Retroflexion in IA: The case of .s

[Vidhyanath Rao]

First, let us look at Iranian changes: I will use S2 and
S3 for shibilants of unknown value that become s and s-hacek in Iranian
and "s and .s in Indian.

PIE s  ---------------------------------------------> h
         ruki ---------------------S3---------------> s-hacek
PIE k' ----------------------------S2---------------> s
And k' must have become an affricate before becoming a shibilant.

I don't see anyway of making S3 disappear and still explain this.
If we keep s/S2/S3, what were the sound values and how do you explain
both the Indian and the Iranian developments?

Beekes, in `Historical Phonology of Iranian', JIES vol 27 (1997) pp
1--??, puts ruki rule in pre-IIr days. He adds that, on phonological
grounds, we can expect .s => s-hacek, and not the other way around. I
don't know enough diachronic phonology to agree or disagree with this.
But it is an issue to deal with.

I still don't know if it is words that matter or stems, for establishing
phonemes. If t and .t were still unified in PIA, what do we do with
.sa.t vs sat?

The second problem here is the total lack of sibilants in
proto-Dravidian. Sibilants seem to come from borrowed words or are from
affricates (PDr c or out of PDr stops). [It is doubtful that PDr `zh' or
whatever you want to call it was a sibilant. Also, it does not seem to
lead to retroflexion, in Tamil at least : there are words such as
`vaazhttu', `vaazhtal'.] Contact in such cases lead to simplification,
not splitting.

>He informs me that he has measured increasing retroflexion as we move
>from the family books of the RV to the later books.

But you cannot leave it dangling there. Why did the s/.s split, if it
was really due to Dravidian influence, not survive into Pali and
Prakrits? Note however that s/"s/.s merger is what we should expect
from convergence with Dravidian languages.

>Diachronically, the reading mo-Su-NaH is clearly *later* than the "daily
>pronunciation" mo-su-naH.

Before I get to this, a digression is needed.

The distinction between external and internal sandhi grows with time.
This is clear in case of sandhi of as+k/p. In RV, s often stays,
especially if there is no metrical or syntactic break. But as time goes
on, it becomes restricted to smaller and smaller group of compounds.
But, -as+k- stays -ask- if k begins a suffix.

A similar thing happens with is/us+k/p. RV has agni.s.te etc. Later this
becomes restricted to (two-member) compounds and within words. [Panini
does allow .s to stand (optionally), even in sentence sandhi, if the
first word is neuter sing. of an is/us stem. But this is rather uncommon
and is unknown in modern editions (and manuscripts?).]

The two situations seem quite parallel. Rules of internal sandhi apply
to closely connected words, if the second word is unaccented, and in
standing phrases. This is steadily reduced in later texts. Why should
we consider the second case as due to `mechanical recitation', but not
the first?

Now to `mo.suna.h'. It is not at all clear to me that `mosuna.h' is
earlier, or even that it was more than marginal. What is usually printed
as `mo .su .na.h' is actually maa + (uu .su) + nas. uu.su is, AFAIK,
always retroflexed. su does not occur as an independent word in
Post-Vedic; not even in Vedic Prose, according to Macdonell. Taken
together, I don't see how `mosuna.h' could be older, anymore than Pali
bhavissati be older than Sanskrit bhavi.syati. (In fact, before ruki,
laryngeals were probably still there which means that I won't even try
to guess the probable form of pre-retroflex mo.su.na.h).

There may have been other recensions to RV than "Saakalya. But it is not
at all clear that they had `mosuna.h'. In fact, according to AA,
Maa.n.duukeya too used .s and .n.

Another point worth noting is that nothing is said concerning .t. Is it
just a coincidence that t and .t remain distinct in MIA but s/"s/.s
merge into one phoneme as do n/.n? [I suspect that what is written as .n
in MIA was originally intermediate between _n and .n. This is what I
would conclude from IA-Dravidian convergence leading to the merger of
n/_n/.n.]

Regards
-Nath