Â鶹Éç

In many settings, the outcome an agent obtains depends on the decision taken by other agents. When analyzing such settings, we are often led to consider hierarchies of beliefs which specify what an agent thinks, what he thinks that others think, what he thinks that others think he things, and so on. Because such hierarchies can be difficult to work with, it is common to use a recursive description due to Harsanyi (1967-68). However, in the important case where agents are boundedly rational, it is unclear whether this recursive description is sufficiently expressive. I provide sufficient under which the recursive description does not impose any restrictions on the types of reasoning that can be modeled.