The intriguing suggestion by Tselentis and Melis [1996], that the tolerance limit in predicting the epicentral location should depend on earthquake magnitude, is discussed.Tselentis and Melis applied the procedure of Mulargia and Gasperini [1992] to an ideally perfect earthquake prediction method, which achieved to predict successfully all the seventeen earthquakes with Ms ≥ 5.3 that occurred within the area 36–41°N, 19–25°E during the three years period 1983–1985, and found that these ideal predictions can be ascribed to chance. Their application clearly demonstrates that the procedure of Mulargia and Gasperini [1992] is questionable. In this Reply we also give a precise example of statistics, which indicates how Mulargia and Gasperini's [1992] procedure strongly violates Poisson restrictions, and hence leads to unacceptable results.