[YG Conlang Archives] > [jboske group] > messages [Date Index] [Thread Index] >
-- In jboske@yahoogroups.com, "Jorge Llambías" <jjllambias@...> wrote:
>
> On 9/26/06, John E. Clifford <clifford-j@...> wrote:
> > --- In jboske@yahoogroups.com, "Jorge Llambías" <jjllambias@> wrote:
> > > But the
> > > new model is not a model containing only John and elephants in its
> > > domain of discourse, the new model contains unicorns too.
> >
> > OK, do it that way. But then notice that we go back to the old model
> > almost immediately, retaining the new sentence as justified.
>
> I would say that for as long as we retain the new sentence, we have
> not gone back to the old model, which did not include the sentence.
> Model = Domain+Theory.
And "Theory" is just the sentences held to be true? This is, of
course, a different sense of theory from the usual one, so part of our
problem may be merely teminological. But I suspect there is more to
it than that. You really do hold that any referring expression (in a
fairly narrow sense apparently -- in Lojban, descriptions and what are
tied to them somehow by identification) whereever it appears in a
sentence has to have a referent, even if the purport of the whole
sentence is that it has no referent. And that referent has to be in
the main (you would say the only?) developing situation (a better word
than "model" for a limited domain and set of sentences taken as true),
not in some temporary and suboridinate situation which may be part of
the truth of some first level claim. I obviously think the opposite
position here (well, I agree that referring expressions refer --
somewhere). This is a matter of how to comnstruct a linguistic -- or
a logical -- theory. All I can say is that several people seem to be
doinhg so in some variation on the position I've sketched here and no
one I know of has taken your approach, which seems to me to just not
take the linguistic facts into consideration.
>The new sentence cannot be in the theory if
> it's about something not in the domain, and in any case a change in the
> theory is a change in the model.
In your sense, granted. In the more familiar sense, we have merely
exposed another sentence true in the usual sort of model, but have not
changed it. (I suppose you could call it a change after all, since we
have added a new condition on the model into which the situation we
are developing must be embedded.) The point is in any case that we
can add or expose this sentence without adding anything to the domain.
Even that takes a bit of rethinking, I see, since we have added an
event (well, the up of an event probably) to the domain, although --
as you know -- I assume that these are present in every domain and so
this is really not an addition, only exposing something there so
automatically as not to be worth mentioning before hand -- it puts no
new restrictions on the (ordinary) model into which the situation is
to be embedded.
> > Every
> > model contains inherently other models to which it makes passing
> > reference in the course of dealing with its situation. The word
> > "want" (or {djica}) calls up such a subordinate model, a wantworld.
> > Tha lasts for a short time in the course of a conversation and then
> > folds back out of the original model, but it leaves the expression
> > ofthe wish behind as a truth of the original model (just as a reductio
> > proof leaves ~p brehind after closing off the model built on p).
>
> Model_0: Does not contain p in its theory.
> Model_1: Includes p and turns out to be inconsistent, so the model
is ditched.
> Model_2: Includes the theory of Model_0 plus ~p. The examination of
Model_1
> was helpful to determine that Model_0 could be expanded to Model_2.
"Had to be expanded" is perhaps more to the point. A situation
ultimately must contain all the logically entailled + claims. But
that is not entirely relevant here.
> Model_0: Contains nothing about unicorns.
> Model_1: Contains unicorns in its domain and the sentence "John wants
> unicorns" in its theory.
> Model_2: Adds the sentence about unicorns to the theory of Model_0 but
> not the unicorns to the domain?
Yes, that is about the way it goes. The internal makeup of Model 1 is
a bit more complex since contains not necessarily a unicorn directly
but rather a further sub model -- or several -- that contain unicorn.
Let's take a case
Model 0 as above.
Model 1: adds a unicorn to the domain and the sentence "John has a
unicorn" to the theory -- the "a unicorn" taken as referring to the
unicorn in the domain.
Model 1a: Model 1 plus the sentence "John's want is satisfied".
Pause
Model 2: Model O plus "John's want is satisfied"
Model 2a: Model 2 plus "John has a unicorn", with the referent unicorn
added to the domain.
Pause
Because of models 1a and 2a, we come to model 3
Model 0 + "John wants a unicorn" The unicorns more or less referred to
here are buried in the subordinate models as a basis for the present
one (I think the structure of models 1a and 2a are a bit more complex
but this will do for now).
