The Great Recession has shattered the consensus on the benefits of capital account liberalization. Capital account controls have been introduced in several countries and have even been supported by the International Monetary Fund. In this paper we investigate whether capital account policies in the post-Bretton Woods era can be explained as a process driven by learning by policymakers, who update their beliefs on the basis of their own experience and of the policies adopted by other countries. We emphasize the impact of financial crises on the learning process. The learning model developed in the paper explains more than 90% of the variability of capital account policies. We find that over time beliefs about the growth effects have changed slowly and not smoothly from negative to positive. However, at the outset of the Great Recession beliefs on the positive growth dividends from capital account liberalization were still affected by a significant degree of uncertainty, which suggests that reversals in external liberalizations in the aftermath of the Great Recession are consistent with rational learning by policymakers. Finally, in evaluating the potential benefits and costs of capital controls in a given set of countries, contagion effects through changing beliefs of other countries should be taken into account.