Message-Id: <199512182143.XAA13222@xiron.pc.helsinki.fi>
Date:         Mon, 18 Dec 1995 13:30:00 -0800
Reply-To: "John E. Clifford" <pcliffje@CRL.COM>
Sender: Lojban list <LOJBAN@CUVMB.BITNET>
From: "John E. Clifford" <pcliffje@CRL.COM>
Subject:      TECH:opaque (ex mass and le/lo)
To: Veijo Vilva <veion@XIRON.PC.HELSINKI.FI>
Content-Length: 9066
Lines: 167

the cmavo list:
tu'a      LAhE     the bridi implied by                      extracts a
concrete
sumti from an unspecified abstraction; equivalent to le nu/su'u [sumti]
co'e

I wondered about the grammar of this, which I was using, I
noticed, as grammatically transparent, like a UI restricted to
appearing before sumti.  But it turns out to be a sumti to sumti
function, transparent in Lojban grammar (because of the general
reduction of all sumti to one level) but complex in Lojban
semantics, where it converts a non-event sumti into an event sumti,
with the predicate derived by convention from the visible sumti
and the predicate to which it attaches.  So _mi djica tu'a lo plise_ is
presumable a compressed form of _mi djica le nu mi ponsu/citka
lo plise_.  Since x2 of _djica_ is specified as taking an event, both
of these forms are totally acceptable, but _mi djica lo plise_ would
be at least questionable -- not ungrammatical  (since their is no
way to specify event references grammatically -- except that _lenu
..._ always counts) but dissonant in some way, since an apple is not
-- Whorf's (unconfirmed) view of Hopi metaphysics
notwithstanding -- an event (Please, let's leave this one for another
time and another thread, because I already know most of what the
shitkickers are going to say that is true).

So, while the presence of _tu'a_ marks an opaque context, it is not
an opaque-context marker, as I was trying to use it.  It rather
creates a standard opaque context (intentional event description) in
miniature.  Thus, too, the use of the cancellation convention with
the "external reference" marker (?xe'e?) is inappropriate, since it
was based on that wrongheaded understanding.  But the _xe'e_,
which presumably is transparent and appears only before sumti
(and so is YACC syntactically in LAhE, too, whatever its more
profound grammar may be), still functions correctly.

That _tu'a_ is not an opaque marker does leave the problem of
dealing those opaque contexts that are not at the surface but well
up from the internal structure of brivla semantics, x2 of _kalte_,
for the running example.  If we ignore that regular feature of the
meaning of _kalte_, that a person can be hunting a lion even if
there is no lion they are hunting (either because there are no lions
at all or because no special one is required, any-old lion will do),
then we get illegitimate inferences looking legitimate, a  major no-
no in logical languages.  If we do not ignore it, how shall we mark
it, since _tu'a_ introduces an event sumti which is already
(virtually) there, not in need of being mentioned again (and it
would make it seems that the hunter were hunting an event, rather
than the main patient in the event)?

These problems need not arise with the artistic subject opaque
places, "picture of", "book about" and the like, since it is at least
plausible such things are always about events: Madame X standing
by a table, not just Madame X, for example.  The plausibility thins
a bit with compositions on abstract subjects, but we can probably
circumvent any problems that arise.  If we decide we need to.

But the problems reemerge when ordinary predicates, ones that do
not usually engender opacity, come to be used in ways that do.  On
the one hand, there are the "professional" uses: he is a poet/writes
poems, when there may be no poems he has written to point to: he
only professes write poetry but his practice has not yet come up to
snuff.  And similarly for many other professions that would lead to
a product but have not yet (or a target -- but that brings us back to
hunting and the like).  But here it is not just the sumti which is
intentional but the whole predication, which ought be marked with
something in the general modal class, thus rendering opaque all the
places in it.  And xe'e to shine the light through where appropriate.

