The Joe Lott Tuff Member of the Mount Belknap Volcanics is the largest rhyolitic ash-flow tuff sheet in the Marysvale volcanic field. It was erupted 19 m.y. ago, shortly after the changeover from intermediate-composition calc-alkalic volcanism to bimodal basalt-rhyolite volcanism. Eruption of the tuff resulted in the formation of the Mount Belknap Caldera whose pyroclastic intracaldera stratigraphy parallels that in the outflow facies. The Joe Loft Tuff Member is a composite ash-flow sheet that changes laterally from a simple cooling unit near the source to four distinct cooling units toward the distal end. The lowest of these units is the largest and most widespread; it is 64 m thick and contains a basal vitrophyre. Eruption of the lower unit led to the initial collapse of the caldera. The lower unit is followed upward by a 43 m middle unit, a 26 m pink-colored unit which is separated by a prominent air- fall layer, and a 31 m upper unit. The Joe Loft Tuff Member is an alkali rhyolite with 75.85-77.31 wt. % silica and 8.06-9.32 wt. % K2O+Na2O; the agpaitic index (Na2O+ K2O/Al2O3) is .77-.98. The tuff contains about I% phenocrysts of quartz, sanidine, oligoclase, augite, apatite, zircon, sphene, biotite, and oxidized Fe-Ti oxides. The basal vitrophyre contains accessory allanite, chevkinite, and magnesiohastingsite. The main cooling units are chemically and mineralogically zoned indicating that the magma chamber restratified prior to each major eruption. Within each of the two thickest cooling units, the mineralogy changes systematically upwards; the Or content and relative volume of sanidine decreases and An content of plagioclase increases. The basal vitrophyre of the lower unit has a bulk composition that lies in the thermal trough near the minima of Or-Ab-Q at 1 kb PH2O. Microprobe analyses of feldspar and chemical modeling on experimental systems indicate that pre-eruption temperatures were near 750?C and that the temperature increased during the eruption of the cooling units. The chemical gradients in the apatite and whole-rock data in the Joe Loft Tuff Member and the consistent mineral assemblages throughout the ash-flow cannot be explained by crystal settling. The fractionation of the Joe Lott Tuff Member appears to closer fit the model of convection-driven thermogravitational diffusion.