The time series of sulphur dioxide (SO2) emissions during the continuing eruption of Soufrière Hills Volcano, Montserrat, yields insights into conduit permeability and driving pressures, the role of the hydrothermal system and changes in magma flux both at depth and to the surface. On a time scale of years, an effectively constant supply of sulphur from a more mafic magma at depth permits evaluation of changes in the permeability of the plumbing system between 1995 and 2002 (due to magma rheology changes and hydrothermal sealing), most of which take place in the upper few hundreds of metres (dome and upper conduit). A broadly increasing SO2 emission rate from 1995 to 1997 can be attributed to a constant or increasing supply of exsolving sulphur from depth, combined with a broadly increasing magma discharge rate at the surface. Decreases in SO2 flux over three orders of magnitude, from July 1998 to November 1999, were due to a corresponding decrease in permeability of the upper conduit and dome due to cooling and 'sealing' by the precipitation of hydrothermal minerals and the closure of fracture and bubble networks. The second phase of dome growth, from November 1999 to the present, April 2002, has been associated with a similar range of SO2 fluxes to the first phase. Large dome collapses in 1997 and during a period of zero magma flux in 1998 were associated with instantaneous SO2 emissions of >10 kt, which indicate a capacity for significant SO2 storage in the conduit and dome prior to the collapses. SO2 data suggest that the second phase of dome building, despite a similar sulphur budget in terms of supply from depth and mean SO2 emission rate at the surface (around 500 t/d), is characterised by a higher bulk permeability at shallow depths and is a more 'open' system with respect to fluid through-flow than the first phase of dome building from 1995 to 1998. The lack of large SO2 emissions after large dome collapses, in 2000 and 2001, suggests limited storage of SO2 in the conduit system. The data suggest that the likelihood of a switch to explosive activity after a large collapse is more unlikely now than during the first phase of dome building. Over shorter time scales, permeability changes may be recognised from the SO2 flux data prior to the onset of dome growth and during cycles of small explosions in 1999. On time scales of minutes to hours, pulses of SO2 -rich gas emissions occur after rockfalls and pyroclastic flows, due to the release of a SO2 -rich fluid phase stored in closed fractures and pore spaces within the dome. Long period and hybrid seismic events may be associated with changes in SO2 emission rate at the surface at various times of the eruption, although only when the temporal resolution of SO2 monitoring is improved, will it be possible for these short-term changes to be correlated and evaluated effectively. Monitoring SO2 emission rates from Soufrière Hills Volcano is, at this stage, of primary value in the long run, on the time scale of years, where the relationships between deep supply and surface emissions can be used to evaluate whether the eruption might be waning, or has merely paused, which is of considerable value for hazard assessment.