This talk will review some results and conjectures about history dependent random walks. For example, edge reinforced random walk (ERRW) is a random walk which prefers to visit edges it has visited in the past. Diaconis showed that ERRW can be expressed as a random walk in a random environment. This environment is highly correlated and is described in terms of statistical mechanics. Phase transitions for closely related models are believed to occur in three dimensions. One phase corresponds to diffusion and the other phase to localization. This talk is based work of Merkl and Rolles on ERRW and my recent preprint with Disertori and Zirnbauer on a hyperbolic sigma model.