Marcus:
> Jesse Bangs wrote:
>
> > > That depends on your theory of "Gen". Prince and Smolensky
> actually provide
> > > two different models. One - the one nearly everybody uses - generates all
> > > possible forms, then selects the most "optimal" form based on the
> > > constraints. The other only generates the forms relevant for the next
> >
> >Huh?  This implies that Native Speaker X, before he says anything, stops
> >to generate every single conceivable combination of words and then applies
> >a bunch of constraints to rule out all of the illegal forms.  That gets
> >impossible pretty fast.  Do I misunderstand "Gen"?
>
> No, you have the right idea. That disturbed me at first too, but it really
> is very simple, and not at all impossible. All you have to do is realize
> that there is a finite number of things that can be done to change a word,
> though these changes can be combined in an infinite number of ways. Once
> you determine that a given change is suboptimal to another one, you can
> immediately dismiss any candidate that involved that change. Dismissing one
> or two such suboptimal processes reduces the possibilities to a managable
> size very quickly. Given the proper algorithm, it is possible to deal with
> infinite sets without any difficulty whatsoever. (And there are many such
> algorithms already in use.)

I ought to defer to Marcus, on the grounds that he surely knows better than
me what is and isn't the right idea, but all the same, I think you both have
the wrong idea.

What happens is that Native Speaker X, before they say anything, stops to
choose one out of all the well-formed structures (=sentences) that comprise
the language. These well-formed structures are those that, out of the
infinitude of things generated by Gen, satisfy -- or best satisfy -- the
constraints of the language.

In other words, (i) Gen has no explanatory role in the theory, and (ii) the
theory of what defines the well-formed structures of the language is
distinct from a theory of how a speaker chooses among these structures.

As Marcus and I seem to have been agreeing, a serious mentalist ought to
reject (ii). But I think it is a fact that (ii) is the standard position in
principle, and that deviations from it would tend to be lapses into confused
thinking or else fundamental reinterpretations of the foundations of the
theory.

> And if you assume a theory of Gen that does not allow you to generate the
> most rediculous candidates, then this problem doesn't even arise.

You then get a potentially arbitrary division of labour between Gen and
the constraints.

--And.