This chapter builds and estimates a macroeconomic model that includes learning. Learning was introduced in the model by assuming that agents do not believe in time-invariant inflation targets nor in constant equilibrium real rates. Given these priors, the optimal learning rule was derived in terms of a Kalman gain updating rule. The results show that including learning improves the fit of the model independently of the type of information included in the measurement equation. Although learning models improve on the rational expectations models, they are not fully satisfactory. Autocorrelation in the errors was found to be significant. Introducing learning in a standard New Keynesian model generated sufficiently volatile stochastic endpoints to fit the variation in long-maturity yields and in surveys of inflation expectations. The learning model, therefore, complements the current macrofinance literature linking macroeconomic and term structure dynamics.