Previous studies of Pinatubo have shown that basaltic magma was injected into what was thought to be a homogeneous dacitic magma reservoir just before the 1991 eruptions. Mingling of the basalt and dacite resulted in a volatile-charged hybrid andesitic magma that ascended through the overlying dacite reservoir, triggering initial andesitic eruptions and the climactic eruption of dacite a few days later. Using mineral and whole-rock chemistry and isotopic data, we show that the dacitic magma was not homogeneous, but was instead heterogenous, produced from heterogeneous dacitic magma, which is approximated by at least two distinct silicic end-member magmas. Our data show that these silicic magmas were not produced by fractional crystallization of basaltic magma, such as that which formed the basalt enclaves of the 1991 and Buag period basalts (similar to 1.0-0.5 ka). Instead, we argue that the 1991 dacite magmas were produced independently but simultaneously by the melting of heterogeneous crustal materials. Based on these results, we propose the following sequence of events. First, basaltic magma distinct from that of the Buag period underplated and heated the crust to form silicic melt pockets of various compositions in a crustal mush zone. During the period between 0.5 ka and 1991, these silicic melts separated from the crustal mush zone and accumulated in the crust, forming the 1991 dacitic magma reservoir. Finally, during the climactic phase of the 1991 eruption, heterogeneous dacitic pumices were erupted, having a composition range that is representative of their variable crustal origins.