A unified Bayesian framework for the econometric evaluation of dynamic stochastic general equilibrium (DSGE) models is presented. The evaluation is coherent under misspecification, that is, low posterior probability of all DSGE models in a candidate set, as well as no misspecification. The framework encompasses many of the existing evaluation schemes as special cases, including Kydland and Prescott's (1996) informal calibration and the traditional macroeconometric approach of judging models according to their ability to track and forecast aggregate time series. A detailed illustrative application of the framework to a standard cash-in-advance model and a liquidity model is provided. The models are evaluated according to their predictions of co-movements between output growth and inflation, and responses to discretionary changes in the growth rate of money supply.