We describe silica-rich (up to similar to 69.5% SiO2) glass inclusions trapped as grain boundary films and within-grain pockets in ultramafic xenoliths hosted by Pliocene basalts from the Sierra Nevada, California. The ultramafic xenoliths are Iherzolites which equilibrated in the Sierra Nevada upper mantle at 1150-1180 degrees C and similar to 1.4-1.8 GPa. The glass inclusions have trachytic compositions, similar to previously described silicic melts from mantle xenoliths [1-9]. We have determined the Sr and Nd isotope compositions of the grain boundary films using a leaching technique, and calculated the glass isotopic compositions. The glass Sr-87/Sr-86 (0.7077-0.7085) and Nd-143/Nd-144 (approximate to 0.51244) ratios are higher than in the ultramafic xenoliths and distinct from the host basalt ratios. Glasses are characterized by Nb depletions (Nb-n/Nb-n*similar to 0.15), enrichment of light rare earth elements (La-n/Yb-n approximate to 50), and the presence of negative Eu anomalies (Eu-n/Eu-n*approximate to 0.7-0.86), indicating a crustal origin for the melt source. The Nd isotope ratios (epsilon(Nd)approximate to-4) are inconsistent with an oceanic crust as the source for these former melts. The source rocks must have been continental materials recycled in the mantle, either foundered lower crust or subducted sediment. Low Rb/Sr (0.036-0.077) and high Sr/Nd (>35) ratios observed in the grasses are suggesting a lower crustal source. The Sierra Nevada lowermost crust (amphibole-bearing garnet pyroxenites and other dense Mesozoic cumulate mafic-ultramafic rocks), as defined by studies of older, Miocene xenolith-bearing volcanic rocks from the same area [M.N. Ducea, J. Saleeby, J. Geophys. Res. 101 (1996) 8229-8244], has isotopic compositions similar to the glass inclusions. Geologic [M.N. Ducea, J. Sareeby, J. Geophys. Res. 101 (1996) 8229-8244] and geophysical [G. Zandt, S. Ruppert, EOS Trans. AGU 77 (1996) 831] evidence indicate that the Sierra Nevada has lost its eclogitic are root, probably by foundering in the mantle. We propose here that the silica-rich glasses were formed by low percent partial melting of the dense, cold Sierran batholithic lowermost crust during root delamination. Further tests need to be aimed at addressing the viability of the main alternative to our interpretation, i.e. derivation of glasses from melting subducted sediments. (C) 1998 Elsevier Science B.V. All rights reserved.