The central question I explore is to what extent realism should be metaphysically informed. On the one hand, if the realist eschews metaphysics she runs the risk of being accused of doing little more than regurgitating the relevant science; but on the other, if she accepts too much she runs the kind of epistemic risk associated with metaphysical underdetermination. Here I suggest that a balance should be struck, by pulling away from certain metaphysical devices in order to avoid underdetermination, whilst deploying other such devices to take one towards a deeper form of realism. This meshes with an instrumentalist approach to metaphysics according to which the latter provides a kind of ‘toolbox’ from which such devices may be selected and further honed so that they can be deployed more precisely to help us understand what it is that science is telling us.