The paper by Burton [1996>, besides its unusual title, contains a lot of points that are inaccurate, or erroneous. A clear misuse of statistics is made by Burton, because he handles values of one parameter (i.e., magnitude) without paying attention to essential points like the following: the magnitude value of a prediction referring to a main shock in an area, which becomes very rarely active with such events, cannot be treated equally with a prediction that may have the same magnitude value, but refers to an aftershock in another area. By using simple rules of statistics, we show that the statistical treatment used by Burton is obviously wrong and can ''reject'' even an ideal prediction method. Furthermore, we show that Burton's criticism on the magnitudes predicted by VAN, is based on a clear misinterpretation of the true meaning of the experimental error and of Gutenberg-Richter relation; the related Burton's claims, contradict the physical expectation. The appropriate spatial resolution of an earthquake prediction is also discussed. For magnitudes between 5.5 and 6.0, the spatial resolution usually achieved by VAN is around 50km; such a value is reasonable from practical point of view, if we also consider the dimensions of the source. For larger magnitudes, i.e., around 7.5, a spatial resolution of ~150km would be accepted. Therefore, Burton's [1996> claims for VAN to increase the spatial resolution are unjustified, as they have some basis only for smaller events (~5.0-units) which, however, have no much practical interest. Last but not least, Burton uses unusual phrases like ''a posteriori demonstration of predictions prevailed,'' ''unclear retrospective filters'' or ''too much is often claimed from too little,'' but does not mention facts, e.g., that during the three years period under discussion (i.e., 1987--1989) VAN issued only one public alarm, which was followed by the most strong and destructive activity during this period. ¿ American Geophysical Union 1996