Leveling surveys across Medicine Lake volcano (MLV) have documented subsidence that is centered on the summit caldera and decays symmetrically on the flanks of the edifice. Possible mechanisms for this deformation include fluid withdrawal from a subsurface reservoir, cooling/crystallization of subsurface magma, loading by the volcano and dense intrusions, and crustal thinning due to tectonic extension (Dzurisin et al., 1991 [Dzurisin, D., Donnelly-Nolan, J.M., Evans, J.R., Walter, S.R., 1991. Crustal subsidence, seismicity, and structure near Medicine Lake Volcano, California. Journal of Geophysical Research 96, 16, 319-16, 333.]; Dzurisin et al., 2002 [Dzurisin, D., Poland, M.P., Bürgmann, R., 2002. Steady subsidence of Medicine Lake Volcano, Northern California, revealed by repeated leveling surveys. Journal of Geophysical Research 107, 2372, doi:10.1029/2001JB000893.]). InSAR data that approximate vertical displacements are similar to the leveling results; however, vertical deformation data alone are not sufficient to distinguish between source mechanisms. Horizontal displacements from GPS were collected in the Mt. Shasta/MLV region in 1996, 1999, 2000, 2003, and 2004. These results suggest that the region is part of the western Oregon block that is rotating about an Euler pole in eastern Oregon. With this rotation removed, most sites in the network have negligible velocities except for those near MLV caldera. There, measured horizontal velocities are less than predicted from ∼10 km deep point and dislocation sources of volume loss based on the leveling data; therefore volumetric losses simulated by these sources are probably not causing the observed subsidence at MLV. This result demonstrates that elastic models of subsurface volume change can provide misleading results where additional geophysical and geological constraints are unavailable, or if only vertical deformation is known. The deformation source must be capable of causing broad vertical deformation with comparatively smaller horizontal displacements. Thermoelastic contraction of a column of hot rock beneath the volcano cannot reproduce the observed ratio of vertical to horizontal surface displacements. Models that determine deformation due to loading by the volcano and dense intrusions can be made to fit the pattern of vertical displacements by assuming a weak upper crust beneath MLV, though the subsidence rates due to surface loading must be lower than the observed displacements. Tectonic extension is almost certainly occurring based on fault orientations and focal mechanisms, but does not appear to be a major contributor to the observed deformation. We favor a model that includes a combination of sources, including extension and loading of a hot weak crust with thermal contraction of a cooling mass of rock beneath MLV, which are processes that are probably occurring at MLV. Future microgravity surveys and the planned deployment of an array of continuous GPS stations as part of a Plate Boundary Observatory volcano cluster will help to refine this model.