Re: THEORY: Ergative syntax

While not exactly a "real linguist," I'll try my hand at these questions
anyway.

> My question is: in an ergative language, which case does the verb assign to
> its internal object and which to the external?
Well, your question is a bit tricky to answer because the current Chomskyan
syntactic theory doesn't have case assigned from verbs anymore (I'm not sure
about other theories, but soon, I'll know--Lexical-Functional Grammar's
coming up next semester).  The other problem is that there is not a whole
lot of work on case in ergative languages in the formalist theories,
especially in the older frameworks.

Now interestingly, the Chomskyan framework did end up positing a hybrid of
your two approaches (but for a nominative-accusative language)
A sentence like 'John sees Mary' is based-generated as:

S[NP[_] VP [John V'[sees Mary]]]

and the subject moves to the higher NP spot by transformation to get the
surface structure:
S[NP[John] VP [t V'[sees Mary]]]

This, the VP-internal hypothesis, was created to account for sentences where
the subject stayed in the VP like in:  S[There is VP[someone knocking on the
door]
(There's another class of sentences that also supported the argument, too,
but I'll have to gloss over them, because they would require too much
background explaining)

The problem of having an empty NP dominated by S also ended up not being
that big of a problem. Some sentences (like: There is someone knocking on
the door) were known to have semantically empty subjects--expletives.
Expletives were deemed to be insertion, especially since the position could
also be filled with semantically full subjects (Someone is knocking on the
door).  So the theory was changed to say that you have to have something in
that NP node (which came to be Spec, IP as S became a phrase [IP] like the
others) only on the surface.  The rest of the time, that position could
merrily be empty.

So in the end, due to the VP-hypothesis, the subject was called the external
argument and the object was called in the internal argument, but whether
such distinctions can or should apply to ergatives and absolutives, I'm not
sure.

[snipping Jesse's suggested analyses, as to keep the length under control]

> Also, are the above trees correct?  I wondered if it might be more
> correct to have CP > S C and S > VP NP in a head-last language.
The trees are only correct for whatever data the language provides.  If you
have complementizers coming at the end of the sentence, then to correctly
capture that fact, the rule would have to be CP->S C.  And yes, to be truly
head-last (in theory) you'd have to go CP->S C.  But languages aren't always
purely head-final or head-initial--German, for example is claimed (by some,
at least) to be head-final, yet has complementizers before the sentence.

> Also, if CP > S C *is* correct, then how should I explain the ergative-first
> feature of Hiksilipsi?
My advice to you would be to not worry so much making Hiksilipsi fit a
theory so much.  Theories can be wrong, and they are prone to change.  My
advice would be look at the data--plenty of languages have the order
Erg-Abs-V, so in that data, there seems to be plenty of inductive evidence
for what you already have.  If you want an answer that "yes, it has to be
that way," then you'll have to start looking the theoretical literature, but
I doubt that there currently exists any definite answer that is "Yes,
verb-final languages have to be Erg-Abs-V and here's why..."

But a good reference to start looking at Ergativity would be:
Manning, Christopher. 1996. _Ergativity: Argument Structure and Grammatical
Relations_. CSLI Publications, Stanford, California.

Although it does require a bit of syntactic background, you could probably
get something at least out of looking at the examples.

If you're at a university and can use one of the wonderful databases that
can find newer books/articles that older books/articles are cited in, you
might check that out, using the Manning book and Dixon's Ergativity book as
the older books.  The reference on the Dixon book is:
Dixon, R.M.W.  1994. _Ergativity_. Cambridge Studies in Linguistics 69.
Cambridge University Press.

> Now, in Hiksilipsi this is relevant because I want to restrict
> NP-exraction in relative clauses, but I was flummoxed as to which
> argument should be extractable.  If I'm correct, and the absolutive is
> the external argument of the verb, then it makes sense that only
> absolutives are extractable in relative clauses.  Is this correct?
Keenan and Comrie (1977) found that subject were the more commonly extracted
than direct objects, but my source (not the actual Keenan and Comrie
article, but _Describing Morphosyntax_ Payne 1997) doesn't mention anything
about absolutives and ergatives.  In looking at Mulder (1994) (_Ergativity
in Coast Tsimshian_), she mentions that Yidiny and Yup'ik Eskimo only allow
absolutive extraction, but there aren't any example sentences and it's
pretty late (for me), so I don't want to make any definite pronouncements.
But you might want to check out grammars of those languages to see the real
story.

Doug Ball
http://tsketar.tripod.com/skerre/
CLI v1.1 !l§ cN:R:S:H a++ y n2d:7 B? A-- E L* N1 Is/m/p k-- ia+ p+ s++ m--
o+ P- S Skerre