There are different positions regarding the definition and the function of replication which lead to several alternatives when setting out to repeat a study. Classifications proposed for replications are of heuristic value but they do not take into consideration all the important conditions. Often replications are regarded as a means to control for errors or as a means for generalizing findings. But these aims of replications are only of secondary importance to the main function of replications: the demonstration of "scientific facts". Only direct or "identical" replications may serve this purpose well.