|From:||David J. Peterson <dedalvs@...>|
|Date:||Friday, March 31, 2006, 20:53|
In designing a language, one could systematize the relationship
between word order and grammatical meaning by defining such particles
alongside pronounced lexemes.
This is true. In fact, if one isn't bound by movement, one could
define hundreds of particles which do any number of things. So,
for example, if (a) is an assertion, then you could have a
null particle at the end of the sentence which forms the question
(a) ten me ga loka.
/I saw a duck/
"I saw a duck."
(b) ga me ten loka?
/a saw I duck Q/
"Did I see a duck?"
You could also give the particles a phonological form. The result
would be that word order would be meaningless until you heard
what the particle was (which, I suppose, would probably necessitate
placing those particles initially--practically, at least).
It seems that the underlying premise of theories of movement is that
grammatical relations are local.
I believe, if I'm not mistaken, that the assumption underlying
this premise is that grammatical relations are assigned (or inherited,
or achieved with some sort of merge operation, etc.). This is why
grammatical relations have to be local. Given a tree structure, there's
no way that a verb can assign accusative case to an object in a
VSO language. Therefore, there must be a movement operation,
so that at some level the verb can assign accusative to the object.
It can only do this if it's in a command relationship, and it can only
be in a command relationship if it's, in some way, next to the verb
(either a sister to it, or somewhere further down in the tree). If
you take away the assumption that grammatical relations are
assigned, though, the movement becomes unnecessary.
I suppose I'm kind of leading myself to construction grammar,
without ever having formally studied it, so I guess I'd better
leave off. I think the general point remains true, though: any
theory that makes any prediction about language can be implemented
in a conlang to sort of test the predictions made (it's a wonder
that linguists haven't figured this out, by and large). Thus, if
particles that drive movement are proposed for a language in
a limited environment, it's interesting to see what they can do
in an expanded environment in a created language, where you
can create the data, as opposed to having to find it.
For example, one unfortunate prediction that was made by
Marit Julien's analysis of Northern Sami causative constructions
is that a subject licenser (this would be AgrS) could never be
generated without also generating an object licenser (AgrO).
This handled the problem she created for herself with her
analysis, but also predicts that every intransitive verb is underlyingly
transitive. Presumably, they would all have to undergo some
sort of object deletion process. Either that, or it can be argued
that all intransitive verbs have an object, most of the time it's
null, and it gets accusative case. Thus, there'd be no difference
between transitive and intransitive verbs. The interesting thing,
then, would be to create a language where all intransitive verbs
take objects optionally--or even mandatorily. For example,
imagine a language that agreed with its subject and object in
person, with markers like this:
1:2 = -me
1:3 = -ta
2:1 = -ki
2:3 = -lo
3:1 = -nu
3:2 = -sy
3:3 = -fi
Then for all intransitive verbs, you'd have to decide on the
person of the object, even if one didn't exist. A language like
this might lend support to Julien's claim.
Anyway, point well taken. The Dutch examples have been
very interesting. I bet the analysis would be that Dutch is a
verb-final language, and that there's something which requires
the verb to move to the second position, stranding the preverb
(or preposition) behind the NP. This is how you get differences
between "in" being behind "de stad" and in front of "de stad".
I imagine the verb will have to move to get tense and/or
"A male love inevivi i'ala'i oku i ue pokulu'ume o heki a."
"No eternal reward will forgive us now for wasting the dawn."