Re: THEORY: phonemes and Optimality Theory tutorial

On Thu, 9 Nov 2000, jesse stephen bangs wrote:

> dirk elzinga sikayal:
>
> > [snip]
> >
> > A way of thinking about the phoneme is to consider it to be the
> > minimal unit of sound which serves contrastive function. Thus any
> > feature or property of a sound which does not function contrastively
> > may not be part of the phoneme. This actually allows quite a bit of
> > latitude. In Shoshoni for example, the 'phoneme' /p/ is realized
> > variously as [p], [b], [B], and [F] (the last two voiced and voiceless
> > bilabial fricatives, respectively). So what is necessary for the
> > Shoshoni speaker? Not the fact that /p/ is voiceless, since there is
> > no /b/ which contrasts. Not the fact that it is a stop, since there
> > are no /B/ or /F/ which contrast. Only the fact that it is bilabial
> > and oral (rather than nasal; there is a contrasting /m/) seems to be
> > relevant.
>
> How is this different from the normal feature-marking theory.  We would
> just say that the features [+labial, -nasal] are enough to distinguish
> /p/, with the rest of the features like [-voice] specified by
> language-specific rules for filling out a pronounciation.
It isn't different from the "normal" feature-marking theory. The
underspecification debate once upon a time was about how fully
specified underlying phonemes really are. Are all features and their
values (even redundant ones) present underlyingly, or just
unpredictable ones? If not all features or their values are present
underlyingly, then how are they filled in? Are only the redundancies
which have to do with noncontrativeness omitted (contrastive
underspecification; see Steriade 1995), or are *all* redundant
features and values omitted, even ones that are active phonologically
(i.e., can be potentially contrastive: see Archangeli 1988)? The idea
isn't as simple as it seems, and it may rest on a faulty understanding
on how humans actually encode information. In many biological systems
there are numerous redundancies built in; why should language be
different?

> > [snipped discussion and examples]
> >
> > Since the choice between underlying /p/ and /b/ doesn't seem to
> > matter, either one's status as a phoneme in the structuralist or
> > generative sense seems to be questionable. Thus Optimality Theory
> > elevates the "non-uniqueness problem" to the status of a grammatical
> > principle by making requirements only on surface forms and takes the
> > wind out of the phonemic sails.
>
> This seems silly, though.  Are you suggesting that the native speaker
> really doesn't know whether the sound is [p] or [b]?  That the speaker
> doesn't know whether the word is "really" /nampa/ or /namba/ since the
> surface forms would be the same anyway?
Not at all. Native speakers will definitely have intuitions about
this. I *think* my consultants considered the underlying sound to be
/b/. I say this because I asked them to write out a couple of words
for me, and they consistently used <b> rather than <p>, even
initially where the phone is voiceless. I'm hedging because I suspect
some English orthographic interference. Initial voiceless stops are
unaspirated and therefore resemble English "voiced" stops more than
they do English voiceless stops. I haven't trained them to transcribe
their own speech using a consistent system.

When I say that either [p] or [b] could be selected as the phoneme,
what I mean is that the choice of [p] or [b] isn't forced by the
theory; the theory can allow either. It's up to you whether you
consider this to be a failure or fortunate result. One the one hand,
it offers a potential explanation for sound change: some speakers
"phonemicize" [p], others [b]. Whichever group gains linguistic
dominance gets to "determine" the next generation's grammar. Under a
strict phonemicist position, this explanation for sound change is not
available, since one or the other *must* must be chosen.

> I don't see the advantages over
> the traditional generative view here.  How does Optimality Theory deal
> with the famous writer/rider problem?
I have never seen the 'writer/rider' contrast verified instrumentally,
and I doubt that it exists now (if it ever did). But the potential
problem that it raises for OT is a real one; it is the problem of the
intermediate representation. There are several ways of dealing with
it, none of them very satisfying.

Dirk

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