The authors show that the optimizaton domain must be at least as large as CP and include a discussion of the cyclic linearization approach favored by Fox and Pesetsky (2005), showing that this approach is less well-equipped to account for the set of data that is addressed in this chapter.