Re: THEORY: Active case-marking natlangs

Marcus wrote:

> Only if you define "active" by the fact that the subject can change "case"
> based on various parameters. If you define "active" by which cases are
> used, then this is not wierd at all. And if you re-read what I wrote, that
> is exactly how I am defining it. There are four proto-typical roles
> (ignoring ditransitives and experiencers): Subject of a transitive (S),
> object of a transitive (P), subject of an "active" verb (A), and subject
of
> a "stative" verb (O). (Defining "active" and "stative" is not easy, and
I'm
> not even going to try here since it is irrelevant.) Accusative languages
> group these four roles as S/A/O (nom) vs P (acc). Ergative languages group
> the roles as P/A/O (erg) vs S (abs). Active languages group them as S/A
> (active) vs P/O (stative). Tokana and Nur-ellen pull S and A apart and
have
> a system like S1/A1 vs S2/A2 vs S3/A3 vs P/0. This is why I do not
consider
> them active. They are most certainly Split/Fluid-S; but Active is only a
> subset of Split/Fluid-S.
Wow, this is interesting! How come you've never said this before, or rather,
how come you've never said it like this before? I think this is a very
nice though-out distinction. For some reason, I've been stuck with S/A/P
in my head. I've read about S/A/P/O (or similar) but never really thought
about it. I really need to think about this.

> >Interesting!  After all, "to work" is a typical active verb, and "to be
> >big" is as stative as it can be!  These examples make the entire thing
> >stand on its head ;-)
> That's why they're fun. :)
Yay! :)

> >So it boils down to: the choice of A- or P-marking for the subject of a
> >particular intransitive verb is completely arbitrary (fixed, but
> >arbitrary)!?!  At least that is what I read out of it.  It doesn't
> >matter *why* some verbs use A-marking and others use P-marking: all
> >that's important is that both types of markings occur in intransitive
> >sentences ;-)
> Close, but not exactly. There are very strong tendancies based on
> event/state, volition, etc. However, none of these tendancies are without
> exception.
True, but these exceptions can be explained in ways that makes you see
that they were once used regularly according to the overall system.
Lexicalization, grammaticalization, borrowing, analogy.

> This is the root of our disagreement. I define active based on the case
> pattern, you define it on variability. Now, for why I think the case
> pattern definition is superior. Chickasaw is fairly variable in its
marking
> as a result of volitionality, but Mohawk is fairly rigid. Yet both display
> the same pattern of case. Under your definition, you would (most likely)
be
> forced to conclude that Mohawk is not active. This cannot be correct. My
> definition allows Chickasaw and Mohawk to be grouped together (as they
> should be), but only to the exclusion of languages like Tokana and
> Nur-ellen. This is not a bad thing for two reasons. First of all, we can
> use the term Fluid-S to classify all the languages that have variable
> subject case. Also, even nominative-accusative and ergative-absolutive
> languages allow for variability in subject marking. A great many of them,
> for example, allow for dative subjects with psych-predicates. Also, many
> ergative languages are nominative-accusative in the pronoun system or in
> the perfect. Thus, if you define "active" according to variability in the
> case of a subject, it is not clear how you would separate Hindi from
> Chickasaw, which absolutely must be done if the classification system is
to
> be meaningful at all.
I think I agree. This looks nice. Or perhaps it's just because I've
just read an article on fishname taxonomy (don't ask me why), that I
like having a hyperonym "fluid-S" and a hyponym "active".

