Five ocean bottom seismometers recorded seismicity near the Mid-America Trench offshore Guatemala for 27 days in 1979. The array was emplaced in the lower slope region, just above the topographic trench, in the area investigated during Deep Sea Drilling Project legs 67 and 84. Approximately 170 events were recorded by three or more seismometers, and almost half were located with statistical hypocentral errors of less than 10 km. Most epicenters were located immediately landward of the trench axis, and many were further confined to a zone northwest of the array. In terms of depth, most events were located within the subducting Cocos plate rather than in the overlying plate or at the plate-plate boundary. Their apparent concentration in the lower crust and upper mantle may suggest that the upper crust does not have the strength to support earthquake-generating stresses. The data permit construction of a magnitude-duration scale, calibrated with m_b magnitudes for events located by the World-Wide Standard Seismograph Network (WWSSN) and recorded by our array and by the network recording foreshocks and aftershocks of the 1979 Petatlan earthquake. Most magnitudes ranged between 3.0 and 4.0 m_b, and the threshold magnitude of locatable events was about 2.8 m_b. Two distinct composite focal mechanisms were determined. One appears to indicate high-angle reverse faulting in the subducting plate, in a plane parallel to trench axis strike. The other, constructed for some earthquakes in the zone northwest of the array, seems to show normal faulting along possible fault planes oriented quasi-perpendicular to the trench axis. The normal faulting is consistent with the segmentation of the Cocos plate that has been proposed from land evidence. Such segmentation might be evidenced offshore by normal faulting along planes subperpendicular to trench strike. Alternatively, the seismicity zone and associated normal faulting mechanism may be the subsurface expression of the tectonics responsible for the San Jose Canyon, a prominent submarine canyon located farther upslope. Finally, projection of our seismicity sample and of well-located WWSSN events from 1954 to 1980 onto a plane perpendicular to the trench axis shows a distinct gap between the shallow seismicity located by our array, and the deeper Wadati-Benioff zone seismicity located by the WWSSN. We tentatively ascribe this gap to inadequate sampling, but we suggest that it requires further investigation.