Philippe wrote:

<<That's very interesting. I read something like that
before, but I didn't quite realize the implications.
Do you mean for ex that for these two sentences:
- I'm sneezing
- I sneezed
the ergative is used only for the secund case (there
was a sneeze from/to me ? Or is it a bad example ?
Maybe the 1st person behaves another way ? Do you have
a better example ?>>

No, not quite.  The ergative/narrative case marks the subject of a
*transitive* verb in the...uh....series 2 screeve (I'll explain in a minute).
Here's an example:

(1)
k'ats-i  roman-s ts'er-s
/man-NOM. novel-ACC. write-3obj./
"He writes/is writing a novel" (a third person subject is indicated by
null marking in this screeve)

(2)
k'ats-ma roman-i da-ts'er-a
/man-ERG. novel-NOM. preverb-write-3rdobj./
"He wrote a novel."

As you can see, the nominative case doubles as the absolutive case in the
past tense.  This is exactly what you'd expect from a split-ergative system
where the split is based on tense.

Now let me explain a little about what's above.  I just gave a presentation
in my grad. morphology class on the verbal agreement patterns of
Georgian, so it's fresh in my mind.

Georgian has eleven tense/aspect distinctions, but they're grouped together
into three "screeves" (which means nothing more than "rows" in Georgian)
based on their agreement morphology.

Going a bit further, Georgian has four classes of verbs, all grouped together
based on *their* agreement patterns (and loosely on their semantics).  The
verb above, "to write", ts'er, is a Class I verb.  Most transitive verbs are class
one.  Class one verbs show three different agreement patterns, based on
screeve series.  So, the first series of screeves include:

(1) The Present
(2) The Imperfect
(3) The Present Subjunctive
(4) The Future
(5) The Conditional
(6) The Future Subjunctive

So that's six of the eleven.  If you want to conjugate a verb with one of those
six tenses, then you use a particular set of verbal agreement markers.  Each
screeve, of course, will also have its own morphology, though it's not pre-
dictable.

Once you've got that set, you need to find out how many specified arguments
the verb has.  If a verb has an indirect object present, it will agree with the indirect
object.  If it doesn't, it will agree with the direct object.  Finally, if there's no direct
object, it will agree with the subject.  The agreement markers are various suffixes
and prefixes.  And if it's the case that, say, the indirect object agreement marker is
a prefix, and the next in line (the direct object marker) is a suffix, then you can use
both.  But an indirect object prefix will block the application of a direct object prefix.

Next, you have to deal wtih nouns.  The series I screeves are simple, because nominals
get case-marked in a traditional nominative/accusative pattern.  In other words, the
subject gets the nominative, the direct object the accusative, and the indirect object the
dative.  (Oh, remember: The dative and accusative are identical.)

So that's series I.  Series II is different.

Series II is comprised of the following screeves:

(7) The Aorist
(8) The Optative

The aorist is used as the completed past tense.  Why's it called the aorist then?  I don't
know.  Anyway, that's what we wanted to show above, and example (2) is in the
aorist tense.

First, one of the things about this particular class I verb in the aorist is that it takes
the preverb "da-".  The preverb is meaningless: It just appears in certain screeves.
It's not an aorist tense marker, because it appears in lots of other screeves.  Basically,
if you look at the whole Georgian verbal paradigm, what's striking is that the same
exponence, or phonological forms, are used over and over again over all the tenses.
The important thing, though, is that each screeve is distinct from each other screeve.
So it doesn't matter that "da-" doesn't mean anything or mark anything.  All it does is
make one tense different from the next (combined with other markers).  Anyway, so
first you add "da-" to the verb "ts'er".  It will appear before all agreement markers.

Next, we need agreement morphology.  The given example above has a direct object
that's present and a subject that's present.  The third person direct object (in this series)
is marked with an /-a/ suffix.  The third person subject is not marked, so that's all the
agreement morphology you get, and you get /dats'era/.

Next, we have to figure out how the nominals get case-marked.  A peculiarity of the
second series is that it's marked in an ergative/absolutive way.  So the indirect object
is still marked with the dative, but the subject is marked with the ergative/narrative,
and the direct object is marked with the nominative.

And there you have it.  The third series gets even more complicated.  And then there
are class II verbs (which aren't bad), class III verbs (which are *very* confusing), and
class IV verbs (which are even worse).

Anyway, it's a really neat system, and I think a good argument for Word and Paradigm
Morphology, as opposed to whatever Chomsky's doing right now.  (I forget the name.
It starts with a "d".)

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

-Jim Morrison

http://dedalvs.free.fr/