Re: Phoneme analysis. (Motherhood statements?) (Long!)

John Atkinson wrote:

I agree completely with these analyses, which are, as we both agree,
equivalent functionally.

I don't agree with that. (3) requires a distributional
constraint on /r/ which (4) does not require. It is not
economical (for in a complete analysis I would need to
give a rule to describe that constraint). This is the
only point I was making.

....
One reason why I don't automatically favour type (4) rules over type (3)
rules, as you appear to:

No, I was not contrasting the insertion and deletion rules.
Perhaps there is a contrast to be made there, but that
would be another topic. Perhaps one can find effects of
the lost r on the preceding vowel, but that's also another
topic. I was only contrasting distributional constraints on
the underlying system.
....
since as as black-box models (3) and (4) are completely equivalent, it's

Not.

really not necessary to decide between them.

Greg

.

Quantcast