From:    John Cowan <[log in to unmask]>
> Thomas R. Wier scripsit:
> > He did indeed.  Which just goes to show that number agreement in
> > English cannot be purely determined based on either semantics or
> > morphology, since more recent American English has changed in this
> > respect.
> I don't think this counts as evidence: the behavior of proper names is
> always going to be lexically idiosyncratic, because some features are
> sealed off whereas others bubble up. [...]

Okay, so then how do you account for the count/mass noun distinction,
or pluralia tanta?  In English, there is no way to predict why, e.g.,
"lettuce" is a mass noun and "scissors" are pluralia tantum, but
something like "army" is countable and has both singular and plural
forms.  The "United States is" is just part of a larger phenomenon,
and happens to be proper as well.

> > It's possible that all lexical items may simply lexically
> > specify what "number" they are, and failure to agree is simply a
> > feature-clash, i.e., there is no trigger or controller of agreement.
> That strikes me as wildly unlikely, unless you want to say that the
> head of "dogs" is "-s".

Ah, but you're assuming that morphemes have independent existence!
There are quite a number of morphologists that would disagree with
that claim (Anderson, e.g.). Anyways, there are also people who
believe in morphemes who would say something very much like -s is
the head of "dogs", since the features of -s override whatever
features "dog-" may have.

(An analogous question:  is "the" the head of the NP "the dog"?
Many, perhaps most, syntacticians now seem to think so.)

>  To my mind, claiming that American (at least)
> English doesn't have syntactic agreement in number is like claiming that
> it has grammatical gender on the grounds that ships, bells, countries
> (sometimes), and a few other inanimates are called "she".

But I wasn't saying that Am. English doesn't have syntactic agreement --
indeed, quite the opposite.  My point (which was really more speculative
than assertive) relied on a different conception of the role of syntax
than is commonly supposed, one which is monostratal, i.e., all modules
generate structures simultaneously, and constraints serve to weed out
contradictions. So, if all nouns are lexically marked for number,
singular or plural, and are not actually assigned such by the syntax,
then agreement is reduced to insisting on "coherence" (to use an LFG
turn of phrase).  As the tree-structure is built up from the lexical
entries, at each node a unification of features occurs, which percolates
up to a yet higher node. Grammaticality (or the lack thereof) depends
on whether clashes occur at any level.

So, getting back to the "United States is", a view of grammar such as
that which I outlined above would unify English dialects.  The difference
of dialect between British English and American English would be localized
in the lexicon, which would have a different rule mapping the properties
of formal semantic or conceptual structures with morphological structure
that produces new lexical items.  The kind of generalization that you
want produced by the syntax would be captured in this kind of interface,
but with different modules.  (Syntax would still be involved, since it
would still have to govern whether features clash or not. It just wouldn't
actually determine those features.)

Thomas Wier	       "I find it useful to meet my subjects personally,
Dept. of Linguistics    because our secret police don't get it right
University of Chicago   half the time." -- octogenarian Sheikh Zayed of
1010 E. 59th Street     Abu Dhabi, to a French reporter.
Chicago, IL 60637