Message-Id: <199512130300.WAA01790@locke.ccil.org>
Date:         Tue, 12 Dec 1995 14:35:23 -0800
Reply-To: "John E. Clifford" <pcliffje@CRL.COM>
Sender: Lojban list <LOJBAN@CUVMB.BITNET>
From: "John E. Clifford" <pcliffje@CRL.COM>
Subject:      Re: tech harangue on le/lo
To: John Cowan <cowan@LOCKE.CCIL.ORG>
Status: OR
Content-Length: 7300

Sorry guys, opacity is out of the bottle again; time to get back to
your McCawley -- not that he or anyone else helps much.  The
places where quantifying in doesn't work -- intensional contexts,
intentional ones, and with propositional attitudes -- are also the
places where substitution of identity doesn't work (and add
quotation for this latter case).  They are all cases where sentences,
variously buried, refer to situations other than the real one (being
deliberately vague here because there is no settled answer about
just what these buried sentences do refer to, except in the case of
literal quotations -- even indirect quotations are controversial).
While the two debates are not the same, it is usually assumed that -
- at least for the overlapping cases -- an adequate solution must
solve both problems.

I note in passing that the referents of the buried sentences are not
propositions in the sense of intensensions of sentences, since all
univerally true sentences have the same intension but cannot be
substituted for one another in these contexts, salva veritate.  The
best guess is that they are structures like sentences constructed of
the intensions of words as sentences are constructed of words.  But
no one knows what that means except that it seems to cover the
known relations of these critters to propositions on the one hand
and sentences -- direct quotations -- on the other.  Of course it does
not deal with the problem that we can often do moves about totally
different languages, with -- one assumes --different structures of
different words with different  intensions.  All of these problems
have been touched on in these discussions on le/lo. As has the
question about the existence of these critters (or some related ones
-- events) when they are not realized (however that may be) in the
real (or the discursive) world.  Lojban does not answer any of these
questions but has a largely uninterpreted maker for these critters
and alows the le/lo of them to instantiate quantifiers -- whatever
_that_ portends.

On opacity, I remember (I am catching up from two weeks of end-
of-semester work which just gave me time to pull things down, not
organize or read them, so I do not know where I saw this) someone
jumping from the claim that someone was hunting loi <unicorn> to
the claim that there were unicorns.  I pass over whether a loi is the
right thing to be hunting, to remind (because we did go through
this less than a year ago) all that "hunt" and its Lojban translations
create an opaque object position, which strictly in Lojban ought to
have a "subject raising" mark (_tu'o_ ?  not what my list says but I
seem to recall that the list I have is wrong) unless cancelled (as it
is not in this case) by an external reference mark (one of those
_xe'V_ from that discussion).  "Hunt" and the Lojban as well have
the deep structure of a verb ("get," "kill," "catch," or some such) in
the scope of a "strive to bring to pass" main verb, so the surface
object actually comes up two semantic layers, even if no syntactic
ones.

On whose choice is it anyhow, for the _le_ in a belief, say, the
uniform practice in natural languages is that it is the speaker who
reports the belief (which he might, by the way, figure out in a
number of ways other than getting the believer to state his beliefs -
- indeed, might find out more reliably than by asking).  In most
languages, these reported beliefs follow the linguistic pattern
(whether or not the logical one) of reported speech, with the
corresponding shifts.  So, a person who believes what he might
express as "I am handsome" is reported as believing that he is
handsome, i.e., from the reporter's, not the believer's, point of
view.  Of course, if the reporter is to report correctly (one of the
rubs), he must, with a _le_, match the believer's own selection with
his own.  And that may be hard to do, since the believer may have
things available to him that the reporter lacks (and -- less crucially
in this case -- vice versa).  In fact, the problems that circle 'round
these expressions largely come just from our trying to frame in our
framework what the believer frames in his.  Is it McCawley that
talks about Commissioner Gordon knowing that Batman is a
millionaire?  Where, of course, the Commissioner knows only that
Bruce Wayne is a millionaire and does not know that (or even
believe) that Bruce Wayne is Batman.  Now, Lojban could, I
suppose, try to keep it in the believer's frame, but the results would
be sometimes unnerving and sometimes just about unintelligible
(but at least not wrong): consider "He believes that I am
handsome" or trying to figure out who there to be selected in a
_le_.

But back to the central point.  There really is not any controversy
about _le_ and _lo_.  Well, almost none: some still hold that _lo
broda_ makes sense when there are no brodas, and at least _ro lo
broda_ may.  That aside, _lo broda_ always refers distributively to
some (generally unspecified at that point) subset of the set of all
brodas, things that really do have the property _broda_ refers to.
On the other hand, _le broda_ refers distributively to all of a
selected set of things, which selected things the speaker choses to
label "broda."  Semantically, there are no restrictions on what
things are in the selected set, they can be any kind of things and
indeed can be any mixture of various kinds of things.  They are
referred to by _le broda_ simply by the speaker's decision so to do
(well, the first speaker's, since other may follow the set
expression).  In short, _le broda_ is not an analysable expression, a
function of _broda_, but a block, even more so than _la djan_,
since the content portion does not have a related independent use
(unlike the vocative use of _djan_).   Howmsoever.  If the speaker
expects to be understood by a cooperating audience (and expects
to be taken as cooperating and.... through all the Gricean stages),
then
if he introduces _le broda_ out of the blue, he'd better be talking
        about things that are brodas or that the audience thinks
        are brodas
if he enters into a situation where brodas are already in the
        universe of conversation, he had better be talking about
        some them -- and make it pretty clear which ones as
        well (_le broda_ is a good pick up for an earlier _lo broda_
        with the same referent -- a second _lo broda_, like a second
        quantifier, allows a new referent)
if he enters a situation where brodas are excluded, he had better
        intend something recognizably like a broda for his
        audience.  He can escape these strictures unscathed only either by
a long, open and notorious practice of misusing words in a systematic way
(and best if he has already abused _broda_ in the same way often) or by
pulling the punchline before people get puzzled by the discussion (roughly
three sentences -- 2/3 of one of mine).  JCB's "That woman is a man" is
safe on all counts.  Most other cases of using _le broda_ for non-brodas
are not.  But they are grammatically and even sematically legal.  (You
might want to look at the discussions of Dave Kaplan's "dthat," the
logical device closest to _le_ but, being Vulcanian, free from pragmatics
(I just threw that in to raise And's hackles). )