Gravity data over Ontong Java Plateau reveal deep crustal structure supporting the plateau's topography. At long wavelengths (>500 km), new shipboard gravity and bathymetry profiles indicate local isostatic compensation. At intermediate wavelengths (250–500 km), however, Bouguer anomalies have large amplitudes, too high to be explained by models of Airy isostasy or volcanic loading on the surface of an elastic lithospheric plate. This finding is also evident in admittance functions, generated from maps of bathymetry and satellite-derived Bouguer anomalies, which show high values at intermediate wavelengths. In addition, coherence between two-dimensional Bouguer gravity and bathymetry spectrums decreases to zero, beginning at intermediate wavelengths. Two possible models can explain the high admittance and low coherence at intermediate wavelengths. The first model explains the high admittance by large-scale lithospheric folding and explains the low coherence by sedimentation and erosion that is uncorrelated with the igneous crustal structure. Lithospheric folding might result from tectonic compression imposed on Ontong Java by the solomon arc and trench system. The second model involves a multiple-stage accretion history with surface magmatism on a thin elastic plate and magmatic underplating beneath a thick plate. The thick plate during underplating is required to explain the high admittance values at intermediate wavelengths, while decorrelation between underplating and the surface loaded magmas can explain the low coherence. The thick plate also suggests that the lithosphere cooled appreciably since the first magmatic stage and therefore magmatic underplating occurred at a significantly later time than the initial eruption. Calculations predict that the underplated material could comprise as much as 50% of the total crustal volume, suggesting that the later eruptive stage(s) may have been as, or more voluminous than, the first.