Rhoades and Evison [1996> indicated a technical flaw in the procedure of Mulargia and Gasperini [1992> and also made many useful remarks and fundamental suggestions on the correct way for performing a statistical evaluation of an earthquake prediction method. We think that these suggestions should be carefully followed by statisticians in the future. However, we do not agree with Rhoades and Evison's [1996> opinion that objective tests of the performance of the VAN-method, using independent data, cannot begin until the method and the null hypothesis have been fully formulated. In our opinion the main question in the present debate is whether our predictions can be ascribed to chance. Such a test has already been carried out in this issue by Aceves et al. [1996>, although they make it clear that they have tested the significance of our predictions and not the overall success of our method. ¿ American Geophysical Union 1996.