Explosive eruptions involve mainly silicic magmas in which sulfur solu
bility and diffusivity are low. This inhibits sulfur exsolution during
magma uprise as compared to more mafic magmas such as basalts. Silici
c magmas can nevertheless liberate large quantities of sulfur as shown
by the monitoring of SO2 in recent explosive silicic eruptions in are
settings, which invariably have displayed an excess of sulfur relativ
e to that calculated from melt degassing. If this excess sulfur is sto
red in a fluid phase, it implies a strong preference of sulfur for the
fluid over the melt under oxidized conditions, with fluid/melt partit
ion coefficients varying between 50 and 2612, depending on melt compos
ition. Experimentally determined sulfur partition coefficients for a d
acite bulk composition confirm this trend and show that in volcanic er
uptions displaying excess gaseous sulfur, the magmas were probably flu
id-saturated at depth. The experiments show that in more reduced silic
ic magmas, those coexisting only with pyrrhotite, the partition coeffi
cient decreases dramatically to values around 1, because pyrrhotite lo
cks up nearly all the sulfur of the magma. Reevaluation of the sulfur
yields of some major historical eruptions in the light of these result
s shows that for oxidized magmas, the presence of 1-5 wt % fluid may i
ndeed account for the differences observed between the petrologic esti
mate of the sulfur yield and that constrained from ice core data. Expl
osive eruptions of very large magnitude but involving reduced and cool
silicic magmas, such as the Toba or the Bishop events, release only m
inor amounts of sulfur and could have consequently negligible long-ter
m (years to centuries) atmospherical effects. This redox control on su
lfur release diminishes as the melt composition becomes less silicic a
nd as temperature increases, because both factors favor more efficient
melt sulfur degassing owing to the increased diffusivity of sulfur in
silicate melts under such conditions.