The volumes and estimated total lifetimes of the Tharsis shield volcanoes on Mars imply that they were built with an average magma supply rate within a factor of 2–3 of ∼0.05 m3 s−1. The morphologies and positions of the summit calderas of these volcanoes, specifically the overlapping relationships between the calderas, are evidence that multiple magma reservoirs existed within them, centered on distinctly different locations at various times. The requirement that the magma in an older reservoir must have cooled below its solidus temperature in order to change the local stresses in such a way that a new reservoir grew at a different location argues for long time gaps (probably tens of Myr) during which the supply rate of magma from the mantle was much smaller than at other times. Conversely, the need to offset conductive cooling to the surface once an active magma reservoir was established implies that the magma supply rate was large, ∼1–10 m3 s−1, for times of order 0.1 Myr. An initial pulse involving rates of more than 150 m3 s−1 acting for at least a few weeks was needed to initiate each new reservoir. The lengths of numerous long lava flows on the flanks of the Tharsis Montes imply that lava must have been erupted at rates of ∼100–300 m3 s−1 for periods of a few years to a few tens of years. The volumes of many of these flows exceed the amounts of magma that could be released from overpressured magma reservoirs of the sizes implied by the dimensions of the summit calderas with the surrounding rocks behaving elastically. A maintained high magma supply rate from the mantle (leading to buffered eruption conditions) is almost certainly required for such flows, whereas inelastic caldera collapse events (unbuffered eruption conditions) can explain the emplacement of other flows. Taken together, the various lines of evidence suggest that the Tharsis volcanoes were built episodically with active phases lasting less than 1 Myr alternating with ∼100 Myr quiet phases. We comment on the implications of these issues for the episodicity of the generation of magma in the mantle and for the mode of transport of the magma in dikes or diapirs.