Query (medium long)

[chris-culy at csli.stanford.edu]

Hi all,

	Recently I've been looking at pronominal systems, and trying to
find some regularity in them. In particular, I've been looking at the
personal pronouns and trying to predict their properties from the
properties of the anaphors. Here are some ideas.

	In looking at pronominal systems, it is helpful to see them as an
amalgamation of subsystems. For example, English has only objective
reflexives, but no possessive reflexives, or nominative reflexive forms.
Danish has three types of third person object r eflexives, but only one
possessive reflexive. Other languages, for example French and German, have
distinct (non-emphatic) reflexives only in certain persons and numbers. It
seems that every subsystem will have one form that either has no binding
conditio ns (e.g. the non-emphatic pronouns in Old English and Bambara),
or has only disjointness constraints. These forms are the personal
pronouns.

	One assumption is that for any pair of NP positions, where one
commands (in the relevant sense) the other, there must be some
referentially dependent form (reflexive or personal pronoun) which can
occur as the lower NP and corefer with the superior NP. W e can refer to
this as Effability.

	Three natural principles for determining the binding properties of
personal pronouns are given in (1-3). 

1. Overlap The overlap of personal pronouns with reflexives in the core
position pairs distinguished by the binding theory must be minimal. (The
core position pairs distinguished by the binding theory in Dalrymple 1993
(Mary's book) are:
	X, subject coargument of X
	X, non-subject coargument of X
	X, subject in same minimal complete nucleus (mcn) as X but not a
coargument of X
	X, subject in same mcn as X but not a coargument of X
	X, subject in same minimal finite domain (mfd) as X, but not in
same mcn as X
	X, non-subject in same minimal finite domain (mfd) as X, but not
in same mcn as X
	X, subject in same root S as X, but not in same mfd as X
	X, non-subject in same root S as X, but not in same mfd as X

The number of pairs in which either a reflexive or a personal pronoun from
the same substem can occur with the same antecedent must be kept to a
minimum.)

2. Domain The disjointness domain must be minimal. (The domain in which
the disjointness conditions on the personal pronouns hold must be as small
as possible.)

3. Simplicity Binding specifications must be minimal. (The binding
specifications of personal pronouns should have at most one non-logophoric
disjointness condition and one logophoric disjointness condition.)

	The idea is that ranking these principles in different ways (a la
Optimality Theory) gives us different types of personal pronouns. For
example, English object reflexives and Danish non-third person reflexives
have the same binding properties (bound to a n argument in the mcn), yet
the corresponding personal pronouns do not. English personal pronouns
overlap with reflexives (disjoint from coarguments), while Danish
non-third person personal pronouns do not (disjoint from arguments in the
mcn). Thus, in En glish, Domain outweighs Overlap, while for Danish
non-third person forms, Overlap outweighs Domain. Danish third person
forms, however, overlap with reflexives, i.e. Domain outweighs Overlap.
Interestingly, Danish and Norwegian third person reflexives hav e
different properties, yet the personal pronouns have the same properties.
This fact also follows from the principles here.

	There are a few cases where Simplicity outweighs the other
principles. Toro So (Dogon) non-third person pronouns are an example: they
simply have no binding constraints (the simplest possible) even though
there are non-third person reflexives. It is also possible that Effability
is not an inviolable principle (e.g. in Marathi, Malayalam), in which case
Effability would be a principle ranked lower than Overlap, Domain, and
Simplicity.

	So far, I have looked at Bambara, Danish, Dogon (Donno So, and
Toro So), English (Old and Modern), Malayalam, Marathi, Norwegian, and
Turkish. Malayalam, Marathi, and Turkish are the hardest to do. Based on
these languages, we can propose an unmarked ran king of the principles: 

	(Effability) < Overlap < Domain < Simplicity

What I would like is information about potential counterexamples to the
notions here, especially languages in which Effability does not seem to
hold. 

Thanks,

Chris
chris-culy at csli.stanford.edu