* Sunday, 2011-09-18 at 16:55 -0300 - Jorge Llambías <jjllambias@gmail.com>:
> On Sun, Sep 18, 2011 at 2:29 PM, Martin Bays <mbays@sdf.org> wrote:
> >
> > Let me expand on that slightly. To reiterate and amend, I'm suggesting
> > that we understand {zo'e} as follows:
> > * All omitted numbered places are filled with {zo'e}
> > * By a "zo'e expression" I mean an instance of {zo'e} along with any
> > attached relative clauses, e.g. {zo'e noi broda}.
> > * Let us assume the Nirvana Conjecture:
> > When interpreting lojban, other rules reduce to the problem of
> > determining the truth value in a given possible world of a bridi
> > whose sumti are all either elements of the universe or are {zo'e}
> > expressions (or are anaphora to the latter, but let's ignore that).
> > So reordering, we have selbri(c_1,...,c_n,zo'e_1,...,zo'e_m).
> > * Interpret this as
> > \exists (x_1,...,x_m) \in C. (selbri(c_1,...,c_n,x_1,...,x_m))
> > * C here is a glorked subset of the mth cartesian power of the universe;
> > it depends on the current context, in particular on any quantified
> > variables the current formula is in the scope of.
> > * Importantly, C is required to be such that any (x_1,...,x_m)\in
> > C satisfy all relative clauses in the zo'e expressions.
> > * Note that {noi} and {poi} have the same effect for existential
> > quantifiers, so {zo'e noi} is the same as {zo'e poi}.
> > * Handling plurals: take our universe to contain pluralities as well as
> > atoms, as discussed elsewhere and as presented nicely in
> > Chierchia98 section 2.1.
> > * Handling kinds: also handled, if handled we want them to be, just by
> > having them in our universe, as in beloved Chierchia98.
>
> I'm not sure I see the point of having zo'e be a quantification over
> the members of C, instead of a direct reference to those very same
> members. What do you gain with the intermediate set C?
Consider:
A: xu do pu klama su'o friko gugde
B1: mi pu na klama [zo'e]
B2: mi pu klama [zo'e]
To get the right readings without an intermediate C, and without using
kinds, we'd have to interpret the first as being the sum of all African
countries, and the second as being a particular country which witnesses
the existential.
With an intermediate C, we can give both {zo'e}s the same
interpretation.
Moreover, the Cless interpretation of B1 relied on the distributivity of
klama's x2. How, without using a C and without using kinds, could we
handle {lo nanmu na bevri lo ti jubme}, if it's intended to mean that no
group consisting of men carries the table?
> Would you also want to say that "ta" is a close-scope quantification
> over a set T whose members are glorked from context in much the same
> way as we glork the referents of "ta" in an explanation that doesn't
> involve quantification?
I don't think so, no. I'd rather handle {ta} and {mi} and so on as
commonly understood constants. In reality, a listener who fails to make
sense of a sentence with their current understanding of {ta} might try
changing it - but I think that's another level of interpretation.
{zo'e} and {lo} are different, I think, as can be observed by their
behaviour under negation.
> As to the equivalence of noi and poi:
>
> A: xu do nelci ta
> B1: mi nelci zo'e poi zunle
> B2: mi nelci zo'e noi zunle
>
> Assuming B1 and B2 mean the same, then "zo'e" could have different
> referents in each answer (or your C could have different members). In
> B1, "zo'e" would have the same referents as "ta", (or C would have the
> referents of "ta" as its members) and "poi" will select from those the
> ones on the left. In B2, the referents of "zo'e" would have to be
> already restricted to those on the left presumably by the context, so
> that all of its referents end up satisfying "zunle". I find B1 more
> natural, and I would say B2 is appropriate only if "ta" already
> pointed to only things on the left.
Yes, that does seem sensible.
In terms of the formalisation suggested above, we would have that C is
required to have its members satisfying all noi clauses, while any poi
clauses get put into the scope of the existential.
So {zo'e noi broda zi'e poi brode cu brodi} ->
\exists x\in C:broda(x). (brode(x) /\ brodi(x))
where C:broda(x) is notation I just made up meaning that C should be
glorked such that \forall x\in C. broda(x) .
> > But sadly this doesn't seem to handle e.g.
> > {ca lo nicte lo cinfo cu kalte lo cidja}, which should be something like
> > Gen (w, n:nicte_w, cin:cinfo_w) ( cabna_w(n) ->
> > \exists cid:cidja_w(cid, cin). kalte_w(cin, cid) )
> >
> > "For generic (contextually relevant) worlds, nights and lions, with the
> > world cotemporaneous with the night, the lion hunts for something which
> > is food to it."
> >
> > I don't see how any of the current understandings of {lo} could get that
> > existential scoped within that generic...
> >
> > doi xorxes, if you're still listening: how do you get the right meaning
> > there?
>
> I have a domain of discourse with three (relevant) members: {Nights,
> Lions, Food}, and a simple three argument predicate: "at x1, x2 hunts
> for x3".
I should have guessed!
I don't think this is satisfactory in the long run - i.e. I think the
truth conditions for that predication should involve the properties of
actual nights, lions and food.
But generic phrases are complicated, and seem to still be a point of
contention among formal semanticists...
(www.press.uchicago.edu/ucp/books/book/chicago/G/bo3631829.html)
so maybe we shouldn't expect to reach firm conclusions any time soon.
Perhaps for now I should be willing to agree that your predication of
kinds there is something highly ambiguous one possible meaning of which
is that I gave, and leave it at that for now.
> The order in which the three constant arguments are presented
> is irrelevant (as long as they are properly tagged). All further
> explanations in terms of particular instances of Nights, Lions or Food
> are beyond this level of abstraction. In this sentence, all we are
> told is the answers to "when do lions hunt for food?", "who hunts for
> food at night?", "what do lions hunt for at night?", "what do lions do
> to food at night?". We are not told anything about particular nights,
> particular lions, or particular instances of food.
>
> (That's assuming all "ca" does is create a new predicate with an
> additional argument, which is the usual Lojban explanation of tags.
> This can in fact be expanded a bit more: ca ko'a ko'e broda ko'i ->
> ko'a se cabna lo nu ko'e broda ko'i, but I don't think this affects
> the present issue, in the expansion we are just showing how to create
> a three-place predicate out of the two two-place predicates.
Agreed on both points.
> )
Can I get your opinion on the "Nirvana Conjecture"?
I think that if we ignore anaphora, which could really be horribly
complicated, we should be in agreement that it should be true. Are we?
Martin
Attachment:
pgpJXpfb_YOZq.pgp
Description: PGP signature