Binding and ellipsis are empirically and theoretically symbiotic: each reveals otherwise hidden facts about the other. Here I investigate a case where a theory of binding is entwined with a problematic ellipsis-licensing mechanism, with the result that there are strong reasons to abandon both. The ellipsis-licensing mechanism in question is Referential Parallelism (Fox 2000), according to which a bound pronoun may support strict identity under ellipsis. Jettisoning this mechanism forces us to abandon theories of binding that involve what I call compulsory binding, which encode a grammatical preference for binding over coreference and for local over nonlocal binding (Reinhart 1983, Grodzinsky and Reinhart 1993, Fox 2000, Büring 2005). In their place, I suggest that we adopt what I call the violation equivalence approach to binding (Heim 1993, Reinhart 2006, Roelofsen 2010) and a Fox-style ellipsis-licensing mechanism based on formal alternatives (Katzir 2007, Fox and Katzir 2011).