Phase transformations in end-member olivines have been investigated in the temperature range comparable to the interior of subducting slabs. This work constitutes the experimental evidence that the kinetics of transformation of silicate olivine (• phase) to modified spinel ([•) and spinel (T) phases is enhanced by shear deformation. Natural fayalite (•-Fe2SiO 4 ) subjected to a pressure gradient from 0 to 25 GPa at 380øC in the diamond anvil cell (DAC) developed a ring of T phase where the pressure was in the stability field of the y phase and shear stress was large enough to promote the •--•y transition. The sample inside the ting, despite being at higher pressure, remained' dominantly as • phase. The outermost, lower-pressure region of the sample also remained as • phase. In the Mg2SiO 4 system, the transition from • to [• was observed at 575øC in runs in which pressure covered the stability fields of [• phase, y phase and mixed oxides. These results show that the characteristic transformation temperature TCh can be lowered as much as -200øC by shear deformation. On the basis of these observations, we propose a nonhydrostatic kinetic boundary for the •--•[• and •--•y transitions in mantle olivine. For temperatures below this boundary, the transformations are kinetically inhibited, while above it, the transformations can be promoted by shear deformation. Therefore, olivine carried to a depth of several hundred kilometers in a subducting slab can remain as metastable • phase until shear deformation causes its transformation. We suggest that this region of shearpromoted transformation in the cold interior of the subduction zone is responsible for the generation of deep earthquakes. invoked to explain deep earthquakes [Bridgman, 1945; Honda, 1957; Evison, 1963; Dennis and Walker, 1965; Liu, 1983]. However, seismological evidence, which indicates a strong double-couple or shear component in most deep earthquakes [Isacks and Molnar, 1971; Frohlich, 1989], does not support this implosive mechanism because implosion is expected to result in isotropic motion. To avoid this difficulty, Vaisnys and Pilbeam [1976] and Sung and Burns [1976] proposed mechanisms that are initiated by a phase transition that subsequently induces a shear instability. The first experimental evidence of high pressure shear instability was found by Burnley and Kirby [1982] in tremolite experiments which exhibited faulting and acoustic emissions as a function of increasing pressure. Evidence of an anomalous shear instability was also observed in ice [Durham et al., 1983; Kirby et al., 1985]. In a later study, Kirby [1987] recognized that the shear instabilities were associated with high pressure phase transitions in both the tremolite and ice experiments. In the Mg2GeO 4 analog of olivine, Green and Burnley [1989] observed faulting associated with the a•¾ phase transformation in a narrow pressure-temperature range that corresponded to the onset of the transformation. Finally, a similar type of faulting was found to be associated with the •[• transf...