This article suggests to understand the debate around the appropriate level of how ideal a theory of justice should be in terms of whether, and at what point, empirical facts should enter the theory. How is this meta-theoretical question to be settled? The article argues that there are two criteria that play a key role. First, a theory of justice should avoid inadmissible idealizations, that is, it must not presuppose particular empirical facts as universally given. This would unduly restrict the scope of the theory. Second, a theory of justice should not be too abstract by abstracting away from facts that are necessary or constitutive for problems of justice. This would make it too utopian. The article suggests that there are two classes of facts that are constitutive of problems of justice: practices of moral dissent and structures of social power. It concludes that Rawls’ distinction between ideal and non-ideal theory should be replaced by a critical and negativistic approach to theorizing justice.