Jackson wrote:
It strikes me that perhaps the original coup of generativism wasn't  
so much the conceptual framework but the theory of movement found in  
'Syntactic Structures' - indeed, that the bravura of this achievement  
is what lent force to Chomsky's polemics.  But generative models do  
lend an intuition of the range of syntactic permutation available, in  
the sense that they render a series of potential landing sites for  
syntactic constituents and features which trigger movement.  One  
could even tweak them so as to construct movement not found in any  
natural language.  It would only remain to translate such rules into  
learner-friendly terms.  Has anybody looked into this, or is there a  
general sense that variable word order is unwieldy?

A couple things.  First of all, saying that it *is* movement is a
theoretical claim.  In order for it to be movement, there must be
a basic word order from which others are derived.  As far as
I'm concerned, there's no atheoretical reason to assume that this
is the case.

Consider an analog from Arabic morphology, for example.
You have triconsonantal roots that fall into various patterns to
derive words


kaatib "writer"
kitaab "book"
kutub "books"
kataba "he wrote"
maktab "office"
maktaba "bookstore"

Then the various other triconsonantal roots fit into these patterns
(and others) to derive semi-similar words (so t/?aalib is a student
[human agent], ZuTuT is corpses [plural noun], Gidaa? is lunch
[non-human singular noun], etc.).  Applying the same movement
principles to this system (which is essentially what distributed
morphology aims to do), one would expect to find a basic word
from which all others are derived.  This seems to be counterintuitive
both from a theoretical perspective, and a language-internal
perspective.  It's clear that all the words are related to writing,
but saying, for example, that all words are derived from the verb
"to write", and trying to create movement rules to do so, would
kind of be missing the point.

Back to syntax, it may be the case that there is no basic word order
from which all others are derived, but that there are configurations
which are used to achieve particular meanings (for example, putting
a verb first in English indicating a question).  While it seems  
that, say, an assertion is basic, and a question is somewhat more
"marked" (whatever that means), what about other cases?  Consider
the following (note: I apologize that some of the following
sentences are a bit crude):

What did you do for the fireman?
What did you do the fireman for?

These two mean *very* different things, but the following two don't:

I took the trash out.
I took out the trash.

The following can mean two different things, but don't have to:

The electrician screwed the lightbulb in.
The electrician screwed in the lightbulb.

And with pairs like the following, it can seem like one is better
than the other to certain people on certain days, but which one
that is seems to vary without rhyme or reason:

I locked the store up.
I locked up the store.

And to me, at least, it seems like the first is in the following pair
is preferable to the latter:

I locked the box away.
I locked away the box.

In the cases where the meaning differs, Chomskyan syntax must
propose a different syntactic structure, and in the cases where the
meaning is the same, the syntactic structure must be the same.
Either that, or one needs to propose different lexemes.  Can it
solve the problem?  Of course.  There is no problem that a syntax
with movement like, say, the Minimalist Program can't solve.
For any problem that arises, you can either propose a movement
operation, or, if it isn't licensed, create a projection and/or feature
which licenses the movement, and, if you don't want to do that,
simply create a new lexeme.  In fact, I believe this was the original
solution to all passives: every verb that could be passivized had
two entries: one, an active entry, and one a passive.

With this much machinery, I'm often left wondering what such
analyses are explaining, and if there's any real insight at all.  It all
seems to be based on the idea that there is a single underlying
form for all utterances, and that certain utterances (like questions)
are based on underlying non-questions.  With the examples above,
which is the underlying form: preposition/adverb first or last?

Also, as to your last question, there *are* languages where word
order is permuted much more to indicate various things.  I don't
have any ready-made examples, unfortunately, but they do exist,
and may be in my typology notes somewhere.  Of course, there
are the Australian languages, some of which are claimed to have
no basic word order whatsoever, but what seems more interesting
are languages where specific phrasal constructions are used to
in particular contexts.  For some reason, Chinese is coming to
mind.  Any Chinese speakers have any insights?

Anyway, it's my guess that the idea of an underlying word order
is a byproduct of the fact that Chomsky and those that followed
were primarily familiar and concerned with Indo-European
languages.  Indo-European languages aren't wrong, or anything
like that, but they are alike structurally in many ways, and if
one tries to describe the structure of language in general based
on a typologically impoverished sample, one will run into
problems when encountering unrelated languages.  At that point,
the decision must be made: Do we change the theory, or make
the data fit the theory?  More often than not, it's the latter route
that syntacticians take.

"sunly eleSkarez ygralleryf ydZZixelje je ox2mejze."
"No eternal reward will forgive us now for wasting the dawn."

-Jim Morrison