Cyclicity in Agree: Maximal projections as probes
Cyclic Agree (Rezac 2003; Béjar & Rezac 2009) has been proposed to account for why a probe can agree with DPs in both its complement and specifier in agreement displacement. When an unsatisfied probe reprojects, its search space (i.e. its c-command domain) is cyclically expanded to include the specifier. The result is that Spec-Head agreement is still Agree under c-command -- the specifier is in the c-command domain of the intermediate level projection. In Bare Phrase Structure (BPS; Chomsky 1995), there is no distinction between head, bar, and phrase levels. Therefore, a prediction of this type of account is that if a probe remains unsatisfied when it reprojects to form a maximal projection, that maximal projection should be able to probe its c-command domain through the same kind of cyclic expansion that makes Spec-Head agreement possible.
In this talk, which draws on data from original fieldwork, I argue on the basis of a pattern of an agreeing adjunct C in Amahuaca (Panoan; Peru) that this prediction of cyclic expansion is borne out. Adjunct C agrees with DPs in its own complement but also with matrix DPs. This is possible because the maximal projection of this high adjunct C can probe its c-command domain -- the matrix TP. This account based on cyclic expansion provides a straightforward way to capture this apparently non-local pattern of agreement without loosening the conditions on locality in Agree.
The Amahuaca data therefore provide support for a Cyclic Agree model and suggest that probe reprojection is fully generalizable and need not be limited to (what in non-BPS terms are) intermediate level projections. The strongest conclusion of such an account is that Agree always requires that the probe c-command the goal, with Spec-Head agreement and the type of apparent long-distance agreement seen in Amahuaca simply indicating cyclic expansion of the probe's domain.