We have developed a new isostatic model for the East Pacific Rise (EPR) that explains the origin of both the axial topographic high and the rise crest gravity anomaly. Our model involves a thin elastic plate, with a consant thickness crust, that is broken at the rige crest. We assume that a buoyant force beneath the plate bends the free edge of the plate upward at the rise axis forming the topographic high and the rise crest gravity anomaly. We have used this model, and two conventional isostatic models, to examine the isostasy of three segments of the EPR: at 9¿-14¿N, 6¿-11¿S, and 16¿-21¿S. Rather than modeling individual profiles we have stacked (summed) the gravity and bathymetry data in each area and computed a single, 200-km-long, mean, mirrored profile for each different ridge segment. These stacked profiles are a remarkable improvement over even the best individual profiles and clearly show both the axial topographic high and its associated free air gravity anomaly. In all three areas the best fitting model parameters are similar and give average compensaiton depths of about 6-7 km below the seafloor (near the base of the oceanic crust) and an upper limit on the flexural rigidity of the plate that increases from about 1018 N m at the ridge axis to about 3¿1019 N m 5 km from the axis and 1020 N m 25 km or more from the rise crest. These results indicate that the EPR crest is compensated at much shallower depths than suggested by previous investigators. Our results are consisent with the presence of a small, low-density magma chamber in the lower crust, although at least part of the compensating root for the axial high must extend into the upper mantle. We interpret the low upper mantle densities below the rise crest as a zone of partial melt which accumulates at the base of the oceanic crust between episodic rifting events when the overlying crustal magma chamber is replenished.