Rubach (2000) proposes a modified version of Optimality Theory (OT) that features derivations. While Prince and Smolensky's (1993) original formulation requires some modification, argue here that, rather than reintroducing derivations, the correct approach is to take fuller advantage of OT's inherent parallelism. I propose that outputs must be related not only to inputs, but to other, “neighboring” representations as well—a feature that is shared by both the output-to-output faithfulness approach and the theory of targeted constraints developed by Wilson (2000, to appear). I show that all the cases cited by Rubach that seem to support derivations are in fact handled by the latter two related theories, and that both ofthese have significant advantages over derivations.