Message-Id: <m0tIiDA-0000ZUC@xiron.pc.helsinki.fi>
Date:         Thu, 23 Nov 1995 20:23:10 +0000
Reply-To:     ucleaar <ucleaar%ucl.ac.uk@UKACRL.BITNET>
Sender:       Lojban list <LOJBAN@CUVMB.BITNET>
From:         ucleaar <ucleaar%ucl.ac.uk@UKACRL.BITNET>
Subject:      Re: TECH: lambda and "ka" revisited
To:           Veijo Vilva <veion@XIRON.PC.HELSINKI.FI>
Content-Length: 9289
Lines: 207

mi joi la djan joi la xorxes cusku di'e

1. On my proposals for the use of prenexes in formal rules for interpreting
   ka abstractions and relative clauses.

John:
> > I was proposing that the official rule be that instead of:
> >   tavla fa le nanmu poi cliva fa le ninmu poi kea xi pa cimba kea xi re
> >   ("the man such that the woman such that she kisses him leaves spoke")
> > the official guaranteed-unambiguous unvague way of doing it is:
> >   tavla fa le nanmu poi kea xi xy zou cliva fa le ninmu poi kea xi zy
> >   zou kea xi zy cimba kea xi xy
> This is certainly a satisfactory device; I would tend to use "poi ke'a
> goi ko'a/ko'e zo'u" and then use "ko'a" and "ko'e" thereafter.

Yes, that is nicer, isn't it:

   tavla fa le nanmu poi kea goi koa zou cliva fa le ninmu poi kea goi koe
   zou koe cimba koa

Jorge:
> > I think that given my prenex proposals, I agree with Jorge that {kea}
> > can and should do the job. I don't support these fancy subscripting
> > conventions - or at least I think they shd be no more than conventions,
> > violable without being ungrammatical.
> I don't have a problem with that. Any subscripts you use will be
> parser-grammatical anyway, whether they follow some convention or not.
> Everyone can make up their own convention and produce grammatical text.
> I think that what convention we chose, if any, is irrelevant. In real
> texts multiple or embedded {ke'a}s are very rare, and you can tell
> from context what is what in any case. (If not, it is probably better
> to rephrase. Subscripts will only add to the confusion, be it with xe'u
> or with ke'a). I totally agree with your prenex proposal, which is what
> anyone would understand without there being a convention behind it.

Lojban has, on the one hand, mere conventions, e.g. that the omitted
kea in a relative clause fills x1, and, on the other hand, grammaticalized
rules of interpretation, e.g. that scope goes left-to-right, but selbri
tcita scope over sumti tcita, or that variables not explicitly in a
prenex go to the localmost prenex.

Subscripting should, I opined, be at most a mere convention, while my
prenex proposals should be honoured with the status of Grammaticalized
Rule.

John:
> > > PA is a wastebasket category, but the maxim of minimum mutilation
> > > tells me that "xe'u" has to go there, considering how late in the
> > > game it is being introduced.
> > I think that maxim would lead to the use of {kea}, as advocated by
> > Jorge & now by me.
> Using "ke'a", though, disallows the neat form "xe'u broda" for "xe'u da
> poi broda", as in:
>        le ka xe'u nanmu cu cinba la djein.
>        the property of being a man who kisses Jane

Well if you're that taken by the device, you could alter the grammar
of kea, to allow

        da poi kea nanmu cu cinba la djein
        that which is a man who kisses Jane

If it's good for {xeu} it's good for {kea}.

> In addition, a special rule for binding "ke'a" implicitly in a prenex
> would be needed, rather than piggybacking on the existing rules for
> da-series. Lambda quantification binds variables, after all, and we
> already have machinery for bound variables.

{kea} is a bound variable, I think. Logically it is pretty much like
a da-series, and moreover, I did propose well-motivated rules for
prenexing kea in relative clauses.

> I am, however, considering a separate selma'o XEhU as proposed by lojbab
> in another message; this would act as a quantifier_300 but wouldn't
> participate in MEX activities.

Well, you could make this selmao KEhA, with one member {kea}, possibly
with grammar such that {kea [ku]} is a legitimate construction.

> > > This fnord is obviously not the way natural fnord fnord languages
> > > operate fnord.
> > Exactly. No grammarians' rules have ever included one for fnord-insertion.
> What about the grammarian Hagbard Celine?

Am I showing my ignorance here?

> > >        bridi le ka xe'u da cinba xe'u de kei la djan. ce zo'e
> > >        Something-is-a-proposition-formed-from the-relation (X kisses Y)
> > >                and-the-set {John, Whoozis}.
> > > or even
> > >        le ka xe'u da cinba xe'u de cu zilbre la djan. ce zo'e
> > > which means the same thing.
> > There's not much call for {zil-}, since the proposition is still there
> > is you have saturated predicate. Another example:
> >     zo cinba valsi le ka xeu da cinba xeu de
> Interesting, although I'm not sure it's quite true. Using "zil-" makes
> for economy; remember that I read "zi'o" as simply removing the place
> from the relationship. IOW, "mi dunda zo'e do" entails "mi dunda zi'o
> do" but not vice versa.

