* Saturday, 2011-08-13 at 15:02 -0300 - Jorge Llambías <jjllambias@gmail.com>:
> On Sat, Aug 13, 2011 at 9:47 AM, Martin Bays <mbays@sdf.org> wrote:
> >
> > OK. I would like to semi-formalise this understanding as follows:
> >
> > The interpretation of a sumti (or more accurately: a sumti-6) consists
> > of a set of referents and a distributivity flag. The distributivity flag
> > has three settings: Distributive, Collective, and Ambiguous. When used
> > in a bridi, the bridi is respectively claimed of each referent, or of
> > the referents as a gunma, or ambiguously between the two options.
> > {lo broda} and {loi broda} both return Sumti6 whose referents each
> > satisfy broda; the difference is just that the flag is set to Ambiguous
> > in the first and Collective in the second. lu'a and lu'o reset the flag,
> > but have no other effect. When quantifying (non-fractionally) over
> > a sumti, the flag is ignored.
> >
> > Does this accurately capture the intention of xorlo?
>
> Not really. xorlo is about "lo", it has absolutely nothing to say
> about "loi" or about masses.
Fair enough. I was using 'xorlo' as shorthand for 'the bpfk gadri
proposal'. But it seems that the definition there of {lo} *does*
interact with masses/groups, because its referents are explicitly
allowed to be groups.
> An ordinary sumti-6 has referents, and that's all it has. The ordinary
> sumti-6 says absolutely nothing about how its referents will
> distribute when used as the argument of a predicate. Ordinary sumti-6
> are things like "ta", "mi'o", "lo gerku", "la djan", "li ze", "lo ka
> ce'u barda", "zo bu'u", "lu mi klama lo zarci li'u", "lo'i manti", and
> so on.
(Sadly there are also extraordinary sumti-6, such as {lu'o mi .a do}...
but I guess this is what you mean to exclude with 'ordinary'.)
> There is no consensus on what a sumti-6 headed by "loi" does, since
> some people think it doesn't just have referents, but it also says how
> its referents play in a sentence when the sumti is used as the
> argument of a predicate. That to me seems like too much to ask of a
> gadri, since there are so many possibilities of how they could
> distribute: one by one, all together, in pairs, in twos and threes,
> some individually and others in pairs, and so on. Having a special
> gadri for the "all together" case seems to me like the wrong way to go
> about it, but it's a popular description of "loi" (yet not the only
> one).
>
> In the particular case of sumti-6 of form "lo broda", we know of its
> referents that they broda. It is probably not required that they broda
> one by one, they may broda together, but this is an open question. The
> point of xorlo was mainly about removing any implication of how or
> whether the referents will distribute when used as an argument, not
> about how they broda.
My feeling is that the level-mixing ambiguity which allowing group
satisfaction of broda in {lo broda} would introduce - {lo besna} could
have its referents being neurons, and generally {lo gunma be lo broda},
with {gunma} as per your definition below, could have its referents
brodaing - would be an ambiguity too far.
Disallowing it does make getting at the individuals which collectively
broda more difficult - {ro lo selgunma be lo sruri be lo dinju cu krixa}
wouldn't be too bad, but see below where I argue that this use of gunma
is too ambiguous.
For what it's worth, my understanding of CLL was always that group
satisfaction was disallowed - though probably it wasn't stated
explicitly.
> As for the meaning of the gismu "gunma", my take is that it means "x1
> consists of x2" or equivalently "x2 constitute x1". x1 is a whole and
> x2 are its constituents. (Note: "x2 are *the* constituents of x1", not
> "x2 is/are among the constituents of x1", which is a different
> predicate.)
And more generally, would you drop CLL's upwards closure axiom for
"masses", such that you can't have {lo plise} having as referent a group
whose constituents are an apple and a badger?
However, I don't think that it is so useful to have the 'constituent'
relation as an ordinary selbri. A magic cmavo might be better.
Consider : we would always have {ko'a gunma ko'a}, where we interpret
collectively on both sides.
Similarly, the referents of {lo selgunma be lo sruri} could be the
individuals which as groups sruri, but they could also be the sruris
themselves - depending on whether {lo sruri} is taken to sumti
distributively or collectively.
> Thus if "loi broda" does mean "lo gunma be lo broda" then "loi broda"
> is an ordinary sumti-6, and its referent then does not necessarily
> broda but rather the constituents of the referent are the ones that
> broda. (In this case "loi broda" will typically have a single
> referent, just like "lo'i broda" will typically have a single
> referent.)
This seems very sensible.
> But nothing of this is stipulated by xorlo, which is about
> "lo", not about "loi" nor about the semantics of "gunma". I rather
> keep that can of worms separate from "lo".
I understand that (and I note that your .alis never uses {loi}/{lei}),
but these issues do need to be settled sometime. And if you want
arbitrary sumtis to be able to sumti collectively, then getting a good
theory of collectivity is necessary even to understand {mi}, nevermind
{loi}.
So allow me to recap how I would like to understand all this:
The data in the interpretation of an ordinary sumti-6 is just a set of
individuals, its referents. When it sumtis, whether it does so
distributively or collectively is ambiguous. {lu'a} and {lu'o} can be
used to disambiguate (forming extraordinary sumti-6). Non-fractional
quantification unambiguously quantifies over the referents.
The referents of {lo broda} are such that
ONE OF (currently unsettled)
(i) each referent satisfies broda, i.e. {ro lo broda cu broda} is
tautologous.
(ii) the referents either each satisfy broda, or they collectively do,
i.e. {ro lo broda cu broda .ija lu'o ri cu broda}
This constitutes the definition of {lo broda}, i.e. no further
information about the referents can be deduced (modulo usual assumptions
of contextual relevance).
A group is a kind of individual, so a possible referent of a sumti-6.
A group has as data a set of individuals - its constituents. Things
"collectively broda" iff the group whose constituents are those things
brodas.
(I think John Cowan would want to disagree here, and say in particular
that the group with only one constituent should not be distinguished
from that individual. Is that right, John? But this seems not to fit
with your doi xorxes account of loi.)
The exact semantics of when a group brodas depends on broda. Perhaps the
x1 of sruri is upwards closed, but the x1 of plise certainly isn't.
In what ways is that summary inaccurate?
Yours,
Martin
Attachment:
pgpHh8UADqcJZ.pgp
Description: PGP signature