In this paper, I analyze the particular conception of reciprocal justification proposed by Nelson Goodman and incorporated by John Rawls into what he called “reflective equilibrium”. I propose a way of avoiding the twin dangers which threaten to push this idea to either of two extremes: the reliance on epistemically privileged observation reports (or moral judgments in Rawls’ version), which tends to disrupt the balance struck between the two sides of the equilibrium and to re-establish a foundationalism; and the denial of any privileged status to such reports (or judgments), which makes the equilibrium into a theoretical monolith.