Re: ago24 & replies

[ucleaar <ucleaar@ucl.ac.uk>]

Jorge:
> > I take your point of course, that the "head convention"
> > is not the only one that is useful. I suppose that one way to go
> > is to have two {xohu}s.
> Aaaargh! No! I already find it hard to understand the meaning of one.
> (I understand it once I've thought it in prenex form, but the idea is
> that it should be immediately understood in afterthought.)

You find it hard to understand on a conscious level. But you use
similar devices in English - namely, "any" - with no trouble. So
you'd get the hang of it, intuitively - and, you being you, also
consciously.

> > We argued about {nitcu} last year, and you eventually came up with a
> > definition of it (which I forget) such that the x2 can coherently be
> > an object.
> If I remember correctly, you accepted that a definition could be based
> on {vajni} (important), which accepts objects and events. Do you agree
> that an irrealis event can be vajni?

I'm sorry that I can't properly recall the relevant discussion.

I now think that "be important" in the sense of "matter" ought always
to have an irrealis x1, and for the reasons I've outlined, I
think only duhu can be irrealis. [I fully realize that I'm probably
igniting a new thread in saying this, & fully hadn't intended to.]

> Even if we can't agree on nitcu specifically. Do you think that there
> is (or should be) certain tersumti that accept _exclusively_ irrealis
> events?

I think there should be certain tersumti that accept exclusively
irrealis sumti: i.e. le+anything, or {lo bridi} or {lV duhu}.
English transitive BELIEVE is like that.

> > (That is, is it the case that
> > for all events there is some time such that the event occurs at that
> > time?) If it is, then I don't think we can have irrealis events.
> Yes we can, just like we can have irrealis objects. {lo'e mlatu} is not
> a realis cat, so I would say {lo'e nu mi klama le zarci} is an irrealis
> event.

I don't have any intuitions on the relationship between (ir)realis and
genericity. In a sense I think generics are purely mental objects, and
so are neither realis nor irrealis. That is, ontologically, lohe mlatu
is not a cat but a conceptual cat-archetype; but properties ascribed
to lohe mlatu as a sumti are understood to be properties not of the
conceptual archetype but of instances of the archetype. {Lohe nu}
is not an actual event, but nor is it a potential event, or an event
such that we don't know whether it happens.
And I'm not comfortable with a cat being irrealis - as I've said, I
think the (ir)realis distinction is a property only of propositions.

> > If it is not, then we can have irrealis events, but claims about
> > nu broda are truth-conditionally vacuous.
> Right. I think {lo nu broda} has to be realis, and for the irrealis case
> I'm inclined to suggest {lo'e nu broda}.

See my comments above. Also, to me (and nothing from LLG leads me to
think I'm right on this) {lohe} is like a default universal quantifier.
So if {lohe mlatu cu xekri} then a white cat is an exceptional, atypical
cat. Say my views on {lohe} prevail: then it would work for "I want
to go" translated as {mi djica lohe nu mi klama}, since it would not
be true that the typical event of my going is wanted by me - e.g. it
is not claiming that if I don't want the event of my going for a walk
on May 5 1971 then that event is atypical.
The way I'd now say "I want to go" is {mu djica le duhu mi klama}, with
"djica" meaning "want x2 to be the case", and with implicit ca tense
on klama.

> > Supposing we therefore said all events are realis. We could get at
> > the meaning approximating to an irrealis event by using {duhu da nu broda}.
> > Cumbersome as that may seem, it seems the most logical position to take.
> > This would mean that all irrealis sumti, e.g. x1 of cumki, should in
> > fact be duhu. ({lo nu broda kei cumki} would be true iff lo nu broda
> > ever occurs.}
> I don't like it, but I can't really say why. They just don't seem to be
> du'u places. I'll try to think of a better argument.

One reason not to like it is that it invalidates vast quantities of
established usage. Another reason not to like it is that the relevant
gismu need to be reglossed. {Cumki} goes from "x1 is possible" to
(if we're being ultra clear) "it is possible that it the case that
there is an x1". My reason for liking it is that it seems to be right
logically.

I can blearily see one logical inconvenience with my proposal.
This is the uniqueness of propositions. You'd generally have
to use overt tense indicators. There is exactly one {duhu mi
klama} (which is why I wish duhu didn't need a descriptor, or
could be {li duhu mi klama}, with {li} labelling uniques), and
it happens to be true. To say "I need to go", I would have to
say {mi nitcu le duhu mi ba klama}.
In fact, I rather fear that might not work. Consider {troci}:
{koha pu troci le duhu koha klama} (koha tried to make it the case
that koha goes). If koha had previously gone, then {koha klama}
would already be true. So we need a tense on klama meaning
"after the trying", "thereafter", {ba *tense of superordinate bridi*}.

> > If you meant {nu broda} as a needee, then by my def. of nitcu, yes
> > it must be irrealis: I suppose we shd have to say {nitcu} means
> > "it is necessary for x1 (i.e. it is to x1's advantage) that x2 be
> > the case", and thus say {nitcu lo duhu broda}, {nitcu lo duhu da nu
> > broda}.
> We already have {sarcu} for that. I think {nitcu} should be primarily
> a relationship between two objects, extended to events only as long
> as the relationship remains analogous to the object case.

{sarcu} is: x1 is necessary for *continuing state/process x2*.
So it differs from my nitcu in the nature of the sumti something
is necessary for.

I am inclined to accept the view that (on grounds of utility) {nitcu}
should be a (post-sumti-raising) relationship between two objects.
But I hold that this would rule out opaque or irrealis readings.

pc:
> Going back to last year's opacity thread, I thought we finally agreed
> that 1)all those cases of opacity were in fact case of subject-raising
> from event descriptions, 2) within event descriptions there were some
> expressions which were inherently transparent (proper names?, epressions
> with local reference, e.g., _levi_ descriptions -- I don't think we got
> the list) and others that could be flagged as such (with whatever the
> other thing than _xo'u_ came out of _xe'e_) and which could thus appear

I think {xohu} was to be this marker of transparency (quantified
outside the event description). {xo'unai} was to confirm the flagged
sumti is to be quantified with the event description.

> raised without the mark of raising, which would have cut them off from
> higher level reference.

I think this indeed what we (including at minimum you & me) finally
agreed. I have now changed my view in two ways. First, I now think
that we should not be so specific as to talk of "event descriptions":
rather the necessary condition for opacity is a subordinate proposition.
Furthermore (and here I am tentative), since I've concluded that
all events happen, it appears to me that the quantification of sumti
within an event description can necessarily be external to the event
description. In other words, event descriptions are necessarily
transparent. It is bridi abstractions that can (and perhaps necessarily
do) give rise to opacity (and irrealis effects).

> I am less sure that we agreed that all events "exist" in the
> generalizable-by-_da_ sense (which may even be _zasti_, though I am
> not sure about that) but differ in whether they "obtain."

We discussed the issue, but I don't recall it being resolved.
I think I am now claiming that all events "exist", and that
*therefore* the bridi that they subordinate "obtains".

> Presumably it is the latter that is irrelevant or denied when when the
> notion of "irrealis" turns up and affirmed when "factive" comes along.

Exactly.

---
And