schwa-raising and related

[Rankin, Robert L]

----------------------------Original message----------------------------
> 1. Why does a token have to 'belong' to some specific phoneme at all? Are
phoneme inventories necessarily exhaustive? Couldn't it be the case that in
noisy or obscure environments you can get values that hover between what
might in other environments be norms for other phonemes, and that are simply
not assignable? It could be a kind of structuralist error to assume that
every utterance token belongs exhaustively to some determinate class.
Perhaps, since hearers are looking for meaning, a lot of detail is simply
allowed to be fudged, as long as context gives enough clues for
interpretation? There are alternative models of phonological structure (like
the one in Bybee 2001) that don't in fact have fully determinate
inventories.

There are several different questions here.  I suppose everyone would agree
that individual tokens can be and are "fudged" (i.e., in production) and/or
misperceived because of noisy channels, etc.  I take that to be an aspect of
"parole", "performance" or whatever, and assume that the underlying or
"stored" forms of these same words do have some sort of determinate
phonological shape.

Beyond that, I suppose one's answer depends on the version of phonological
theory one subscribes to.  Praguean orthodoxy specified one answer to the
neutralization question and American Structuralism another.  Various
generative phonologies have tried to combine the two to different degrees.
But I think all agree on some sort of distinctive (underlying in some sense)
phonological representation for all words.  And with the exception of some
instances of sound symbolism and loanword phonology I think most would agree
on an underlying inventory.  Interestingly though, in the development of
generative phonologies of the '60's and '70's (i.e., before I became bored
with them and went back to strictly diachronic work), there was no place in
the phonology of a language itself where an inventory of phonemes,
underlying or otherwise, was specified.  There were only underlying
representations and rules that operated on them.  The matrices with all the
little pluses and minuses in the various textbooks were not, in fact, part
of the "phonological grammar" of the language.  They were just there for the
linguist's reference.

This said, I'd be surprised to find a native English dialect where [I] and
[schwa] were really allophonic.  Neutralization in "parole", yes, but
probably not in "langue."  I could be corrected on that point however,
since, as I've said, I'm not an Anglicist.  I just speak the language by
accident of history.

Bob Rankin
Linguistics Dept.
University of Kansas
Lawrence, KS