Re: THEORY: Case mismatches (was: Re: Viko Notes)

On Wed, 26 Jun 2002, John Cowan wrote:

> Marcus Smith scripsit:
>
> > This topic was part of my MA thesis, so I have lots more I could say about
> > this.
>
> Cough up already, then!
My thesis is broadly concerned with instances of 'unbalanced
coordination', 'partial agreement' in particular. By 'unbalanced
coordination', I mean contexts where [X & Y] is grammatical, but [Y & X]
is not. The example I start off with is in Pima:

Napt aapi c heg Eric hen~-ees heg hen~-kalit tako.
Q-2sg-pf you and DET Eric me-steal DET my-car yesterday
`Did you and Eric steal my car (from me) yesterday?'

but just by swapping the order of the two subject conjuncts, you get an
ungrammatical sentence.

*Napt heg Eric c aapi hen~-ees heg hen~-kalit tako.

The reason it is ungrammatical, is because the initial auxiliary must be
inflected to agree with the first conjunct of the subject. That condition
is satisfied in the first example, but violated in the second.

The case facts that started this thread are another type of unbalanced
coordination, because the case of the conjuncts depends on the ordering of
conjuncts (among other things). "him and I" is good, but "I and him" is
bad.

There seem to be two broad patterns to the ordering you find across
languages. In some languages, the "deviant" conjunct (the one that does
not participate in triggering agreement or gets the incorrect case) is
typically ordered with respect to the conjunction in the same way as the
object to the verb. That is, in OV languages, the conjunct immediately
before "and" is deviant, but in VO languages, it the conjunct immediately
after "and" that is deviant. This pattern was first noticed (to my
knowledge) in a 1993 dissertation from Oslo by Johannessen.

Corbett, in his work on gender and number, notes that agreement with the
closest conjunct is very common. So unlike the previous pattern, the
"well-behaved" conjunct is the one that is closest to what it is
interacting with. That is, if the verb (adjective, etc) agrees with just
one conjunct, then it agrees with the closest; if only one conjunct has
the proper case, it is the one closest to the case assigner.

Some languages follow the Johannessen pattern (like Norwegian), but others
follow the Corbett pattern (like English, probably). But in most
languages, you can't tell which pattern is being followed. If you look
back up at the Pima example, the conjunct triggering agreement is the
closest (as Corbett predicts), but the deviant conjunct is where I would
expect it to be in Johannessen's analysis.

The way to distinguish the two patterns for Pima, then, is to change the
word order, which is normally perfectly legitimate: it is a
"non-configurational" language. The surprising result is that it is not
possible to have partial agreement and a subject initial sentence. If both
the Corbett and Johannessen pattern's are not obeyed, the entire sentence
is ungrammatical. Other languages that appear to pattern like Pima are
Arabic and German.

I'll spare you guys the theoretical explanation I give. It's couched in (a
modified version of) the Minimalist Program, mainly for political reasons.
(66.6% of my committe are rabid minimalists.) My analysis is closer to
Categorial Grammar, though. (MP, CG, what's the difference these days? The
older Chomsky gets, the more his stuff looks like CG and TAG.) And
obviously I've glossed over tons of details: there are animacy heirarchy
effects (which I don't really discuss, just note), different semantic
interpretations, and all kinds of other nutty confounding factors. This
gives the basic ideas, though.

Marcus

Reply