We propose a framework in which expectations have a rational and a learning component. We describe a solution method for these frameworks and provide an application to the Volcker disinflation with the New Keynesian model. Although the model with rational expectations does not seem to account for this episode, results improve when a small and empirically plausible proportion of private agents are learning. The learning component is argued to be more robust and plausible than the rule-of-thumb expectations present in the hybrid Phillips curve.
