Description: To examine the processes and histories of arc volcanism and of volcanism associated with backarc rifting. 130 samples containing igneous glass shards were taken from the Plioccne-Quatemai^ succession on the rift Hank (Site 788) and the Quaternary fill in the basin fill of the Sumisu Rift (Sites 790 and 791). These samples were subsequently analyzed at the University of Illinois at Chicago and Shizuoka University. The oxides determined by electron probe do not account for the total weight of the material; differences between summed oxides and 100% arise from the water contents, probably augmented by minor losses thai result from alkali vaporization during analysis. Weight losses in colorless glasses are up to 9%; those in brown glasses (dacitcs to basalts) arc no more than 4.5%; shards from the rift-flank (possibly caused by prolonged proximity to ihc scafloor) generally have higher values than those from the rift-basin fill How much of the lost water is magmatic, and how much is hydrated is uncertain; however, although the shards absorb potassium, calcium, and magnesium during hydration in the deep sea, they do so only to a minor extent that does not significantly alter their major element compositions. Therefore, the electron-probe results are useful in evaluating the magmatism recorded by the shards. Pre- and syn-rift Izu-Bonin volcanism were overwhelmingly dominated by rhyolile explosions, demonstrating that island arcs may experience significant silicic volcanism in addition to the extensive basaltic and basaltic andestic activity, documented in many arcs since the 1970s, that occurs in conjunction with the andesitic volcanism formerly thought to be dominant. Andesitic eruptions also occurred before rifting, but the andesitic component in our samples is minor. All the pre- and syn-rift rhyolites and andesites belong to the low-alkali island-arc tholeiitic suite, and contrast markedly with the alkali products of Holocene volcanism on the northernmost Mariana Arc that have been attributed to nascent rifting. The Quaternary dacites and andesites atop the rift flank and in the rift-basin fill are more potassic than those of Pliocene age, as a result of assimilation from the upper arc crust, or from variations in degrees of partial melting of the source magmas, or from metasomatic fluids. All the glass layers from the rift-flank samples belong to low-K arc-tholeiitic suites. Half of those in the Pliocene succession are exclusively rhyolitic: the others contain minor admixtures of dacite and andesite, or andesite and either basaltic andesite or basalt. In Contrast, the Quaternary (syn-rift) volcaniclastics atop the rift-flank lack basalt and basaltic andesite shards. These youngest sediments of the rift flank show close compositional affinities with five thick layers of coarse, rhyolitic pumice deposits in the basin fill, the two oldest more silicic than the younger ones. The coarse layers, and most thin ash layers that occur in hemipelagites below and intercalated between them, are low-K rhyolites and therefore probably came from sources in the arc. However, several thin rhyolitic ash beds in the hemipelagites are abnormally enriched in potassium and must have been provided by more distal sources, most likely to the west in Japan. Remarkably, the Pliocene-Pleistocene geochemistry of the volcanic front does not appear to have been influenced by the syn-rift basaltic volcanism only a few kilometers away. Rare, thin layers of basaltic ash near the bases of the rift-basin successions are not derived from the arc. They deviate strongly from trends that the arc-derived glasses display on oxide-oxide plots, and show close affinities to the basalts empted all over the Sumisu Rift during rifting. These basalts, and the basaltic ashes in the basal rift-basin fill, arc compositionally similar to those erupted from mature backarc basins elsewhere.