Erupting volcanoes commonly exhibit characteristic ground deformation that is typically interpreted in terms of pressure changes of magma reservoirs within the crust. However, other processes may also be significant. Since 1995, the Soufrière Hills Volcano, Montserrat, has erupted about 1 km3 of magma over five discrete extrusive phases with clear cycles of associated ground deformation, recorded by GPS. Here we consider the contribution to deformation by loading of the ground surface with erupted deposits. We estimate topographic change and net deformation for the whole eruption to date, between November 1995 and February 2010. We derive the surface load distribution using differenced digital elevation data, which additionally enables us to constrain the budget of erupted lava. About a third of the lava erupted from Soufrière Hills since 1995 remains in subaerial, onshore deposits; more than previously thought. Another third is emplaced immediately offshore, and the remaining third has been transported further afield. We combine the deposit thickness map with representative deposit densities to calculate surface load and model the deformation response using finite elements. Our results show that net displacements accumulated over 14 years on Montserrat (tens of centimetres) could be explained by loading of erupted deposits on the flanks. The proportion of the observed deformation that can be explained by loading alone depends on crustal rheology. Using rheology structures favoured in the literature, our forward modelled displacements are remarkably similar to long-term observations, down to detail that we ascribe to localised load-topography interaction. Results suggest that the shallow crust beneath Montserrat is more compliant than usually assumed in geodetic models, with more rigid rheology at depth. Loading is largely accommodated by elastic strain in the shallow crust (top few kilometres) with negligible contribution from intra-crustal viscous flow over the time period investigated. We thus infer that the role of stress transfer from the surface load in metering magma reservoir behaviour must be negligible but may influence degassing in the shallow conduit, for example. Our findings suggest that, when volcanic ground deformation accompanies a voluminous eruption, geodetic model inversions will be misled if data are not appropriately corrected for the surface loading effect.