Nine principal layers of tephra have been distinguished in the vicinity of Mount Katmai and Novarupta. These were products of the 1912 eruption, and five of them represent major explosive events. Isopachous maps for four of the five layers show conclusively that they originated at Novarupta; while the thickness and coarseness of the other — the bottom layer — suggest that it could not have come from Mount Katmai, and that Novarupta is very probably its source also. The great tuff flow in the Valley of Ten Thousand Smokes may have started before the explosive phase of the 1912 eruption, but a major part of it was emplaced after the first cataclysmic eruption of rhyolite tephra. The banded pumices of the tephra deposits indicate the mixing of two magmas — rhyolitic and andesitic. The rhyolitic chamber underlies Novarupta, while the andesitic chamber underlies the chain of Mounts Katmai, Trident, Mageik, and Martin. Only the topmost layer of very fine tephra originated from the conduit at Mount Katmai. Its ejection was accompanied by the outpouring of the Katmai River tuff flow from a source within the crater and by the outpouring of a smaller, but similar, tuff flow down the western slope of Mount Katmai over glacier 3. A small steeply dipping welded tuff of the 1912 eruption occurs within the crater along the western wall. Its presence there, together with the outpouring of tuff flows down the flanks of the volcano at a very late stage of the eruptive activity, indicate that the summit of Mount Katmai collapsed slowly during the entire period of the 1912 eruption and that the crater only reached its final dimensions after almost all activity ceased. The summit of Mount Katmai was supported during most of the eruption by a column of magma, whose upper part must have been at or close to the surface, but which did not erupt until all activity ceased at the principal source, Novarupta. This suggests that the two conduits were intimately connected throughout most of the 1912 eruption, but were finally separated near the end. The sequence of rhyolite, mixed magma, and rhyolite from the conduit at Novarupta also supports this hypothesis. The 2.6 cubic miles of tuff flow in the Valley of Ten Thousand Smokes reduces to 1.5 cubic miles of solid andesite, almost exactly equal to the volume of the crater and missing summit of Mount Katmai. This equivalence is probably fortuitous, however, because there is no apparent compensation anywhere for much of the 4.75 cubic miles of tephra (approximately 1.9 cubic miles of solid rock).