We have studied the distribution of light and volatile elements (Li, Be, B, H2O, F, Cl) in matrix glasses and melt inclusions in quartz and sanidine phenocrysts from tephras of the Cerro Toledo Rhyolite, which were erupted from the Bandelier magma chamber between two caldera-forming eruptions at 1.608 Ma and 1.225 Ma. Since the small-volume tephras of the Cerro Toledo Rhyolite record chemical evolution of the upper parts of the magma chamber during this 383-kyr period, we can track volatile evolution in the magma during its differentiation, specifically, the development of a separate fluid phase from the magma, the partitioning of volatile and light elements between this fluid and the magma, and the timing of the second caldera eruption at 1.225 Ma. The melt inclusion data reveal that Li, Be, and B were progressively enriched with time at the top of the magma chamber, showing a similar three fold enrichment. By contrast, F, Cl, and H2O show more variable behavior. F was enriched 3.6 times in the magma and Cl increased 2.5 times, while water concentrations remained constant. Evidence for a separate fluid phase includes (1) comparatively small magmatic enrichment factors for Cl and H2O, elements which are known to partition into fluid, (2) decline of the Cl/Be and Cl/B ratios in melt inclusions with time, indicating preferential loss of Cl, and (3) maximum Cl contents of 2900--3000 ppm in melt inclusions, which are concentrations at or close to the saturation limit for Cl in a two-phase gas-saturated magma. The lack of temporal H2O enrichment in the magma suggests that it became gas saturated at an early stage. The Cl fluid magma partition coefficient was initially low, due to the presence of a low-salinity single-phase vapor. The fluid is calculated to have increased its salinity from 0.76 wt % equivalent NaCl to 3.1 wt % NaCl by crystallization and Cl enrichment in the magma. Consequently, the fluid was transformed to two immiscible phases. As a result, the Cl partition coefficient increased, and the Cl content in the magma was buffered at 2900--3000 ppm. Overpressures in the magma chamber have been modeled as a function of crystallization, total pressure and depth, H2O solubility in rhyolite, resurgence, crystallization contraction, and initial fluid mass. For the large amounts of crystallization (40--70%) required by the geochemical relations, unrealistically large overpressures (>75 MPa) are indicated unless (1) the magma existed as a foam or (2) the chamber was able to degas passively, and thus buffer the buildup of pressure. To maintain realistic overpressures in the magma chamber (<50 MPa), an equilibrium was required between gas buildup by crystallization and volatile loss by passive degassing. The change from a one-phase vapor to a two-phase immiscible fluid before the Upper Bandelier Tuff eruption may have promoted partial or complete sealing of the magmatic-hydrothermal system, permitting significant overpressures to develop. The result could have been a cataclysmic caldera-forming eruption. ¿ American Geophysical Union 1996