On the Units of Spell-out and Long Distance Agreement
Within generative syntax, there is a debate on: (i) the units of spell-out, e.g., CP and vP in Chomsky
(2000, 2001 & 2005) and TP in Uriagereka (1999); (ii) the role of covert movement ( Chomsky 1995) vs.
Long-Distance Agreement, AGREE (Chomsky 2000, 2001 & 2005). As far as (i) is concerned,
Uriagereka claims that TP is a phase in rich agreement languages as Spanish. Still, the question arises as
to why this should be the case. With regard to (ii), one wonders what empirical evidence would draw the
line between both alternatives. Traditionally, an argument for the covert/overt movement distinction is
found in the fact that in a number of languages in situ elements do not trigger agreement as opposed to
elements that have moved into a Spec,Head configuration overtly (e.g., Arabic verbal agreement,
Hungarian adpositions, French participles, Spanish unaccusatives, etc.; see Kayne 1994 and Hornstein et
al. 2005, among others) . This can be captured naturally in a framework including covert/overt movement
because the relation between the Probe (P) and the Goal (G) at the point of Spell-out is different in the in
situ and the movement counterparts. Hence, it is natural to consider the Spec,Head relation the domain of
(morphological) agreement (Kayne 1994) at least in the relevant languages. However, as AGREE is
adopted, there is no distinction between the overt and the covert version in terms of the relation of P and
G at the point of Spell-out. Surely enough, things move or do not move but the relation between P and G
is the same no matter what. Therefore, such Agreement Asymmetries do not follow from the system as
opposed to the overt/covert approach. In fact, as the role of Spec,Head relations is diminished, this calls
for a reconsideration of a number of issues, Agreement being among the most relevant ones (Chomsky
2005).
In this context, it is argued that a unified treatment can be given to the mentioned Agreement
Asymmetries and to the phase nature of TP in rich agreement languages, while keeping the system
compatible with the AGREE approach. Specifically, both sets of facts follow from computational
dynamics of the interfaces, in particular, from the incremental nature of the parser (Levelt 1989) and its
interaction with the syntax (e.g., Bock et al. 1992, Phillips 1996 and Ferreira 1996). According to this
view, whenever the grammar allows it, syntactic elements are uttered as they become available. In the P­
G word order, where G is in situ, a P showing morphological / rich agreement cannot be uttered till G has
been coded. This is the case because the agreement causes P to `wait' for G to become available. Only
then can P be spelled-out. As a consequence of this wait, both would come to be spelled-out as a unit.
Inasmuch as such a `wait' goes against the spirit of incremental / efficient production, one option to avoid
it is to drop agreement markers or adopt default agreement. With regard to the G­P order, the Spec,Head
configuration, the `wait' for G takes place anyway because it precedes P. Hence, there is nothing to be
gained by dropping agreement markers.
Let's see this in Spanish, where agreement is optional with unaccusative verbs, (1). In the VS order, the
verb has to wait for the subject to be coded against the spirit of incrementality. This fact together with the
existence of V-to-T movement provides a rationale for the phase behavior of TP in this language. In an
attempt to achieve incremental production, default agreement might be adopted, (1a). In the case of the
SV order / Spec, Head configuration, (1b), the `wait' is irrelevant because the goal has to become
available anyway for the verb to be uttered. Hence, dropping the agreement is not justified.
Whereas syntactic treatments in terms of agreement with an expletive, first conjunct agreement or a
unitary interpretation of the conjoined elements in the default agreement cases (e.g., see Aoun et al. 1994,
and Mohammad 2000, among others) are most relevant in this context, the Agreement Asymmetries are
not restricted to verbal agreement. The role of the parser, therefore, should be combined with such
syntactic treatments of the verbal agreement facts to achieve a general treatment of Agreement
Asymmetries. The crosslinguistic variation regarding agreement (and lack of thereof) with in situ
elements (and even cases of optionality as in (1b)) come from the fact that the strategies of the parser are
defeasible: e.g., central embedding in English is disfavored due to its costly nature, but it is not banned by
the parser. This analysis explains the Agreement Asymmetries in a way compatible with AGREE
(Chomsky 2005) and provides a rationale for the phase nature of TP in rich agreement languages
(Uriagereka 1999).

Examples:
      (1)   a. El dinero y la leche faltan         / *falta.
                the money and the milk are missing / * is missing
            b. Falta     / Faltan      el dinero y la leche.
               is mising / are missing the money and the milk