[1] The evolution through time of the oceanic lithosphere is a substantial, incompletely resolved geodynamical problem. Consensus remains elusive regarding its thermal structure, physical properties, and the best model through which to unify observational constraints. We robustly reevaluate all three of these by (i) simultaneously fitting heat flow, bathymetry, and temperatures derived from a shear velocity model of the upper mantle, (ii) using the three main thermal models (half-space, plate, and Chablis), and (iii) analyzing five depth-age curves, wherein contrasting techniques were used to exclude anomalous features from seafloor depths. The thermal models are updated to all include a temperature-dependent heat capacity, a temperature-and pressure-dependent thermal conductivity, and an initial condition of adiabatic decompression including melting. The half-space model, which lets the lithosphere thicken indefinitely, cannot accurately fit the subsidence curves and requires mantle potential temperatures, T m , that are too high. On the other hand, the models including a mechanism of basal heat supply are able to simultaneously explain all observations within two standard errors, with best-fitting parameters robust to the choice of the filtered bathymetry curve. For the plate model, which imposes a constant temperature at a fixed depth, T m varies within 1380-1390 C, the equilibrium plate thickness a within 106-110 km, and the bulk thermal expansivity a within 2.95À3.20 Á 10 À5 K À1 . For the Chablis model, which prescribes a fixed heat flow at the base of a thickening lithosphere, the best-fitting values are T m = 1320À1380 C, a = 176À268 km, a ¼ 3:05À3:60 Á 10 À5 K À1 . Driven by more accurate ocean depths, the plate model provides better joint-fittings to the observations; however, it requires values of a lower than experimentally measured, which can be explained by a reduction of the apparent expansivity due to elastic rigidity of the upper lithosphere. The Chablis model better fits the data when a is set close to or above the experimental values. Although statistically consistent within two standard errors, a tendency toward incompatibility between observed depth-age curves and seismically derived temperatures is revealed with new clarity, because the latter do not exhibit a clear steady state whereas the former flatten; further work is needed to identify the origin of this apparent discrepancy.This work opens the way to investigations fully independent of particular solutions of the heat equation.