Re: Tagalog & trigger idea: I'd like comments. :)

On Wed, Nov 17, 2004 at 05:20:18PM -0500, Kit La Touche wrote:
> On Nov 16, 2004, at 9:32 PM, H. S. Teoh wrote:
> >Hehe, be prepared for a misfit. :-)
> yeah - i never expected a perfect fit, i just thought i'd see if i
> could describe any of it as if it were a natlang - no intent to fit it
> into boxes that don't work.
I know. But it's fun to see how close/far Tatari Faran is to a natlang
nevertheless. (Although in my mind, nothing about it excludes it from
being a possible natlang... I happen to be of the opinion that the
representative natlang types we know about today are only a very small
sample of the set of *possible* natlang types.)


> >There are only 3 core cases. Well, maybe 4, if you consider the
> >unmarked absolutive a case (it is used only in stative constructions
> >and zero-valent "verb" constructions). Or 5, if you count the
> >vocative, but the vocative is weird in the sense that you can tack on
> >one of the 3 core cases to it and turn it into an NP simultaneously
> >vocative and originative, for example.
> i'd definitely consider an unmarked one a case.
So do I. The question, however, was whether it qualified as a *core*
case.


[...]
> >Yeah, the cases are semantically rather than syntactically selected.
> >However, it doesn't quite use the agent/patient paradigm, but instead
> >a source/destination/transferee paradigm.
> well, the agent/patient paradigm is sort of mappable to
> source/transferee, with destination being the goal, in theta-role
> terms, but all these things are pretty mushy and imprecise, relatively
> speaking, so a mapping doesn't say much...
Yeah. Keep in mind that agent/patient doesn't quite map as easily as
it might appear at first glance onto TF's paradigm. For example, the
equivalent of "subject" in a verb of motion (to go, to walk, to run,
etc.) is always in the conveyant. However, the patient of "to kill"
is also in the conveyant. The eater in "to eat" is in the *receptive*,
and the thing eaten in the conveyant; but the biter in "to bite" is in
the *originative*.


[...]
> >Hmm. I think the difficulty stems from trying to map Tatari Faran
> >cases on the traditional subject/object paradigm. If one is willing to
> >stretch the definition of "object" to cover all verb arguments (i.e.
> >everything following the verb) and call the fronted NP the "subject",
> >then the difficulty disappears: the "subject" of the 1st clause is the
> >fronted NP, which is identical to the "subject" of the 2nd clause, and
> >hence it is elided. A simple case of identical subject deletion. The
> >case marker is retained, however, because otherwise it becomes
> >ambiguous which semantic role the "subject" plays in the 2nd clause.
> >
> >(My explanation for the retention of the case marker in the 2nd clause
> >is that historically, the case markers were pronouns.)
> no, subject/object has nothing to do with this, i don't think: the
> pivot need not be subject or object, per se, but need be in the less
> marked syntactic case - usually nominative,
Ahh, "pivot". Well, in this case the pivot happens to always be the
fronted NP.

But the three core cases (and I'm beginning to be a bit wary of using
the term "case" now that I think about it) are all equal in marking
and in treatment. The closest thing to nominative is the unmarked
absolutive, but that is almost never used in verbal sentences.


> but consider dyirbal, which
> is syntactically ergative and therefore allows:
>
> 1) the man kissed the woman and left.
>
> to mean only that the woman left - "woman" would be have the same
> syntactic case as object of kiss and subject of leave, therefore it
> pivots.
Cool.


> with TF, though, i was thinking that perhaps whichever NP is fronted is
> a topic, and the rest of the sentence is a comment, so you can pivot
> around the topic, as it were, by something like absent pronominals.
Wait, so is the fronted NP a topic or a focus? "Topic", if I
understood correctly, refers to the (possibly implicit) context of the
conversation, which need not appear in every sentence. The TF fronted
NP does not behave like this; it is required in every sentence, even
if it is syntactically elided. In this sense, I prefer the term
"subject", but then "subject" has other connotations which do not
apply to TF, so I'm at a loss as to what to call it.


