The Arabic quantifier kull displays a Q___NP and NP___Q alternation. Shlonsky (1991) argues that in both patterns Q heads a QP projection with the NP as a complement that may undergo movement to [Spec, QP] or beyond to yield the NP___Q pattern and Q-float structures. On the contrary, I argue on the basis of evidence from reconstruction, Case, and agreement that the two patterns are radically different. In the Q___NP pattern Q is indeed the head of a QP projection that contains the NP. In the NP___Q pattern, however, Q heads a QP adjunct that modifies the NP and in some cases the VP.