This article defines in a precise manner three different mechanisms to achieve impartiality in distributive justice and studies them experimentally. We consider a first-person procedure, the Rawlsian veil of ignorance, and two third-party procedures, the impartial spectator and the ideal observer. As a result, we find striking differences in the chosen outcome distributions by the three methods. Ideal observers that do not have a stake in the allocation problem nor information about their position in society propose significantly more egalitarian distributions than veiled stakeholders or impartial spectators. Risk preferences seem to explain why participants that have a stake in the final allocation propose less egalitarian distributions. Impartial spectators that are informed about their position in society tend to favor stakeholders holding the same position.