We have relocated earthquakes along the Colombia-Ecuador coast with the method of Joint Hypocenter Determination in order to examine the nature of fault heterogeneities that controlled northward propagation of plate-boundary rupture from the source region of the earthquake of 14 May 1942, to the source region of the 19 January 1958 earthquake, and eventually to the source region of the earthquake of 12 December 1979. The entire region of the 1942, 1958, and 1979 shocks had previously ruptured in the single great earthquake of 31 January 1906. The relocated hypocenters lie on the same plane to within the approximately 20-km uncertainty of the focal depths. Also, the main shocks nucleated at nearly the same distance from the Colombia trench. These observations suggest that the heterogeneities between the 1942 and 1958 ruptures and between the 1958 and 1979 ruptures do not correspond to a major distortion of the downgoing slab but rather to either minor distortions of the slab or to regions of high friction or low available strain energy on a continuous fault surface. In particular, the heterogeneity between the 1958 and 1979 rupture zones seems to have been a high-strength barrier with dimensions much smaller than the dimensions of either of the rupture zones. The 1942 and 1958 earthquakes had source dimensions no larger than the 1979 main shock, but they had stronger aftershock sequences than the 1979 earthquake. This observation suggests that the arrest of earthquake rupture in 1979 left the plate-boundary segment that had ruptured in 1906 in a state of lower stress than it had been following the 1942 and 1958 earthquakes. Long-term seismicity in the decades preceding the 1979 earthquake occurred mostly outside or on the boundaries of the rupture area defined by the distribution of 1979 aftershocks. The intense aftershock activity that followed the 1958 main shock within tens of kilometers of the eventual 1979 hypocenter may correspond to a long-term precursory seismic swarm for the 1979 earthquake.