Alternatively we add an event sense and the unicorn turns up, if at
all, in the world in which that sense is realized.
> I find this Model_2 unsatisfactory.
I hope the intervening models will help you out -- though I am not
sanguine.
>
> > > If you are happy interpreting that sentence as involving hidden
> > > propositions, that might work. I prefer an analysis that does not
> > > make use of hidden things. For me all that the interpretation of
> > > {la djan djica lo pavyseljirna} requires is a domain of discourse
> > > with (at least) two members, John and unicorns. The sentence
> > > can then be added as a true sentence of the model. It does not
> > > entail that {lo pavyseljirna cu zasti} must also be a true sentence
> > > of the model, but then why should it? I do not accept that
> > > {ro da zasti} has to be a true sentence of every model.
> >
> > We largely agree, except that I hold (with most other folk who do this
> > sort of thing) that {la djan djica lo pavyseljirna} can be true in a
> > model that has no unicorns in it, just John even.
>
> I assume you mean "I hold that 'John wants unicorns' can be true
> in a model that has no unicorns in it", i.e. you hold that the English
> sentence cannot be formalized (in the relevant interpretation) as an
> ordinary FOPL formula "Fab". Or do you really mean that the Lojban
> sentence {la djan djica lo pavyseljirna} cannot be so formalized?
As matters now stand, the Lojban sentence is ambiguous. Under one
reading it can be formalized in that way, with {lo pavyseljirna} in
the "b" position even. But the, {lo pavyseljirna} refers to some
particular unicorn(s), "any one will do." Under the other -- and more
common -- reading, it cannot be, precisely because this would entail
the particular generalization, which need not hold -- even if there
are unicorns (having unicorns in the domain just doen't make that much
difference when wants are the issue). The point of this exercise is
to find some way to disambiguate these two readings in a simple
fashion. Now, as a help in this endeavor, we are trying to figure out
what is different about the two different readings in terms of what
went into their deep structures to give rise to the identical (though
unofficial) surface forms. We could, I suppose (and, indeed, often
decide) just make up a mark to show that one of them is different and
mark this case and all other similar cases (opaque contexts) the same
way. But it would be nice to have some notion what the mark stood for
in the history of the sentence. So far, you have only explained one
reading and, as such, your explanation is surely about right. The
question is ow how to get the other reading. but your present
positions seems to be that there is no orther reading, which flies in
the face of everyboy's linguistic experience.
> > I am not a pro
> > enough to lay out all the arguments for this nor, indeed, to explain
> > all the details of how and why it works (even one version of it, let
> > alone the several that come to the same point along different paths),
> > I merely note the consensus -- which, I admit, agrees with my
> > intuitions and so ranks ahead of possible other approaches that have
> > different results.
>
> None of the relevant articles in the Stanford Encyclopedia site seem to
> be aware of there being such a consensus.
Do any othem talk about this approach or offer alternatives. I would
be especially interested in the latter, since I have not come across
any such theories.
> > But, on the other hand, I don't know of any
> > professional approach that goes your way (of course, I stopped looking
> > when I found one I liked). The main line among the people I like has
> > been that doing it your way 1) ignored linguistic data -- like the
> > scope problems and 2) was clearly influenced by a surface sructure
> > which clearly had a more complex deep structure (from that data
> > again). As I have pointed out many times, I agree with that
> > assessment and, thus, think your proposal is merely a lsst-ditch
> > effort to save what is simply malglico in the context of a logical
> > language.
>
> For my part, I find nothing in logic that settles the ontological
issue, and
> would thus force a preferred ontological perspective on a logical
language.
> Here's another interesting article that bears on the matter:
> <http://plato.stanford.edu/entries/nonexistent-objects/>
I don't think this is an ontological issue. We are not in any case
(except a strict real world model) concerned with what exists, only
with what the variables range over, what is. And even that is not
about reality in any sense but merely about how a particular theory
fits together to produce certain results. It turns out that, to get a
sentence like {mi djica lo pavyseljirna} (narrow reading), we don't
need to have unicorns in our domain. Indeed, having them there
complicates matters slightly. So, what is the best derivation of the
narrow scope want for Lojban and how might we best indicate that,
differentiating it from the broad scope case?