Yes, but since it'll always be true even without the {zio}, it seems
a bit redundant to use it. But what the hey - why not use it, since
it gets up eveyone's noses.

> > > Besides, how can you be sure that you've filled every argument?
> > Only by filling them, at all levels of subordinacy.
> And therefore the existence of a separate cmavo "du'u" is useful,
> because it guarantees 0-adicity of the intension.

Yes, true. But it pains me to concede this. If it were possible to
make sumti obligatory, I'd say the "lambdas" within ka bridi should
be obligatory.

> > So duu is really a one place predicate, denoting a class of sentences.
> > Or rather, duu is used in a range of one-place predicates, each of
> > which denotes a class of sentences.
> Or rather still, "du'u" is used in a range of two-place predicates, each
> of which relates a class of propositions (not sentences) to the sentences
> which express the proposition.  (Again, "proposition" = "0-adic
> intension".)

This is what I was denying, I think. Here goes again: {duu} is used in
a range of 2-place predicates, each of which relates a proposition to
the sentences that express the proposition. This is tantamount to
a 1-place predicate, just as "cousin of John" is effectively a 1-place
predicate - the x2 of "cousin of John" is of course John, so it seems
a bit pointless to have "le se cousin-of-John" to refer to John.

Enough of this. I'll kill this topic unless there's some change of
getting a construct that takes a bridi and yields a sumti.

> > It is very naughty of Lojban to exhibit type-token ambiguity. It of
> > all languages should be well-behaved. It's why I say such-and-such
> > a selbri shd be a sumti, and vice versa.
> Well, consider:
>        mi cusku zo djan.
> where "djan." is a token, versus
>        mi se cmene zo djan.
> where "djan." is more like a type: I do not mean that my name is a
> specific >instance< of the word "djan."

I see what you mean. If {zo} were a selbri, I'd say:

     mi cusku suo zo djan
     mi se cmene loe zo djan / mi se cmene ro zo djan

> > I think I'd like to argue that "abstraction" has no meaning, at least
> > not beyond the n-adic ka/duu.
> > I don't know which book you're talking about.
> "The Crucifixion of Jesus Christ Considered As A Downhill Motorcycle Race."
> (Arguably my translation doesn't render the word "considered".)
> The point is that every other abstraction can be expressed as a "su'u"
> with an appropriate x2:  "nu" is "su'u ... kei be lo fasnu", "jei" is
> "su'u ... kei be lo niljetnu", etc.

I think I get it.

   lo suu broda kei be lo ganxo

is equivalent to

   lo ganxo poi kea duu broda
   lo duu broda kei poi kea ganxo

- Or? Since a bridi is not a fasnu, or a ganxo, or whatever, these would
always fail to refer, unless used with a nonveridical gadri.

> > > > Right. A major property is temporality - nu are associated with times
> > > > and duu arent.
> > > Well, that's shaky.  "le nu li re su'i re du li vo" is probably the same
> > > as "le du'u li re su'i re du li vo", although both are equally temporal
> > > or atemporal or totitemporal or what you will.
> > I don't think such atemporal or omnitemporal things can be nu. For me,
> > all nu must be terminable.
> I think that this point need not be prescribed: "nu" certainly covers
> space-time events, and its application to abstract events can be left
> open.

I'm not sure what you mean by "abstract event". To me, the nature of
situationhood is approximately as clear and as definable as the nature
of, say, applehood or virginhood, or whatever. I know intuitively what
they mean.

> > > "Physical" is a sticky notion. There is no problem with "nu" objects
> > > that aren't actualized, like "le nu le djordj. .ualas. cu merko gugde
> > > ralju" even though George Wallace wasn't ever U.S. President.
> > There is every problem with such nu objects. {nu la djordj ualas cu
> > merko gugde ralju} is false.
> I assume that by "false" you meant that it is predicated of {noda}.

Yes. And since {zoe} can be anything except {noda}, my usage of
"false" was legitimate, right?

> In that case, how do you say "I desire George Wallace to be etc."?
> Certainly "mi djica le nu ..." is traditional here. There is a difference
> between what contingently didn't happen, but could have, and what is
> not a happening at all.

We've debated this before. I insist that {lo nu} refers to an actual
happening, not something that could have happened but didnt. So I would
say {mi djica lo dahi nu}, here, where {nu} is extensionally the set
of all actual events and {dahi nu} is extensionally the set of all
imaginable events.

---
And