Re: THEORY: phonemes and Optimality Theory tutorial

On Wed, 15 Nov 2000, jesse stephen bangs wrote:

> dirk elzinga sikayal:
>
> > My analysis directly violates the Biuniqueness Condition. As I see it,
> > the phoneme--whatever its definition--had the following properties:
> >
> > 1)      it was segment-sized; that is, it was not decomposible into
> > features, prosodies, or elements (though it was characterized by
> > having certain properties such as labiality, voicelessness, etc);
> >
> > 2)      it was the unit of speech which enabled the expression of
> > opposition and contrast, and it was embedded in a system organized by
> > such oppositions and contrasts;
> >
> > 3)      it was part of representations which uniquely determined
> > phonetic forms; likewise, phonetic forms were analyzable into
> > sequences of phonemes (this is the Biuniqueness Condition).
>
> These all describe the classical phoneme, which no one is defending.  To
> me, the failures of these descriptions isn't evidence for the death of the
> phoneme, but proof for a need of the redefinition of a phoneme.
This is exactly the problem I was trying to avoid--terminological
looseness. 'Phoneme' has a specific definition (several) with well
established properties. If these properties fail to hold of linguistic
representations, then what you have are not phonemes but something
else. If you want to *call* them phonemes, well, okay. But that wasn't
the original definition, and it's probably better to come up with
another term to avoid just the kind of definitional problem we're
having. I'm not advocating the abolition of underlying or abstract
levels of structure; I'm just saying that the term 'phoneme' is no
longer applicable for current views of such representations.

> My argument is simply that the phoneme should be
> redefined.  I consider the phoneme to be a
> collection of features which characterize its particular phones,
Characterize or constitute? The American structuralist phoneme was
characterized by "features" of pronunciation without be composed of
them.

> the
> unit from which lexical and grammatical entries are made,
I'm assuming you'll allow more complex (non-linear) structures as
well...

> and the starting
> point for phonological rules.
In a theory without rules this wouldn't characterize the phoneme;
American structuralism was largely a theory of representations, not
rules; Early Generative Grammar was a theory of rules.

> It is *not* atomic (i.e. indivisible), not
> unconditionally unique, and not the lowest level of information.
I don't think the phoneme was ever considered to be the lowest level
of information; there were also morphophonemes which are even more
abstract than phonemes. As for phonemes not being atomic, this is true
of segments at all levels of representation; the biuniqueness
condition was implicitly or explicitly an ideal of phonemic analysis.

> It is
> one of the many intermediate levels between thought and speech, and one
> which is particularly useful for study.
It *is* useful, since there are potentially interesting things to be
said based on a phonemic analysis. I wouldn't argue that. Nor would I
argue that there are no underlying phonological representations (well,
I actually argued just that in a little paper I wrote on Fula
consonant mutation, but that's another story). I would argue though
that underlying segments are not phonemes, since 'phoneme' has several
explicit definitions, all of which share well understood properties in
the American structuralist tradition. These properties have since been
shown not to be true of phonological representations at any level,
therefore the term 'phoneme' is inapplicable for elements of such
representations.

Dirk

--
Dirk Elzinga
dirk.elzinga@m.cc.utah.edu