In this talk, I argue for a new level of representation, _primitives_, which combine into sets and replace prominence scales and feature geometries in capturing person-based prominence effects in agreement. I argue this new theory has both a conceptual and empirical edge over previous accounts.