Re: THEORY: OT Syntax (Was: Re: THEORY: phonemes and Optimality Theorytutorial)
|From:||Marcus Smith <smithma@...>|
|Date:||Thursday, November 16, 2000, 9:02|
Nik Taylor wrote:
>Marcus Smith wrote:
> > So when I say those theories are
> > unlearnable, it means that a child would not be able to learn the language
> > if those theories are accurate models of grammar.
>How can one prove that a theory is learnable or unlearnable?
I don't know the exact details. I'm not taking the course until next
quarter. But as I understand it so far, it has to do with how many ways the
grammar allows one to create the exact same sentence. Take for example,
transformational grammar. Under this approach, the entire sentence must be
constructed in its "underlying form" then various movements occur to
satisfy the necessary requirements and word orders. The problem is that
there are countless ways you could do the movements to achieve the exact
same end result. These movements actually have consequences in areas where
you would not believe, and that are too subtle to be noticed without
somebody explicitly pointing them out. Even worse, there are no principles
to specify why many of the changes take place -- they just do.
For example, there are two interpretations of "Someone loves everybody" and
four ways to achieve them. Each interpretation is achieved by different
movements, though the sentence that is actually pronounced does not reflect
those movements. How is that supposed to be learned? Chomsky (and many
others) says this is proof that Language is innate. (If he doesn't, he has
to admit that he is wrong, and Chomsky never admits to being wrong, even
when he is.) Others say it is proof that the theory doesn't work as stated.
I tend to fall in the latter camp.
So basically, if there are two many possible ways to parse a given
sentence, then the child will have no way to figure out which one it is.
The amount of work it would take to solve one sentence that has more than 6
words would be extraordinary -- not even taking into account the fact that
they don't know what the individual words are yet! Then semantics on top of
the syntax. Ouch!
Minimalism, TAG, CG, and OT severely restrict the possible derivations, and
they give principled reasons for each step. Nothing is random, unlike the
"When you lose a language, it's like
dropping a bomb on a museum."
-- Kenneth Hale