This paper seeks to explain how thought experiments work, and also the reasons why they can fail. The paper is split into four sections. The first argues that thought experiments in philosophy and science should be treated together. The second examines existing accounts of thought experiments, and shows why they are inadequate. The third proposes a better account of thought experiments. According to this account, a thought experimenter manipulates her world view in accord with the â��what ifâ�� questions posed by a thought experiment. When all necessary manipulations are carried through the result is either a consistent model, or contradiction. If a consistent model is achieved the thought experimenter can conclude that the scenario is possible, if a consistent model cannot be constructed then the scenario is not possible. The fourth section of the paper uses this account to shed light on the circumstances in which thought experiments fail.