Whether the volcanism of the Columbia River Plateau, eastern Snake River Plain, and Yellowstone (western U.S.) is related to a mantle plume or to plate tectonic processes is a long-standing controversy. There are many geological mismatches with the basic plume model as well as logical flaws, such as citing data postulated to require a deep-mantle origin in support of an “upper-mantle plume” model. USArray has recently yielded abundant new seismological results, but despite this, seismic analyses have still not resolved the disparity of opinion. This suggests that seismology may be unable to resolve the plume question for Yellowstone, and perhaps elsewhere. USArray data have inspired many new models that relate western U.S. volcanism to shallow mantle convection associated with subduction zone processes. Many of these models assume that the principal requirement for surface volcanism is melt in the mantle and that the lithosphere is essentially passive. In this paper we propose a pure plate model in which melt is commonplace in the mantle, and its inherent buoyancy is not what causes surface eruptions. Instead, it is extension of the lithosphere that permits melt to escape to the surface and eruptions to occur—the mere presence of underlying melt is not a sufficient condition. The time-progressive chain of rhyolitic calderas in the eastern Snake River Plain–Yellowstone zone that has formed since basin-range extension began at ca. 17 Ma results from laterally migrating lithospheric extension and thinning that has permitted basaltic magma to rise from the upper mantle and melt the lower crust. We propose that this migration formed part of the systematic eastward migration of the axis of most intense basin-range extension. The bimodal rhyolite-basalt volcanism followed migration of the locus of most rapid extension, not vice versa. This model does not depend on seismology to test it but instead on surface geological observations.