It has long been speculated that the subduction of seamounts may have profound mechanical consequences, particularly on the generation of large subduction zone earthquakes. We consider this problem as a special case of the seismic coupling model of Scholz and Campos. The model predicts that the subduction of a large seamount will increase the normal stress across the subduction interface and hence will enhance seismic coupling. In the case of coupled arcs, this increased coupling will result in a large increase in the recurrence intervals of earthquakes. Such events will therefore appear rare in the instrumental record, as is observed. For decoupled arcs, the subduction of a large seamount should cause them to become locally coupled. Three of the latter localities have been found with satellite altimetry data, two on the Tonga-Kermadec Trench and one on the Izu-Bonin Trench. The diagnostic feature is a shallowing or obstruction of the trench axis adjacent to a chain or group of large seamounts on the subducting plate. In each case, the small to moderate plate-interface seismicity typical of decoupled seismic zones is almost completely absent in a 100–200 km arc segment, indicating that the subduction zone is locally locked, as predicted by the model. The occurrence of large interplate earthquakes at both Tonga sites confirms that they are seismically coupled.