Earthquakes beneath the Himalayan collision zone occur at depths between near surface and around 100 km below sea level. After relocating earthquakes with two one-dimensional (1-D) velocity models, we found a clear bimodal depth distribution for earthquakes in the Himalayas of eastern Nepal and the southern Tibetan Plateau and evidence that some earthquakes originate at upper mantle depths. Seismicity in Nepal shows an accumulation of earthquakes along the front of the Himalayan arc, with a seismic gap between longitudes 87.3¿E and 87.7¿E. Although upper crustal seismicity along the topographic front of the High Himalaya is consistent with a region of high strain accumulation associated with convergence on the Main Himalayan thrust fault, microearthquakes do not necessarily occur on this fault. Instead, they concentrate in the hanging wall. Seismic activity in the sub-Himalaya and the Terai Plains is almost exclusively limited to the vicinity of the location of the magnitude 6.5 20 August 1988 Udayapur earthquake, with most of the earthquakes in the lower crust and the upper mantle. Clusters of earthquakes in the Lesser and High Himalayas and south Tibet (Tethyan Himalayas) mark very well defined zones of seismicity at depths between 50 and 100 km, confirming the presence of earthquakes in the upper mantle in the region of continental collision. The occurrence of earthquakes at sub-Moho depths favors the idea that the continental upper mantle deforms by brittle processes.