Wh-Agreement as the Morphological Reflex of Agree
In pre-minimalist syntax, it was natural to analyze wh-agreement as the morphological reflex of specifier-head agreement between the moved wh-phrase and a functional head, such as C (see, e.g., Watanabe 1996). Such an approach dealt well with some language-specific patterns of wh agreement, but not others. (The complexity of the Chamorro pattern is problematic; see Chung 1998, 259.) Minimalist syntax offers a promising alternative. Instead of appealing to the specifier-head relation, we can analyze wh-agreement as a morphological reflex of Agree, the relation of feature valuation that must hold for movement to occur (see also Bruening 2001, 209–210).
In the minimalist view of movement, in order for one category (the goal) to move to the vicinity of another (the probe), the two must stand in the Agree relation: the probe must c-command the goal, the relation between the two must be local, and each category must have active (i.e., uninterpretable or unvalued) features whose values can be set by the corresponding features of the other (see, e.g., Chomsky 2001, 4–5). Movement is then implemented by merging the goal in the vicinity of the probe, an operation that leaves the original copy of the goal un touched.
In minimalist syntax, access to previously constructed constituent structure is regulated by the Phase Impenetrability Condition (PIC), which permits operations outside a phase Ph to have access only to the head of Ph and to its edge (Chomsky 2000, 108). If we assume the PIC and take CP to be a strong phase, then it becomes clear how to analyze the two flavors of wh-agreement that are manifested in long-distance wh constructions.
Flavor (i) (illustrated earlier by ‘‘wh-agreement’’ in Chamorro) is the morphological reflex of Agree holding between a probe that bears the active feature that drives wh-movement (henceforth [X]) and a goal that bears the active feature [wh]. Morphological agreement of this type signals the Agree relations indicated in the schematic tree in (1) (where u signifies ‘‘uninterpretable’’ or ‘‘unvalued’’).
(1)
Flavor (ii) (illustrated earlier by ‘‘operator-C agreement’’ in Chamorro) is also a morphological reflex of Agree holding between a probe that bears the feature [X] and a goal that bears the feature [wh], but here the morphology also reflects the fact that both probe and goal bear an operator feature—[Q] in questions, [pred] in relative clauses, [emphasis] or [contrast] in focus constructions (see Rizzi 1990, 1996, 1997)—that is active (i.e., unvalued) on the probe but inactive on the goal. Morphological agreement of flavor (ii), then, signals the higher Agree relation indicated in the schematic diagram in (2).
(2)
Notice that there is a second Agree relation in (2), one that holds between the lower F and the lowest instance of the wh-phrase. The point is that this Agree relation, though syntactically present, is not signaled morphologically by wh-agreement of flavor (ii).
More generally, we propose that the two types of wh-agreement spell out the abstract Agree relation holding between a wh-goal and two types of probes. While the feature composition of the probe determines the flavor of wh-agreement, it is the Agree relation itself that ultimately lies behind the occurrence of this special morphology. All that matters here is that the setup of the computational system does indeed permit the Agree relation to have phonetic effects; on this point, see Chomsky 2001, 5.
When this view of wh-agreement as the spell-out of Agree is combined with the copy theory of movement, a clear prediction emerges: it should be possible for a language to exhibit wh-agreement as well as wh movement in overt syntax, followed by pronunciation of only the lowest copy. The result should be ‘‘apparent’’ wh-in-situ whose participation in syntactic wh-movement leaves visible morphological effects. This prediction is borne out by empirical evidence from Passamaquoddy and Coptic Egyptian, as we now show.
