This paper analyzes how the formation of expectations constrains monetary and fiscal policy design. Economic agents have imperfect knowledge about the economic environment and the policy regime in place. Households and firms learn about the policy regime using historical data. Regime uncertainty substantially narrows, relative to a rational expectations analysis of the model, the menu of policies consistent with expectations stabilization. When agents are learning about the policy regime, there is greater need for policy coordination: the specific choice of monetary policy limits the set of fiscal policies consistent with macroeconomic stability - and simple Taylor-type rules frequently lead to expectations-driven instability. In contrast, non-Ricardian fiscal policies combined with an interest rate peg promote stability. Resolving uncertainty about the prevailing monetary policy regime improves stabilization policy, enlarging the menu of policy options consistent with stability. However, there are limits to the benefits of communicating the monetary policy regime: the more heavily indebted the economy, the greater is the likelihood of expectations-driven instability. More generally, regardless of agents' knowledge of the policy regime, when expectations are anchored in the long term, short-term dynamics display greater volatility than under rational expectations.