Joseph Fatula wrote:
>    If two phonemes have the same phonetic realization in a particular
> environment, how do you determine which a word has if it has that
> environment?
> Consider "fair, bear, chair, hare, very".  In my idiolect at least, short
> E
> and long A are two separate phonemes, but before R they have the same
> phonetic realization.  I realize this isn't the case for many of you, but
> it's just an example.

The problem is that in some dialects of Engl. (mine and apparently yours,
and others) the tense/lax (or call it long/short) distinction _neutralizes_
before /r/, principally in monosyllables (and polysyllables too). Thus it's
a six-of-one situation w.r.t assigning the vowel sound to a given phoneme.
Generally I think the lax variant is chosen because (1) the actual sound is
closer to it and (2) of the two, the lax variant is less "highly marked".

Those dialects that distinguish the vowels of _Mary, marry, merry_ may be
able to assign the phoneme differently, provided a derivative form exists
where the non-neutralized variant shows up.

IMO it's questionable whether it's economical in a description of English to
posit an archiphoneme in just this one environment. If we do, then we ought
also to posit them for all the vowels that can alternate with [@], and
pretty soon we end up with an extremely large and unnatural vowel inventory.

> Or consider the T phoneme being realized phonetically as a D after a
> stressed vowel and before another vowel (again, in my idiolect, I realize
> it
> doesn't apply to many of you).  In the word "atom", it's phonetically D,
> but
> in "atomic", it's phonetically T.  This one I can confidently assign to
> the
> T phoneme, in that I have a variant of the word with the D-realization
> environment and a variant without.

It depends on the stress placement, of course.

What about a word where I don't have a
> varying pair like this?  (And where the spelling doesn't indicate a
> historical pronunciation before merger...)

Are you thinking of pairs like _latter, ladder_? where some people have a
flap in both.  I don't. "Latter" has a flap, "ladder" has...well, [d].