> That is because you do not work with them. You take generalizations that
> were made by others, and have tried to make sense of the generalizations.
> You have not looked at the raw data and tried to put together a coherent
> picture of your own. I was bothered by these at first too; but now they
> make sense to me. They feel right. When I say "be red" or "be tall" is a
> state, this is because these predicates describe a property of their
> subject argument. When I say "work" or "laugh" is an event, I am saying
the
> subject is temporarily in the midst of some action. Now, what about "be
> three"? This is *not* a property of the subject, this is a temporary
> situation that the subject is in. Thus, in some ways it is more similar to
> an event and it deserves to be treated like an event. But it isn't really
> an event, because nothing happens; therefore it could reasonablly be
> treated like a state. Languages, then, should differ on this point. They
> probably do, though Daniel would probably know more about that than I.
Well, as I've said before, I think the basic distinction in Chickasaw
is control vs. non-control. However, Marcus is the one working on that
particular language, so I'm probably dead wrong. Although, my idea (well,
not _my_ idea, but the one that I see as most plausible) is to have only
_one_ basic distinction for every language: control or P/I/E or event, and
occasionally empathy or significant affectedness for further precision.
I'm interested in knowing why you, Marcus, think "that the marking is
determined by event vs. state, control vs. non-control, and individual vs.
stage level predication." Anyway. I think the explanation of "temporary
situation" is plausible too. The important thing is that you _can explain_
it.

I'm thinking of Central Pomo (again ;-) ). The basic distinction is
control: "to be lazy" and "to live" take AGT and "fall" takes PAT.

Now for the cool part. If you possess an inherent state such as
'tall', 'good', 'beautiful' and 'deaf', you receive an AGT pronoun.
But if an inherent state is marked by an inchoative, the state suddenly
becomes *significantly affected* and so marked as a patient. Confer the
sentences below:

     'I:AGT am old.'
     'I:PAT have gotten old.'

Come to think of it, this is rather an event/state distinction. I've
never really thought of it like that before. Though the weird part is
that it is _the event_ that takes PAT and the state that takes AGT.
The explanation is that inherent states are seen as controlled, but
inchoatives are not. This might be explained by some sort of
lexicalization, but I'd rather want to see it as a basic semantic
distinction, namely significant affectedness.

Another thing that follows from this is that only human beings can be
marked as patients, since non-humans cannot be significantly affected
and one cannot feel empathy for them. Two examples will clarify:

     'I killed him:PAT.'
     'I killed it:AGT (the bee).'

The same also holds for intransitive sentences:

     'He:PAT died.'
     'It:AGT (the bee) died.'

Yet another fact based on empathy is that empathy (and thus the patient
case) can only be assigned to first persons. One does not want to claim
to feel what other people feel. Again some examples to explain:

     'I:PAT feel warm.'
     'He:AGT feels warm.'
     'I:PAT am afraid.'
     'He:AGT is afraid.'

One more thing to notice about the active system of Central Pomo is
that one can also choose to use the AGT, if one wants to just state
a fact, without paying any attention to the affectedness of the person(s)
involved in the action. To sum it up, this means that in Central Pomo a
person is marked as a patient only if it is simultaneously out of control
and significantly affected and the speaker chooses to express empathy with
him or her (because he of she is affected in such a way.)

What I mean by all this -- sorry if it was a bit longish -- is that
the system might seem weird at first, you might think "why are numerals
marked as AGT?", but when you sit down and go through the system using
the terms control, P/I/E, event, significant affectedness and empathy,
you can explain what at first seems odd.

> Not to mention the occassional structural requirements, but you
> functionalists don't like to hear that. :)
AAAAAAAAAAAAAAAAAAAAAAAAAAHHH!!!!!! :)  Hmmm... I think Andreas Johansson
had a point when he said that there were universal unarticulate sounds
we scream when in extreme agony or pain. ;-)

> >Well, case inflection is *one* way to do active marking; another way to
> >do it is verb agreement, and the latter may of course go together with
> >case marking of a different kind (examples: nom-acc in Chickasaw;
> >erg-abs in Amman-iar).
> A sentence can have multiple cases for the same argument? Now that's
wierd.

I think Jörg is referring to so-called split ergative or accusative systems.
It's not different cases, it's just two names for the same case, although
it *is* two ways of alignment.

daniel

--
"Lea eica waenaidh mae bwochath waenë, ja jordhëchaidh mae gothëje
jordhëchë."
"Discovery consists of seeing what everybody has seen and thinking what
nobody
 has thought."  -- Albert Szent-Györgyi