We examine bathymetry and gravity observations acquired by surface ships and deflection-of-the-vertical data derived from the Geosat radar altimeter for the purpose of determining if the bathymetric relief along the Marquesas Fracture Zone, in the central Pacific Ocean, is caused by differential thermal subsidence across a locked fault. If the fracture zone can maintain between 20 and 40 MPa of shear stress, we would predict that a ridge and a parallel trough would develop to the south and north, respectively, of the fracture zone on account of flexural modification of the step in bathymetry locked into the plate along the fracture zone at the ridge-transform intersection. We find that only 2 out of 32 deflection-of-the-vertical profiles from the Geosat altimetric mission are consistent with this model, while none of the 27 bathymetric profiles or 9 shipboard gravity profiles fits the theoretical predictions. In cases where bathymetry and gravity can be modeled simultaneously, it appears that the gravity anomaly is simply not large enough in magnitude to be consistent with the Moho flexing conformably with the observed bathymetry, as required if the relief is produced by deeper thermal variations across the locked fracture zone. Rather, the potential field data allow shear stresses only 25%of those required to lock the fault and are more consistent with a model in which the relief on the Moho is opposite in sign to that of the surface topography, as would be produced by local or regional compensation of surface loads arising from volcanic activity along the fracture zone. Thus we conclude that the Marquesas Fracture Zone, in contrast to those other Pacific fracture zones which fit the predictions of the locked-fault (high stress) model, may be unusually weak along almost its entire length. ¿ American Geophysical Union 1992