This paper deals with the problem of specifying a general learning model, the rationality of which is not situation dependent. I propose a very general model of adaptive learning suitable to study learning problems in games with imperfect monitoring, such as extensive form games. In this context...