From lojbab@access.digex.net  Sat Mar  6 22:44:50 2010
Date: Tue May 23 03:13:53 1995
From: Bob LeChevalier <lojbab@access.digex.net>
Subject:      Quantifiers (was Re: A modest proposal #2: verdicality)
To: DPT@HUMA1.BITNET
X-From-Space-Date: Tue May 23 03:13:53 1995
X-From-Space-Address: lojbab@access.digex.net
Message-ID: <HsXluJBtd1D.A.JdE.it0kLB@chain.digitalkingdom.org>

I'm still behind on this thread, but notice that Jorge wasn't dealing
with all Dylan's questions exactly as I would have.  So here is my take
so-far (I'm still back on last Wednesday's traffic, so pardon me if some
of theis has been rendered moot.)

Dylan:
>[1] Which leads me to ask:  how would I say:  "two of the man, the woman
>and the child" (as a sumti)?  "The man, the woman, and the child" is
>
>        le nanmu joi le ninmu joi le verba
>
>How do I select two of them?

Ah.  The old "coffee, tea, milk, or water" problem!  Where logical
connectives break down, and non-logical ones don't have appropriate
grammar to meet the specific need.

So we invented a specific solution:  lu'i ... lu'u which has now been
expanded into 3 choices lu'a|lu'i|lu'o ... lu'u, of which you want to
use lu'a for the purpose you have in mind.  You also probably want "ce"
rather than "joi" for the non-logical link - joi creates an indivisible
mass with components and not members.

re lu'a le nanmu ku ce le ninmu ku ce le verba


>> That's a sumti, it's an extension of the above: "one of the two of the
>> three women". Of the three women you have in mind, you are selecting
>> two, and then saying something about one of them. But notice that
>> the last selection is not the same as the others, you are claiming
>> something about one of the two, but not selecting which one.
>> {le pa le re le ci ninmu} on the other hand, does select which one.
>> Of course, all this nesting of selections would be quite confusing in
>> actual use, so it probably won't be very common.
>                    ^^^^^^^^
>
>{pi'e} This is the only good news in this letter:  I take it that this
>form has not been used much?  Good, let's ditch it.

Alas, it is useful (maybe necessary) for some albeit rare circumstances,
and in any case is a logical consequence of the grammar, which permits
many constructs that mighht not be necessary - but since the grammar
generates them, we tend to try to find use for them (often by
serendipity finding them to be useful indeed - the lu'i ... lu'u
grammar, for example has proven much more logically useful than its
original intent because it is now used to convert sumti types among the
three:  masses, sets, and individuals - very important for logical
precision regarding anaphora. 

>To me, {pa le re le ci ninmu} looks like a total bastard.  If {re le ci
>ninmu} is a sumti, what are you doing putting {le} in front of it?  I
>thought {le} was the form for converting a selbri to a sumti--what's it
>doing twice here?

Because many elements of the grammar have multiple - generally parallel
uses.  Yes, "le" converts a selbri into an intensional (i.e.
non-veridical) sumti.  Grammatically it turns out that the result - an
intensional sumti with certain default quantifiers, is the more
important property than the fact that in the simplest case one starts
with a bare selbri and pulls out the x1 of that selbri as a description.

The relevant grammar portion is thus:

<description_110         :  LA_558  sumti_tail_111  gap_450
<                        |  LE_562  sumti_tail_111  gap_450
<                        ;
<
<sumti_tail_111          :  sumti_tail_A_112
<                           /* inner-quantified sumti relative clause */
<                        |  relative_clauses_121  sumti_tail_A_112
<                           /* pseudo-possessive
<                              (an abbreviated inner restriction);
<                              note that sumti cannot be quantified */
<                        |  sumti_F_96  sumti_tail_A_112
<                           /* pseudo-possessive with outer restriction */
<                        |  sumti_F_96  relative_clauses_121  sumti_tail_A_112
<                        ;
<
<sumti_tail_A_112        :  selbri_130
<                        |  selbri_130  relative_clauses_121
<                           /* explicit inner quantifier */
<                        |  quantifier_300  selbri_130
<                           /* quantifier both internal to a description,
<                              and external to a sumti thereby made specific */
<                        |  quantifier_300  selbri_130  relative_clauses_121
<                        |  quantifier_300  sumti_90
<                        ;

And you can see that the selbri, hidden in a portion of 4 of the 5
lowest level rules (BTW, this last phrase is an example of nested
quantification in English), is a critical but non-essential core of a
description.  You can always use the last rule (which may never expand
to include a selbri at all) to get such phrases as "re le ci do" (2 of
the 3 of you).


