We present a reappraisal of Columbia River basalt petrogenesis based on an internally consistent X-ray fluorescence and inductively coupled plasma–mass spectrometry data set for major and trace elements plus new and existing isotopic analyses of the Imnaha, Steens, Picture Gorge, and Grande Ronde Basalts. Source materials for the main-phase Columbia River Basalt Group are upwelling ocean-island basalt source–like mantle, depleted mantle variably fluxed by slab-derived fluids, Phanerozoic arc crust, and ancient North American cratonic crust. The mantle upwelling may be a deep-seated plume or material displaced and mobilized by fragmented sinking slabs. We endorse the conclusions of earlier workers that the Imnaha, Steens, and Picture Gorge Basalts represent different mixtures of upwelling mantle, depleted mantle, slab-derived fluids, and crust. Cratonic crust of the Idaho batholith has a minor role as a contaminant of Imnaha basaltic magma and a major role in the petrogenesis of the Grande Ronde basaltic andesites, which we model as contaminated Imnaha basalt. We emphasize the geochemical continuity of the Imnaha and Grande Ronde Basalts and propose that they derive from a single central crustal magma system (or chamber) that lasted from ca. 16.7 Ma to 16.0 Ma. The Imnaha–Grande Ronde magma system was centered beneath the location where the western Snake River Plain, Oregon-Idaho graben, and Chief Joseph dike swarm converge, and probably transgressed the cratonic boundary to the east. Magma from this system was injected into dikes and traveled hundreds of kilometers northward to erupt and feed the gigantic Grande Ronde lava flows. In contrast to previous studies, we question the idea that the dike swarm provides a geographic map of the magma source regions.