The large-signal performance of electrostrictive materials, such as lead magnesium niobate-lead titanate, is of critical importance to sonar and actuator designers. However, obtaining these large signal parameters properly, particularly under compressive prestress, is expensive and time consuming. The complexity of these measurements precludes them as a method for quickly and easily screening materials. Traditionally, resonance measurements, which otherwise are relatively simple to perform, have been used, but they suffer from the drawback that the material parameters obtained are at the incorrect frequency and under no prestress. Furthermore, the significance of the results of resonance measurements for nonlinear materials is unclear. It has recently been suggested [Hom and Shankar, IEEE Trans. Ultrason. Ferroelectr. Freq. Control UFFC-46, p. 1422, 1999] that dc biased resonance measurements on electrostrictive ceramics would be an accurate predictor of the coupling factor and optimum bias point. Here, dc biased resonance measurements on several formulations, with varying dielectric maximum temperatures, are analyzed to determine which composition has the highest predicted coupling factor. This prediction is compared with large signal quasistatic measurements conducted on NAVSEAs SDECS (stress-dependent electromechanical characterization system). The predictive ability of the resonance measurements is analyzed as a function of temperature. [Work supported by ONR.]