The lack of field evidence for significant crustal stretching during the main phase of flood basalt eruption rules out a passive-mantle origin for the Columbia River Basalt Group. The short duration (1.1 m.y.) of the main-phase lavas, their large volume (~195,500 km³), and their high eruption rate (~0.178 km³/yr) in an area of minimal extension (<<1%) are better explained by an active-mantle upwelling origin. Such a scenario could be satisfied by the arrival of a lower-mantle plume, or by the ascent of upper mantle through a slab gap or around the truncated edge of the Farallon plate. These competing ideas have been based on the interpretation of seismic data beneath the Yellowstone region that appear to resolve either a sheet of mantle upwelling extending downward into the mantle transition zone, or an irregular corridor of upwelling extending through the transition zone and into the lower mantle to ~900–1000 km depth. A lower-mantle origin is supported by the seismically resolved upward deflection in the ~660 km discontinuity at the base of the transition zone, which in turn is consistent with a lower-mantle chemical signature reflected in high ³He/⁴He ratios found in derivative basalts from both the Snake River Plain and the Columbia River flood basalt province. Mid-Miocene arrival of the Yellowstone plume was centered beneath the Oregon-Nevada border region, where there are direct spatial and temporal connections among the initiation of flood basalt volcanism at ca. 16.7 Ma, the initiation of rhyolitic volcanism at the western end of the Snake River Plain hotspot track at ca. 16.5 Ma, and the initiation of a major period of uplift and crustal extension in the northern Basin and Range between ca. 17 and 16 Ma. Rapid uplift and the propagation of volcanism away from this region are consistent with models of plume emplacement. The inability of seismic studies to image a mantle-plume fabric of radiating flow is due to its destruction by corner flow, which is evident in the contemporary east-west flow fabric derived from shear-wave splitting measurements.