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

At 8:02 AM -0400 10/26/98, Mathias M. Lassailly wrote:
>Nik wrote :
>
>David G. Durand wrote:
>> >   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.
>Regarding the number of arguments (A, P, S...) a specific system (nom/acc,
>erg/abs, age/pat) implies, I think it's precisely a matter of how many
>'applicatives' the definition of the predicate implies. For example, you
>say in English 'I protect my chidren from the wolf' and you think that
>'from the wolf' is 'added' as a third argument, not necessarily implied
>('applicated') in 'to protect'. Now in Japanese you have '(watashi wa)
>kodomo ni ookami wo fusegu' so Japanese consider that the verb 'fusegu'
>implies a third argument, which they never feel as 'third in ranking'
>since  being accusative it is actually the second so 'fusegu' it often
>mistaken for 'to fend off' by foreigners. 'accusative' or 'indirect
>object' is only a deixis, a feeling as to what argument we think is CLOSER
>to the process. A mere question of spatial reference and nothing more.
>Another African language I can't remember which one would always consider
>the instrument to protect (shield, weapon, shelter) as!
>  a!
> fourth 'applicative' argument implied in the predicate. So this is only a
>matter of semantics (yes, here I go again. ouch ! not on my glasses ! :-).
I think that we're in the throes of a misunderstanding, which my post of
last night may point the way to unravelling. I'll repeat it, since
restating the argument helps me to understand it.

In the Anglo-American school of comparative linguistics that Nik and I are
promulgating, there are 3 levels of description. [I'd like to note that in
my case, at least, I'm explaining this theory because it helps me make up
languages, not because I have deep theoretical commitments to the theory. I
have noted several places where the analysis is incomplete, and I think
most working linguists, even in this tradition, would agree with me.]

The 3 levels are semantic roles, syntactic functions, and grammatical cases.

Semantic roles are things like: Actor (sometimes "agent" in confusing and
sloppy writers like myself), Destination, Instrument, Location, Patient,
etc. These are defined by the semantics of an action (as interpreted by a
human) and are invariant across processes like passivization, etc. They are
the underlying reality being expressed by the language. I have very light
presuppositions at this level, as I believe that the philosophical and
empirical ground is still so unclear that even lengthy discussion will
never converge on an answer. If this were a forum on linguistic philosophy,
then I might get into it.

Syntactic functions are an abstraction that allows us to understand the
realities of many of the most common case systems of human languages. They
recognize 3 basic roles A, P, and S (the relationship to the terms agent,
patient, and subject is suggestive, but they're not the same thing as
either cases or semantic roles). A and P apply to the two arguments of
transitive clauses, S to the one argument of intransitive clauses. The
virtue of these constructs is that it lets us talk about the relations
between case systems of different languages.

Grammatical cases are things like nominative, accusative, ergative, etc.
The term case is also used (in this kind of comparative work) to refer to
other ways of marking "case" such as a distinguished position in the
clause. For instance, in English (excepting pronouns), the "Nominative" is
really marked by a distinguished position on front of the verb. Cases
represent a concrete language's method of realizing the syntactic
functions. Those functions map to underlying roles, by means of some
language-specific set of voice distinctions.

There's much less agreement on the best way to describe variations in voice
systems, or indirect cases. Some comparative linguists do add an extra
letter "I" to represent the syntactic function of _any_ indirect case, and
they can then express the fact that demoted arguments (e.g.,. Former
subjects in passive sentences, usually appear in some indirect case).
Trigger and inverse languages don't fit this framework so well, nor do
inverse systems like those in the Algonquian languages.

I don't want to engage in a discussion of whether this is a "true" theory
of meaning, or the extent to which language determines, or reflects
fundamentally different conceptions of the world. I haven't seen any such
discussions make progress, so I stay out. I will say that your accusation
that this analysis is based on English is a bit bizarre. For one thing, the
functions A, P and S are not distinct in English (subject and object are
the only two terms needed for English grammar). For another, this is a
partial descriptive framework used by field linguists, who examine a great
variety of different languages, and are distinctly non-theoretical in
Anglo-American linguistics. This kind of explanation serves their purposes
as language describers (and mine as a constructor of artlangs) while trying
_not_ to make strong commitments about semantics and grammatical theory.

I recommend, again, Thomas Payne's Describing Morphosyntax as a painless
introduction to this area. (He does _totally_ mangle his description of
predicate calculus and logic, which annoys the mathematician (and
philosopher) in me no end, but actually speaks well of the a-theoretical
nature of his descriptions).

  -- 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                    \__________________________