The February 24--25th, 1981, Gulf of Corinth earthquakes, ruptured only the western end of the South Alkyonides Fault Segment (SAFS) whose length has been confirmed by study of spatial variations in fault throw, footwall uplift and hangingwall subsidence, and fault slip directions: this contrasts with what we would expect of the surface ruptures from characteristic earthquakes. Specifically, the eastern end of the 1981 ruptures, where the smallest coseismic throws and northwest plunging coseismic slip vectors were reported, coincides with the position where the greatest values for uplift/subsidence and fault throws exist along the SAFS, and where fault slip data record northwest, north and northeast plunging slip vectors from previous earthquakes. The SAFS appears to have grown by recurrence of noncharacteristic earthquakes. A new model of earthquake recurrence is proposed which accounts for ruptures that are shorter than the host fault segments, spatial variations in cumulative throw and fault slip directions along a fault segment, and temporal variation in the coseismic slip direction for successive earthquakes. The model implies that recurrence intervals vary both along single fault segments and with time for single localities. Thus palaeoseismological data from one site may not constrain earthquake recurrence at another along the same fault segment or fault slip rates over time periods containing numerous earthquake cycles. ¿ American Geophysical Union 1996