We model broadband seismograms containing triplicated S, S2, and S3 along with ScS to produce a pure path one-dimensional model extending from the crust to the core-mantle boundary beneath the East Pacific Rise. We simultaneously model all body wave shapes and amplitudes, thereby eliminating depth-velocity ambiguities. The data consist of western North American broadband recordings of East Pacific Rise (EPR) affiliate transform events that form a continuous record section out to 82¿ and sample nearly the entire East Pacific Rise. The best fitting synthetics contain attenuation and small changes in lithospheric thickness needed to correct for variation in bounce point ages. The 660-km discontinuity is particularly well resolved and requires a steep gradient (4%), extending down to 745 km. We find no discernible variation in apparent depths of the 405- and 660-km discontinuities over ridge-orthogonal distances on the order of 1000 km (or 20 Ma lithosphere). Body waveform comparisons indicate that we can resolve discontinuity depths to less than ¿10 km, providing an upper limit to transition zone topography. These depth estimates, in conjunction with the fan shot nature of the ray paths, lower the detection limit from S2 precursor analysis of the lateral length scale over which short-wavelength topographic variation could occur and indicate the sub-EPR Transition Zone and upper mantle are remarkably homogeneous. The lower mantle beneath the East Pacific Rise is well modeled by PREM, with the greatest variation occurring in ScS, reflecting strong heterogeneity along the core-mantle boundary. Together, these observations require that the East Pacific Rise spreading ridge cannot be actively supplied from the local lower mantle and that tomographically imaged lateral variation beneath the ridge likely reflects lateral smearing of outlying velocity gradients. Dynamically, the transition zone therefore appears vertically decoupled from the overlying East Pacific Rise spreading system.