Serpentinite has been proposed as a cause of both low strength and aseismic creep of fault zones. To test these hypotheses, we have measured the strength of chrysotile-, lizardite-, and antigorite-rich serpentinite gouges under hydrothermal conditions, with emphasis on chrysotile, which has thus far received little attention. At 25??C, the coefficient of friction, ??, of chrysotile gouge is roughly 0.2, whereas the lizardite- and antigorite-rich gouges are at least twice as strong. The very low room temperature strength of chrysotile is a consequence of its unusually high adsorbed water content. When the adsorbed water is removed, chrysotile is as strong as pure antigorite gouge at room temperature. Heating to ???200??C causes the frictional strengths of all three gouges to increase. Limited data suggest that different polytypes of a given serpentine mineral have similar strengths; thus deformation-induced changes in polytype should not affect fault strength. At 25??C, the chrysotile gouge has a transition from velocity strengthening at low velocities to velocity weakening at high velocities, consistent with previous studies. At temperatures up to ???200??C, however, chrysotile strength is essentially independent of velocity at low velocities. Overall, chrysotile has a restricted range of velocity-strengthening behavior that migrates to higher velocities with increasing temperature. Less information on velocity dependence is available for the lizardite and antigorite gouges, but their behavior is consistent with that outlined for chrysotile. The marked changes in velocity dependence and strength of chrysotile with heating underscore the hazards of using room temperature data to predict fault behavior at depth. The velocity behavior at elevated temperatures does not rule out serpentinite as a cause of aseismic slip, but in the presence of a hydrostatic fluid pressure gradient, all varieties of serpentine are too strong to explain the apparent weakness of faults such as the San Andreas.