The timing and duration of surface uplift associated with large igneous provinces provide important constraints on mantle convection processes. Here we review geological indicators of surface uplift associated with five continent-based magmatic provinces: Emeishan Traps (260??million years ago: Ma), Siberian Traps (251??Ma), Deccan Traps (65??Ma), North Atlantic (Phase 1, 61??Ma and Phase 2, 55??Ma), and Yellowstone (16??Ma to recent). All five magmatic provinces were associated with surface uplift. Surface uplift can be measured directly from sedimentary indicators of sea-level in the North Atlantic and from geomorpholocial indicators of relative uplift and tilting in Yellowstone. In the other provinces, surface uplift is inferred from the record of erosion. In the Deccan, North Atlantic and Emeishan provinces, transient uplift that results from variations in thermal structure of the lithosphere and underlying mantle can be distinguished from permanent uplift that results from the extraction and emplacement of magma. Transient surface uplift is more useful in constraining mantle convection since models of melt generation and emplacement are not required for its interpretation. Observations of the spatial and temporal relationships between surface uplift, rifting and magmatism are also important in constraining models of LIP formation. Onset of surface uplift preceded magmatism in all five of the provinces. Biostratigraphic constraints on timing of uplift and erosion are best for the North Atlantic and Emeishan Provinces, where the time interval between significant uplift and first magmatism is less than 1??million years and 2.5??million years respectively. Rifting post-dates the earliest magmatism in the case of the North Atlantic Phase 1 and possibly in the case of Siberia. The relative age of onset of offshore rifting is not well constrained for the Deccan and the importance of rifting in controlling magmatism is disputed in the Emeishan and Yellowstone Provinces. In these examples, rifting is not a requirement for onset of LIP magmatism but melting rates are significantly increased when rifting occurs. Models that attempt to explain emplacement of these five LIPs without hot mantle supplied by mantle plumes often have difficulties in explaining the observations of surface uplift, rifting and magmatism. For example, small-scale convection related to craton or rift boundaries (edge-driven convection) cannot easily explain widespread (1000??km scale) transient surface uplift (Emeishan, Deccan, North Atlantic), and upper mantle convection initiated by differential incubation beneath cratons (the hotcell model) is at odds with rapid onset of surface uplift (Emeishan, North Atlantic). The start-up plume concept is still the most parsimonious way of explaining the observations presented here. However, observations of surface uplift cannot directly constrain the depth of origin of the hot mantle in a plume head. The short time interval between onset of transient surface uplift and magmatism in the North Atlantic and Emeishan means that the associated starting plume heads were probably not large (??? 1000??km diameter) roughly spherical diapirs and are likely to have formed narrow (??? 100??km radius) upwelling jets, with hot mantle then spreading rapidly outward within the asthenosphere. In cases where rifting post-dates magmatism (N Atlantic Phase 1) or where the degree of lithospheric extension may not have been great (Siberia), a secondary mechanism of lithospheric thinning, such as gravitational instability or delamination of the lower lithosphere, may be required to allow hot mantle to decompress sufficiently to explain the observed volume of magma with a shallow melting geochemical signature. Any such additional thinning mechanisms are probably a direct consequence of plume head emplacement. ?? 2007 Elsevier B.V. All rights reserved.