There is now su¯cient information to attempt an integrated model for melt generation, transfer and storage beneath subduction-zone volcanoes. Fluid release from the subducting oceanic crust into the mantle wedge may occur over a period ranging from a few hundred kyr, to as little as less than 1 kyr, before eruption. This supports models in which ®uid addition is closely linked to partial melting, though there may also be evidence for a component of decompression melting. The timing of the onset of ®uid addition may be linked to the rate of subduction (i.e. water supply rate) and the angle of subduction, and, consequently, the thermal structure of the mantle wedge. In contrast, contributions from subducted sediments to subduction-zone lava sources appear to occur some 350 kyr{4 Myr before eruption. Evidence for partial melting of the sediment component, combined with the short ®uid transfer times, phenocryst equilibration temperatures and other observations all point to quite high mantle wedge temperatures close to the interface with the subducting plate. New 226 Ra data permit only a short period of time between ®uid addition and eruption. This requires rapid melt segregation, magma ascent by channelled ®ow and minimal residence time within the lithosphere. Typically, the evolution from basalt to andesite occurs rapidly during ascent or in magma reservoirs, inferred from some geophysical data to lie within the lithospheric mantle. Mineral isochron data suggest that some andesitic magmas subsequently stall in more shallow crustal level magma chambers, where they can evolve to dacitic compositions via fractionation, typically combined with assimilation, on time-scales of a few thousand years or less.