Message-Id: <m0r4WTu-00005bC@xiron.pc.helsinki.fi>
Date:         Mon, 7 Nov 1994 10:49:03 EST
Reply-To:     bob@gnu.ai.mit.edu
Sender:       Lojban list <LOJBAN%CUVMB.bitnet@FINHUTC.hut.fi>
From:         bob@GNU.AI.MIT.EDU
Subject:      {lo} excellent as is
To:           Veijo Vilva <veion@XIRON.PC.HELSINKI.FI>
Content-Length: 6313
Lines: 156

{lo} is excellent as is, but first, we must accept a change to the
universe....

pcliffje@crl.com cusku di'e

    ... it is amazing how much we can say about something just on the
    basis of its veridical description, ...  We can say almost all the
    historically interesting things about the assassin of Archduke
    Ferdinand without a clue that it was Gavrilo Princip.

Yes, but only when we constrain the possible universe.  If, as the
`sumti' paper misleadingly suggests, {lo} marks the whole universe,
then we can say nothing meaningful.  After all, the whole universe may
be infinite in time and space; the laws of nature may vary; it may
include parallels, like those told by science fiction writers.

We have opinions about the true nature of the universe, but we do not
know.  In the `whole universe' it might be true that the assassin of
Archduke Ferdinand stopped as well as precipitated the First World
War.

The `sumti' paper says:

    ... examples like

    7.7)    [su'o] lo ci gerku cu blabi
            [some-of] those-which-really-are three dogs are-white

    are semantically anomalous; Example 7.7 claims that some dog (or
    dogs) is white, but also that there are just three dogs in the
    universe!

This should be changed.

We can avoid paradoxes of parallel worlds or limited numbers of dogs
(unless we wish to talk about them) by presuming that the universe
against which {lo broda} is veridical has constraints.

Consider the following:

    .i lo mlatu cu grusi
    that-which-really-is/are cat(s) is/are gray

The current default outer quantifier is {su'o}:

    .i su'o lo mlatu cu grusi
    at least one of that-which-really-is/are cat(s) is/are gray

Now consider an inner modifier:

    .i su'o lo mi mlatu cu grusi
    at least one of that-which-really-is/are my cat(s) is/are gray

I don't think there is any argument against this usage.  The cat or
cats really are, and they or it are really mine.

Now for the controversial usage:

    .i su'o lo mi ci mlatu cu grusi
    at least one of those-which-really-are my three cats is/are gray

My contention is that this is a good use of {lo}; I am saying that the
universe of my cats consists of three of them.  On the other hand, if
{lo} must refer to the whole universe, then the utterance

    .i su'o lo mlatu cu grusi

is meaningless since, of course, the utterance is very likely true.
*Everything* is likely true in an infinite universe with variable
laws, including contradictions.  We don't know.

In order to be able to utter meaninfully, the universe of {lo} must be
constrained.

Having said that, I can say

    .i pa lo ci mlatu cu grusi
    one of those-which-really-are three cats is gray

Here I am not telling you in this utterance that these cats are mine,
but I am telling you that the universe of this utterance contains
exactly three real cats.


Now for {le}.  Suppose I say

    .i pa le vo mlatu cu zirpu
    one of those-which-I-am-designating as four cats is purple

What is this telling you (besides the color of one entity)?  This is
telling you that what I am designating as a cat may not be `for real'
in the epistemology of our dialogue.  It could be for real, but it
need not be.  Veridicality is not part of the categorization.

Until I provide you with more information, you don't know whether the
one cat is a real cat, now dead of old age, that had a purplish tinge
to its fur as a kitten or whether it is a picture of a cat drawn by my
three year old niece.  Had I used {lo}, you would have figured I am
referring to the real, but dead cat, not the picture.

( A while back, Jorge said:

    I don't remember any use of {le broda} where the referent isn't a broda,

Here is an example where the referent may not be the broda, at least
not in the epistemology I am using [which is, a picture of a cat is
not a real cat, even though one suggests the other]. )


Hence in terms of communication, {lo} is more specific, i.e., refers
to a more closely delineated referent, than {le}.  (Using the recent
jargon definition, substitute `more definite in the mind of the
listener' for `specific'.  (And, of course, the referent may have
fuzzy boundaries, hence the phrase, `more closely delineated '; this
subsumes the case where the boundary is sharp.))

As for `specific' and `definite'.  *Both* {le} and {lo} may be both +
and - both qualities.  How is that?  Normally, {le broda} is
considered specific in the mind of the speaker, if nowhere else.  But
the speaker is only *trying* to make a designation.  He or she may not
be succeeding....  The designation has the *goal* of being specific,
but may not be.

As a practical matter, a designation is likely to be specific, but it
is false to claim a designation is always specific or definite.
Indeed, I often take time in my own mind to create a specified
designation.  This is partly a matter of cognitive style; I mostly
don't think in words, so converting a non-verbal thought to a
designation is like composing a parable.  (There are various side
effects of cognitive styles: of people I know who think in words, one
has a tendency to over-specify; another does not, and writes
wonderfully definite descriptions.)

Getting back to the controversy over default quantifiers for {lo}: the
current default quantifiers, {su'o} outside, {ro} inside, work very
well, but only if you accept that in every discourse you have an
implicit or explicit context, which is exactly what has always been
said about Lojban utterances.

As Lojbab said:

    ... LO would not exist in Lojban without the veridicality
    criterion.  "lo" is our version of JCB's "lea" (= all of those in
    the set that veridically are) which in Lojban is "ro lo".  It was
    my suggestion to pc that "lo" be contrasted with "le" and thus
    usable with individuals out of that veridical set if appropriate,
    ENABLING non-specific selection from the veridical set to be the
    default, because "all broda" statements really aren't that useful
    in language when people are really concerned with truth values.
    Hence the outer quantifier "su'o".

I see this construction as inspired.

    Robert J. Chassell               bob@grackle.stockbridge.ma.us
    25 Rattlesnake Mountain Road     bob@gnu.ai.mit.edu
    Stockbridge, MA 01262-0693 USA   (413) 298-4725