A new precursory measure called active zone size (AZS), based on the development of spatial correlations, was recently introduced in the context of predicting large events in dynamical models of faults. The new measure counts the fraction of the total seismic area which has exhibited earthquakes recently. In each of the models tested, AZS was found to be more strongly correlated with an increase in the likelihood of a large event than the more conventional measure activity (N), which is defined to be the total number of earthquakes. In this paper we report the first application of AZS to seismicity catalogs. We compare measurements of AZS and N both individually and in the context of the intermediate-term earthquake prediction algorithm M8 applied here specifically to data from the western United States. To test the performance we replace functions in M8 which are based on N with similar functions based on AZS. When this is done, at least on this limited data set, the performance of M8 is somewhat improved: false alarm time is reduced, and the Landers earthquake is successfully predicted. ¿ American Geophysical Union 1995