Why syntax isn’t about word order.
Please note, before we start, that I think syntax is primarily about word order. I’m just explaining what some people think.
The point in syntax, more or less, is that some word orders are okay, and some aren’t. (Even in so-called free word order languages, they’re never entirely free.) In English declaratives, for instance, it has to be subject first, then the verb, then the object(s). So we call English an SVO language. Japanese is SOV, Irish is VSO, and Malagasy is VOS. SVO and SOV have around 45% of the languages each, VSO 5% and VOS also 5%. There are, I believe, OVS and OSV languages, but they’re rare (and often debated about). (I’m pretty sure my numbers are wrong here, but I don’t remember the right ones, exactly — SOV is probably more common than SVO, and VSO than VOS. I’m more or less right, though.)
So let’s use English, because we all know English, and at this level, we can describe pretty much what we need with just English.[1]
Some theories of syntax involve movement. GB (government binding) and its latest incarnation, MP (the minimalist program) are crucially about movement. Movement is involved (in these theories; there are also movement-free theories) in things like questions.
Let’s look at a basic obligatory transitive verb. I’ll stick with tradition and use hit. I’ll break with tradition a little bit, though, and not use John and Mary — I’ll use Jaime and Lee. An asterisk at the beginning ofa sentence means it’s ungrammatical.
(1) a. Jaime hit Lee.
b. *Jaime hit.
c. *Jaime hit Lee Pat.
You need the object, and you need just one object.
What happens when we make it into a question?
(2) a. Who did Jaime hit?
b. *Who did Jaime hit Lee?
We can’t have an object anymore. But the who (I don’t say whom except directly after prepositions) is clearly meant to be the object of the verb — if you have case there, it is accusative, and not nominative. So the argument is that who started, at some level, where the object would have been, and moved to the place where we see it.
Now there’s tonnes more movement than this in GB/MP syntax. Some is, I think, useful, but much just seems like movement for movement’s sake. However, my reservations about these aside, I do like the idea of movement (or anything to the same effect) in syntax.
The other part that matters here is that we have two parts to sentences — the sound part (PF) and the meaning part (LF). The names don’t matter so much. But we need those two, on some level: sentences sound like something, we say them out loud (or can), and they mean something, because that’s sort of the whole point.
The idea is that sometimes, the words are in a different order in the sound part and the meaning part. So you can say something like Everyone didn’t eat cake and that could mean that every single person ate food that wasn’t cake or it can mean that not every single person ate cake. (Trust me. People use it to mean the second a lot. I listen for it.) So in PF, everyone comes before not; at LF, it might be a different order.
The theory then says that there’s something like quantifier raising, but it happens only at LF, not at PF. Fine; we have a fairly decent idea of the restrictions for movement at LF, inasmuch as there ever is. We want PF to mostly involve expletive insertion, and all of phonology, and possibly also morphology. But — importantly — not syntax. (Now, you might not like this theory. That’s fine. Lots of people don’t. I haven’t explained it well, either.)
When I was writing my honours thesis back at Canada U, I went to a professor, S (not my advisor!), to ask about something, because at that poitn I could get either the scope facts or the word order facts, but not both. (I eventually got things to work out okay, mostly by changing my topic.) And her response — a syntactician, mind you — was “don’t worry about word order, you can always move it around at PF.” No. No you can’t. You really can’t.
Not that she isn’t in good company: Chomsky’s Minimalist Program suggests that “[t]he best answer would be that the order is really (205a) throughout the . . . computation. . . . We thus take the output to be really (205a), irrespective of what is observed at the PFoutput.”(from Ora Matushansky’s site)
[1] At this level. I don’t care what Chomsky and MIT grads say, English — or any language — does not exemplify every single interesting structure in language, and we can’t figure everything out from it, no matter how smart we might be.