My dissertation centers on the epistemology of computer simulations and modeling. Examining how computer simulation models are built in cases where there is important empirical information lacking, I argue in my first paper that computer simulation modeling is ampliative. That these computer simulation models are built with computer simulations weighs on the argument that representational similarity is the crux of simulation modeling. To that end in my second paper I examine accounts of simulation methodology and similarity accounts of modeling to see whether they can account for the kind of learning that takes place with computer simulation models. Finally, the solution of inverse problems (which is done via computer simulation) is nonunique and the models are underdetermined by data. In this third paper I consider in what sense we think we learn about the target system by simulation modeling. In the face of the nonuniquness and underdetermination, is it possible to maintain a realist position?