Joint epicenter determination of earthquakes that occurred in northern Algeria near Ech Cheliff (named Orléansville in 1954 and El Asnam in 1980) shows that the earthquake of 9 September 1954 (M = 6.5) occurred at nearly the same location as the earthquake of 10 October 1980 (M = 7.3). The 1954 main shock and earliest aftershocks were concentrated close to the boundaries of segment B (nomenclature of Deschamps et al., 1982; King and Yielding, 1984) of the 1980 fault system, which was to experience approximately 8 m of slip in the 1980 earthquake. Later aftershocks of the 1954 earthquake were spread over a broad area, notably in a region north of the 1980 fault system that also experienced many aftershocks to the 1980 earthquake. The closeness of the 1954 main shock and earliest aftershocks to the 1980 segment B implies that the 1954 earthquake involved either (1) rupture of segment B proper, or (2) rupture of a distinct fault in the hanging wall or footwall block of segment B. The first interpretation cannot be accommodated in the characteristic-displacement model of fault behavior but instead requires a model, such as the time-predictable model, which would permit slip in 1954 to have released only a fraction of the elastic strain that was stored on segment B and which would later be released in 1980. The second interpretation, specifically with rupture in the hanging-wall block of segment B, accounts more directly than the first interpretation for differences in the focal mechanisms and ground deformations associated with the two main shocks; this interpretation is therefore preferred. The second interpretation is consistent with a characteristic-displacement model of the 1980 fault system. No matter how it is modeled, the Algerian episode provides an example of an earthquake (1954) that was large enough to perhaps be judged the culminating event of the seismic cycle in a regionally important fault zone but that was followed instead by an even larger shock (1980). The space-time characteristics of the episode are unusual enough that they cannot represent the most likely mode of strain release on reverse-fault systems in general, but the episode should be viewed as one possible scenario for the evolution of seismicity on other such fault systems.