Cross-country evidence suggests that during recent years a large fraction of developing countries seem to began to overcome fear of oating, i.e., a lower relative volatility of exchange rates to monetary policy instruments. To explain this trend, we build a model that describes the behavior of Central Banks in developing countries under uncertainty and fear of misspecication about the eects of exchange rate depreciations. The Central Bank is uncertain about two sub-models which dier in that exchange rate depreciations can cause output either to expand (textbook eect) or contract (balance sheet eect). Optimal policy within the second sub-model is consistent with fear of floating. A feature of fear of oating is that, by preventing sizeable exchange rate swings, Central Banks could loose valuable information useful to distinguish among models. We describe how the Central Bank's the evolution of the prior depends on the optimal policy and viceversa. We conclude that the trend towards less fear of floating may not be explained by Bayesian or robust policies because it would have been too quick to explain the data. However, if there was a parameter change affecting many countries during the early 2000's, the model generates the observed pattern.