Re: THEORY: phonemes and Optimality Theory tutorial

Marcus:
> And Rosta wrote:
>
> >One can't really argue with what you find more intuitive and realistic.
> >But does not all rationalistic theorizing strive for a reductionist
> >approach, in the sense of one that, other things being equal, is maximally
> >simple? Or by 'reductionist' do you mean 'simplifying to the extent of
> >ignoring significant differences and contrasts'?
>
> What I took a difference with is the conclusion that E and O are formed
> from A+I and A+U respectively. Under your analysis, there is A, I, U which
> can occur anywhere. In stressed syllables, the language allows A+I = E and
> A+U = O. So you have a phonotactic constraint that says A can combine with
> I and U only in stressed syllables. I, on the other hand, would say that
> the language has A, E, I, O, U, but does not allow E and O to occur in
> unstressed syllables. Which approach is better? I don't see any objective
> way of deciding in this case.
The essential idea behind my proposal is that E and O are more complex than
A, I, U. Only strong syllables can support this additional complexity.
(That, btw, is a familiar finding in languages -- that extra complexity
(such as double moras) is supported more strongly in strong than in weak
syllables.) Paradigmatic complexity is reanalysed as a kind of simultaneitous
syntagmatic complexity.

> But try the classical problem in English that /N/ and /h/ are always in
> complementary distribution. The only sensible thing to say is that N and h
> are separate entities that accidentally are not contrastive. You would not
> want to derive either one from other units. You simple state that N cannot
> occur word initially and h may not occur word finally. This partially
> parallels the case above, where E and O do not occur in unstressed
> syllables. I would suggest that you should use the same methodology in both
> cases.
My analysis does say "E and O do not occur in unstressed syllables", but has
the added virtue of explaining why.

As for the /N/ /h/ problem, this is an artefact of that problem with trad
phonemics that tries to establish contrastive segments irrespective of
their syntagmatic position. Once you in effect treat position in the syllable
as a distinctive feature, there ceases to be any temptation to treat /-N/
and /h-/ as the same entity.

> Basically, I don't think "maximally simple" is desirable. I think
> simplification should be taken to a reasonable level, which is, of course,
> hard to define and different for everybody. If you could find good evidence
> that E and O are A+I and A+U, then I would go along with it. But under the
> simple situation you presented, it is reduction for the sake of reduction;
> the system is not simpler or better because of it, except on a purely
> subjective level.
I would like to think that it is possible to have a fairly objective measure
of simplicity, even if it doesn't always accord with first impressions.

Anyway, I understand your point. As for evidence that E and O are A+I and A+U,
that would take us off into an entirely new thread, so I won't go there.

--And.