|Abstract:||Two conceptions of crucial non-ranking have been proposed in the literature: i) one that produces variation (Anttila 1997) and ii) that of equal ranking (Crowhurst 2001, Crowhurst and Michael 2005, Topintzi 2005, Rice in press). I show that the former is erroneous in that it predicts unattested variation or fails to account for certain cases altogether. It is also unnecessary since it can be subsumed by the Gradual Learning Algorithm (Boersma and Hayes 2001). I thus conclude that a single notion of crucial non-ranking exists, that of equal ranking. Since alternatives such as constraint conjunction cannot replace equal ranking, I argue that it must be recognised as a genuinely distinct ranking relationship.