> actually, what you say about ambiguity of theta-role for elided
> "subjects" gets me thinking - perhaps, in a natlang paradigm, one could
> say that TF has voice, but it's never phonologically distinct.
How does this "voice" explain what TF does, though?


> again, this is all an attempt to describe TF as a natlang just to see
> if i can get something approximate.  not all my explanations fit
> perfectly, thank goodness :-)
[...]

But what constitutes a natlang, though? Until (relatively) recently,
it was thought that all languages are accusative because the languages
well-studied at the time were accusative. Then along came ergative
languages, which added a whole new level of complexity to the natlang
model. And then Tagalog with its radically different syntax came along
and broke all the "rules" known at the time. (Or at least, they tried
to fit it into the accusative model, but it was like a square peg in a
round hole.) Who knows, maybe next year we'll discover some hitherto
obscure language that sports a totally different typology than either
accusative or trigger languages. (And I've already seen theses
proposing that perhaps the Sinitic languages aren't as accusative as
they're made out to be, but are more heavily based on topic-comment
constructions.)

I'm not trying to start a flame war, of course, but I'm genuinely
curious about what constitutes a natlang (other than the fact that
some community on earth speaks it).  What does it mean to say that
conlang X is a plausible natlang or "like" a natlang? Does it mean
that X fits in the set of natlang typologies currently known? Or does
it involve some other requirement/heuristic that all known natlangs
satisfy? (Note that many the Greenburg "universals" are inapplicable
to languages like Tagalog, such as the word order universals, unless
you really stretch the definition of S and O like I tried to do.)

If it is the former, are there generic heuristical restrictions of
what natlangs can be, or are all bets off? E.g., is there anything
that makes it impossible to take, say, a programming language like
Java, remap its "nouns" and "verbs" suitably to refer to things
natlangs typically refer to, and get something that looks like a
plausible natlang? (I.e., change the referents of its constituents,
but retain the syntax.) If this is not possible, why? What changes to
syntax would be needed to make it a plausible natlang? Or, if this is
possible, what implications does it have on existing typological
theories?

Or, to start from first principles, let's say that speaker S wishes to
communicate some event E to some audience A. Let's say, for the sake
of having a concrete example, that the event E is that person T gave a
gift G to person U. So here are some questions:

(a) What mental model would speaker S have of the event E?  S could
think of E as consisting of an action ACT, qualified by the arguments
T, G, and U. Is this the only mental model possible? Are other models
possible? And what model(s) are possible for describing the roles of
T, G, and U? A possible role assignment would be T: agent, G: patient,
U: beneficiary. Another assignment might be T: active-entity, G:
changed-entity, U: passive-entity. What are the restrictions, if any,
on what kinds of models constitute a plausible basis for a natlang?

(b) Are there any restrictions or general guidelines on how this
mental model gets mapped to the syntax of the language L that S
speaks? For example, is it possible to map the second model (T:
active, G: changed, U: passive) onto, say, an accusative syntax? Say,
by mapping active->subject, changed->object, passive->indirect object.
Or even, changed->subject, active->object, passive->indirect object.
What is it about the second mapping that makes it seem less realistic,
if it does seem so? Is syntax completely independent of semantics, or
are they somehow connected in some way? If they are connected somehow,
what is the nature of this connection?

(c) What constitutes a "realistic" syntax? For example, are
accusative, ergative, active, trigger, the only possible types of
syntax? Why? What is it about these types of syntax makes them more
amenable to human communication? If there were another type of syntax
that could serve as a natlang, what would it look like? What
properties must it have/not have?

Perhaps there's no absolute answer to any of these questions. But I
think it would be very interesting to explore the possibilities, and
very instructive to understand, if perhaps only a little, the
impossibilities. In a purely theoretical sense, "everything is
possible", but to me that's just too dismissive. What are people's
"gut feelings" about what is possible/impossible in a natlang, and
why do they feel so?


T

--
It is of the new things that men tire --- of fashions and proposals and
improvements and change. It is the old things that startle and intoxicate. It
is the old things that are young. -- G.K. Chesterton