Message-Id: <199410092252.AA00807@nfs1.digex.net>
Date:         Sun, 9 Oct 1994 15:34:49 -0700
Reply-To: "John E. Clifford" <pcliffje@CRL.COM>
Sender: Lojban list <LOJBAN@CUVMB.BITNET>
From: "John E. Clifford" <pcliffje@CRL.COM>
Subject:      mehr Licht? opacity, etc.
To: Bob LeChevalier <lojbab@access.digex.net>
Status: RO

        Let me try this again (@#$%*&! machines -- and instruction books)
             Opacity (strictly "referential opacity") seems to be an
        inherent property (like needing an animate referent or an event
        one) of certain places of certain predicates and other
        expressions derived from predicates by (dare we call it)
        predicate lowering (like "possibly" from "is possible").  It is a
        logical property, affecting inferences from the one sentence
        which contains an expression in such a place to other sentences.
        In particular, it prevents us from exporting referential and
        quantifier expressions which occur in the opaque place out into
        the wider, non-opaque, sentence, and thus from further inferences
        which hinge on that move.  In particular (using [] to mark an
        opaque context), i prohibits all of
             [some x] => some x [x]
             [all x] => all x [x]
             [referring_expression] => referring_expression [anaphora]
        The last of these prevents our binding an opaquely placed
        referring expression with an outside quantifier, the usual
        referring_expression => some x, and prevents the referring
        expression from being available for universal instantiation (All
        x => referring_expression) outside the opaque context.
             The Lojban discussion (which goes beyond the usual logic one
        because of the greater expressive power of Lojban) suggests that
        opacity is closely tied with taking event arguments.  Most of the
        examples in the recent (and continuing) discussion have been of
        "subjects" raised from event-referring items.
             It is clear in any case that such event-referring items,
        when fully present, do generate by themselves opaque contexts.
        Referring expressions and quantifiers in event-referring
        expressions cannot be exported from those expressions to the rest
        of the sentence, pretty much regardless of what predicate the
        event-referring expression is argument to.  The only exception
        (if it is one) I can think of is the intransitive "obtains" and
        its synonyms (I am unsure of the Lojban for this, but it
        presumably is not _zaste_, for events exist even if they do not
        obtain, are not facts, are not real {How's that for metaphysical
        neutrality? ;)}): If that everything is so-and-so obtains, then
        everything is such that that it is so-and-so obtains.
             Apparently each event-referring expression creates its own
        world and its referring expressions and quantifiers deal within
        that world, which need not be the real one outside.  It is
        possible, however, that the external world overlaps that of the
        event-referring expression, so we may include in an event-
        referring expression variables bound outside the expression and
        referring expressions whose bona-fides have already been
        established. Those external expressions behave normally in opaque
        contexts where they end up: we can go from all x [x] to [a],
        where a is an externally established referring expression, for
        example. And even [a] => some x [x] for such an a.  But we cannot
        be sure how much more than stated the real world and the event-
        referring world overlap.  Thus we cannot -- in the absence of a
        reference to a in the opaque context -- do [all x] => [a],
        although this is controversial and it may be that different
        predicates place different restrictions on event-referring places
        ("need," for example, seems to require that all the real things
        be in its event-world, since any of them would satisfy the need
        as well as some dreamed-of one -- better probably; "dream" has no
        such requirement).
             Thus, many opaque contexts can be taken to just cases where
        be have an abbreviated event-referring expression, only the
        principle referring expression or quantifier (often NOT the
        subject, "subject raising" to the contrary notwithstanding)
        standing for the whole.  The rest is then "understood" -- or
        left hopelessly vague (the "that I have ..." of the "I need a
        ---" discussion, where "have" goes beyond -- and falls short of
        -- possession in any sense, depending on what --- is.) The
        invisible event-referring expression provides the opacity.  But,
        being invisible, it does not provide the warning that the normal-
        looking referring expression or quantifier is not to be treated
        normally.  At best, that information comes from the knowledge
        that the place where that expression occurs is one reserved for
        event expressions and the anomaly raises the warning flag.
             But, unfortunately, not all opaque contexts are obviously
        places for event-referring expressions. For example, the subject
        of a work of art (broad sense) is opaque: the chryselephantine
        statue of Aphrodite existed even though Aphrodite did not,
        similarly books about Sherlock Holmes and tone-poems about fauns.
        In these and many other similar cases, we might argue that what
        the work portrays is not just the named object but some event
        (maybe very generic - Aphro standing, say) in which the object
        participated.  And so we could insist that these places too are
        for events, with the "subject" habitually raised.
             But some cases seem to resist that.  It is hard to imagine
        that I am seeking an event, for example, when I am seeking a
        unicorn.  However, even here we have an event-centered out,
        namely that "seek" is a complex expression that has already
        absorbed the event into its semantics.  "Seek" is "strive to
        see", where "strive (for)" is an event-governing predicate, a
        more active version of "want" ("want and taking steps to bring it
        about") and "to see" is the heart of the event striven for, that
        <subject of striving> see <object overtly referred to>.  Thus,
        it may be that all opaque contexts can be covered by event-
        referring expressions -- at some level of the grammar.
             But we will just have to learn when these event-referring
        expressions are to be understood and that is not easy, as noted.
        It would be nice, then, to have a mark to remind us that we have
        occluded such an expression in getting to relic of it that
        actually occurs.  Since subject raising is a transformation and
        any transformation may have a visible mark, we might use say xe'e
        (or, rather, something out of xVV space once it is not
        experimental) to mark that transformation, even when it is buried
        in the lexicon, as in "seek".
             While we are in that area, we should note that English has
        afterthought quantifiers that can be uttered in contexts but
        which jump to the head of the utterance in which the context
        occurs, regardless of the messinesses -- opacity, negations, and
        the like -- in between.  "Any" does this for universals, "A
        certain" for particulars and there are various devices for doing
        it for other quantifiers and for straight referring expressions.
        It would be nice to have these in Lojban, too (and I think we did
        once decide we would but I do not know what became of that).
        They make a nice complement to the afterthought connectives --
        though we have never gotten to RPN.