We experiment with backprojection migration processing of teleseismic receiver functions from the Snake River Plain (SRP) broadband seismic experiment. Previous analyses of data from this experiment have used a common midpoint (CMP) stacking approach, a method widely applied for analysis of P‐SV converted phases (receiver functions) to obtain high‐resolution imaging of upper mantle discontinuities. The CMP technique assumes that all P‐SV conversions are produced by flat‐lying structures and may not properly image dipping, curved, or laterally discontinuous interfaces. In this paper we adopt a backprojection migration scheme to solve for an array of point scatterers that best produces the large suite of observed receiver functions. We first perform synthetic experiments that illustrate the potential improvement of migration processing over CMP stacks. Application of the migration processing to the SRP data set shows most of the major features as in the original CMP work, but with a weaker 410‐km discontinuity and a more intermittent discontinuity at 250 km apparent depth. Random resampling tests are also performed to assess the robustness of subtle features in our discontinuity images. These tests show that a 20‐km elevation of the 660‐km discontinuity directly beneath the Snake River Plain is robust, but that the variations in 410‐km discontinuity topography that we observe are not stable upon resampling. “Bright spots” near 250 km apparent depth are robust upon resampling, but interpretation of these features is complicated by possible sidelobe artifacts from topside Moho reverberations.