as retroflex
> also?
In PIE this, the accent in this type of was on the first element
(despite its phonetic reduction, cf. also *sm.'-logHo-s > Gk. álokHos
'wife, concubine', *h1sú-g^Huto- 'well offered' > Ved. súhuta-, etc.).
I'm not sure why the accent was shifted to the medial syllable in
Indo-Iranian just in *n.-mr.to- (but we know it was, see Ved. amr.'ta-);
it normally remained on the initial *n.-, cf. Ved. á-ks.ita-
'undecaying' (= Gk. ápHtHitos), etc.
> What about k@...; is the accent on r there too?
Yes. I assume you mean <k&hrp-> 'shape'; it's a root noun, after all, so
the accent has nowhere else to fall on in the strong cases! Of course
PIIr. (and even PIE) syllabic liquids were accentable. If you want to
see the rule at work in this particular word, compare the acc.sg.
k&hrp&m < *kr.'pam with the same form followed by enclitic -c^a 'and'.
The accent was attracted to the syllable directly preceding the
enclitic, and so we have:
*kr.'pam- + *-c^a 'and the form' --> *kr.pám-c^a > Av. k&r&p&m-c^a
Do you believe me now? :-)
> This rule occurs in Avestan, not Old Persian. In OP
> there are forms like mars^iyu- and ars^a- (syllabic r
>
>>ar in most positions) without any possible
>
> coalescence of hrt; since Avestan has r.>0 before s.
> in a following syllable looking at only that language
> could be misleading.
>
> There are enough optional rules and conditioned
> variation in stems without analogical leveling in
> Avestan that I didn't think my rules would create
> controversy; I didn't know other theories were
> accepted as standard. I'm sorry that this combined to
> make my response seem seem as if I were trying to be
> argumentative for no reason.
It's all right, I'm only trying to present the orthodox view as best I
can. I hope I have shown that the Avestan alternation is governed
accentually, and that the alternation of <s^.> vs. <rt> is parallel to
that of <hrp, hrk> vs. <r&p, r&k>. Whatever rules you formulate for
Avestan or for its ancestor, they ought to take these facts into account.
> What about m@-r@-z^di-ka-/marz^-di-ka-? I can't
> think of any other possible explanation and it fits
> with all other evidence. Also ar is the standard
> result of r_ in most environments for most of Iranian;
> r_ > ar in one specific environment in Avestan doesn't
> seem far-fetched.
Well, Avestan manuscripts are not entirely consistent in distinguishing
<&r&> from <ar&>. Perhaps some degree of free variation can be admitted
here, though I'm no expert.
Piotr