In a similar way, the goals in processes which are not completed
are not yet existent, by definition, so that at least quantifiying on
them seems illegitimate.  However, identity substitution does seem
to work for these and, in the context of a discussion of a process, it
seems likely that the product is in the domain of discourse, so
quantification (which is over that domain, rather than reality)
would apply as well.  In this case, the aspect markers, other than
the completive and perfective, might serve as warnings of possible
opacity, even if rarely needed.  Indeed, the problems with
"professional" uses might be absorbed to this case, an unfulfilled
profession being, I suppose, at least the inchoative of a process of
fulfilling it.

So, all of the cases of opacity except those for where the
intentional predication is buried in the semantic of the main
predicate -- the "hunt" case -- have natural solutions within the
present system, using _tu'a_ as an event sumti creating operator
and allowing _xe'e_ to function to show external reference where
it is not given (and just what gives it is still in need of
specification, but unrestricted _lo_ and _da_ pretty clearly will not
do).  The remaining problem could be handled by a lexical note,
similar to the ones we already have about places that require event
sumti, for example.

Still, it might be useful to have a single way of indicating opacity
across all these cases, the more so if it would also simplify dealing
with transparent sumti in opaque contexts.  Using _tu'a_ in the way
I have been would do this.  But it combined with the _xe'e_
convention would also create the situation where it is very easy to
create a misstatement: the absent _tu'a_ being equivalent to _tu'a
xe'e_ and so transparent, where the absence may just be due to
carelessness.  On the other hand, it seems unnecessary (and
unlikely to happen) to write two words -- four syllable -- to
indicate that a sumti behaves as sumtis usually do, especially when
the need for this indication is buried away in the definition of a
late place in a gismu and has no overt signs in the utterance.  Here
is a nice Whorfian test; can a speaker become so aware of the
demands of logic that they overcome the practicalities of easy
expression.  Or it is a test of our ingenuity to find a workable
compromise.

x(in le/lo)
I would say that pretty much every place of a gismu can be opaque, yes.
Another example discussed here once was "that is a human head". I would
say {ta stedu lo'e remna}. You may want to say {ta stedu tu'a lo remna},
pc:
Why would I want to say that?  If it is a human head then there is a
very definite -- though presumably dead -- human whose head it is.
No opacity problems (basic rule is that the dead are always with
us, though the not-yet-born may not be).

x:
The question is, do we
want/have a general way of treating the opacity that every place
structure could have? I say that we do:
        ko'a kalte lo'e cinfo
        She hunts lions.
        ko'e zbasu lo'e zdani
        He builds houses.
        ta stedu lo'e remna
        That is a human head.
        mi nelci lo'e plise
        I like apples.
        do pilno lo'e valsi le nu ciksi lo'e logji
        You use words to explain logic.
        ti katna lo'e pelji
        This cuts paper.
        lo'e jinci cu katna ta
        Scissors cut that. (That is cut by scissors.)

        lo'e junri jajypre cu mutce djica le vi pixra
        Serious collectors want this painting very much.
        This painting is much wanted by serious collectors.
        This painting is much wanted by the serious collector. Here I am
not saying that there are some serious collectors that want the painting,
but rather I am mentioning a property of the painting, which is "being
serious-collector-wanted". The x2 is in this case transparent:
pc:
I take it that the first two of these are simply false and the next two may
be true but are not at all opaque.  Third group, of three, I am not sure I
understand but none seems to involve opacity, except perhaps after
_ciksi_.  But then it is just false, for even if I did explain a typical
bit of logic (I am not sure what that would be), I would need very
specific words to do so, not typical (or whatever) ones at all.  As for
the final case, x2 is transparent only by the convention (which is surely
right) about _levi_ (indeed, maybe _le_ altogether) giving external
reference.  And x1 is not opaque, since quantification and identity
interchange present no problems for it (though just what the quatification
of _lo'e_ is is a problem), the referent is not outside the universe of
primary discourse. You are not saying there are some collectors who want
it because of _lo'e_, not because of the context of the sumti.

I think most places can become opaque in certain circumstances, but I do
not think they can simply be taken as opaque at will (I am not even sure
what that means).  And I do think we should mark them when they are
opaque, even over the markers in the context (aspect, professionality,
vocabulary items, etc.), but we also need to be able to circumvent the
implications of opacity when the road round is available.
pc>|83