>{le ci ninmu} is almost as bad.  When I first saw it, I thought {ci
>ninmu} was a unit--but no, this is really indecomposable.  Consider:  to
>say "there are three men in the room" I can say
>
>.i lo ci nanmu ne'i le kumfa
>
>as a sumti by itself; 

Lots of problems here - you probably want a "pe" before "ne'i", or else
"cu nenri".  In any event - if there is no relevant restrictions on the
selbri "nanmu", then "lo ci nanmu" has the implicit claim that there are
only 3 members of the set of nanmu in the universe of discourse (You, me
and who else?)

A correct way to express your English is 

ci [lo] nanmu cu nenri le kumfa

>as a sumti by itself;

to be a sumti that must be:
ci [lo] nanmu poi nenri le kumfa  
or
ci [lo] nanmu pe ne'i le kumfa   

Later:
<My misunderstanding of the grammar led me to make a mistake; I meant
<        .i lo ci nanmu pe ne'i le kumfa
<which _does_ mean "there are three men in the room", right?

No - that is a sumti: (at least one of) the exactly 3 men which can be 
identified by the phrase "inside the room").

>but to make a bridi, I need to rewrite this, in
>one of the following ways:
>        le ni loi nanmu ne'i le kumfa du li ci

Ungrammatical since ni requires a bridi inside of it - or rather it is
grammatical and the entirety is a strange sumti:

>   le ni          {loi nanmu            ne'i le kumfa    du       li ci} [KEI]
    The amount of  the mass of mankind  inside the room  being
                                                         equal to the number 3
         
I don't know about you - but I ain't part of any mass that is equal to a
number zo'o

Later:
<Sorry, forgot a {cu}: should be
<        le ni rolo nanmu pe ne'i le kumfa cu du li ci

I had actually missed the missing "cu" in the above argument.  The
missing cu makes the li ci simply a third sumti inside the leni clause
which then has no selbri.)

>or
>        loi nanmu ne'i le kumfa cu cimei

Again, I think you want "pe ne'i" but this is more reasonable - x1 of
cimei is a mass.

>or maybe
>        da noi te cimei nanmu ne'i le kumfa

Something which is incidentally (a 3-membered-set man
inside-the-room-ly).  That ne'i phrase is a tagged sumti attached to the
bridi "te cimei nanmu" and has nothing necessarily to do with "da".

<Damn those {cu}s!  I meant {da noi te cimei cu nanmu ne'i le kumfa}

Because of the "noi" instead of the "poi" you have an incidental
qualification, and you are thus saying that something that is a member
of a set of three-things is a man (inside the room).  You need cida, and
poi.  And I think nothing about this claims that there cannot be some
other men in the room besides "da" - but it is a bit murky to me.

>or
>        ci da nanmu ne'i le kumfa

3 somethings are men inside-the-room-ly.
That one sounds pretty good.

Umm.  I'm not sure which of these mean what I want. 

Well, they all mean something, two of them being particularly reasonable
interpretations of the English.

>In any case, my point is that there's this syntax for implicitly making
>a claim about the number of ways to fill in variables to make a sumti
>true, without any corresponding syntax for filling in the places of a
>bridi.

I'm not sure I understand the antecedent.  There are lots of implicit
claims about quantification.  But logic basically demands that any
explicit claim be a bridi.

ci da zo'u da nanmu joi nenri le kumfa

is the explicit quantificational claim as most logicians would write it
(I think).

and "lo nanmu poi nenri le kumfa cu te cimei

is a bridi that states the cardinality of the set of men in the room.

> Why
>can't
>        ci nanmu ne'i le kumfa
>mean "There are men in the room (three of them)" or some such? 

