Procedurally Rational Experimentation in In ̄nite-Horizon Games

A systematic departure from optimizing behavior is incorporated into a notion of equilibrium in in ̄nitely repeated 2£2 games. In \Experimental Equilibrium", players test (excatly once) every threat that causes optimal response to be non-myopic. Otherwise, they adhere to best-replying. I derive necessary conditions for mutually non-myopic behavior to be sustained in equilibrium. These conditions strongly restrict sustainable payo®s and play patterns. the equilibrium concept is extended to a larger class of in ̄nite-horizon games. ¤I worked on this project while a research fellow at Nu±eld College and a member at the Institute for Advanced Study, Princeton. I am deeply grateful to both institutions. I also thank Dirk Bergemann, K ̄r Eliaz, John Genakopolos, Shachar Kariv and Ariel Rubinstein, as well as seminar participants at Yale and Princeton, for helpful comments.