Re: ergative? I don't know...

At 9:33 PM -0400 10/25/98, Sally Caves wrote:
>       But then what is S?  Does the S category still depend on
>       transitivity as it does in ergative systems?  So that
>       S can function as an A in active languages if it's volitional,
>       but also only if it's intransitive?  And likewise,  it can
>       function as a P in active languages if it's non-volitional,
>       but then also only if it's intransitive?
I accept Tim's correction about active languages, and his revision of my
terminology (to agree with the rest of the world!)

S is the _one_ argument of an intransitive verb: this is often unmarked as
to agent status (volitionality, I think, in your terms). Active languages
distinguish agent status in intransitive verbs (verb with one core
argument).

A and P refer to two poles of "agentness" in a two argument verb. Since
typical transitive verbs are usually asymmetrical in this way, it's easy to
tell. You add an additional distinction of agentness on the most agentive
argument.

Some languages split the S into two cases (S<sub>A</sub> and S<sub>P</sub).
Those are the active languages. The different systems of core cases are
just a matter of how A, P, and S are grouped:

  all separate: rare, tri-partite system, what I called active.
  A, S together, P separate. Nominative/accusative system.
  P, S together, A separate: ergative system.
  A, P together, S separate: Unattested, seems basically silly.
  A, S, P together: rare system, no case or dependable syntactic marking of
argument roles. I've never seen examples of this, but it's claimed to
exist, and depends heavily on context or paraphrase to distinguish agent
and patient.


>       I think I have finally identified my confusion about
>       active languages and the use of S.
One issue is that S is just a term for the argument of a one-argument
(intransitive verb). It may or may not have a separate case, and in the
case of active langauges, it may have two cases (agent-like and
not-agent-like) (volitional, and non-volitional, I think, in your terms).

>       I thought I had understood this, but I guess I'm still
>       operating under a misconception about S.  The term "S"
>       HAS TO MEAN INTRANSITIVE SUBJECT, doesn't it?
yes.

>  Now I
>       finally understand where I have gone wrong in labeling
>       my terms for T.  Under the system that I now have, it
>       is incorrect to call the non-volitional subject an
>       "S," because this only really properly refers to
>       intransitive subjects.  I then lead everyone astray when
>       I apply the term to non-volitional transitives, as in
>               Galleyli li zef oua--"the man (S) hears music."
right, You have two cases for the A role, and two cases for the S role (and
they're the same, right?) That's what makes it basically a
nominative/accusative system. What confuses that standard typology is that
you distinguish two levels of agent status in more agent-like of the
arguments of transitive verbs. This is like what active languages do for
intransitive verbs, but is original.

>       What *I* mean by S is "participant," but a participant
>       is not necessarily an intransitive subject in Teonaht.
Nor is it necessarily in other languages, but usually the distinction is
not marked in absolute terms, but in relative ones. So "the door knocked
over the candle" marks door as agent, and candle as patient, even if the
door was blown by the wind, because it's _more_ an agent than the candle is.

>       It looks as though the active languages that you describe
>       above, Tim, derive from or partake of ergative structures that
>       they have modified to include a volitional/non-volitional
>       or dynamic/stative component ON TOP of a distinction made
>       between a transitive and intransitive subject.
You can make an active language on top of an ergative or a nominative
system, I think.

You can split the S role by just re-using both cases from the transitive
system -- That's the traditional active language I think. Whether you call
the two cases nom/acc or arg/abs seems to make no difference, because both
cases can be used to mark the S role. So you can say that there are
accusative subjects of intransitive verbs (if you prefer to think of
nom/acc), or that there are ergative subjects of intransitive verbs --
there are two cases, that mark 4 kinds of role.

You add additional markings to your nominative case to distinguish the
volitional and non-volitional Ss, and so create a clearly nom/acc
split-intransitive system. You also split the nominative case for
transitive clauses.

