Pronunciation of "r" - again?

Hi cybalist members,

I cannot remember whether I already posted the question I am about to
post now. I know I posted it on the "Ask-A-Linguist" site, and I
believe it may have been this question for which I was directed to
Cybalist by the "Ask-A-Linguist" moderators. I tried to do a search
for this question in our message database, but I could not narrow the
parameters sufficiently to weed out extraneous messages. Having said
this, I probably did submit this question before, perhaps about a
year or more ago, but I will submit it again because I still don't
have an answer in my mind:

What is the consensus view on the original pronunciation of the Indo-
European phoneme /r/? Most modern IE languages have a "rolled"
or "trilled" r, whether by the tip of the tongue or the uvula, which
suggests that by numerical probabilities alone, IE /r/ was probably
trilled too. But then where did the prevailing untrilled English /r/
come from (the /r/ of almost all English dialects and accents with
the chief exception of Scottish)? Is it a "softening" of an earlier
trilled /r/? And why then does William Dwight Whitney say that
Sanskrit /r/ "is clearly shown by its influence in the euphonic
processes of the language to be a lingual sound, or one made with the
tip of the tongue turned up into the dome of the palate. It thus
resembles the English smooth r, and, like this, seems to have been
untrilled" (He goes on to say that Panini reckons it a "lingual"
while other grammarians define it as being made "at the roots of the
teeth" which might suggest a trilled r, but no grammarian makes any
mention of vibration). If an ancient language like Sanskrit had an
untrilled r, would that point in favour of IE /r/ being untrilled? I
know some have said that IE /r/ probably varied from place to place
and person to person. But surely one pronunciation must have
dominated?
I have pointed out before (to "Ask-A-Linguist") that the development
of /z/ to /r/ in Germanic and Latin suggests that the /r/ that
developed such was untrilled, because /z/ and such /r/ would be
linked in pronunciation by their continuant (non-obstructive)
pronunciation, unlike trilled /r/ which has obstruction. This
suggests that perhaps all /r/ were untrilled originally in these
languages, since the /r/ from /z/ merged completely with other /r/.
However, countering this is the observation that the initial
sequence /wr/ which existed suggests that /r/ was trilled, since an
untrilled /r/ would make the preceding /w/ practically inaudible (try
to pronounce an English /w/ immediately followed by an English /r/
(plus a vowel) and you will see that it sounds too much like
simple /r/. Trilled /r/ on the other hand would make the /w/ more
distinct). Also countering the first argument is the observation
that /sr/ in Germanic and some other languages became /str/, which
suggests that the obstruction of a trilled /r/ became manifested as a
plosive /t/ after /s/. An untrilled /r/ would not be likely to cause
such a change (and an English /r/ would probably tend to change
the /s/ to /sh/ (/s^/), similar to how English /tr/ and /dr/ sound
like /ts^r/ and /dz^r/ (chr and jr in English letters)) in my
opinion.) Lastly, certain words in English suggest that English /r/,
at least in some positions, was originally trilled: words such
as "father", "gather", "weather", "mother" originally had /d/
before /r/ not /dh/ (voiced interdental fricative). To me this
suggests that it was a trilled /r/ that occasioned the change to /dh/
(if you try pronouncing a dental (not alveolar) /d/ plus trilled /r/
you will see that the /d/ tends to sound like /dh/. An alveolar /d/
tends to get absorbed by the trilled /r/, in my opinion). So maybe
English /r/ actually was trilled once like all the other modern IE
languages', and only recently softened. Countering this possibly is
the recognition that Old English back diphthongs before /r/ plus
consonant ("heard", "heorte") suggest a pronunciation of /r/ that
produced a back glide - probably a retroflex approximant, in my
opinion. At least before consonants Old English /r/ might have been
an approximant, while elsewhere it might have been trilled.

Of course, these are only my personal theories. You may have seen
them before, I probably did post these arguments previously on
Cybalist. But I would really like to know what is the majority
opinion of the original pronunciation of IE /r/: was it untrilled as
in general modern English and as claimed for Sanskrit, or was it
trilled as in the majority of modern languages? I am curious to know
whether English is conservative in this respect as it is conservative
in its retention of the phoneme /w/ in initial position. I will save
any responses to this posting to my hard drive, so that I will not
ask this question yet another time.

Andrew Jarrette