Four vertical-component seismograph stations were operated in western Crete, southeastern Peloponnesus, the island of Milos, and south of Corinth on Mount Didimon from 1981 to 1984. Combining arrival times from this network, the Greek National Network, and VOLNET, we relocated hypocenters within Greece and the surrounding area. The difference between the relocations and the locations published in the National Observatory of Athens Bulletin in most instances were minor, except for earthquakes located in the vicinity of the western Hellenic arc. In this area, we found that epicenters thought to lie south of the plate boundary, within the African plate, relocated an average of approximately 60 km to the N30°E, which places them at or north of the plate boundary. We suggest that this difference is a minimum estimate of the location error and the true locations may lie still further to the NNE. We propose that anomalously slow seismic velocities in the Aegean back-arc upper mantle cause travel-time delays to the seismograph stations located to the NE of the source volumes, resulting in location errors in a direction away from these stations. Extrapolating from our observations to other segments of the Hellenic plate boundary and to periods not covered by our operation, we hypothesize that most or all earthquakes thought to have originated on the African plate, within about 100 km of the Hellenic arc, were actually located at the plate boundary or on the Aegean plate.