Against the Split-CP Hypothesis

The purpose of this paper is to argue against the hypothesis that CP is
split into a number of projections with distinct interpretations and which
trigger A-bar movements to those projections.
	The Split-CP Hypothesis has received its primary support from what
appear to be strict ordering restrictions on the left-periphery (see
especially Rizzi 1997). For example, preposed topics follow wh-relative
pronouns, but precede wh-interrogative pronouns:

(1)	a.  the man to whom liberty, we could never grant
	b.  *the man, liberty, to whom we could never grant
(2)	a.  On the table, which dishes are you going to put?
	b.  *Which dishes are, on the table, you going to put?

Such facts would seem to support the idea of a wh-relative projection
above a topic projection, in turn above a wh-interrogative projection.
These facts, however, follow from independently-needed principles, in
particular the Nested Dependency Constraint (which space does not allow me
to demonstrate here).
	The Split-CP Hypothesis is directly challenged by the fact that
preposed wh-elements themselves have information structure properties. I
will show that this fact leads to logical conflicts. For wh-questions, in
order to derive the syntactic facts, the wh-phrase has to be in a low
projection, but to derive the illocutionary force facts, the wh-phrase has
to be in a high projection
	A focus projection, of course, entails the movement of focus to
that projection. I will go on to challenge such a movement. The strongest
argument for a movement approach to focus is its quantificational nature
and associated weak crossover effects. But, as noted in Reinhart (1991),
anything can be in focus. The stressed NP in (3) is contained in a strong
island, so its extraction should be impossible:

(3)	a.	[ [ That Linda argued with THE CHAIRMAN] is surprising].
	b.	[ [ Even the paper that LUCIE submitted to our journal]
		was weak].

Furthermore, as observed in Chomsky (1971), any subconstituent of a phrase
containing the sentence's intonation center can be the focus, as in (4):

(4)	Was he warned to look out for [an [ex-convict [with [a [red
	[shirt]]]]]]?

But not all of the bracketed constituents are phrasal: the movement of the
noun 'shirt' to Spec, FocP, therefore, would violate constraints on
movement going back to Emonds (1976). Indeed, contrastive focuses (which
Rizzi assimilates into his system) need not be either on right branches
(5) nor be formally continuous (6):

(5)	Q. Has John read Slaughterhouse-Five?
	A.  No, John doesn't READ books.
(6)	Q: Did Mary wash the car?
	A: No, TOM washed the WINDOWS.

For certain A' movements, there is no plausible interpretable attracting
projection at all. Consider Heavy-NP-Shift, which is often assumed to be
leftward movement to Spec, FocP. However as (7) shows, the heavy NP need
not be in focus:

(7)	a.  To whom did you give all your books on the phonetic
		foundations of conventional implicature?
	b.  I gave [TO MARY][all of those unbelievably ridiculous
		publications].

Such facts suggest that the more traditional analysis of Heavy-NP-Shift as
adjunction to the right is on the right track.
	For Rizzi (unlike in more mainstream minimalism), the features are
interpretable as well as the projections containing them. This leads to
serious problems for the wh-feature, which would have to vary wildly in
its interpretation. For example, appositive relatives differ from simple
questions in not manifesting an operator-variable configuration:

(8)	Ralph Nader, who several million Americans voted for t, lost the
	election.

As pointed out by Lasnik and Stowell (1991) there is no weak crossover
effect with appositive relatives. Likewise, fronted wh-elements do not
reliably act as scope markers, as is demonstrated by partial wh-movement
in German and Romani. Also, fronted wh-phrases in different constructions
play very different roles in terms of information structure. In simple
questions, the wh-phrase is a focus; in restrictive relatives, it is a
topic; in free relatives, it fills the semantic functions of the missing
head noun; and in wh-clefts, it prepares the hearer for the focused (new)
information in sentence-final position.
	It is not the purpose of this paper to provide a comprehensive
alternative to the Split-CP Hypothesis. However, I will point to the fact
that several of the problems with it can be avoided given Erteschik-Shir's
level of Focus Structure. I conclude with some very general remarks on how
the MPs rejection of optionality in the syntax leads inevitably to
building more and more of discourse into the syntax, and with it the
otherwise unmotivated need for 'interpretable' higher projections and the
movements that accompany them.