The purpose of this paper is the clarification of some methodological problems concerning Rawls’ theory of justice. The first part seeks to make more precise Tugendhat’s distinction between 1st-person-theory and 3rd-person-theory. Rawls’ theory fulfills all criteria for 1st-person-theories. In the second part Rawl’s coherence model for the justification of norms („reflective equilibrium“) is critically analyzed and opposed to the hypothetical decision which individuals are to make in the original position (contract model). It is shown that the conception of reflective equilibrium (...) is in various aspects mistaken. In conclusion a problem is indicated which Rawls has not satisfactorily resolved: The veil of ignorance is supposed to guarantee that the decision for the basic principles of social justice is unanimous. Nevertheless it would appear that the individuals in the original position either have too little empirical knowledge in order to make a rational decision, or they have too much knowledge in order to come to an unanimous decision. The veil of ignorance is either too fine or not fine enough. (shrink)