Mid-ocean ridge topography is modeled as the flexural response to loads using a thin plate approximation and setting thermal structure of the lithosphere to allow, but not require, a region of rapid cooling near the axis. Loads on the lithosphere arise from the presence of low-density melt, densification due to cooling with distance from the ridge axis, and thermal contraction stresses. We find two end-member classes of temperature and melt structure that can produce axial high topography and gravity observed at the East Pacific Rise (EPR). One class is very similar to previous models, requiring a narrow column of melt extending to at least 30 km depth within the mantle and lithosphere which cools and thickens very gradually with distance from the ridge axis. The other is a new class, predicting lithosphere which cools rapidly within a few kilometers of the axis and then slowly farther from the axis, with melt which is contained primarily within the crust. The latter solution is consistent with tomography and compliance studies at the EPR which predict rapid crustal cooling within a few kilometers of the axis that is attributed to hydrothermal circulation. This solution also allows the melt region to be coupled to crustal thermal structure and requires no melt anomaly within the mantle. Model fits predict 0--30% melt in the lower crust, depending on how temperatures are distributed within the lithosphere and the degree to which thermal contraction stresses are assumed to contribute to topography. The model generally predicts a wider axial high for lithosphere which is thin over a wider region near the axis. This is consistent with previous correlations between large cross-sectional area of the high and indicators of higher melt presence or a warmer crustal thermal regime. For a slightly slower rate of lithospheric cooling at distances more than ~5 km from the axis the model predicts a trough at the base of the axial high. Such troughs have been previously observed at the base of the high on the western flank of the southern EPR, where subsidence rates are anomalously low. Finally, thick axial lithosphere reduces the amplitude of the high, making it sometimes difficult to distinguish from long-wavelength subsidence. This morphology is comparable to that of some intermediate spreading ridges, where topography is relatively flat, suggesting a transition from fast to intermediate style morphology. ¿ 2001 American Geophysical Union