Re: solutions to sumti opacity

[ucleaar <ucleaar@ucl.ac.uk>]

Jorge:
> la and cusku di'e
> > >         ko'a viska lo'e gerku
> >
> > This means koha saw the generic dog, & probably (I don't know how
> > the scope of the genericity is decided) means that if X is a typical
> > dog then koha saw X.
>
> There is no X that is a typical dog, at least not a {lo'e gerku}, I'm
> not talking about ordinary dogs {lo fadni gerku}. {lo'e gerku} is not
> a transparent reference, it doesn't have an identifiable referent.

"Lohe gerku" does have an identifiable referent. There is only ever
one "lohe gerku", namely the "ideal/prototypical" dog - which is not
actually a dog but rather a concept or something transcendental.

When I said "if X is a typical dog" I meant "if X is conceptualized as
an instance of the ideal/prototypical dog".

> > > > (2) the x2 of djica, nitcu, troci & other intentional gismu should be
> > > > of siho-type.
> > >
> > > I don't like that at all. I prefer to be able to want and need objects
> > > rather than ideas. (And I like being able to say "this is needed", "this
> > > is wanted", without circumlocutions.)
> >
> > We've been through this before. If you want/need objects, then there's
> > implicit sumti raising, & the x2 will have to be transparent.
>
> Certainly the x2 will have to be transparent unless otherwise marked, but
> why is there implicit sumti raising in the transparent case?

Because wanting involves two bridi: one for the desiderative attitude
and one for the desideratum.

> If I say "I go to the market", is there implicit sumti raising? After all,
> what I really mean is that my location changes to coincide with the location
> of the market. Is there any predicate that doesn't involve implicit sumti
> raising?

"Gerku" doesn't involve sumti raising. "Klama" probably does, but this
never causes problems because there is no intentionality.

> > I suggest
> > that you content yourself with lujvo from djica, nitcu etc., with
> > transparent x2, & if you want an opaque reference, use djica/nitcu
> > with siho-type x2.
>
> If you have no problem with such lujvo, why do you have a problem with
> a gismu that would have the same meaning? Especially since such gismu
> allows you to say the same things you could say with the other
> interpretation but in a simpler manner.

I don't really care which is the lujvo & which is the gismu. What I
do care about is (a) if an object can be the x2 then reference must
(and should) be transparent, and (b) we need a rational way to handle
opacity.

> How would you say "this is needed" with the siho-type x2? Something
> like {le si'o du ti se nitcu}, instead of {ti se nitcu}.

I'm not sure what the possible referents of "ti" are. If it can
refer to a thought then "ti" could be a siho-type x2. If you
want to refer to a book you need to read, then you would say
"lo siho mi tcidu ti kei se nitcu".

The lujvo "xxxxx zei djica" would be defined as:
   x1 xxxxx zei djica x2:
   x1 djica lo siho x2 cohe

> Is there anything that is easier to say with the siho-type interpretation?

No. But it makes things more rational: the syntax is a more faithful
reflection of the meaning, with the useful consequence that logical
problems of opacity go away.

> To make the siho-thing work, you'd have to be able to identify each and
> every gismu that likes opaque referencial sumti, otherwise you are not
> being very consistent.

I quite agree. This is really something that's already been underway,
since there's been a long-standing attempt to get rid of sumti-raising
and have syntactic structure correspond more accurately to semantic
structure.

> Why not use a solution that is general and not
> dependant on particular sumti?

Because the problem stems from faults in sumti structure of particular
gismu. We have to accept that these "faulty" gismu can't have opaque
sumti, or we have to redefine their sumti structure.

> And why do you allow lujvo to have
> meanings that you don't allow for simple gismu. There shouldn't be any
> difference in the types of predicates that are allowed for gismu and
> for lujvo.

I didn't mean this. I just feel it makes more sense for the gismu
to be the basic general expression, with lujvo derived from it
for purposes of brevity.

> > >         do djica la'e lu mi ponse le cukta li'u
> > >
> > > Otherwise, what you want is the sentence "mi ponse le cukta".
> >
> > But we don't want the referent, either, do we? Or can we say that
> > the referent of an utterance can be a thought? If so, then yes to
> > "lahe".
>
> Well, you don't want the thought itself either, you want what the
> thought expresses, and that is what I understood {la'e} to mean.

It is confusing to use "want" as a gloss: it obscures the problem.
The x2 of djica should refer to an idea whose realization pleases me
and whose nonrealization displeases me. Given this, I think (in a
muddled sort of way) that "lahe" is wrong for our purposes, and instead
we need either:
  (i) "lu broda lihu" means either "the sentence 'broda'" or
      "the thought 'broda'", or
  (ii) We need an analogue of "lahe" that means "the idea of", e.g.
       "xahe" in "xahe lu broda lihu", or
  (iii) We need an analogue of "lu" that marks quoted thoughts, e.g.
       "xuhu" in "mi djica xuhu mi viska do lihu"

---
And