A natural experiment in landscape evolution is a case study of landform development in which only one element varies significantly, and for which the driving forces, initial conditions, and/or boundary conditions are well constrained. Natural experiments provide a means of testing landscape evolution theory on the large space and time scales to which that theory applies. Natural experiments can involve either steady or transient conditions. Cases with steady conditions allow one to test predictions about the relationships among topography, erosion rates, and various attributes related to climate and material properties. Transient cases are valuable for distinguishing between models whose predictions might be similar, and therefore indistinguishable, under steady conditions. Essential ingredients of a natural experiment include minimal variation in all but one factor, good constraints on timing and/or rates, well-characterized processes, and high quality topographic data. Other useful ingredients include information about intermediate topographic states (such as a former valley profile revealed by strath terraces), and knowledge of the time history of erosion rates. In order to deepen our understanding of the physics and chemistry of longterm landscape evolution, there is a pressing need to identify natural experiments and develop the necessary databases to take advantage of them.