Bounded reasoning about rationality can have important implications for behavior. These bounds are typically viewed as an artifact of limits in the ability to engage in interactive reasoning, i.e., to reason through “I think, you think, I think, etc …” However, in principle, these bounds need not be determined by limits in ability. This paper develops a novel identification strategy to show that bounded reasoning about rationality is not determined by limitations in ability. It goes on to show that non-degenerate beliefs about rationality can be an important determinant of behavior. This has important implications for out-of-sample predictions.