fate of SPEC and mutual selection

[Frank Van Eynde]

>
> Ulli is right about the motivation for SPEC is from German, as far
> as syntax goes. But we also needed for English to enable determiners
> to have their hands on the CONTENT of their N-bar sisters. See Pollard
> and Sag 1994, p. 50.

I think that one could also construct a syntactic argument for SPEC
on English specifiers. Take the agreement data in

a/one hundred cars/*car
a great/good many cars/*car

Let us make the reasonable assumption that these NPs are plural and
that they consist of a plural noun (`cars') and a specifier with internal
structure (resp. `a/one hundred' and `a great/good many').

Suppose now that we use SPR on the noun to model the number agreement.
Then `cars' has to be said to require a plural specifier, but
neither `one hundred' nor `a good many' qualify as plural
(cf. the presence of `one' and `a').

If we use SPEC to model the agreement, we do not have this problem.
`hundred' can be said to require a plural N' without being plural
itself; its own number value can be left underspecified, so
that it is compatible with both `one' and `two'. Along the same lines,
`many' can be said to require a plural N', while being singular
itself.

In short, number agreement within the NP can be modeled
straightforwardly in terms of SPEC, but not in terms of SPR.

If SPEC is eliminated, then this kind of agreement has to be
modeled in some other way. I wonder how.

Friendly greetings,

Frank Van Eynde