Essay on Subcategorization

[Colin Fine <C.J.Fine@BRADFORD.AC.UK>]

ESSAY ON SUBCATEGORISATION
========================

(Notes and explanation of lojban terms are at the bottom)

There are two features +/-mass and +/-set which every terbri
and every sumti may have; they may however be unspecified
(~mass, ~set)

Thus most terbri, I suggest are
        -set
but
        se cmima is +set
and many (eg 'cinri') are ~set

most terbri are ~mass. (see note 1)
Some, eg gunma, are +mass.
I can't think of any that are -mass, but there
may be a few.

In addition, gadri may have these features.
lei/loi are +mass ~set
le'i/lo'i are +set -mass (note 2)
le/lo are ~mass ~set

When you construct a gadysu'i
it accumulates features from both the pamoi terbri
of the selgadri and the gadri.

Examples:

badna   -set ~mass
se cmima        +set ~mass
gunma   -set +mass

le badna        -set ~mass
le se cmima +set ~mass
le gunma        -set +mass

lei badna       -set +mass
lei se cmima +set +mass (note 3)
lei gunma       -set +mass

le'i badna      +set -mass
le'i se cmima   +set -mass
le'i gunma      +set -mass (note 3)

When you construct a bridi by filling the terbri with sumti,
the subcategorisations must match.
If the terbri is
+feature
the sumti must be +feature
If the terbri is
-feature
the sumti must be -feature
If the terbri is
~feature
the sumti is free.

Examples:

le badna cu cmima       = the banana(s) is/are member(s) (of some set)
(-set ~mass)
le'i badna cu se cmima = the set of bananas is a set (has members)
(+set -mass)

If they do not, I am unsure whether what you get is meaningless,
or simply necessarily false:

*ti cmima le badna
(remoi terbri is +set, le badna is -set)

Other cmavo
-----------------------

la, lai, la'i clearly have the same subcategorisation as
lo, loi, lo'i, but I suspect that they are opaque to the
subcategorisation of their selgadri. Thus

la gunma = 'the thing(s) named "Mass"' is ~mass, not +mass
and
lai se cmima = 'the mass of things named 'Set'" is +mass ~set


li, lu and the other quote marks are -set -mass.

lu'a, lu'i, lu'o have the same subcategorisation as
lo, lo'i, loi respectively, and in general perform a
constructive operation (note 2).
Note that lu'a is ~mass ~set, not -mass -set: just
because the individuals have been extracted, does
not mean that they themselves can't be sets or masses.

All pro-sumti (cmavo be zo ko'a) are neutral, or more
specifically take the categorization of what they refer to.

I believe that a similar scheme applies to abstractions, but
I have not worked out the details yet.

Notes
-----
1. I am not completely certain about this (ie that most terbri
are ~mass). The claim rests on the proposition that

lo badna cu gunma

is meaningful. I believe it is, and that both the following are
true:

loi badna cu gunma lo badna ('The mass of bananas is a mass composed
                                        of bananas')
lo badna cu gunma lo selci  ('A banana is a mass composed of cells')

(as well as other claims such as 'loi badna cu gunma lo selci').

If this is correct, then 'badna' must be ~mass.  Then 'lo badna' is
~mass, and 'loi badna' is +mass, and both satisfy 'gunma'.

If I am incorrect about this, then most terbri (including 'badna')
are -mass, and 'loi badna' involves a construction (see note 2),
but
*lo badna cu gunma
is either meaningless or necessarily false.

2. There is a difference in 'level' between sumti with individual
gadri (le/lo) and set gadri (le'i/lo'i). One way to see this is
to observe that
        lo broda cu broda       necessarily
but     lo'i broda na broda     in general

I conceive this as being a 'construction' - lo'i 'constructs' a
set from the items fitting the selbri - but I am hard put to
be any more precise than this (and I don't want to get into
any arguments about whether the set of items is already
there or not!)

I think that this 'construction' is implied whenever a +feature
in the gadri overrides an explicit -feature in the selgadri, but
not when the feature is unspecified in the selgadri.
Thus, since most terbri are -set, most "lo'i broda" are
constructive.
If I am right that most terbri are ~mass (note 1) then most
"loi broda" are not constructive, but if on the contrary they
are mostly -mass then loi will normally be constructive.

3. It may seem odd that a sumti can be +mass +set, but I believe
this works:
'lei se cmima' means 'the mass of sets', and can satisfy
terbri that subcategorise for +set, +mass or both. It may
or may not construct a mass, depending on the view of
the ~mass feature, but will not construct a set.

"le'i gumna" on the other hand means 'the set of masses' and
is a set, not a mass. It is constructive of the set, but I
don't think there is any special consequence of overriding the
+mass with -mass.

Glossary
--------
badna              banana  x1 is a banana/plantain (fruit) of species/breed x2
bridi     bri      predicate    x1 (du'u) is a predicate relationship with
 relation x2
                        among arguments (sequence/set) x3
broda rod          predicate var 1  x1 is the 1st assignable variable predicate
cinri         ci'i interesting   x1 (abstraction) interests/is interesting to
 x2;
cmima mim cmi      member   x1 is a member of set x2/ belongs to group x2;
gadri gad          article    x1 is an article/descriptor labelling description
 x2 in
                        sentence x3, language x4, semantics x5
gadysu'i = gadri sumti = description argument.
gunma gum          jointly  x1 is a mass/team/is together composed of components
 x2,
                        considered jointly;
pamoi terbri = 1st terbri
sumti sum     su'i argument   x1 is a/the argument of predicate/function x2
 filling
place x3 (kindanumber)
selgadri = se gadri = description (which I take to mean the selbri introduced by
                        the gadri, using gadysu'i for the whole sumti)
terbri = te bridi = is an argument role of predicate x3