S and A and P are not cases, but a names for roles that underlie case
systems. They are related to, but not the same as semantic roles, however,
since A, S, P don't _necessarily_ map to actor, experiencer, and the like,
since passivisation, antipassivisation and other valence-adjusting
operations can map the semantic roles into different syntactic functions
(and those into surface cases).

As far as I can tell, A, S, and P are most useful for explaining how
systems work _across_ languages: you don't need them to explain the mapping
of cases to semantics in one language, but only to understand the
relationships between how different languages use the cases that they
already have.

>       This explains why David Bell and others who have tried
>       to make me understand active and failed always appealed
>       to examples in ergative languages.
>
>       The question then remains:  if that is what S means, then
>       how is *S* functioning in nom/accusative languages?  It
>       would seem that S is non-existent in these, if NO distinction
>       is made between transitive and intransitive subjects.
Right, S is not a case, but an underluying function, that can help you to
understand the relations of cases in the various systems. If you think in
terms of nominative/accusative, an ergative langauge or an active language
are very confusing -- because those concepts don't map properly onto the
other system.

A, P, and S, seem to be useful fictions (or analytic concepts) for seeing
what is common between the argument marking systems. S is _not_ a syntactic
category in ergative or accusative languages, the concept of "subjects of
intransitive verbs" (Ss) makes the difference easier to talk about (they
pattern with Agents of transitive clauses in one kind of language, and
patients of transitive clauses in the other kind). They pattern completely
separately (with unique cases) in the tripartite languages. They are
further split into two types in the active languages (where the S term is
useless internally not because it's a hidden distinction, but because it's
too coarse).

>       WHAT IS "S"???  That's my big question.  It sounds to me
>       like it's an elusive thing... just a concept, a wildcard
>       that is either handed over to the nominative or to the
>       absolutive.  Does it have any real status outside of its
>       recognition as a "place holder" by linguists?
As far as I can tell not. It's a way to summarize a wide range of language
variation very simply. It may have psychological meaning and it may not,
and we may never know for sure. It is exceptionally helpful for
cross-language comparison, however.

>       I need some new terminology.  Agent, Participant, and Patient
>       in my books will have to go by new abbreviations:  Ag, Pt, and
>       Ob perhaps.  Is there a system I can borrow that does this
>       or do I have to make it up?
If you're in the semantic realm (underlying Agent, for instance, regardless
of passivisation or the like), then there are _too many_ systems. Semantics
is an area of little agreement. Pick terms that make sense to you and
define them.

>               Teonaht is definitely tripartite in its use of
>               articles to define these argument types:
>
>                       Le zef, li zef, and il zef
>                       The man, (agent) the man (participant)
>                       and the man (object or patient)
>
>               But it is nom/acc. in its use of PRONOUNS--and I
>               want to keep it that way.  Nothing does more violence
>               to a beloved and long established conlang than to
>               dicker with its pronouns.  I have, however, introduced
>               a variation on the emphasized pronouns, as some of you
>               may have noticed when I set up the incomplete pronoun
>               chapter on my web:  Yry firrimby  "I grateful" is
>               now the participant marking.  But the variant "yryi"
>               indicates the agentive.  Yryi kerreca, "I disapproving."
>               You can be grateful because of circumstances outside
>               your agency... but you actively disapprove of something.
>
No reason to change it. It's a neat system, and it works for you. The point
of this discussion (I think) is that we all learn more about how all this
stuff works, and how to explain our inventions, not that we tell you how to
"fix" your language.

In my latest language, I'm playing with the idea of egative bound pronouns,
and no case markings (or preferred word orders) at all. Context (and
marking of the pronouns for animacy) should be enough to disambiguate, and
if not, there must be another way than just case or word order.

   -- David
_________________________________________
David Durand              dgd@cs.bu.edu  \  david@dynamicDiagrams.com
Boston University Computer Science        \  Sr. Analyst
http://www.cs.bu.edu/students/grads/dgd/   \  Dynamic Diagrams
--------------------------------------------\  http://www.dynamicDiagrams.com/
MAPA: mapping for the WWW                    \__________________________