The disastrous earthquake of Sept. 19, 1985 (Ms=8.1) broke a segment of the plate boundary along the Mexican subduction zone known as the Michoacan gap for which the seismic potential was uncertain. This earthquake was followed by another major earthquake on Sept. 21 (Ms=7.5). Teleseismic, long-period seismograms show that the first earthquake consisted of two subevents separated in time by 27 s; the local strong-motion data indicate that the second subevent occurred about 95 km SE of the first one. Teleseismic records of the Sept. 21 shock can be modeled by a single source. Aftershock areas, slips, and stress drops of the two earthquakes are 170¿50 and 66¿33 km2, 220 and 330 cm, and 19 and 43 bars respectively. In hindsight, it seems very likely that the earthquake of June 7, 1911 (Ms=7.9) occurred in the same region; if so, its rupture area was probably only 60% of that of the Sept. 19 earthquake. Furthermore, the earthquake of Sept. 21 appears to have broken the updip part of the interface which ruptured in the 1979 Petatlan earthquake (Ms=7.6). These observations suggest a variable mode of rupture in the region.