John Cowan wrote:

> Paul Bennett scripsit:
> > >You might consider the possibility that some of your examples contain
> > >an
> > >/r/.
> >
> > Not in my non-rhotic lect they don't. And really, I'm trying to aim for
> > a
> > set that's rhoticity-agnostic.
> I think Mark's point is that it may make sense to analyze some of those
> non-rhotic sounds with an underlying /r/ *phoneme*, even if no r-ish
> sound appears in the *surface* realization.  Doing this actually makes
> things rhoticity-agnostic for the most part.
I agree, but Paul's most recent comment-- that "r" could be an
archiphoneme -- is very apt, IMO, since it has so many realizations:

[r\] always, in rhotic speech

[r\] in onsets and (sometimes [4]) between vowels
finally, offglide [@] or 0
vowel length ~changed V quality in other environments
(and maybe others....)

And Tristan's Australian shows that it may be entirely absent in closed
syllables-- I think he's reported things like [bI:d] 'beard', thus giving
rise to anomalous (for English) contrastive long lax vowels.  Whether the
V-length is a remnant of //R//, or it has vanished underlyingly in such
forms in that and similar dialects, is open to debate.  IIRC Tristan doesn't
agree, but personally, I'd say it's present-- assuming that Australian is
part of pan-dialectal English but in need of many "adjustment" rules, just
as one needs for Southern US. The fact that we can all understand each other
(more or less) suggests we're all using the same _underlying_ forms.

Depends on whether one believes in Archiphonemes :-)

"r" is the bane of English phonemics (not to mention foreign learners).