* Thursday, 2014-10-30 at 19:05 -0300 - Jorge Llambías <jjllambias@gmail.com>:
> On Wed, Oct 29, 2014 at 10:33 PM, Martin Bays <mbays@sdf.org> wrote:
>
> > Well, CLL is quite explicit in considering forethought and afterthought
> > to be equivalent. And it seems natural to me at least.
> It is reasonable, but I don't know how natural. With forethough
> connectives, the surface form places the two sentences clearly under the
> scope of the tag.
I see what you mean.
Actually, CLL prescription has a related wrinkle: it mandates in 10.16
something like the interpretation you're suggesting for forethought
tense connectives when they're connecting bridi tails or sumti (or, one
might suppose, the other few in-sentence places such connectives can
appear). Forethought tensed connectives of *sentences* are declared
equivalent to afterthought, however.
Meanwhile, 9.8 seems to have *all* kinds of forethought connection with
"modal" tags be equivalent to afterthought.
> With afterthought the second sentence is under its scope, but it's not
> so clear for the first. CLL says that both sentences are independently
> claimed when connected with a tag, which can be strange for some tags,
> such as "no roi", or say "fi'o natfe". .
Right, I'm understanding it as meaning that the two sentences are
claimed - but one has the tag applied first. When the tag is a tense,
the first sentence is claimed as-is, but only the tagged version of the
second is claimed. For non-tense tags, it then has to be the other way
round. This is counterintuitive given the surface form, certainly.
For the vast majority of non-tense tags, we have
(*) {[tag] [[sumti]] broda} always implies {broda}
, so this still has {broda .i [tag] bo brodu} implying {broda}.
But there are exceptions. That use of {fi'o natfe} is one, as is e.g.
{fi'o selbalvi} used to replace {ba}. I assume those have no usage
outside of metalojbanics; but {de'i}, {di'o}, {koi}, {ti'u}, and {tu'i}
(BAI4 in the BPFK classification), and probably also {va'o} and maybe
also {fau}, seem to have tense-like semantics despite being in BAI.
I wonder whether it wouldn't be neatest to declare these BAI, and all
fi'o, to be tenses for the purposes of tagged connectives - and declare
that all other tags satisfy (*).
But for now, perhaps we can just survive e.g. {broda .i ti'u bo brodu}
not implying broda?
> I find the forethought case easier in that I would read it as just
> a binary predicate with two bridi/event arguments,
So then a tag would be able to play three roles:
(i) combining with an optional term to form a modal operator;
(ii) acting in certain connectives as a binary relation on events;
(iii) acting in the {pe [tag]} construction as a binary relation.
It would be nice to unify (ii) and (iii). The problem is that the
relation in (iii) isn't restricted to events(/facts) like that in (ii)
is, so whereas in (iii) the obvious choice for a BAI tag is a relation
based on the first two places of the underlying selbri, that might not
make sense in (ii).
I don't see a natural way to fully reduce (ii) to (i) or (i) to (ii).
But I'm not convinced we need (ii) at all.
I'm reading CLL as having e.g. {broda [tag] gi ko'a gi ko'e} be entirely
equivalent to {[tag] le nu broda ko'a kei broda ko'e}. So that explains
tensed forethought connectives without having to go via (ii).
Except that again, considering negation makes me suspect it shouldn't be
{le nu}, nor {lo nu} nor even {zo'e noi nu}, but rather {su'o nu}. So
then the only difference between forethought and afterthought would be
in whether there's a claim that the seltcita sumti event fasnus ("now").
> The mix with logical connectives is the tricky case, but it becomes easier
> if we think of the tag as modifying only the second connectand, in which
> case "broda .i je tag bo brode" is just (broda) .ije (tag(brode)).
For tenses, yes. The only tricky bit is deciding what exactly the
seltcita sumti of the tag should be.
> > > (1) broda .i ba bo brode -> broda .i ba la'e di'u brode
> (3) broda .i ki'u bo brode -> broda .i go'i ki'u lo nu brode
Right... I guess that's one solution.
It doesn't seem very useful, though. Since {ki'u}, like almost all BAI,
satisfies (*), there's no reason to explicitly claim broda as well as
{broda ki'u lo nu brode}; and the latter expression is no longer than
{broda .i ki'u bo brode}.
> In my mind it was only PU (and presumably FAhA as well) that behaved oddly
> as connectives, but I see now that CLL mentions PU, ZI, FAhA, VA and ZAhO.
> For CLL however, ZAhO is irregular as sumti tcita, so in fact CLL and I
> would agree again on how ZAhO behave as connectives. We would both say that
> "broda .i co'a bo brode" means "it brodas, starting when it brodes", but
> we'd get there by two different roads.
Right, whereas using CLL rules for the connective but the BPFK section
meaning of co'a as a sumtcita, we get something like "it brodas, and
meanwhile starts to brode".
> carvi .i jo ki'u bo glare
> It rains iff because it's warm.
> (Either it rains because it's warm, or it doesn't rain because it's not
> warm.)
Good.
OK. So it currently seems to me that:
* quantifying over events is the right thing to do;
* for tenses: the second sentence gets tagged, with the seltcita sumti
being (the variable for) the event of the first sentence, which is
declared to occur (unless the connective is forethought and below
sentence level);
* for non-tenses: the same, but with the roles of the first and second
sentences swapped;
* when a logical connective other than {je} is involved: we have to
separately consider the cases that the seltcita sumti involved is an
event of the sentence and that it's an event of the sentence's
negation.
("event" may be read as "event/fact".)
Martin
Attachment:
signature.asc
Description: Digital signature