Reid's elastic rebound theory implies that regional stress should increase in the interval between characteristic earthquakes. Based on this, it is postulated that the size of the largest annual earthquakes should also increase in this interval. To test this, the variability of seismic activity in the New Madrid, Missouri, and the San Francisco and Fort Tejon, California, regions is examined using sequences of magnitudes of largest annual earthquakes. The calculated probability of a magnitude 6.5 or 7.5 earthquake based on past occurrences varies significantly with time. It is currently growing in the New Madrid and San Francisco areas and declining in the Fort Tejon area. The calculated probability frequently increases in the five years before earthquakes 1.5 magnitudes or more greater than the regional average maximum annual magnitude. The center of significant activity is moving northeastward in the New Madrid area.