The mantle is the most massive carbon storage reservoir on Earth. Interactions between surface and mantle reservoirs of carbon strongly govern atmosphelic chemistry and the habitability of our planet on geological timescales. The petrological and dynamical behaviour of carbon in mantle rocks promotes chemical differentiation and extraction of heat-producing elements through the migration of low-degree melts. Additionally carbonate melts mobilise hydrogen from nominally anhydrous minerals, which helps lower mantle viscosity and maintain active tectonics. I present four studies, each of which focusses on a different aspect of carbon's behaviour under high-pressure conditions. I study a suite of Brazilian diamonds and their mineral assemblages (Chapter 2). The isotopic signature of the host diamonds and mafic chemistry of their inclusion cargo reveals that they are fragments of recycled crust that has been exhumed from mid-mantle depths. In Chapter 3 I determine the melting curves of MgSi03- MgC03 and MgC03-CaC03, analogues of carbonated mantle material, between 15 and 80 GPa by performing experiments in the laser-heated diamond-anvil cell. Results from these experiments demonstrate that carbon-bearing lithologies will melt, potentially creating an isolated carbon reservoir, in the lower mantle. Chapter 4 presents multi anvil experiments that investigate the melting phase relations of carbonated MORB throughout the upper mantle. These reveal a solidus ledge that implies the transition zone can act as a carbon filter to downwelling material. This causes a narrow interval of melt production that could be geophysically observable and responsible for diamond formation. Finally, in chapter 5 I present experiments and geochemical modelling that tests whether diamond-hosted inclusion assemblages are compatible with a transition zone growth mechanism where slab-derived melts interact with ambient mantle material.