Because that is two sumti, one tagged with ne'i, and no selbri linking
them.

Jorge answered this, interpreting the question differently:
:I remember asking this very same question to John Cowan once, since the
:sumti {ci nanmu} can already be said {ci lo nanmu}, why not let {ci
:nanmu} be a selbri meaning "x1 is three men"?  The answer is that for
:historical reasons it is what it is.

The specific historical reason is the "SE SORME" = ze mensi issue that I
described the other day.  JCB thought it loose and illogical to allow
the construction at all, but it kept creeping back into actual usage
(i.e. the little Loglanders in his head always used it %^), and he
finally decided that it was more natural than "SE LE SORME".  When we
set up the lo/le dichotomy, we had to choose between ze mensi meaning ze
lo mensi and ze le mensi and chose the former as being likely more
useful.  But at no time do my knowledge has it been proposed that "ze
mensi" be an abbreviation for "lo zenei cu mensi".  In fact, such a
usage would go against another canonical Loglanism:  to say that a table
has 4 legs, it is classic (the closest thing we have to an idiom) to use
the form "le jubme cu se tuple voda".

> (Or
>maybe that should be
>        cimei nanmu ne'i le kumfa
>and we should just use this tanru instead of the inside quantifiers.

That is a bridi, with a tagged sumti attached.  Converting it into a
descriptive sumti would give

         le cimei nanmu be ne'i le kumfa  

This would be valid, though I would myself use "nanmu cimei".  But any
use of tanru is inherently somewhat ambiguous in meaning, and some
people (And?) would want to know how to expand your tanru into its
non-tanru equivalent.

In any case, "le cimei nanmu" is longer than "le ci nanmu", and probably
more grammatically complex in that the "cimei" as a tanru unit is a more
complex component of a sumti than a quantifier.  In either case, you can
muck up the waters with some strange quantifiers that would baffle the
mind, but it is a little harder with the quantifier than the tanru unit.

>Yes, I'm starting to like this.)    

You can use it, and probably be understood.  But it will get long winded
when you say that "the three men each carried forty of the thousand
library books from the first room to the second room via one of the two
hallways using two of the (unspecified number of) library carts.

>I've been studying the syntax of sumti.  It is, to put it mildly, a
>mess.  Did you know {le mi do se cusku} is ungrammatical?  Neither did
>I.

I can't imagine what it would mean - my mind balks at "le mi do" - "My
you" (???)  Maybe you want "le mi se cusku be fi do" = "le se cusku be
mi bei do"?

Let us analyze this:

"mi do se cusku" is a bridi meaning "I am expressed by you"
"do se cusku" is a bridi meaning "you are an expression"

How do you want "mi" and "do" to relate to the selbri?  As joint
possessives?

 le mi joi do se cusku

<Well, I thought it meant {le se cusku pe fe mi zi'e pe fi do} -- "that
<which I expressed to you".  But why can't it mean {le se cusku pe mi
<zi'e pe do}?

Why should it mean that in particular instead of all the other
possibilities.  You want use to infer a "pe" AND a "zi'e", and the
result of the "zi'e" ain't all that clear anyway.  There are other
possibilities.

<(There's got to be a technical reason the {zi'e} is necessary, yes?  I
<know {le se cusku pe mi pe do} is grammatical with a different
<meaning, but {le se cusku pe mi ge'u pe do} is not grammatical.)

At one time zi'e was part of a set of logical connectives that could go
there and link relatives.  No one ever used anything but zi'e and even
then it seemed to be more of a non-logical "joi" than a logical '.e'.
Perhaps your alternative could have worked if it had been though up
first, but I am not sure - zi'e can mix both clauses and phrases, and it
would take a bit of testing to make sure that a grammar that allowed
your suggested form would have all the capability.  The most common
original reason was to allow restriction, and then allow the whole mess
to be "goi ko'a".



There are some limits to what you can put before the bare selbri in a
simple description.  We have expanded this to allow for preposed
relative clauses after the Finnish model, and this took considerable
work and debate.  A recent proposal to allow preposed "be/bei"
constructs was embedded in last week's discussion - it might work, and
might be useful, to people with a preposed grammar native structure.
Veijo???  But no guarantees we could make it grammatical, and it is not
importnat enough to justify a change if it causes anything more than a
trivial expansion rule (if even then).

