The authors investigate an international monetary business-cycle model in which agents face monetary policy processes that incorporate regime shifts. In any given period agents cannot directly observe the policy regime, but instead form beliefs that are updated via Bayesian learning. As a result, expectation adjustment displays inertia that adds persistence to the effects of monetary shocks. Monetary policy process for the U.S. and an aggregate of OECD countries are estimated using Hamilton's Markov-switching model. The authors then solve and calibrate a version of the model and examine its quantitative properties.