Drilling results from several western Pacific atolls indicate the long-term subsidence of these islands is much more than would be expected from the cooling and thickening of the underlying lithosphere. This excess subsidence cannot be satisfactorily explained by isostatic adjustments to the weight of the volcano or the coral reef cap. It appears to be related to island formation atop unusually shallow areas of sea floor, like the Hawaiian swell, associated with midplate hot spots. The excess subsidence is caused by the gradual return of these shallow areas to normal depths. Several authors have suggested that the Hawaiian swell is supported by upward flow in the asthenosphere. However, this model offers no reasonable explanation for the shape of the swell or the observed rates of atoll subsidence. The regional gravity anomaly over the Hawaiian swell indicates an average depth of compensation within the lower half of the lithosphere, not within the asthenosphere, as would be expected if the swell were maintained by asthenospheric flow. While the compensating mass may extend to greater depths, most of the density changes appear to occur within the lithosphere. We propose that the Hawaiian swell is formed by lithospheric thinning over the Hawaiian hot spot. Since the asthenosphere is less dense than the lithosphere, replacement of the lower portion of the lithosphere by asthenospheric material causes isostatic uplift of the surface of the plate. After the lithosphere moves away from the hot spot, it cools and thickens, and the swell subsides. The subsidence histories of the Hawaiian swell and several Pacific atolls are in quantitative agreement with this mechanism. The main problem with this model is that it requires fast rates of lithospheric heating.