The focal process of the Sanriku earthquake of March 2, 1933, is discussed in relation to the bending mechanism of the lithosphere. On the basis of the P times obtained at more than 200 stations, it is confirmed that the hypocenter of this earthquake is within the lithosphere beneath the Japan trench. The P wave fault plane solution, the amplitude of long-period (100 s) Love and Rayleigh waves and two near-field observations suggest, almost definitely, that the Sanriku earthquake represents a predominantly normal faulting on a plane dipping 45° towards N 90° W. A fault size of 185 × 100 km 2, in agreement with the size of the aftershock area, is required to yield a slip dislocation of 3.3 m, a value consistent with the tsunami data. This result suggests that the fracture took place over the entire thickness of the lithosphere, thereby precluding the possibility that the Sanriku earthquake merely represents a surface tensile crack due to the bending of the lithosphere. This large scale lithospheric faulting is presumably due to a gravitational pull exerted by the cold sinking lithosphere. The fracture probably took place on an old fault plane which had once fractured and healed up. The existence of this fracture zone which decouples, to some extent, the oceanic lithosphere from the sinking lithosphere accounts for the sharp bend of the lithosphere beneath oceanic trenches and also the abrupt disappearance of seismic activity across oceanic trenches. The sharp bend of the lithosphere is therefore a result, not the cause, of great earthquakes beneath oceanic trenches.