Retroflexion

[Vidhyanath Rao]

"David L. White" <dlwhite at texas.net>

> The idea that only developments that cannot be explained internally should be
> explained externally (which seems to be lurking here) is nicely dismissed in
> one of the earlier chapters of Thomason and Kaufman.

What then is to stop us from proposing external influence in every
single case (or bacteria/viruses from comets for biological
explanation)? For example, what is to stop me attributing the emergence
of front rounded vowels to substratum influence of pre-IE European
langauges, or the Great Vowel Shift of English to whatever?

If memeory serves me right, Thomason and Kaufman use retroflexion in IA
as a prop for this dismissal. Since the adequacy of substratum
explanation for that is precisely what we are discussing, it seems
circular to invoke them as authorities here. What we need are
>independent< criteria for comparing internal and external explanations
and testing the latter in the case of retroflexion.

For me, the criterion is simply which explains more and requires least
amount of special
pleading to explain >all< relevant data. Substratum explanations for
retroflexion simply ignore all the troublesome details about Prakrits.]

> > Turning now to substratum explanations:
>         What is the supposed origin of the contrast between palato-alveolar
> and aleopalatal sibilants in the first place?

Huh? The claim is that the distinction is rare to non-existent. If you
mean why laminal vs apical distinction should come about, the question
is badly phrased: ruki-s ( -> apical) and affricate (?) outcome of PIE
k' were distinct to begin with. The origin is the assibilization of k'
which lead to ruki-s becoming retroflex to maintain the distinction.

> I note (with some trepidation) that L&M assert that having "high-domed
> palates", which renders retroflex articulations unsually easy to make and
> distinctive in sound, is an Indian racial trait.

I missed who L&M are. And is this a statistical trait, or absolute
trait? [retroflexes are absolute trait of Indian languages, not limited
to individuals with high-domed palates.] And what about other
languages/dialects that have/had retroflex sounds?

BTW, there is another change that proponents of substratum influence for
retroflexion don't mention. The dentals seem to have moved forward.
Rk-Praati"sakhya calls the dentals "dantamu:li:ya". [I have heard this
attributed to Kashmiri pronunciation as well.] But today, the dentals
are pronounced as inter-dentals (tongue between the teeth or at the
bottom of the top teeth). Why not attribute this to substratum too? [And
what racial trait would this reflect?] The alternate, neutral
explanation is that Dravidian had the threefold distinction of dental vs
alveolar vs retroflex and the pronunciation tended to make for maximal
differences between them.

>         What is this thing " 's "?  A palatalized /s/ or just a typo?

It was meant to be what in TeX will be \'s: s with an accute mark over
it, the usual transliteration of the palatal (laminal) shibilant. Normal
convention for those using ASCII us "s, which I forgot.

> One of the more common ways of making a [s^] sound is effectively
> semi-retroflex.

But then you are reducing the distinction between "s and .s!

> "Language X" could well have died out in the interval, and would hardly be
> expected to reach out and exert any sort of influence from the grave.

What "Language X"? Till now substratum explanations have tended to
attribute retroflexion to Dravidian influence, with occassional nod
towards Munda. In borrowed words and syntactic matters, their influence
grows till Pali at any rate. Bringing in a "Language X" with unknown
phonology hardly explains anything. [Language X, Para-Munda etc are
posits as sources of words in I-Ir and IA with unexplainable etymology.
It is not even clear how many languages are
involved, much less what their structure was.]

> [...]  Indeed traditional Sanskrit grammar, if I have understood it
> correctly, regards the difference between /r/ and /l/ as being that
> /r/ is retroflex whereas /l/ is dental.

Strangely enough, that is not true. Assigning r to retroflex series is a
comparatively late development. The earliest phonological notices
consider r to be produced at the gum line (dantamu:li:ya) or as alveolar
(barsvya). [Norwegian r was not retroflex, I am told.]

> Again we, or at least I, must wonder, if this sort of thing is normal from
> internal causes alone, 1) why we do not see it more often, and 2) why it
> happened only in India, where the pre-IE population was probably both
> racially and linguistically pre-disposed to favor retroflexes.

Wasn't the whole point Hock's objections that retroflexes are not unique
to India, but are found elsewhere?

>         It would seem better to answer both questions than neither.
> One good question would be why the development was to /s^/ in Avestan and
> /.s/ in Sanskrit rather than the other way around.

Because, in Avestan, PIE k' went to /s/. So there was just one shibilant
(if we ignore the one from *rt).

Actually, once I pressed M. C. Vidal on this point in the Indology list
as to why we should not posit a retroflex shibilant to Proto-Iranian as
well. One can just get around it by making the various relevant changes
occur in a precise sequence (see
http://listserv.linguistlist.org/cgi-bin/wa?A2=ind9808&L=indology&P=R2172
). But that is not accepted by all. Beekes (JIES, late 1990s) ,for
example, attributes a retroflex shibilant to PI-Ir reflex of ruki-s.

> [...]  But please note that I am not questioning that the various
> retroflexes were phonemic in Sanskrit (after a time),

If you accept that real words count as do sandhi variants but not
accent, then u:dhas (udder) vs u:.dhas (<*wgh'tos) is a minimal pair of
unimpeccably IE words. I thought that even one pair is
enough.

> What I would guess is that in the few cases where a retroflex-conditioning
> environment went to zero, this was in part because very many speakers already
> possessed the ability to hear retroflexes as phonemic, so that for them, no
> new contrast was created.

Does this theory apply to every phoneme split? Are they all are due to
external influence? If not, what is special about retroflexion?

Note that retroflex stops from /rt/, ruki-s+t etc become phonemic in
Prakrits due to simplification of consonant clusters. Due to diglossia
that must go back to even Middle to Late Vedic, IA speakers would have
made this distinction anyway.

BTW, Proto-Dravidian /rt/ survives in (Formal) Tamil. Hock gives a map
of the fate of /rt/. It depends on geography, not on language family.
How do you reconcile this with substratum explanations?

Regards
Nath