A realistic mechanical model of the oceanic lithosphere must be able to explain more than just the topography of a typical trench and outer rise system. A successful model should also be consistent with our knowledge of rock mechanics from laboratory studies, with the distribution of earthquakes associated with bending of the plate before subduction, with the amount of strain within the plates as indicated by the seismic moment of the earthquakes, and with the topography associated with other loads, such as seamounts and trenches where plate convergence is very slow. These observational tests are applied to six mechanical models of the lithosphere: models in which the regional, horizontal compressive stresses are of the order of several kilobars are the least successful.