Rolf Noyer: Gender in DM

[Martha McGinnis]

Hello everyone,

This is a bit rambling, but ...

As I understand it, Aronoff *defines* gender as a property of a lexeme
which can be syntactically transmitted while inflectional class is
*defined* as a property of a lexeme which cannot be syntactically
transmitted.  So if we accept this definition then what we mean by "gender"
is precisely those properties of the vocabulary which, at least in more
conventional theories, "propagate" in the syntax in some way.  At least,
let's be clear on the fact that "gender" is *not defined a priori* so we
have to first agree on what it is we have in mind or whether we're talking
about a collection of unrelated phenomena which happen to be taxonomized as
"gender" for lack of a better term.

As Alec points out, it does not seem that the gender of the noun a root
makes has any effect on the verb the same root makes.  If gender were
inalienably associated with the root VI, then we might predict such
situations.  However, we might also construe gender as a contextually
"activated" property of a VI.  Might it not be possible to suppose that the
property [fem] is contextually activated when _caj-_ is inserted under n,
but otherwise this feature remains "dormant"?  But perhaps this is simply
not interesting.  It is more or less the old way of thinking about the
problem.  But I'm not certain it's wrong, either.

If we examine the gender assignment systems in many languages, there are
frequently subgeneralizations of the type, "the fruit is gender X but the
tree that produces the fruit is gender Y".  The question that concerns me
is whether this information is part of the Vocabulary or the Encyclopedia
or part of some system external to both.   Certainly I think the evidence
points to a system of subregularities which relates grammatical gender,
natural gender and inflection class, and this system of rules has to be
situated somewhere (for Spanish, Harris' "The Exponence of Gender in
Spanish" from LI is a good place to start.)

Following up on Karlos' comments, I'm not entirely sure I would consign the
property of gender to the encyclopedic entries.  After all, the fact that
_caj-_ means 'box' in the context of D/n is a property of the encyclopedia
entry for _caj-_, but the fact that _caj-_  is feminine (or, perhaps, has
the distributional requirement of occuring only under [fem] n, if that is
how we do it), does not strike me as "real world knowledge" but rather as
grammatical knowledge appropriate for encoding in the Vocabulary in some
way.  If we do this, the question becomes "how?"

"Licensing" of VIs, as Heidi points out, is an option.  But "licensing", I
have recently begun to feel more acutely, is a kind of "inverse" of
subcategorization of the Aspects-theory, and as such might be subject to
the same objection raised in Tim Stowell's dissertation, i.e. isn't there a
needless replication of the distributional properties of syntactic items
(i.e. syntax) encoded in the distributional privileges of the VIs? Any
comments on this?

I think the fundamental problem here is what precisely the function of the
Vocabulary is.  The role of the Vocabulary would seem to be two-fold: (i)
to provide phonological form to syntactically abstract expressions, and
(ii) to provide a "link" to the Encyclopedia.   Under (i) we can justify
the inclusion of inflectional class information in VIs, since this
information is crucial to the proper "spelling" of terminal nodes.
However, we can't justify "licensing" alone by (i) unless we take a fairly
liberal view of what it means for a terminal to be "properly" spelled-out,
in particular an l-morpheme, in the terms Heidi and I used (i.e. morphemes
for which Vocabulary Insertion is not deterministic).

With respect to pal-o/-a, if the Vocabulary has only the functions (i) and
(ii) above, then there could be a single Vocabulary Item /pal-/ provided
that the /pal-/ in _palo_ and the /pal-/ in _pala_ have all the same
properties that Vocabulary Items should have.  The fact that there are two
Encyclopedia entries "shovel" and "stick" does not affect how many
Vocabulary Items there are.  Unless -- the gender of _palo_ (and, mutatis
mutandis _pala_) is a property of the Vocabulary Item, in which case one
could argue for two Vocabulary Items.  But Separationism, taken literally,
entails that semantic polyfunctionality alone does not require two
Vocabulary Items.

Let's consider this problem:

Is  *_el caj-o_  ungrammatical "in the same way" as *_John arrived the
book_ or (in standard English) *_Mary and John plays the trumpet_?

Also, keep in mind that

*caj-o

could be "wrong" for two reasons: (1) the word-marker /a/ could be
'wrongly' inserted instead of /o/, even though morphosyntactic agreement
(or whatever replaces it) has operated properly; (2) morphosyntactic
agreement has not operated properly but /o/ has been inserted properly.

I suspect that speakers' judgments, when presented with an "ill-formed"
expression, may reflect the fact that derivationally the ill-formedness can
arise in different ways.  So when we try to assess the situation, we have
to be careful to consider all the ways in which a particular expression can
be assigned a structural description.  Because -- linguistic objects are
not phonological forms, they are structural descriptions of derivations!
This problem becomes acute, really, when we think about something like

*John arrived the book

because, assuming the "correct" syntactic structural description we still
have two analytical options: (1) _arrive_ is the _arrive_ we all know and
it has been inserted into the "wrong/funny" environment, or (2) this is
"new" _arrive_, call it "Jabberwocky" _arrive_, which is not part of (most
people's) vocabulary of English, but I just invented it and you don't know
what it means yet but you're trying to guess ("you" in this instance could
just as well be the child learner) and your first guess is that it has
something to do with the _arrive_ you already know.

Scenario (2) strikes me as analogous to Karlos' idea that *_caj-o_ has been
assigned no meaning by the Encyclopedia.  Scenario (1) is like Heidi's
"licensing."

Rolf