Re: LE and VOI

[And Rosta <a.rosta@UCLAN.AC.UK>]

I wrote (with horrible typo corrected):
> >So the following combos are useful:
> >
> >1 nonspecific, veridical
> >2 specific, veridical, "indefinite" (= referent not (necessarily)
> >   identifiable by addressee)
> >3 specific, nonveridical, "definite" (= referent
> >   identifiable by addressee)
> >
> >Function 1 is performed by {lo}. Functions 2 & 3 are both
> >performed by {le}. Both 2 & 3 are useful, & it wd be nice to
> >think of an easy way to distinguish them.

In other words, veridicality correlates with identifiability of
referent.

John wrote:
> The closest analogue is "lebi'u" for 2 and "lebi'unai" for 3.
> "bi'u" signals "new information", so "lebi'u nanmu" =
> "a man, a certain man", whereas "lebi'unai nanmu" =
> "the (previously mentioned) man".

This is only a subset of 2 and 3. For example if we have
been discussing two books, and I then want to refer to
just one of them, the referent is old info yet not necessarily
identifiable. And likewise, if I out of the blue refer to all
the members of a given class (e.g. the messages I posted last
week, or the president of the USA), then the referent is
identifiable but new info. (In this second case, though, it is
at best dubious whether it is specific, so it is probably
correct to think that {specific, "definite"} entails "old info".
However, {specific, old info} doesn't entail "definite" - that
is the first of my points above.

Jorge:
> Your martini example shows that combo 3 is useful, and that's
> what we have. I don't see in what case is combo 2 useful. My
> impression is that when you have specificity, veridicality
> becomes irrelevant. A specific reference is just a tag on the
> referent, much like a name. It doesn't make sense to ask
> whether a name is veridical or not: all it matters is whether it
> succeeds in identifying its referent or not.  When you want to
> make a claim about a specific cat for example, the tag "mlatu"
> is usually the most convenient, and so people will usually
> prefer "veridical" tags, not in order to make a claim but to identify
> the referent easily.

{lo mlatu cu xekri} can be rephrased as {da mlatu gi`e xekri}:
it makes no difference whether {mlatu} is in the sumti tail
or is the brivla. (It does make a diff when there's universal
quantification.)
Now if one wishes to refer to a specific entity, X, and one
wants to say that X is a cat and X is black, how do you say
it? Neither {le mlatu cu xekri} nor {le xekri cu mlatu}
do the job, because of the nonveridicality factor.
As another example, suppose I want you to read a certain book,
though I'm not sure if you know which one it is.
What do I say? Both {ko cilre le cukta} and {le se cilre be
ko cu cukta} are both subject to interference from nonveridicality.

> Even for indefinite specifics (I take that to
> mean things like "the oldest cat in the world", is that right?)

I think that's a "definite" nonspecific. Universal quantification
over a singleton set, or over a set of known extension.

I'm putting "definite" in (s)carequotes because I'm using it
to mean "identifiable by addressee", whereas in most discussions
of logical semantics, definiteness is what "the" signals, and
I do not believe that "the" signals "identifiable by addressee".
(I believe that "the" signals universal quantification over an
identifiable set. But even though I am Right, there is no
consensus on this perennially debated matter among logicians/
semanticians.)

> even for them veridicality is not required: is the description
> enough for your audience to identify which referent you mean,
> at least in principle?

Not for type 2.

> Then there's no need for you to be claiming
> that your referent actually is the oldest cat in the world. You may
> go ahead and claim that as well, but your claim will be obvious to
> your audience anyway if they identified correctly what you referred
> to with your tag.

This is type 3 that you're talking about, really.

> For specific reference, veridicality is mostly irrelevant.
>
> For non-specific reference veridicality is useful, because it
> is a convenient way of informing your audince over which set
> you're running your quantifiers.

I wonder whether we are at cross-purposes over terminology.
You seem to have misunderstood me, and though I am easily
misunderstandable, you are usually the last person to do so.

> You forgot combo 4 in your list above:
>
> 4 nonspecific, nonveridical
>
> which is what we have in things like {su'o le mlatu} = "at least one
> of the cats of which I'm talking about". Here the set over which we
> quantify is referred to nonveridically. (In most cases it will consist
> of real cats, but it doesn't matter as long as our audience
> understands which so-called cats are the members of the set over
> which we quantify.)

I would consider this a case of Type 1. A nonspecific instance
that really-is a member of a certain set X, (the members of which
are described as cats).

So I still feel that Types 2 and 3 are distinct, equally useful,
and yet both covered by {le}.

--And