Santorini caldera has had a long history of plinian eruptions and caldera collapses, separated by 20-40 kyr interplinian periods. We have carried out a study to constrain magma storage/extraction depths beneath the caldera. We analysed H sub(2)O in 138 olivine-, pyroxene- and plagioclase-hosted melt inclusions from plinian and interplinian products from the last 200 kyr, and CO sub(2), S, Cl, F and delta D in various subsets of these. The dataset includes 64 inclusions in products of the Minoan plinian eruption of the late 17th century BCE. All the melt inclusions were ellipsoidal and isolated, with no textural evidence for volatile leakage. Mafic melt inclusions contain 1-4 wt % H sub(2)O and up to 1200 ppm CO sub(2), 1200 ppm S, 2000 ppm Cl and 400 ppm F; silicic inclusions contain 2-7 wt % H sub(2)O, up to 150 ppm CO sub(2), up to 400 ppm S, 2000-6000 ppm Cl and 600-1000 ppm F. The delta D values of 27 representative inclusions (-37 to -104ppt) are intermediate between mantle and slab values and rule out significant H sub(2)O loss by hydrogen diffusion from olivine-hosted inclusions. H sub(2)O, S and Cl behave compatibly in melt inclusion suites varying from mafic to silicic in composition, showing that entrapment of many melt inclusions took place under volatile-saturated conditions. Most Santorini melts are saturated in a free COHSCl vapour phase at depths of less than 10 km; the only exceptions are basaltic melts from a single interplinian eruption, which were volatile-undersaturated up to K sub(2)O contents of 1 wt %. The rhyolitic melt of the Minoan eruption probably contained a free hypersaline liquid phase. H sub(2)O + CO sub(2) saturation pressures were calculated using suitably calibrated solubility models to estimate pre-eruptive magma storage depths. Magmas feeding plinian eruptions were stored at >4 km (>100 MPa) and extracted over depth intervals of several kilometres. Plagioclase phenocrysts in rhyodacitic pumice from the Minoan eruption have cores containing melt inclusions trapped at depths up to 10-12 km (320 MPa), and rims (also orthopyroxene and clinopyroxene) containing inclusions trapped at 4-6 km (100-160 MPa). This records late-stage silicic replenishment of a <2 km thick shallow magma chamber, rather than extraction of melts syn-eruptively over the entire depth range. The plagioclase cores were carried from depth in the ascending melt, then overgrown by the rims in the shallow chamber. Exsolution of volatiles during ascent may have caused the replenishment melt to inject as a bubbly plume, causing mixing prior to eruption. This would explain (1) the homogeneity of the Minoan rhyodacitic magma, and (2) extraction of melt inclusions from the entire pressure spectrum during the first eruptive phase. Most silicic magmas feeding eruptions of the interplinian periods were stored in reservoirs at shallow depths (2-3 km) compared with those feeding the plinian eruptions (>4 km). Melt inclusions from the ad 726 eruption of Kameni Volcano yield a pre-eruptive storage depth of 4 km, which is similar to that estimated from geodetic data for the inflation source during the 2011-2012 period of caldera unrest; this supports a magmatic origin of the unrest. The level of pre-ad 726 magma storage beneath Kameni was deeper than that of earlier silicic interplinian eruptions, perhaps owing to changes in crustal stress caused by the Minoan eruption. Combined with previously published results, the melt inclusion data provide a time-integrated image of the crustal plumbing system. Mantle-derived basalts are injected into the lower crust, where they fractionate to produce evolved melts in bodies of hot crystal mush. Evolved residual melts separate from their parent mushes in the 8 to >15 km depth interval, then ascend rapidly into the upper crust, where they either crystallize or accumulate as bodies of eruptible, crystal-poor magma.