A statistical test of the VAN method compares its prediction rate (number of successful predictions divided by the total number of predictions issued) between 1987 and 1989 to that from randomly sampling the earthquake catalog. Unlike previous tests, this one samples distributions derived from the historic spatial and magnitude distribution between 1960 and 1985. The VAN method yields a prediction rate of 67.9% at a time delay of 22 days, which is achievable in less than 0.06% (P-value≤0.06%) of the realizations in randomly sampling the full NEIC catalog (37.2% mean prediction rate). VAN's prediction rate drops to 66.7% with aftershocks removed from the catalog and prediction list, which is still significant at a p-value of less than 0.1%. With an alternate aftershock model, the prediction rate drops to 48.0%. This latter value is significant only for P-values less than 7.96%. We conclude from these tests that the VAN method is formally significant at a time lag of 22 days. ¿ American Geophysical Union 1996