Geoid anomalies derived from the Geosat altimeter data set have been used to model the change in the geoid height across four Pacific fracture zones. After accounting for geoid signals from the flexural response to differential subsidence and thermal bending stresses, we find that the estimates of geoid steps across the fracture zones are compatible with existing conductive models for the oceanic lithosphere. If flexure is not accounted for, the geoid steps will be underestimated by 30--40%. Departures from the predicted conductive trends may be explained by small-scale convection and uncertainties in the crustal ages. ¿ American Geophysical Union 1989