In these remarks, I have examined the problem of phonological opacity for theories without serial ordering of rules, focusing on Optimality Theory. I have argued in favor of extending a correspondence-based approach to faithfulness to the statement of phonological markedness constraints. The core of the proposal is separate specification of the levels at which featural, adjacency, and linear order conditions must be met. I have compared this approach to two others, noting many similarities and a few differences: the structural approach adopted in Prince and Smolensky (1993) and most other OT work, and the Two-Level or Cognitive Phonology of Koskenniemi (1983) and Lakoff (1993).