>la xorxes cusku di'e
> > That's not grammatical, you mean {le plini be fi le xunre}, however, it
> > does seem to be a reasonable construction. I don't know if it would
> > cause problems, but if it doesn't, perhaps it should be made grammatical.
>
>This should almost certainly be grammatical.

This also is a variation of a preposed relative.  Again - is it useful
enough to bother with it at this late date?  And if it is non-trivial
grammatically, it will not be considered.  There simply hasn't been all
that much usage even of preposed relatives (and their grammar is tricky
- you more often cannot elide the terminators than in postposed units).
At best this will be an "advanced topic" for any language instruction
since it does not actually provide any new capability - just an
alternate ordering for an existing capability.

Still the fact that someone here thought it up means that it IS
conceivable that some language users might find it more natural than the
normal way ...

>  One problem is the
>meaning of {le fa le brode ku broda}, but that already exists: what
>does {le broda be fa le brode} mean?

There are many possible expressions that don't have a clear-cut meaning.
To exclude "be fa" we would have had to put "fa" in a different selma'o
from the rest of the FA members, and double all the other rules that
allow FA just to exclude this one.  Who cares?

Or rather - if you insist, I'll come up with a semantics for it - I
would say that {le broda be fa le brode} means "le broda voi brode" -
the broda which also may be described as a brode.  But I would never use
the former instead of the latter - more confusing and complex - except
perhaps for some poetical parallelism between broda and brode.

I've played this game of devising semantics for all manner of
strangeness, and not just in the sumti grammar, which in spite of your
comments is probably among the cleaner portions of the grammar.

> > BTW, strangely enough, {le pe fi le xunre ku plini} is grammatical, but
> > with {be} instead of {pe} it isn't.
>
>What on earth is {le pe fi le xunre ku plini} supposed to mean?

It means that we wrote the rule for relative phrases to accept any
"term_81" rather than to try to select some limited subset that "made
sense" because in most cases we CAN make some sense out of it.  In this
case, I would see no qualms with interpreting "pe fi" as meaning "be
fi".

But if you want a challenge, try 

le broda pe naku

or worse

le broda pe nu'i fi le blanu naku nu'u joi ne'i le kumfa

To make simple the rare cases when termsets are useful, we use a simple
all-encompassing grammar that allows for all manner of useless garbage.
And probably the only reason we put termsets in the language at all is
that JCB had them, and we wanted to keep Lojban compatible with TLI
Loglan.  There are a lot of fossilized possibilities in the language.
Anyone for "setesevexeseteve blanu"  (especially funny since blanu is 
a 1 place brivla)

>Anyway, IMHO the syntax of sumti needs both rethinking and debugging.
>>From the state of this part of the grammar, I'd guess that it's
>{puta'e} been patched; I think a rewrite rather than further patching
>is in order.

Everything in the language has been patched.  But believe me, what we
have is far simpler a grammar than anything that would be a redesign
that was semantically solid.  And debugging it would take the full 8
years we have spent debugging the current grammar, since any time you
redo a whole chunk of the grammar, you have to look at all the tests
that have ever been performed to prove the solidity of the design - and
there isn't exactly a very good log of those tests.

Indeed it is safe to say that for the last 5 years or so, we have ONLY
considered grammar changes that were expansions to the language - added
rules that opened new possibilities, and changing rules that even COULD
delete a useful and already used possibility have taken months of
verification through usage before they were formally adopted, except in
the case where we could argue that the entire structure of a rule was
logically "broken".

Loglan/Lojban has been patched ever since the initial design in 1954, of
which relatively little is known.  Even the "total redesigns" of the
morphology in 1979-82, and the YACCing of the grammar (1976-present)
have had the implicit requirement of some degree of backwards
compatibility in as many as possible of those aspects not directly
examined for change impact.  I think anything else, and we could not
honestly call the result "Loglan" anymore - and I have fought hard for
years for the moral and legal right to use that name honestly, since
only in that way will this language project responsibly honor the
unmatchable contribution of its inventor.

lojbab