Message-Id: <199411190417.AA06355@nfs1.digex.net> From: "John E. Clifford" Date: Fri Nov 18 23:17:08 1994 Subject: opacity and ol' uncle tom cobleigh Status: RO X-From-Space-Date: Fri Nov 18 23:17:08 1994 X-From-Space-Address: LOJBAN%CUVMB.BITNET@uga.cc.uga.edu Unmarked opacity seems to be at the heart of much of the recent discussion. Because the globally opaque context, fiction, goes unmarked in Lojban (as in most languages), we came to think that _lo_broda_ referred even when there were no brodas and even though we noted that "_lo_ seems to take us to a context where do exist." Based on this primal failure to not notice when we left reality (and a strange notion that "all" did not imply "some" and that "at least one" somehow disappeared) we came up with the default quantifiers on _lo_, which do not fit its grammar and make it the boar-tit of Lojban (duplicating two other forms at least and being less easy to anaphorize than either). Similarly, the unnoticed deep structure of "seek" (etc. - but I do not know the cetera yet) led us to make the second argument of _sisku_ a property rather than an object. This was a clever way to avoid the apparent problem of seeking a unicorn when t here were no unicorns. But it made it very hard to talk about seeking the things we usually do seek, things not properties. We do not yet know all there is to know about opaque contexts but at least we now recognize that event descriptions create them an d that "seek" is a complex that includes one and that one encompasses the second place, "strives to see...", where the "..." emerges to the surface as the second place of "seek". If we can deal with opaque contexts, then, we can regularize _sisku_ to fit in (note that even the usual objection, that raised subjects are so vague, does not apply here, since the predicate of the buried event description is fixed (though we do not say what we will do after we see the object). I am inclined to think that Chasell's puzzling pieces also have to do with opaque contexts or, at least, unmarked moves to leave or restrict the default real world. But I can't push that point, since I do not yet understand where he is going with this. I am stopped by questions like, if a sentence means different things depending on whether it is true or false, how do we find out whether it is true or false, since we have to know what it means to make that determination? (I do not think that possible-world semantics helps here much.) Opaque contexts seem to need three things to fit in with a logical language. First, they have to be marked so that we do not violate logic by inappropriate inferences involving them. Event descriptions clearly carry the mark as do imperatives and quest ions, maybe including indirect ones. I am less sure about the attitudinals, but they are marks, they can be interpreted as blocking inferences if that turns out to be the way to go. The problems are subject-raised sumti and the broad area of fiction (of all sorts). The first has _tu'a_ to indicate that the sumti is raised and thus from an opaque context and not to be treated as in the real world. For the second, I think we just need a fiction marker (or a pair or, better, a fiction and a reality marker). Anything after a fiction marker (I actually think we had these at some point and may still) and the canceling one are opaque to reality. I suppose this takes some of the fun out of things, but it is not without precedent in cultures with rich storytelling traditions. And we are trying for a logical, not a mendacious, language. There may, of course, be yet more opaque contexts, but we should have an array of devices to deal with them by the time we find them. The second thing opaque contexts need is a way to get back in touch with the real world from within. This is part of a general problem of afterthought quantifiers and the like, one of the uses of "any" and "a certain" -- and other context leapers --in English. I was sorry to see that Cowan withdrew his proposal for _lo_, since it was at least moving in the direction of making it -- and its quantifiers, default or expressed -- into such a leaper, and thus useful again rather than redundant. But I think that a better plan would be a local mark to indicate jump out of the context to the highest level available (and maybe others for steps in between?). I realize that I reversed the significance of Xorxes' _xe'e_ in suggesting it for that purpose, but it catches the general idea (and can we please try it quickly and then give it a real word rather than the VERY hard to say experimental _xe'e_?) I am not sure whether such a mark can unambiguously be used both to get out of intensional contexts and out of truth-functional ones like the antecedents of conditionals and negates, the classic home of leapers, but we should try with just one mark first. The leaper mark looks like it might also solve the problem of reference to the real world imported into -- rather than attempting to leap out of -- opaque contexts, looking for my real book, for example. _mi_sisku_tu'a_lemi_cukta_ does not get what we w ant, since the what maybe my book in the opaque realm need not be that in reality nor conversely. Tacking a _xe'e_ (or whatever) on may work and, since that is starting to get awkward, we might decree that the two cancel each other out, giving the very n atural and non-misleading _mi_sisku_lemi_cukta_. The danger with this is, of course, that we forget the _tu'a_ when there is no _xe'e_, when the reference is opaque. But then we just say something we did not intend, a painfully common event even in our native languages. We may also not learn to use the _xe'e_ when we are in other opaque contexts. (Having _lo_ be a leaper would ease a lot of these problems, alas!) Or, perhaps, we can some conventions about at least some of these contexts: no one seems to have problems with the opacity of "me" in the imperative "Give me a hammer", but the "hammer" is a problem, since I can equally we demand a nonexistent, like Santa asking for a unicorn. And it may be that we need a different mark for references that begin outside from those that begin inside -- another experimental term. I hope that we can get on with the experiments, but first that we clean up the definitions of a few words and at least the explanation for _lo_'s default quantifiers, even if we do not change them.