Analysis of a deployment of broadband sensors along a 500-km-long line crossing the Yellowstone hotspot track (YHT) has provided 423 in-plane receiver functions with which to image lateral variations in mantle discontinuity structure. Imaging is accomplished by performing the converted wave equivalent of a common midpoint stack, which significantly improves resolution of mantle discontinuity structure with respect to single-station stacks. Timing corrections are calculated from locally derived tomographic P and S wave velocity images and applied to the Pds (where d is the depth of the conversion) ray set in order to isolate true discontinuity topography. Using the one-dimensional TNA velocity model and a Vp/Vs ratio of 1.82 to map our Pds times to depth, the average depths of the 410- and 660-km discontinuities are 423 and 664 km, respectively, giving an average transition zone thickness of 241 km. Our most robust observation is provided by comparing the stack of all NW back-azimuth arrivals versus all SE back-azimuth arrivals. This shows that the transition zone thickness varies between 261 and 232 km, between the NW and SE portions of our line. More spatially resolved images show that this transition zone thickness variation results from the occurrence of 20--30 km of topography over 200--300 lateral scale lengths on the 410- and 660-km discontinuities. The topography on the 410- and 660-km discontinuities is not correlated either positively or negatively beneath the 600-km-long transect, albeit correlation could be present for wavelengths larger than the length of our transect. If this discontinuity topography is controlled exclusively by thermal effects, then uncorrelated 250¿ lateral temperature variations are required at the 410- and 660-km discontinuities. However, other sources of discontinuity topography such as the effects of garnet-pyroxene phase transformations, chemical layering, or variations in mantle hydration may contribute. The most obvious correlation between the discontinuity structure and the track of the Yellowstone hotspot is the downward dip of the 410-km discontinuity from 415 km beneath the NW margin of the YHT to 435 km beneath the easternmost extent of Basin and Range faulting. Assuming this topography is thermally controlled, the warmest mantle resides not beneath the Yellowstone hotspot track, but 150 km to the SE along the easternmost edge of the active Basin and Range faulting.¿ 1997 American Geophysical Union