Supported by vaious field geologic and petrologic data, the contents of Pb, U, Th, Rb, and Sr and the isotopic compositions of Pb and Sr for upper Cenozoic volcanic rocks of the Yellowstone Plateau volcanic field are consistent with the hypothesis of derivation of the basaltic and rhyolitic magmas by partial melting of distinct source regions in the upper mantle and lower crust, respectively. All the basalt samples analyzed but one have systematically lower values of 207Pb/204Pb and 87Sr/86Sr than the rhyolites. The values of 206Pb/204Pb are smaller, and 87Sr/86Sr are mostly larger than known values in oceanic basalts. In all but one case, the values of 207Pb/204Pb are higher than expected from an extrapolation of known values in oceanic basalts to less radiogenic values of 206Pb/204pb. Because there are no xenoliths, phenocrysts are only moderate to sparse in abundance. REE patterns are low and flat at the radiogenic end of lead isotopic compositions, several values of Rb/Sr are low, and 80% of the basalt samples form a well-developed secondary isochron separate from the rhyolites, we favor an interpretation for basalt genesis wherein isotopic signatures of most mafic magmas were attained in a continental 'keel' of mantlelike character about 2.6 b.y. old or somewhat older attached to the crust, and these signatures were unaltered by magma passage through the crust. At the very least, the current data continue to cast serious doubt as to the inevitability of crustal contamination for basaltic magma intruding the continental enviroment and postulate that much can be learned about the mantle under continents through the study of continental basalts. One basalt unit with an unusually low value of 207Pb/204Pb and an 87Sr/86Sr less than 0.704 may represent subcontinental 'keel'-derived magma that rose unaltered to the surface. Our data also are not consistent with formation of this rhyolite-basalt association primarily by such processes as crystal fractionation, separation of immiscible silicate liquids from a common parental magma, or fractional melting of a homogeneous source. Rather as a conceptual model, we envision large mafic intrusions to have been injected into the lower crust resulting in rhyolite generation through partial anatexis of the adjacent wall rocks which probably had a 206Pb/204Pb<17 and 87Sr/86Sr<0.709; a model that has much in common with that proposed by Holmes (1931). All the other hypotheses listed have the necessary added complication that either the basalt or the rhyolite or both become contaminated after the two magma types separated, have problems accounting for the lack of igneous rocks of intermediate compositions or production of such large volumes of rhyolitc material (~5000 km3), and fail to explain why rhyolitc magma is not a more common occurrence in the ocean basin. We appeal to bouyance of rhyolites to generate a barrier for basalt magma migration and account for the great preponderance of rhyolite relative to basalt at the surface. Furthermore, the complex isotopic picture in the rhyolites indicates than many of these magmas interacted with the upper crustal geologic units that they traversed. The interactions involved diverse processes, probably including reaction with hydrothermal fluids or hydrothermally altered rocks at high levels as well as by contamination with Phanerozoic sedimentary and Precambrian crystal-line rocks at deeper levels. At the very least, we feel our study adds a cautionary note to the currently increasingly popular hypothesis that differentiation of basalt or gabbro magnas to rhyolite or granite (as distinct from tonalite or dacite) is a common occurrence and is therefore an important continential building process. Models for formation of rhyolite and granite predominantly by reworking of crust (anatexis) must still be considered. The primitive Archean mantle of the region was characterized by higher Rb/Sr, U/Pb, and Th/U values than are typical of modern suboceanic mantle. The mantle residuum within the continental subcrustal lithospheric 'keel' that resulted from the Archean crustal differentiation event probably was depleted in Rb/Sr and U/Pb, and the curst was correspondingly enriched in these ratios. The crust probably was futher differentiated by an Archean high-grade metamorphism, during or after the primary event, into a granulitic lower crust depleted in U/Pb and Rb/Sr and a lower-grade upper crust enriched in these ratios.