Sally Caves scripsit:

> I just can't duplicate what John is describing and still pronounce "car"
> the way I do it.  So there's no curling up of your tongue tip towards
> the roof of your mouth?  It stays behind your lower teeth?  Is there any
> curling at all, John?  When I try to duplicate that, without the curl,
> I get not only a sound that changes the quality of my "a," but an "r"
> that sounds like "caw" with "r-coloring,"  If I curl it, with the tongue
> still behind the lower teeth, I get a deeper sounding r, but in order
> to make it sound right, it still points up at the roof of my mouth.

On further investigation, my /r/ (both initial and coda) is a velar
approximant, with the tongue-behind-lower-teeth a secondary gesture.
Or perhaps it is not a gesture at all, but just the physical consequence
of keeping the front of my tongue slack.

> You and I have met at Tim's house (that was a wonderful party!).
> I don't think I noticed that your "r" was different from mine.
> Maybe these distinctions are so subtle that it's hard for others to
> hear it when they aren't listening for it.

Mindful of this, I taught myself to say "car" and "rack" with my own /r/,
with an alveolar approximant, and with a retroflex approximant.  I tested
these as minimal pairs and as the full triplet on two native speakers of
American English, one rhotic and one partly non-rhotic (typical speakers
of NYC English have both rhotic and non-rhotic varieties at command,
and use more rhotics as the register rises).  Nobody could hear any of
the differences.

So I suspect that children learn their American /r/s whichever way,
and suppose that everyone else pronounces it just the way they do,
but if all our mouths had fingers in them, we'd find a wide variety of
different styles of pronunciation.

> I guess I'm frustrated that I don't completely grasp where these areas
> in my mouth are: "post alveolar, alveolar palatal, and retroflex region.
> I have been entrenched in thinking that retroflex means the curling
> of the tongue UP.

The trouble is that the classical POA terms are capturing two separate
facts simultaneously: where the tongue is touching or almost touching,
and what part of the tongue is doing the work.  So retroflex s and
alveolopalatal s are both being fricated against the same part of the
palate (just behind the alveolar ridge), but the first is with the tongue
tip, whereas the second is with the blade so placed that the tongue tip
winds up behind the lower teeth.

> What we need in CXS is a better representation of the variations in the
> American "r."  Judging from what I've heard, these sounds have been
> neglected.

They are neglected precisely because they make little or no difference
to anyone (except us phonetician-geeks).  For all we know, there are
pairs of identical twins out there that have learned and use different
pronunciations of their /r/s.

John Cowan                                <[log in to unmask]>    
Yakka foob mog.  Grug pubbawup zink wattoom gazork.  Chumble spuzz.
    -- Calvin, giving Newton's First Law "in his own words"